Recording natural soundscapes
— Some experience-based tips

Recording over sea from Willapark headland near Boscastle — Recording over a wide sea panorama, using a Røde Dead Kitten windshield on my Sony PCM-M10 recorder, which is on a Hama Mini Tripod.

Preliminary Notes

In June 2012 I bought my first sound recorder — a Sony PCM-M10. As a complete novice, I set out to make my own natural soundscape recordings, learning as I went. In reality, though, I had already listened to many natural soundscape recordings by other people and become aware of the issues that many of them had addressed or needed to address, and thus I had already learnt quite a lot in 'armchair' mode before I ever got my first recorder. In any case, as a (classical music) composer I was already well experienced in thinking carefully about what I was doing when setting up interestingly balanced and satisfying configurations of sounds, so really natural soundscape recording was just an extension from what I'd already been doing some years before.

'Expertise' is a relative thing, and I prefer to keep clear of the term 'expert' relating to me or indeed others; I simply do the best I know how or can work out how, and learn from my 'errors' and unexpected or indeed unwanted outcomes, and, in the process, have usually obtained recordings that have greatly pleased me (and indeed others if my usually 5-star ratings in Freesound.org are anything to go by).

Anyone else, if they're willing to use their 'grey matter' and think about what they're doing, similarly could quickly get making excellent natural soundscape recordings, without significant recourse to supposed 'experts'.

What I'm presenting below is NOT intended to be regarded as any sort of comprehensive guide or handbook, but simply some tips that some people (mostly but not all novices) may find helpful — that's all.

Some really basic considerations

Be as clear as possible about what sort of sounds you want, and their perspective and relationships to each other within a recording. Then you can be purposeful in choosing recording sites and recorder / microphone placements. This obviously improves with experience, so you help yourself greatly by auditioning your recordings critically as though you were somebody else — a complete stranger — coming to each respective recording for the first time (Now, what might I do differently if I were this recordist?, etc….).
Microphones. For natural soundscapes, generally a microphone setup with a wide acceptance angle is best, as this gets a really spacious sound-stage in any recordings. The Sony PCM-M10 internal microphones are not only of excellent quality (to a point) but also give wide coverage*, so if you want to travel light as I do, with a minimum of additional equipment, this model has a lot to recommend it for natural soundscapes. Naturally, for 'solo' recordings of birds, or interviews with people, that recorder would need an external microphone of suitable, more directional type to be plugged in.

* Unfortunately, the mics are more or less omnidirectional, which means they can give a well-defined stereo image only if there is a suitable acoustic barrier between them, which there isn't, and couldn't be in such a small recorder. For my critical comments about their stereo image, please see my comments on the PCM-D100 much further below.

Another consideration is the type of microphone configuration. For natural soundscapes one thing to avoid is the crossed, X/Y configuration, such as you would find in the current Zoom recorders. The problem with that configuration is that it's used in an attempt to eliminate the natural phase difference that your ears would hear from any specific sound source that isn't exactly central in the soundstage. That may be good for making mono recordings or subsequent mono reductions, but it would actually greatly reduce the natural spaciousness and realism of a three-dimensional soundscape such as we're interested in here, and in particular would make listening with headphones a less spacious and natural-sounding experience than it could be.

Ideally the microphones should be separated by the normal distance between a person's ears, and indeed with an object the size and shape of a human head filling that space, with the microphones pointing at a similar angle to that of our own ears. In practice, of course, most of us wouldn't be wanting to cart around an artificial head, and indeed I, for one, cannot even justify the extra weight and space taken up in my pack to be carrying just the requisite two external microphones. However, even with the internal microphones of my little Sony PCM-M10 the little bit of separation between those would be a vast improvement on the 'same axis' X/Y configuration of the Zoom models and no doubt some other makes of recorder.

On the face of it, Ambisonics would be the best setup of all — though, having looked at the configuration of Ambisonics soundfield microphone setups, I see that they don't separate the individual microphone units within the overall capsule, so they'd most likely lose something important that a straightforward stereo setup gives, because they would be unable to record the phase difference that we hear from non-central sounds.

I was impressed by the great spaciousness and apparent realism that is achieved in many people's recordings made using the M-S (mid-side) microphone arrangement, and was hankering after doing this myself (necessarily using external microphones). However, when I researched further I found that, as I expected from my own reasoning, M-S encoding simply doesn't make sense if you want an accurate representation of what you heard. The M-S method actually gives you a FALSE, reconstituted impression of stereo, which has lost a lot of the original stereo details (because it doesn't record left and right separately).

I also found that recordings that were converted from M-S into stereo were nonsensical when listened to with headphones. Whereas the stereo image tended to be bigger than the space between the speakers (good!), through headphones the stereo image was remarkably two-dimensional, centred bang in the middle of one's head! Impressive, yes, but a grotesque distortion of what the actual recorded subject sounded like 'in the flesh'. M-S would be a valid system ONLY if instead of a bi-directional mic for 'side', a wide-angled proper stereo pair were used. Then, post-processing, one could adjust the balance between the stereo image and the central mono image.
Windshield. I carried out a comparative test between four makes of suitable type and size, and came out with a clear winner, the Røde Dead Kitten. It is essential to use such a windshield when recording outdoors, because even small breaths of wind cause microphone disturbance, at least for the recorder's internal microphones, and a very modest wind indeed will very readily cause gross clipping of the waveform — i.e., overloading and consequent distortion.

Using the Dead Kitten, I find that I rarely get intrusive microphone wind noise with wind strength below force 5 on the Beaufort scale. However, I note that for a given wind strength I get significantly more disturbance with the recorder facing more or less into the wind, so I seek to keep the recorder's back to the wind as much as reasonably possible. However, how much wind noise is considered acceptable is going to differ between recordists and of course the types of content that they're recording, so you need to find out for yourself how much wind is (un)acceptable for your purposes.
Correct significant frequency response anomalies with a graphic equalization curve in a suitable sound file editor. I use a batch operation ('effect chain') in Audacity for all my new recordings, using a saved graphic equalization curve that corrects for a slight muffling of the recording caused by the windshield. I carefully worked out the curve to use, assisted by the plug-in program Span from Voxengo, which is the best free graphical frequency spectrum analyser that I have yet come upon.

However, please don't fall into the trap that I did, of actually correcting for frequency response peaks or troughs in your own listening setup, blaming those on the recorder! The need is to use the most accurate (as distinct from pleasing-to-listen-to) listening gear. The vast majority of 'hi-fi' speaker systems are configured for some sort of 'pleasing' response, as were my very nice Castle Harlech speakers. You need a linear response, which is approached by genuine 'monitor' quality listening gear.
My own correction curve for use in Audacity on all new files to compensate for the slight muffling caused by the Røde Dead Kitten windshield. The data needs to be inserted in a file called EQCurves.xml, which in my case (in Windows 8) is in the path: C:\Users\Philip\AppData\Roaming\Audacity\. Change the curve name to whatever is most helpful for you to identify it. Please note that this may need modifying for your own purposes, or indeed a totally new curve might need to be constructed; it's simply what has worked best for me so far. For other sound file editors, most likely you would need to convert the format of the data for them, and I suspect in most cases they wouldn't recognise such precision.

The first data on each line is the frequency in Hz, and the second is the amplitude change setting in dB. Actually I very much doubt whether any of the decimal places beyond the first are significant for practical purposes (though the second might be significant at the very bottom end), so I'm a bit mystified as to why the format uses 12 places!

Please note that this is a revised version, uploaded on 1 March 2014, for I've removed the part of the curve that had actually been correcting for my Castle Harlech speakers and not genuinely for my recorder as I'd been assuming up to just a few days before.
(please note also that the configuration lines given below cannot display properly on narrow screens)

<curve name="NewFiles-Deadkitten> <point f="7000.000000000000" d="0.500000000000"/> <point f="7969.000000000000" d="1.000000000000"/> <point f="10000.000000000000" d="3.000000000000"/> <point f="11108.000000000000" d="3.700000000000"/> <point f="12000.000000000000" d="4.000000000000"/> <point f="15000.000000000000" d="4.000000000000"/> <point f="21000.000000000000" d="3.500000000000"/> <point f="22194.000000000000" d="3.800000000000"/> </curve>

That equates to the following curve:
Transferring the above curve to WavePad — In 2016 I upgraded my copy of WavePad and found that at last that program stores custom EQ curves, which it didn't when I set up my new-soundfile processing procedures. It's a bit awkward to transfer the curve data, because WavePad stores it in the Registry. However, it wasn't really difficult to do, so that I didn't have to set up the curves from scratch in that program. I note, however, that WavePad takes much longer to apply an EQ curve to a file than Audacity does. On the other hand, it's conceivable that the slower speed might be associated with using a better, less sound-degrading process than the Audacity uses, so I wouldn't go out of my way to use the faster option just for the sake of the speed.
Be prepared to embrace the unexpected and serendipity! In my experience, particular field recording sessions often don't come up with what I thought I was after, and I ended up, often quite dubiously, recording something else instead, which I had not previously thought to record. More often than not, such recordings have turned out to be real gems. If I had simply stuck to my original intention, in most of those cases I'd have returned empty-handed and with a sense of failure, whereas, with my more open-minded approach, each recording outing has been something of a joyful adventure, in the vast majority of occasions providing me with something new and wonderful.

On my latest outing while writing this note, I set out to a rugged bit of far Cornish coast path, aiming to record seals on a rock island known as Seal Island — but the seals weren't performing. Instead, just a little further along I came upon some fantastic blowhole activity at the foot of the cliffs, and got some beautiful and indeed quite spectacular recordings of that, which I hadn't been expecting at all.

Placement for optimal sound balance…

What is good balance anyway? Generally speaking, there is no single 'right' answer as to what constitutes 'correct' balance in a recording, because it all depends what you are after. What may look to be a 'failed' or poor recording from one perspective or for one particular purpose may actually be a brilliant recording from a different perspective. For example, when I set out to record wind chimes high up in the Teign Gorge on 30 January 2013, I found that a gale was blowing — far too strong to record in, and where I found some shelter the chimes' sound was still often drowned out by the roaring of the gale through the trees nearby, so I was getting to feel a bit despondent about that.

But then it dawned on me that actually chimes recordings in that workably sheltered situation would be wonderful and spectacular recordings of a gale in the valley woods, with wind chimes going in and out of focus as the roaring gusts came and went — and indeed I did persist and make a series of recordings, which actually turned out wonderful and spectacular.
Different perspectives on a location — In many situations it can be worthwhile making more than one recording of a particular soundscape, with different balance / perspective — in particular, some element of the soundscape more to the fore, or less so. I bought a second recorder so that I could make a concurrent pair of recordings with different perspectives, or indeed, if far enough apart, one with wind chimes and the other without.
Sea hissiness — I have found that sea or other very 'active' water with a strong white noise element in its sound is usually best not recorded at all close, unless there is some form of screening that subdues the hissiness, which latter is hard on the ears for extended listening, and which obscures all sorts of details that one would normally want to hear in a natural soundscape, including bird sounds. This is NOT something you can usefully correct afterwards, because you would simply muffle the overall sound, which isn't at all what to aim for as it would lose detail and impart a dullness to the recording. The point of shielding from the main part of the sea hissiness is that you can then hear a whole lot of details that are otherwise masked by the hiss, and those details need all their high frequencies intact in order to be heard properly.

So, if I'm recording sea sound, usually my best recordings are a bit back from the edge of a clifftop, just enough to keep the recorder out of direct line of sight of the noisy bits of the sea. Generally that does not lose detail in the sea sound unless the recorder is particularly heavily shielded from the sea, and indeed actually greatly increases the detail because of the subduing of the obscuring blanket of hissiness. However, if the clifftop from which one is recording is quite high, one's distance from the sound source may well be enough to attenuate the hissiness enough anyway, so that shielding could then be counter-productive (but again, depending on exactly what one is after).
Manage the relative distance from the recorder of different specific sound sources. — When I'm recording more than one set of wind chimes (always in a natural soundscape), I find that I need to keep aware of which sets of chimes have the more penetrating and 'carrying' power, for they need to be placed further from the recorder than those with a more genteel sound. Thus my Music of the Spheres Gypsy chimes need to be placed well back from the recorder, and, at the other extreme, my smaller Woodstock chimes need to be closest — though with rather an exception for the Chimes of Mars, which have a particularly shrill and penetrating tone, and thus need to be placed just a bit further back than the other small chimes.
Think about the stereo soundstage, and position the recorder to get the best out of, e.g., birds that seem to be answering each other across a valley or between different trees, or simply flying about in a flock. A flock of jackdaws or brent geese having a flyabout and coming overhead is a wonderful sound in a recording, and I've found choughs particularly delightful because of their aerobatic flying and their really zingy calls from one to another.

Authenticity and informality

Edit out or keep in? No rule nor 'received wisdom' can inform you properly about this, for exactly what you edit out or keep in is bound to depend on your particular intentions for the 'finished' recording. However, I do get the impression that there are a lot of recordists out there who appear to be 'expert' and very tech-savvy, and yet who upload to sites like Freesound.org remarkably rough and 'unfinished' recordings. A good start is to cut out the bits at beginning and end (and anywhere else) where you yourself have disturbed the recording. That includes not only 'handling noise' but also any of your body noises, and sounds made by your camera when photographing the recording setup.

For my own natural soundscape recordings, generally I cut out all bits with people's voices, dogs barking, aeroplanes, farm machinery, intrusive sound of people's footsteps passing by, and some motorboats out at sea. Sensible flexibility greatly helps, though, and I do sometimes leave in the odd very quiet distant aeroplane, which can sometimes add a little extra perspective to a peaceful natural soundscape, and I tend to regard reasonably distant and quiet motorboats in that sort of light and keep them in. I always regard bumblebees and other insects flying by as a very welcome part of the soundscape, but I edit out all but the very quietest of the little clicks I often get, caused by small hard insects — I think mostly beetles — colliding with or falling onto the recorder.
The place of wind sound in natural soundscapes — As already remarked, microphone wind noise, and indeed the sound of wind in the trees or whatever, can be a very effective and pleasant part of a soundscape — the microphone wind noise, if softened by a furry windshield, often sounding like the wind in your ears and thus improving rather than detracting from the recording's authenticity of effect.

While I understand that recordists will vary widely as to how much microphone wind noise is acceptable to them, I do think that many recordists of natural soundscapes tend to go too far, and actually diminish their recordings, by always using impressive-looking big Blimp furry windshields, which undoubtedly give a 'professional' look to their operations (hey, a bit of social status, what!) but actually remove something of the authenticity of their recordings — no more so than when they are actually recording windy soundscapes! I myself have found most wind recordings that I have downloaded remarkably unsatisfying, because their lack of any microphone wind noise makes the recorded wind sound seem remote and almost artificial, lacking in a proper sense of perspective.
Beginnings and endings — Abrupt beginnings / ends, typically also with some handling noise and other disturbances, are a real turn-off for the listener, and inform him that the recordist doesn't care a toss (whether or not the latter is really exactly the case!). Clip off from your recording all audible traces of your handling and retreat from or approaching the recorder, and then, use the Fade In or Fade Out functions in your sound file editor to give a smooth beginning and end.

I personally settled on a 3-second fade-in at the beginning and a 6-second fade-out at the end, though after a year or two I reduced the fade-in to one second, finding that much more satisfactory, but have tended to make the fade-out just a little over 6 seconds. Raw, 'unfinished' beginnings and ends may be authentic in the sense that they're part of the record of the recording session, but they're NOT part of the natural soundscape that you're recording! So, a genuinely authentic recording of a natural soundscape requires that you remove as much as possible that gets in the way of experiencing the full authenticity of that soundscape — and that certainly includes excluding anything like rough beginnings and ends, which draw attention to you and the recorder.
Add at least 3 seconds' silence to the end of each recording. Although this isn't essential, generally it's good practice. That, then, would ensure that any CD compilations of your recordings have at least a reasonable 'settling time' gap between tracks, even where you've neglected to set a suitable 'pre gap' duration in your CD burning software. Still longer gap times can be very helpful for the best listening experience. Indeed, for classical music CDs, my preferred gap times between tracks are 6 seconds between movements of a work, and 10 to 12 seconds between whole works. Running tracks together end-to-end may be fashionable, but it doesn't make for aware and healthy listening, and makes the whole listening experience relatively dull and fatiguing.
Keep yourself out of the recordings! — In other words, once you've started the recording, get well away from it, and preferably behind it so that any sounds you do make would be less likely to be picked up.
Make long recordings! — Whenever practicable, make recordings of half an hour or longer. Indeed, most of the time I regard 30 minutes as being a minimum target duration of the final, edited version of a recording. Because of the bits cut out during editing, that means that I aim to record for 40 minutes to two or three hours, or occasionally even longer, depending on how much I think I'm likely to need to cut out.

Long recordings of natural soundscapes enable the listener really to get into the experience. Short recordings never have a strong feeling of authenticity about them, because their very shortness draws attention to the recording, and indeed to a CD or playlist compilation that contains it, rather than to the soundscape itself. I generally aim for whole-CD recordings of an hour upwards, or two-per-CD recordings of 30 to 45 minutes each.

It's always best to record for longer than you're likely to use in a CD or other application, for you can cut the result down to your required length. If you recorded only for 5 or 10 minutes, then you can extend that only by the horrendous practice of looping. Anyone who does the latter has no genuine feel for nor understanding of nature or natural soundscapes, and their work would be best avoided.
Think in terms of recording authentically, in order to get the most satisfying results — i.e., resist the temptation to produce mixes. If you want to record particular combinations of things, then take the extra bit of trouble to find situations where you can get those particular elements right there together in a single recording.
'Authenticity' can be rigidly overdone, like so-called 'political correctness'! The skill to cultivate is that of treading the fine line between a rough and ragged recording that is unfocused / 'scattered' and containing distractions on the one hand, and an over-polished, over-sanitized one on the other, which has lost something of its one-time vibrancy and 'life'. It takes considerable awareness and sensitivity to be able to tread that line at all well. Some people are able to cultivate that quality in themselves, while others appear to be programmed to work by following rules and instructions, and are completely closed to developing any really aware and sensitive approach. Many supposedly 'professional' nature recordings are very poor from my perspective for this reason.

Various practical considerations

Guarding the recorder(s) while making recordings — Because the majority of my recording situations were beside or at least close to fairly well used footpaths, I found it highly advisable to be quietly lingering on the footpath, reasonably near the recorder, though not necessarily in direct sight of it, constantly looking out for approaching people. On spotting anyone approaching, I'd quietly hurry towards them, signalling to them with a finger on my lips (with a lot of smiling to communicate friendly intent and that I'm not telling them off!), and then, when close enough, to explain in a whisper about the recording being made just ahead and the need for the quietest possible passage along that stretch.

Generally the majority of people are co-operative in their intent, though probably about half of them are, sadly, so dim that they clearly haven't understood what I'm going on about and are still noisier than they need be (some even continuing yacking as they pass the recorder), and inevitably I get the odd individuals or groups that clearly resent any request to modify their behaviour, even for just a few hundred metres along the path, and go grumpy-quiet, heavily stomping on their way.

As a rarity, I've had the odd ones who at once went into argumentative / abusive mode on the basis of "It's a free country and we have a right to do what we want! Who are you to tell us what to do…?!". In the latter case all I can sensibly do is shrug my shoulders and turn my back on them, withdrawing from engaging with the particular people, and just accept that I've got a bit to cut out of my recording there.

I have a particular challenge as I'm usually making two concurrent recordings, generally some distance apart, so that one recorder is quite often more or less unguarded. That can lead to disturbances where walkers without the sense they were presumably born with actually go up to the recorder, talking loudly to each other about it, and even pick it up to look at it more closely to try to work out what it is!

I thought that attaching a conspicuous notice by the recorder would work, but my experience so far of doing that has served as a warning as to how irrational perhaps as much as 50% of 'serious' walkers tend to be in responding to something unexpected that they encounter beside the track. I'm not saying that notices are completely ineffective, but rather simply that it would be unwise to put a lot of trust in their effectiveness. Guarding is still strongly advisable if people are likely to pass by within even fairly distant sight of the recorder(s)!
Recording level — For the best sound quality and most natural effect in natural soundscapes, always record the full dynamic range. That means always recording in Manual mode, and setting a suitably low recording level, to ensure you don't get overloading (waveform clipping). You can always bring the level of the recording up to an optimal level afterwards in soundfile editing software. Keep Auto mode for things like impromptu interviews with people, where recording quality and dynamic range are not so important.

For my own purposes, on the Sony PCM-M10 I most usually use number 4 on the recording level wheel (sensitivity always set to High), increasing that to 5 in very quiet situations where no prominent peaks (such as from fairly strong wind gusts) are expected. And of course, conversely, if there is at least a reasonable possibility that I might get clipping with the level set at 4, then I turn that down further, occasionally as far down as 3.

For a thunderstorm (which hardly ever happens for me in Exeter), I'd expect to use level 3, though experience of some close earth strikes might persuade me to set it even a little lower if those strikes caused clipping. You might think that auto level control is the answer for thunderstorms — but actually that would simply give you a degraded representation of the real experience, so it's not for anyone who is seeking to portray nature realistically.
The biggest dynamic range that I've yet had to deal with was from the sea violently booming in a presumably fairly small cave at the end of a cleft in a sort of rock 'apron' extending from Tregerthen Cliff, near Zennor, Cornwall, UK. I think something of a megaphone effect made the booms so intense for recorders placed near the edge of the cleft. In that case I finally got all the booms without distortion by spending several minutes initially watching each recorder's level display, and adjusting the level so that even the loudest booms during that period didn't go above about -9dB.

The levels arrived at were something close to 3 on Low sensitivity — the lowest setting I had ever intentionally used. Then, back at home, once I'd nipped of the beginning and end of each recording to remove handling noise, I then normalized it to 100%. That still left them at an absurdly low level, however, and for playing at a realistic level the playback volume needed to be increased by a whacking (and anti-social) 12dB (the loudest booms then made the concrete floor of my flat slightly shake, and the neighbour immediately below me come up effectively to make an oblique complaint.
For the finished and 'polished' recordings afterwards, I generally do NOT use the Normalize function (i.e., in the sound file editor) — at least, unless there is a particular need to set the biggest peak to a specific amplitude. The 'normalize' function does not and cannot know what is the right listening level for any particular recording. Therefore one needs to use the Amplify function instead, and find out by some tests just what on-screen core waveform size corresponds with the best listening level for that particular soundscape, so that it will sound life-like when the playback volume level is set to a particular standard setting.

If you use the Amplify function to increase rather than decrease the level, then you need to check that no peaks have got clipped. If they have, undo the amplification and repeat with a smaller increase, or normalize to almost 0dB (100%). Or, just possibly in certain situations, allow the particular clipping if there are no more than a very few such clipped peaks and you cannot hear the distortion they have produced — but always regard the allowing of clipping as an exceptional rarity for some special reason, as it does adversely affect the sound even if you don't notice it.

My own yardstick is to make all my recordings play back at something about the original subjective sound level when my hi-fi volume setting is right for a reasonably life-like reproduction of wide dynamic range symphonic music for full orchestra. Thus the vast majority of my recordings can be played at the one playback setting, and at least potentially without need to readjust between (at least classical) music and my own recordings. The exceptions, with particularly large dynamic range, include a few of my booming cave recordings, as noted above, and also the odd close-up blowhole recording and certain thunderstorm ones.

Finally, a caution about what appears to be a bug in the PCM-M10, which I have noticed in both of my recorders. You would surely reasonably expect that if you set your recording level well and avoid clipping that you can see in the waveform, you won't get overloading and distortion. However, I have found that peaks of low frequencies can cause overloading distortion without the waveform of the saved file having reached clipping point, and sometimes indeed the offending peaks being of really modest amplitude.

Also I noticed that the distortion (heard as transient partial dropouts, not as the sort of crackle / 'spitting' that I'd associate with clipping) often doesn't correspond exactly with the largest peaks in the particular sound event. I get this generally from wind noise in the microphones and also from the very low frequency thumps and booms that the sea can make on the more rugged Cornish cliffs.

I generally manage to avoid this by ensuring that the recording level is set low enough to give good headroom, and may reduce the level just a bit further where there are strong booms from the sea or it looks as though there may be fairly pronounced microphone wind noise (i.e., in spite of the furry windshield). However, in the latter case often one would want to edit out the very strong peaks of wind noise anyway, so it's less important than with the sea booms, which are definitely wanted.
Recorder placement — If you travel light, like me, you may want to use just a tiny mini-tripod such as the Hama Mini Tripod, as I've done. However, there are issues about placing a recorder so low. For most purposes that does degrade the sound as compared with what you yourself would hear during the recording session. Putting the recorder so low has various undesirable effects. Proximity of the ground narrows the soundstage (i.e., between top and bottom), so to some extent impoverishing it.

It's also liable, especially when on a rock surface, to bring out an interfering bass resonance, which may manifest as some degree of boominess or rumbliness, and in many of my clifftop recordings it has imparted a somewhat cavernous sounding coloration that I didn't hear at the time. It may also actually attenuate very low frequencies that really need to be heard in the recording, such as in the thundering of large breaking surf, or where the sea is crashing against cliff formations and giving very low frequency booms and whoomphs. Also, such a little tripod as the Hama Mini Tripod is difficult or impossible to use where there is more vegetation than a fairly well cropped turf.

Nowadays I have advanced from that situation. Yes, I do still take one or two Hama mini-tripods with me, because in some situations they are my best option, but now I usually carry also at least one Tamrac Zipshot tripod. I actually have two of the full size and two of the 'Mini' size Zipshot — the 'Mini' size actually being what I'd call 'midi', and actually an excellent compromise size for many purposes, though on my clifftop recordings it's still low enough often to cause a bit of cavernous-sounding coloration.

The Zipshot has the advantage over conventional camera tripods, that it's much lighter (i.e., for similar height) and also very much quicker to set up and pack up. Its two main disadvantages are that, being so light, it's relatively easy for the wind to blow it over, and also, its legs are NOT adjustable, so it can be used only on notionally level spots — though if the ground is sufficiently uneven it's quite often possible to find a sufficiently level placing of the legs even on a fair gradient. I have also been using the Zipshot tripods for suspending wind chimes on clifftops where there are no suitable trees or bushes to suspend them on.

Thirdly, on many outings I also take with me one or two Jobi GorillaPods — tiny mini-tripods with rather stiffly flexible legs, which latter you can curl around a tree branch, railings, top of a fence post (if narrow enough) and various other supports. This enables one often to get excellent impromptu recorder placements, and higher ones than one would get from using a normal tripod.

However, there is an issue about higher placement of recorders / microphones, because the higher you place them (including very much the height difference between Hama Mini Tripod and the Zipshot Mini), the more they are liable to catch the wind. You thus need to think carefully about that when considering higher placements.

In 2019, thinking to order another full-length Zipshot tripod, I found that Zipshots are no longer available, except for the odd eye-wateringly expensive second-hand offers. So I ended up buying remarkably cheap lightweight conventional tripods instead — an Amazon Basics 50" one and a Hama Star 700EF, which latter is also 50". Although the Amazon notionally commends itself through being the lighter, in practice I found that these two in their respective travel carriers were almost identical in weight. As the Hama seems to be the sturdier and their prices are almost the same, on that basis I purchased a second Hama and declared the Amazon redundant as I find it more fiddly to set up the recorder on it.

Although these more conventional tripods are somewhat bulkier and heavier than the Zipshots, at least they have the advantage of much more flexibility in use, with their 4-section legs. The Zipshots, having non-adjustable legs, left me greatly limited as to where I could place recorders, and of course their height was fixed.

However, dissatisfied by their restrictively narrow leg angle and horrible leg length adjustment lock tabs (which tend to painfully nip one's fingers and indeed draw blood), I more recently bought much slicker similar-size carbon fibre tripods that allow independent full angle adjustment for each leg, and with quick lock / unlock action screw-type locks, which don't nip fingers and are much easier to operate. They are a little heavier and much more expensive, though.

N.B. The tripod links given above are Amazon Affiliate ones, which would earn me a tiny bit of money for any items purchased via those links.

Tamrac Zipshot tripods in action on
clifftop near Perranporth, Cornwall, UK; standard size with Woodstock Chimes of Polaris, and the recorder on a Zipshot Mini. — Tamrac Zipshot tripods in action on clifftop near Perranporth, Cornwall, UK; standard size with Woodstock Chimes of Polaris, and the recorder on a Zipshot Mini.
**Warning!** — Great care required in order not to lose equipment over the cliff! — In fact on one occasion I did have a Zipshot Mini blown over a clifftop and lost in the sea — fortunately, with no equipment on it!

Recorder perched on tree branch by means of small GorillaPod — Recorder perched on branch by means of a GorillaPod, for recording of birds and wind in the Teign Gorge, Drewsteignton, Devon, UK

Editing of sound files — To cut out sections, e.g., where I want to remove aeroplane, people, dog or wind disturbances, I use the Crossfade function, but there is a catch, because I find it doesn't work quite right for me with the default options. Audacity has a crossfade plug-in, which I have got to work okay, but I find the crossfade facility in WavePad just a bit more intuitive and helpful, so it's this that I use. In both cases, however, there is a weird little practical fact that needs consideration.

The idea is that you select the passage that you want to cut out, and the crossfade function both cuts out what is in the middle and draws the cut ends together, tapering them both and overlapping the tapers, so that theoretically, with similar-enough sound before and after the deletion, the join would be undetectable. You can set the length of the taper, which needs to be a different length in different situations, and, in WavePad, the degree of overlap. In WavePad the default is always an exact overlap, but this can be adjusted by setting the 'Gap' parameter, which by default is set as equal to the overlap length, which to my mind ought to work fine.

In practice I find that using the default exact overlap results in a dip in playback volume during the passing of that overlap, so I always have to change the Gap setting to about two-thirds of the overlap length. This isn't a WavePad quirk but a peculiarity of overlapping the cut ends. I have tried creating crossfades manually, and I found then that I had to draw in the ends to about a third more than the length of the overlap. In the Audacity crossfade plugin, instead of fiddling with the degree of overlap you set a parameter for increased level within the overlap.

Unfortunately WavePad.s crossfade function has been broken in all versions after 9.01, so I've had to stick with that version. Although the developers did eventually get the crossfade itself to work, the 'gap' function remains broken, so for my purposes the whole crossfade function remains unusable, and the developers are nowadays ignoring my repeated protests about that.

Meanwhile, Audacity has lost its original crossfade plugin. It does have a function called 'crossfade clips', but that has fixed parameters for slope and gap, and works well only with very short selections.
Listening facilities for editing / auditioning — For really effective editing work you need real 'monitor' quality playback. That is, it needs to be as accurate as possible, not necessarily what you find most agreeable. That might seem a rather perverse thing for me to be saying, but the reality is that most people (including professionals), despite any beliefs to the contrary that they may hold about their listening to recorded sounds (including music), don't want genuinely accurate rendition, and actually want particular distortions and colourations that make the sound more impressive or 'agreeable' in some way than the original.

Even if you want to end up with such a 'rose tinted spectacles' version of your recording, it's better to do your first auditioning and editing based on accurate rendition, for that way you have real control over your end product. You can always adjust it afterwards to be the sort of sound that you were after — i.e., if you feel that you really must! However, if you work this way you might well become better attuned to accurate rendition of the original sound, and, having cultivated some better awareness, let go of that obsession with altering the sound to fulfil some actually poorly founded criteria that you had been applying in the past.
A cautionary story — beware!
For ages I'd been struggling along with my decidedly non-hi-fi Creative computer speakers (which nonetheless did a tolerable job as computer speakers!), but I found that I was missing things that way. So eventually, after some careful online research, I obtained a pair of Sony MDR-V6 headphones (and indeed at a remarkably reasonable price for real quality headphones), because I thought there was no way that I'd be able to have full frequency range monitor quality speakers either side of my computer.

I should say that basically headphones and I don't mix — I detest them all — but I really had no other option in order to have accurate sound there at the computer where I'm doing the editing work, with the recording's waveform in front of me in WavePad or Audacity. The particular headphone model is absolutely spot-on for sound in terms of unforgiving accuracy, at least frequency-wise; I'd never before heard such neutral, uncoloured sound from any playback device. It was just unfortunate that they were physically stressful for me to wear, and I was always greatly relieved whenever I took them off.

Because of inherent limitations in listening through headphones, I regarded it as essential to give a full post-editing auditioning to each of my recordings, and for that I was faithfully using my proper hi-fi system. I realized that my Castle Harlech speakers wouldn't pass as full 'monitor' quality, but I thought they were nonetheless very good indeed as far as domestic hi-fi cabinet speakers go, and indeed more neutral and apparently more accurate than many a much more expensive speaker system that I'd heard, including some that were claiming or at least implying that they were of supposed 'monitor' quality (nearly all with an unnaturally aggressive, over-bright treble). Anyway, listening to the recordings like that, with a wide stereo separation, gives a much better impression of the original soundstage than the always very artificial effect of the headphones, and thus is the best indicator of what people could hear when playing the recordings on their own hi-fi setups.

However, then I bought a pair of Audioengine A2+ speakers to replace my old Creative computer speakers. Their performance was so startlingly accurate that I then added in an Audioengine S8 subwoofer. The result was amazing for accuracy, putting the Castle Harlechs to shame, and rendering the Sony headphones redundant. It then became clear to me to my horror and disgust that I had been applying a completely unnecessary equalization to all my recordings to compensate for the bass hump supposedly of my recorders. That bass hump was clearly belonging to the Castle speakers, and thus should not have been corrected for in the recordings, which all thus had been significantly harmed with a fairly draconian bass cut.

Unfortunately, 'undoing' that cut in a test file by applying an inverted version of the original 'correction' caused a jarring roughness in the bass, so it became clear then that for all my recordings where really low bass was important (mostly sea recordings) I'd have to take new copies of the archived unprocessed original recordings and process / edit them all over again, omitting any supposed 'correction' to the bass — an absolutely huge task.

Although this was tiresome for me, at least I had the satisfaction of putting right something that I'd been doing wrong, and ensuring that some incredibly beautiful and in various cases powerful recordings were made available without the previous serious 'dent' in them. Also, I put my Castle Harlech speakers up for sale, with plans then to replace them with a pair of Audioengine A5+ speakers plus another S8 subwoofer, for really accurate listening in the room at large. In the event, however, I went beyond those and ended up buying a significantly higher-grade setup using passive main speakers with a high-grade amplifier, and a pair of higher-grade subwoofers than the so-so Audioengine one.

The bottom line, thus, is that 'monitor'-quality listening is really essential for editing and auditioning natural soundscape recordings — at least, if you want any semblance of accuracy in the end product.
Quirks / bad design of the Sony PCM-M10
- The microphone sensitivity switch is a slider switch, which very readily gets reset from high to low sensitivity without one's noticing, when one puts on the furry windshield — the windshield's elastic coming right over that switch. Nowadays, once I have put the windshield on the recorder I then very carefully lift the elastic that is hiding that switch, and check carefully that it's set to High, and change it if necessary. Indeed, because it can be a bit difficult to see the switch position when viewed like that, I even manually push it anyway, to make doubly sure that it's set to High.
  
  Caution! — It's all too easy for that switch to get changed back to Low when you let go of that raised bit of the elastic! I have thus got into the routine of using a particular way of lifting and letting go of the elastic, which minimizes any possible pull towards the switch's 'Low' position.
- Opposite, on the underside, is the playback volume rocker switch, which sometimes gets activated when the windshield's elastic is pressing on it. Even if, like me, you never use the recorder's pretty useless playback facility, it's a good idea to familiarize yourself with the volume control display in the main window, by operating that control, so that when the volume control is inadvertently activated you will know what has happened when you see that particular display. All you need do when that happens is ease the windshield elastic just a little forward on that side, but not enough for it to pop off.
- Another slider switch that can inadvertently get set to trash a recording, though fortunately for me that happening has been very rare, is the one for selecting manual / auto recording level. I did have one potentially major and great recording trashed by this, because, not being used to that happening, I hadn't checked to ensure that the recording mode hadn't been changed from manual. It was only when I was about to switch off the recorder, happy to see that it was still recording after some 90 minutes, that I noticed in the display the dreaded word 'Auto' in place of the level indicators — and I had no time then to repeat the recording. "Oh, shit!", as they say!
- Inserting a micro-flash card is a bit tricky and disconcerting, because the slot for it's actually too large! It is thus difficult to get any real sense of the card properly locating in anything — but actually with both my recorders I found that when I'd got the card roughly located and then closed the cover over it, it worked. But evidently a bit of bad design there, no doubt because of the slot also being compatible with another memory card type. My immediate neighbour, who also recently got one of this recorder model, had exactly the same issue, so it isn't just me being 'funny'!
- If your recorder has an inserted memory card and you have set it to use that rather than the built-in memory, if at any time you do any file management operations within that inserted memory from a file manager on your computer you will find that the recorder at once automatically reverts to using its built-in memory, and you would have to go through its menu system to re-set it to use the inserted memory card. I've thus learnt always to just copy (not move) files from the recorder to my computer (using my file manager FreeCommander), and then, once all the wanted files have been copied over, to do the deletions of the originals through the recorder's own facilities. Then the recorder remains set to use the memory card and not the built-in memory.
- As a considerable rarity I have had a recorder spontaneously stop recording immediately or at least soon after I'd started the recording (though on one occasion the stop was after 56 minutes), so occasionally losing me a recording opportunity that I'd thought I'd got 'in the bag'. At different times it has happened for both my recorders. A helpful protection against more or less immediate stopping would clearly be to watch the recording time display for, say, five seconds before you move away from the recorder to leave it undisturbed.
  
  Later note (2016):
  I eventually got a suspicion that that obnoxious behaviour was a result of a badly designed / made power on/off and hold switch. Its return spring is very weak, so that after a little use I found that the switch didn't automatically return to its centre (neutral) position. I therefore then made a point of always moving the switch back to the centre after operating it, and with that policy in place I've not had a recording spontaneously stop for the best part of two years now.
- The battery level indicator is extremely misleading for rechargeable batteries, even when it has been 'told' (through the menu system) that you are using rechargeable ones. It looks as though such batteries have a very short recharge life, but that is quite a deception! After just a session or two the indicator would be showing apparent low charge, but generally, if the batteries are any good to start with, you can keep using it like that for session after session.
  
  Unless you have a very long and important recording coming up, you don't need to change the batteries until the battery level indicator is flashing, and even then, it's widely claimed, if really necessary you can still record for at least a further half hour before the recorder automatically shuts down. Best, though, to change the batteries at the first opportunity once you've seen the level indicator flashing. Working like this, I have never yet had a recorder run out of charge during a recording — or indeed at all.
- The distortion caused by non-clipping-level peaks of very low frequencies, as already noted under Recording Level further above. I can only surmise that some component in the recorders gets overloaded by the very low frequencies at levels below the general clipping level. Naturally I suspect the internal microphones, which are generally excellent otherwise — though I guess it could be some other component.

The Sony PCM-M10 discontinued — upgrade to PCM-D100?

One of my two PCM-M10 recorders had a rarely-manifesting fault, in which the level of the left-hand channel would be almost inaudibly low for part or all of a particular recording, but by the time this became frequent enough for me to be moved to raise the matter with Sony (in 2016), I'd already been using it for over 3 years, so it was well outside any warranty, and thus all I could do was to purchase a new one. The catch was that then I found that the model had been discontinued, with nothing replacing it. How absolutely crass can a company be, to do such a thing?! The PCM-M10 has been widely recognised as markedly superior to the other generally available small recorders in the same price range, and had become hugely popular. Sony's lame 'justification' (at least, to me) has been that the PCM-D100 is the replacement, but that is just bullshit and complete nonsense.

The D100 is definitely a wonderful piece of work, but in NO way can it be called a replacement for (i.e., simply an updated version of) the M10. It costs more than twice as much, it's more bulky and is more than twice the weight, and it eats batteries more than twice as quickly. If you buy a D100 you aren't buying a reasonably improved equivalent of the M10 but a very major upgrade to a completely different, more professional-type model, as the price would indicate.

In March 2016 I finally bit the bullet and purchased a D100 — though recognising that it wouldn't necessarily be used in place of the M10 recorder because of its bulk and weight. Here follow some notes on my observations as I test the D100.

The microphones
- One great down-side — the microphones are MUCH more wind-sensitive than those of the M10. Using a DeadKitten furry windshield, I found that even a really gentle breeze disrupted recordings. This was promising to be quite a headache to attempt to resolve, as, on my hikes, I don't have pack space for a 'Blimp'-type windshield, and it would be difficult fitting in such additional kit as external microphones — and in any case a sufficiently good quality stereo microphone setup looks like costing more than this recorder did — not good for me on my meagre State pension income! Also, the recorder isn't directly compatible with most quality plug-in microphones — of all stupid things! You'd need an external phantom power unit to operate most of them, for a start.
  
  I tried experimenting with the Rycote Mini Windjammer and the Gutmann furry windshield for the PCM-D100, but they both proved to be insufficiently effective, so I was still unable to use the D100 in most situations where I could previously use the M10.
  
  Eventually I started making progress when I tried two of the appropriately-sized Movo windshields, which were both rectangular-box-shaped. The furry one (model WS-R30) gave as good a performance as the original version of the Røde DeadKitten, which was a small advance, though still very much inadequate for the D100, while the non-furry one (model WST-R30) was poorer in performance 'as is', but suggested to me a fairly radical solution to the whole issue.
  
  The point is that the latter model, because of its well-defined shape, maintained some still air space around the front of the microphones — which is supposedly an important ingredient of a more effective wind-protection setup. Effectively I could use it as a miniature 'Blimp'. I then discovered Windcut, a UK maker of particularly effective furry windscreens at a fraction of the price of equivalents from the big-name makers, and had them make a custom 'furry' to fit over the Movo non-furry.
  
  I then used that combination through the latter half of 2016 to late February 2017, at which latter point I finally woke up to my need to investigate the poor (rather screamy and abrasive) quality of the treble in all my sea recordings made in that time. When I did investigate I found that the Movo non-furry 'inner' had been imparting a distinct peak to the treble, centring at around 5KHz. The issue wasn't a quirk of that specific make / model, but rather, of that semi-rigid squared design (regardless of whether it was furry or not), which would suffer similar internal resonances whatever the make and model name. So I had to drop the use of those Movo shields like a hot brick and start trying out using furry windshields as candidates for the 'inner', over which the larger custom Windcut furry would fit.
  
  I worked out a temporary solution in April 2017 and thoroughly field tested the following month, with what seemed to be really good results, which would surely be equivalent to using the PCM-M10 with an original version Røde DeadKitten — perhaps even slightly better. I'm still using the custom Windcut 'furry', but am now fitting that over another custom Windcut furry made to exactly the same design but small enough to fit directly onto the D100. Both inner and outer are made, at my special request, with the seam down the middle (top or bottom) and not going round one side as it does in Windcut standard 'furry' for the D100.
  
  However, in the longer term I found the D100 still too wind-sensitive with the two furries, and eventually tried three furries. This does work, but at a considerable performance cost because of the strong EQ curve I had to work out for that. This brought to notice a weakness of the D100 that I haven't seen mentioned anywhere. That model is highly regarded with regard to signal / noise ratio, but actually its noise floor rises considerably towards the high treble. So, the EQ to correct for three furries is inclined to result in noticeable high treble hiss in recordings of very quiet natural soundscapes, because that correction further raises the treble noise floor.
  
  Indeed, it's more than that. The D100 has a serious and fundamental flaw that affects all very quiet recordings, such as nocturnal or remote moorland natural soundscapes in quiet weather conditions. Its built-in microphones, despite many excellent points, simply have too low an output! Whereas with the PCM-M10 the normal level setting to get the right volume for a natural soundscape was about 4 (out of 10), in the case of the D100 that level is between 6 and 7 (again, out of 10). Because turning up the level brings up the noise floor (at least, in the treble), that means that for the recordings that most need a really low noise floor, you have a s/n ratio seriously below the stated specification for this model.
  
  At an early stage I tried a Rode NT4 stereo mic connected to a D100, but, despite it being somewhat less wind-sensitive (great!), to my dismay it needed a level setting on the D100 of at least 8. Hopeless! I didn't keep it. In any case it had only a 90-degree angle, so would have always had me rather unsatisfied.
  
  In late 2018, exasperated by wonderful night recordings on Dartmoor marred by hiss, I researched the Zoom H6 as a possible alternative (it has almost as good a quoted s/n ratio and lots of very positive features, and might need less wind protection anyway) — but found a user review that complained of exactly the same issue with regard to the output level from its XY mics, indeed his having to turn the level up even more than I have to for the D100. It's thus clear that I just have to make the best I can of the D100.
  
  Basically, in very quiet conditions I have to limit myself to occasions when the wind is light enough for me to use no more than two furry windshields, so I don't have to use so much treble-boosting correction for the windshields' muffling effect. And even then, most likely young people who can still hear well above 10kHz, and especially up to 20k, would still hear a hiss. I've tried reducing the amount of treble correction for the windshield muffling effect, but even a small reduction not only blurs detail, but loses much of the breathtaking depth of the recordings, so it's something I'd not accept.
- That leaves me in a dilemma, because in order to avoid noticeable hiss in really quiet soundscapes I simply mustn't use three furries — but the D100 is so wind-sensitive that my opportunities for acceptable recordings with two or less furries are really few, at least for really quiet soundscapes.
  
  If you want to order the relevant Windcut furries for yourself I suggest you ask Windcut for the custom furry with dimensions to fit directly on the D100 (which is a size they've actually been offering on their eBay store), plus the custom furry to fit over that one, both with seam down the middle (top or bottom), NOT round the side (their default design), and both exactly as supplied to Philip Goddard in Exeter. Then hopefully there would be no misunderstanding. For a third furry, to put over those two, you can ask Windcut* for the custom furry design for fitting over a Movo windshield (which I now wouldn't use or recommend) — again, as they had supplied to me.
  
  * When I checked on Windcut during the first COVID-19 lockdown in 2020, it appeared to me that Windcut must have ceased trading — but subsequently, in 2021, I was alerted to their being present but now calling themselves Windcutuk — though their products are still branded 'Windcut'. Their presentation continues to look as though their eBay site might not be genuine, but I've been in correspondence with one individual who decided to take the risk, and it did turn out to be the same business (including same proprietor name), and as helpful as ever for doing custom-made furry windshields. So I'd say 'definitely genuine and good, but simply not very good at presenting themselves'.
  
  The correction curves I'm now using (in Audacity) are generally as follows (please note that this cannot display properly on narrow screens):
  
  <curve name="2Furries 1qtr + 9dBbass"> <point f="5.000000000000" d="-12.000000000000"/> <point f="10.000000000000" d="9.000000000000"/> <point f="35.000000000000" d="9.000000000000"/> <point f="55.000000000000" d="0.000000000000"/> <point f="700.000000000000" d="0.000000000000"/> <point f="4000.000000000000" d="2.000000000000"/> <point f="10000.000000000000" d="4.250000000000"/> <point f="13000.000000000000" d="5.000000000000"/> <point f="20000.000000000000" d="6.500000000000"/> <point f="24000.000000000000" d="0.000000000000"/> </curve>
  
  That corrects for both the treble muffling and the rather deficient very low bass from the D100; the bass correction can be left out if you don't want that . — And then, herewith the other option I use, where I need to minimize microphone wind noise or other boomy bass frequencies; it's shaped still to let through a fair amount of really low, non-boomy bass, as usually it's only the boomy frequencies that one may need to tone down. Note that no more than two decimal places could be significant, even if that! (Again, please note that on narrow screens this won't display properly.)
  <curve name="2Furries 1qtr 9dBbassCut"> <point f="5.000000000000" d="-12.000000000000"/> <point f="10.000000000000" d="-1.000000000000"/> <point f="20.000000000000" d="-1.500000000000"/> <point f="24.222791908617" d="-9.000000000000"/> <point f="70.000000000000" d="-5.500000000000"/> <point f="175.000000000000" d="0.000000000000"/> <point f="700.000000000000" d="0.000000000000"/> <point f="4000.000000000000" d="2.000000000000"/> <point f="10000.000000000000" d="4.250000000000"/> <point f="13000.000000000000" d="5.000000000000"/> <point f="20000.000000000000" d="6.500000000000"/> <point f="24000.000000000000" d="0.000000000000"/> </curve>
  
  I also use variant curves with different strengths of bass boost or cut, and different strength of the treble correction according to whether I used one, two or three windshields. Those variants are all stored in Audacity's EQCurves.xml file for ready access without having to keep fiddling around with the EQ graph.
- For recordings in which at least the treble is really continuous and unvarying (more so than in sea sound, so this really refers to sounds like a waterfall or venting steam), a separate curve needs to be made — the same but with the treble dB values doubled or even further increased.
- The microphone alignments — only the wide stereo setting is of any use for natural soundscape recording — though I suppose a slightly less-wide coverage could be chosen (though surely not sensibly for natural soundscapes!). I tried the 90-degree X-Y configuration, and recordings that used it gave a very narrow sound-stage, sounding as though half-way towards being mono. Those recordings had depth, yes, but it was like the depth impression of looking into a tunnel, but without the tunnel's reverberation. Basically a highly unnatural effect, which has use only for recording things like musical instruments close-up, or getting solo recordings of people. It sounds particularly bizarre through headphones.
  
  I should point out, though, that such an extremely restricted soundstage isn't intrinsic to the 90-degree X-Y configuration, and I found that with the Røde NT4 stereo microphone, for example, its fixed coincident 90-degree X-Y configuration gave a relatively wide and spacious soundstage, albeit not as wide or spacious as that produced by the D100's wide stereo configuration.
  
  The D100's wide stereo setting gives a particularly wide and spacious soundstage, MUCH more vividly and precisely detailed than that provided by the M10. A long recording in Branscombe Landslip (near Beer, Devon, UK) captured with awesome precision the sea echoes on the cliff towering above, with the movement of the echoes precisely captured just as I had heard it — whereas all the M10 could come up with there was a general haze of sea sound, with only the odd vague hint of movement across the soundstage. With the D100 you really got the impression of hearing the cliff towering above you.
- The strongly directional quality of the microphones can result in quite unnatural sides of the soundstage when you're listening with headphones, for sounds coming from hard left or hard right are not heard at all (at least noticeably) in the other ear. The effect is quite disconcerting and distracts your attention from the soundscape itself, at least until you've got well used to it. However, when I play such recordings through my monitor-quality room hi-fi system they produce a really vivid, breathtakingly realistic three-dimensional soundstage, with even the hard-left and hard-right sounds usually seeming to be behind or beside the speakers rather than coming from them.
The 120dB signal / noise ratio function — amazing, at least initially
This is switched off by default, but I did my initial test recordings with it on, rather doubting whether its effects would be acceptable to me, because it uses the recorder's level limiter mechanism in a clever way to increase one's available dynamic range free from system noise. I have to say, I found its effectiveness quite astounding, clearly with the limiter functionality still managing the peaks. NONE of the big peaks of wind disturbance, nor any of the deep boom peaks from the Beeny Cliff cave system got even a hint of distortion, and all the strongest peaks were contained at 'maximum' level, with no clipping at all highlighted in Audacity. Yet in all cases the recordings sounded completely natural, without even a trace of wavering or 'pumping' of the general sound-stage as one gets from all ordinary level limiters.

For the Beeny Cliff recording I had deliberately left the recording level set at a 'normal' value (6 on the dial), whereas with the M10 I'd have set the level ridiculously low in order to avoid distortion of the stronger booms, so I was treating that recording just as a throw-away test one — but the D100 handled all that perfectly. By limiting the strongest peaks it had reduced the dynamic range, but at least it had reduced the range to what was playable without distortion, and sounded really convincing and lifelike.

The lesson here is that if I want the strongest peaks to stand out to their full extent in such a recording in future, I can simply set the level lower and then turn up the playback volume to get the correct overall sound level. The latter is what I was always having to do with the M10, but here the S/N ratio enhancement would mean that I'd get better sound quality for an equivalent degree of under-recording and subsequent compensatory turning up of playback volume.

Later on I became bemused in remembering the apparent peak limiting, because since then I haven't seen that effect, at least much, and 'over' peaks normally did get clipped.

At least, that 100db signal/noise ratio setting presumably would help in very quiet recordings, so I've been leaving it switched on.
Rendition of very low frequencies
Whereas with the M10 I usually had to apply a significant cut to the lower bass to get a reasonably accurate sound, with the D100 I found that it was actually fairly weak in its recorded level of the very low frequencies (and yes, I did check that its low-frequencies-cut filter was switched off). To ensure an accurate playback I applied a correction curve, as follows (please note that this cannot display properly on narrow screens):
<curve name="9dBbass cut"> <point f="5.000000000000" d="-12.000000000000"/> <point f="10.000000000000" d="-9.000000000000"/> <point f="35.000000000000" d="-9.000000000000"/> <point f="55.000000000000" d="0.000000000000"/> </curve>

Please be aware that Audacity cannot handle that curve properly if the lowest point is below 20Hz or / and the highest beyond 20KHz. You have to manually edit the EQCurves.xml file in the Audacity folder to create or edit the particular curve. and if you want to use that curve in a command chain you have to manually create a chain file containing that curve. Trying to create that chain within Audacity would result in a chain called 'Unnamed' being used instead of what you have chosen!

— except that in practice I have it combined with whatever treble correction curve I'm using. I have that curve saved in Audacity, so I can easily apply it to all relevant new D100 recordings. It's extremely important, however, to use that curve ONLY as part of a command chain that includes first an overall level reduction (negative amplification value) so that the boost of very low frequencies doesn't cause clipping.

In fact I've given myself a choice of command chain to use for new files, because not all recordings would really benefit from that bass boost even though it would be technically correct. So, I have an alternative command chain that does the windshield correction but not the bass correction, with only a small prior level reduction to avoid clipping, and I'd use that especially for recordings where it's important to minimize audible microphone wind noise.

Although I'd prefer simply fully linear and accurate recording of all frequency ranges, actually the slight weakness of recording the very low frequencies in the D100 would result in much less pressure on peaks incorporating very low frequencies, so minimizing the possibility of distorted major peaks, at least when one isn't using the limiter or 120dB S/N ratio feature. That weakness therefore does have a positive aspect, provided one doesn't mind having to apply a correction curve to boost those very low frequencies. Without that correction, many soundscapes can lack a certain feel of vibrancy and depth — though obviously that would be noticed only with really good listening setups that reproduce those low frequencies properly in the first place.
Slider switches
Thank goodness, all slider switches, including microphone sensitivity, are recessed and not at all easily moved in error as they are in the M10, and their positioning makes it much less likely that any would get accidentally moved while putting on or removing a windshield — though it has happened for me as an extreme rarity.

Eventually, after making my first really major recording with the D100 — a 7½-hour one very well placed in Branscombe Landslip, near Beer, Devon, UK, in particularly sheltered conditions on 24 April 2016 — I came to the somewhat embarrassing and very inconvenient conclusion that the D100 was showing up quite dramatically how narrow, blurred and foggy the stereo image produced by the M10 really was. Indeed, to such an extent that I realized that I no longer wanted to make further recordings with the M10, at least using its internal microphones.

Because of the blurred-image issue with the M10 and the near-unusability of the D100 because of the extreme wind-sensitivity of its microphones, I reluctantly got investigating external microphone options for both recorders. However, that proved to be a frustrating and hair-tearing experience, and indeed an expensive one, which left me quite relieved when I very quickly came to the clear conclusion that it was simply not workable for me to attempt to use external microphones with either model of recorder.

At that point I'd become clear that I wouldn't use M10s any more, at least for 'serious' natural soundscape recording, and so opted for purchasing an additional D100 to give me my two recorders for really time-efficient sessions. As I have remarked further above, at last I have reasonably resolved the wind-sensitivity issue, so look forward to producing a good number of recordings now with the breathtaking realism that the D100 readily gives with its own microphones in wide stereo configuration.

Using the PCM-D100 to record a thunderstorm

The excellent thunderstorm that came over Exeter during the night of 18/19 July 2017 gave me an educational challenge whose resultant insights I'd like to share with other users.

As I'd done previously for attempts at thunderstorm recordings, I put the recorder on a (diminutive) Hama Mini Tripod on a pile of books (covered with a plastic bag to keep them dry) on my bedroom window sill with the vertically opening window fairly well open. Having learnt from a mishap the previous time I'd attempted a thunderstorm recording there, this time I used a length of string to tether the head of the tripod to a heavy object on my floor, so that the recorder couldn't get pushed out of the (second-floor) window. I did this before going to bed and let it run overnight, at that stage not knowing for sure that a storm would come over, but at least it appeared possible to me with my meteorological experience and insights.

Because of the very big dynamic range that has to be expected if there are any more or less close earth strikes, I set the recording level low, still with high sensitivity, but at just 4.5 on the little wheel (my normal recording level is 6). As usual I was using two custom Windcut furry windshields — a bigger one over the direct-fitting one — though I wasn't really expecting much challenge from wind this time.

I left the recorder running all night, to capture whatever it could. The first noticeable faint murmurs from the storm appeared at about 11.10 p.m., while I was still asleep, and the last that I was aware of were at about 1.44 a.m. This storm had a very long lead-in period, having its anvil-top drawn out a considerable distance in front of the relatively narrow line of heavy cloud base where the earth strikes and heaviest rain were concentrated. During that loudest part there were two relatively close earth strikes, though neither being quite close enough to give a startling sudden bang.

Afterwards I had the expected editing work to do, cutting out various bits of traffic noise disturbance, and various people disturbances (this is in the city centre), but there was a problem with the two close earth strikes. They'd both been strong enough for the main peaks in their waveform to get substantially clipped despite my low level setting. So, the dramatic crack sounds of the lightning were rather obliterated by the clicks and 'spits' caused by the clipping. Oh dear — perhaps I'd have to simply cut those two most dramatic bits out of the recording!

I was aware of a 'clip fix' function in Audacity, but had used it in the far past with little obvious improvement in the sound — and indeed I tried it this time, after considerably lowering the level of the overall recording to allow for any peak rebuilding. The clip fix function itself needs to be applied just to short selections — not the whole recording. Anyway, again it appeared to achieve very little, so I was sadly thinking I'd probably have to cut those two dramatic episodes out.

But then I realized that the quieter thunder was actually getting a bit of room resonance overemphasizing some of the very low bass frequencies, and, rather than do another EQ adjustment on the file, I chose to take a new copy from the unprocessed original and then give it the most appropriate EQ adjustment all in the one initial pre-editing EQ adjustment (including correction for the windshield muffling effect), then having the whole editing task to repeat from scratch. Although doing the editing task all over again was a rather tiresome prospect, at least it would give me the opportunity to make a more effective job of not only those two fairly close strikes but also a couple of other points in the recording where I'd probably been a bit too hasty.

Fortunately, before actually doing that I had a hunch — that very likely the reason why Audacity's clip fix function hadn't worked significantly for me was because the respective sound-files had already been processed, so that then the clipped peaks were no longer recognisable to Audacity as 'clips'. So, before I did the standard processing I lowered the level of the recording by the recommended 10dB to allow for rebuilding the clipped peaks, then ran clip fix on the lightning strikes. To my amazement the clipped peaks were restored to such an extent that they got clipped, so I had to undo that and the negative 'amplification' and use trial and error to find a level for the file at which the fully repaired peaks would be accommodated without new clipping. The reduction I arrived at was 13dB.

The result was greatly impressive and really sounded like the real thing once more. Great! — Except that now, if the edited version of that were put on CD with those peaks intact, the listener would have to turn up their playback volume some 15dB above a sensible normal volume setting. That would be about a 600% increase, and would be liable to cause damage to playback systems without good power handling, and of course most people would simply not turn it up enough to hear the very quiet parts at all and indeed anything properly.

The next challenge, then, was, rather than to degrade my new 1h 55' full version, to make a CD-length selection, which of course included all the loudest part of the storm, making a separate file of that and then to see what I could do about those peaks in that version. In the past I'd found that I could usually cheat with low-frequency peaks and raise the overall sound level just enough for the strongest peak to be clipped by 3dB. Anything more than that, and the clipping would cause artefact effects. So, I tried that, but the dynamic range was still too much, still requiring a playback volume level 12dB above normal (nothing more than 6dB above normal would be acceptable to me). Not only that, but a few slight clipping clicks were audible again, albeit nothing like as strong as before I'd repaired the clipped peaks.

My impression is that most people would just accept that they have to put up with a few slight clipping clicks or 'spits' and leave it at that, but I still wanted to do something about that, and in any case somehow needed to reduce or cut out those peaks with absolutely least possible reduction in the realism and dramatic impact of those strikes — i.e., if the recording was to be fully satisfying to me. Yes, I was truly hankering after a holy grail!

Because simply reducing the amplitude of a peak would cause a corresponding transient amplitude drop in the ambient sound, that wasn't an option for me, so I started thinking about possibly cutting out a tiny slice of each of the offending peaks, done at a very high zoom level so that it could be done very precisely and with absolutely minimal collateral loss. That would NOT be done by directly deleting the selected slice, as that would tend itself to cause a click or discontinuity, but by using a crossfade function so that the cut ends are cross-faded together to leave a smooth join.

Generally I do crossfade editing in WavePad, but when I tried it at the necessary high zoom level it wasn't possible to use the crossfade function at all because one needed to be working in units some 10 or even 100 times smaller than the available milliseconds in order to do this work. Ouch!

But then, having 'smelled blood' so to speak, I investigated another function in Audacity with a very enticing name — 'crossfade clips'. This was a bit surreal actually because my investigation into what it actually does showed that, on the face of it, its intended purpose is nothing to do with repairing clipped peaks but is simply a quick means to smoothly join together two adjacent sections ('clips') of recording. And yet I discovered that it could actually do just what I wanted, because if it was used on any selected part of a recording it would crossfade each half of the selection together, so that when used at a very high zoom level, it enables one to smooth out individual glitches in the waveform.

Because I didn't have to set any values for length of crossfade or any offset, I wasn't constrained by having to set values in milliseconds. I just had to select the central part of the clipped peak, or the other little glitch, being careful in having just the right amount of 'okay' waveform on either side of the unwanted bit, and then use the function to do the smoothing crossfade.

Then, having noticed how this produced reductions in the peaks without significantly degrading their sound in ways that one would notice, I went on to use this method to reduce the amplitude of all the peaks that went above 50% (about -6dB level), whether or not they were actually clipped. Having done that and being satisfied with the sound, I 'normalized' the file so that its highest peak was at 100%. Thus I'd managed to enable myself then to raise the overall sound level to one that required only a 6dB-above-normal playback volume level (i.e., without further clipping) for a realistic overall sound level. Although the lightning strikes were reduced in volume now, actually they still had almost all their original impact, and with virtually no artefacts so that the lightning cracks sounded quite as real and intimidating as in the original. I regard 6db as an acceptable maximum amount above normal playback volume that I'd allow for any of my CDs.

I was chuffed to have discovered this approach to restoring and reducing clipped and excessively strong peaks, because in future I can use this for handling strong peaks from the sea booming in caves or whoomphing in blowholes, quite apart from future thunderstorms. I really don't want to set a recording level below 4.5, so being able to restore and 'manage' peaks that still do clip will be a great boon.

The D100 mics not so good after all — a hiss problem

Exceptionally low noise floor figure quoted for the D100 is seriously misleading

I've already given a fleeting mention of this issue. In June 2018 I had a particularly wonderful recording session at and just to the north of Bellever Tor on Dartmoor, starting early afternoon and continuing overnight till past dawn, to capture in particular the cuckoos and nightjar choruses, in addition to the general evening and dawn bird choruses.

Despite some rather tedious wind disturbance to the recordings made on Bellever Tor itself, the recordings were really great — except for one issue, which I'd noticed in other (mostly night) recordings of very quiet soundscapes. That is, a quite distinct background hiss. I was mortified! This issue had come to notice in some degree since I started using three instead of two furry windshields on each recorder, to extend somewhat the number of occasions when I could usefully record with this insanely wind-sensitive recorder model. I carried out a range of tests, and found that the recorder's overall noise floor rises in a concave curve from about 3K to a peak of c. 13.5dB higher (using the high sensitivity setting, which is essential for all my recordings), regardless of what level had been set for the recording. I did also check the noise floor for a low-sensitivity setting with a lowish level (4 out of 10) set, and that showed the same curve but worse, starting at 2K and rising to c. 18dB at 21K.

Frequency spectra of test recordings — created by Voxengo's Span plugin

PCM-D100 noise floor with high sensitivity (0dB attenuation) and level setting 6-7 — The PCM-D100 with level set to 6–7 (my usual setting) at high sensitivity (0dB attenuation). Note that the overall level part of the graph (from 3K, actually down to around 12Hz), is NOT quite the noise floor, but, as I discovered later (see further below) is nonetheless a major component of the audible sound in the recording. The ambient background sound is very muffled (coming through very effective double glazing), and so has no significant treble component. The bass region was very mobile; peaks were coming and going so that any screen clip makes that region look to be anything but flat. The rising part of the curve, from 3K upwards, does represent the noise floor for that range, however. The excess at 21K is c. 13.5dB.

PCM-D100 noise floor with low sensitivity (20dB attenuation) and level setting 4 — The PCM-D100 with level set to 4 out of 10 (lower than I'd ever set) at **low sensitivity (20dB attenuation)**, which I never intentionally use. Again, the overall level part of the graph contains real ambient background sound, but the self-noise is still a major part of the audible sound in the recording. In this case I amplified the recording by some 42dB to bring its overall level up to that of the previous one illustrated here. The excess noise level here is c. 18dB.

Although those curves produced in the tests don't look very nice in relation to a healthy frequency response curve, they're at an overall low level. I've boosted them here by appropriate amounts post-recording, to bring them all to the same overall level to enable me to listen to the hiss in each of the test recordings and provide the most telling comparison in the frequency spectrum curves (see above).

Actually the hiss doesn't come to notice in any of my very quiet soundscape recordings when listened to at anything like the original soundscape level, except where I've used the third furry windshield. The trouble is, the correction curve I worked out for the muffling caused by the windshields precisely matches and amplifies that noise floor curve, and there's nothing I can do about that except limiting my recording of very quiet soundscapes to those with little enough wind that I can record without that third furry. That means having to miss most opportunities that present themselves. The 3-windshield correction I have to apply for accurate treble restoration rises to a peak of 13dB at 20K, which of course also raises the noise floor by that amount at that frequency. As they say, 'Ouch!'

I tried cutting off the windshield correction curve, to be flat above 10K, but even though I can't obviously hear a sinewave above 9 or possibly 10K (still good going for a septuagenarian), I noticed a quite distinct effect of that cap on the correction — a significant loss of detail, clarity and depth in the soundstage — so I have no plans to use that compromise as a workaround.

I carried out these tests after failing to get clear indications from reviews and forum posts. Unfortunately, the specialist forums, although no doubt having really genuinely knowledgeable and helpful members, are also replete with ignorant and opinionated individuals who spout their opinions and bits of hearsay on relevant issues without having having done any proper homework to establish what the real situation is. Thus we have individuals telling us that any audible hiss heard from D100 recordings couldn't be other than ambient background sound, because both preamps and the mics are just too low-noise to produce noticeable hiss.

And further, some of those (to put it politely) very silly poseurs also claim that for best results one should use a low level setting (4 was particularly recommended), and even the low-sensitivity setting (20dB attenuation). In reality, the only sensible use for such low settings is where you expect to be recording very loud sounds indeed and so need that extra headroom to avoid clipping.

The reality that my own tests show clearly is that within the high sensitivity level range there is a consistent level of hiss that remains the same in proportion to whatever recording level has been set (in my case the excess was about 13.5dB at 21K for each of the three recording levels that I tested). That points to the microphones being the culprit, not the preamps. In my tests, hiss in the ambient background sound was firmly ruled out. Then, my test down in the low-sensitivity range showed that a new level of noise floor was showing there. That presumably was pre-amp noise, but was really not of interest to me because it's so far below anything that would materially affect even the quietest of my recordings.

Now, if the D100's mics were not so desperately wind-sensitive, forcing me to use multiple furry windshields and thus have to apply big treble corrections for all that muffling, this wouldn't be such an issue for me and indeed other natural soundscape recordists. To use two instead of the normally necessary three furries would reduce the necessary correction (and thus treble boost) by some 6.5dB, and if only I could get away with just one furry I'd have a further reduction of some 3.25dB.

I did investigate (online) the Zoom H6 as a possible alternative — but found that users doing quiet natural soundscapes ran into exactly the problem I have with the D100. Well, except that they say nothing about windshields, and I wonder whether they were correcting for windshield muffling at all. Very likely the H6 was actually performing significantly worse than the D100. Also, even a bit more so than the D100, the H6 reportedly requires unduly high recording levels to be set to obtain realistic playback levels from the recordings. On the D100 I have to use the midpoint between 6 and 7 (out of 10) for normal recordings, whereas 4 or 5 would have been much healthier and allow more headroom (as I had with the PCM-M10) — if only that recording level would give a sensible amplitude in the recording itself!

Also, the H6 didn't really commend itself to me because of certain ergonomic considerations — not least, the cluttered and difficult-to-read LCD display, which is said to be particularly difficult to read in direct sunlight. That wouldn't matter if you're using the recorder in studios / concert halls but not in unshaded outdoor situations as I generally am. And for me the H6's confusing clutter of features and adjustments that I'd never use is another contra-indication.

So, it still appears that serious natural soundscape recordists who cannot use (or indeed buy) real professional gear specifically intended for that sort of recording are to a fair degree left out in the cold in the arena of hand-held compact fully integrated recorders, which inevitably are aimed at what most people would use them for — natural soundscape recordists still, sadly, being a very tiny proportion of the potential user base.

Further update, 31.12.18
This evening I thought to attempt a recording of the D100's self-noise — all of it, as clear as possible of ambient sound — so that I could possibly use a clip of it for use by appropriate noise-reduction software. I found I could do this relatively well, because my flat has a small store room that has no windows and picks up extremely little ambient sound. So I started the recorder recording at my usual level (6–7 at high sensitivity), put it in its original manufacturer's box, which is sturdy and padded inside, closed it all in, and left that for several minutes in the store room with door closed and all other doors connecting with my diminutive hallway also closed, and kept as still and quiet as possible myself.

From that recording I clipped out the quietest stretch I could find — some 7 seconds, which would be enough to assist any software noise reduction function. What rather shocked me was that it became clear that all along in my recordings of very quiet night soundscapes I'd been hearing masses of self noise, of which the hiss is only a little part. The very broad midrange of the self noise sounds so spacious that I'd assumed that that had been genuine background sound in my very quiet recordings, even though I didn't remember hearing it quite like that. But my doing this self-noise recording in a fairly padded box in my store room completely rules out any of that being ambient noise, for nothing could sound spacious there!

I tried using Audacity's noise reduction facility. It did work, BUT, just as I expected, whatever settings I used, it removed vital detail from the recording I was testing it with, rendering the spacious very three-dimensional recording almost two-dimensional, and wiping out almost all the reverberation. Bird sounds were audible at different volumes, but there was little sense of them being at different distances away. So, on this basis, noise reduction isn't acceptable for my purposes, and I have to regard the affected recordings as simply unusable except for my own occasional background listening.

This means that I have to pull some delectable recordings from my CD catalogue and list of prospective CDs, and, for the most part, would now sadly not make further recordings in such very quiet conditions — at least, with the D100. Because the issue is now much more than just hiss, I now know that recording in very low-wind situations (so that I don't have to use a third furry windshield) wouldn't save the day, except where there is midrange background sound to mask the similar-sounding midrange part of the self-noise. That situation would tend to apply where I'm recording anywhere in the Teign Gorge, or in coastal situations, for example.

I am thus notionally on a quest for a holy grail — a suitable small portable recorder that does the job that the D100 miserably fails to, though all my researches so far have shown nothing sufficiently portable and within any affordable price range. Most frustrating!

2020 update
Actually in early 2019 I got waking up to a so-far persistent error of mine, in seeking to get such immensely quiet recordings to start with. I developed a new determination to make another attempt on Bellever Tor to capture evening and dawn choruses including nightjar choruses and cuckoos, but doing it differently. I had a graded series of prospecting and rehearsal sessions prior to the very short seasonal 'window' to capture the nightjars and cuckoos optimally. My approach that time was to place recorders close to or just inside the edge of the forestry on Bellever Tor's western flank. That would provide a certain level of background sound from wind in the trees, and also there would be a good proportion of birds reasonably close to, or at least not so distant from, each respective recorder.

Indeed I found that the small clumps of young self-seeded Sitka spruce there provided exceptional wind shelter, to such an extent that I could use all the recorders with only one furry windshield, still with very little wind disturbance, so that there would be minimal raising of the microphone hiss during post-session processing.

That different approach worked a treat, fully justifying the time and effort spent on all the preparatory sessions. It was my most fruitful and productive session ever, with four recorders in widely separated carefully chosen positions, running form mid-evening on 31 May 2019 right through to nearly 8.0 the following morning. I divided that into the evening / dusk and pre-dawn to early morning sections, deleting the really quiescent middle-of-night section of each, having checked that there was nothing in those quiescent periods that I might want to keep.

After a huge editing task, cutting out masses of disturbances, I still had a full CD's worth of five-star evening / dusk recording from each of the four recorders, and enough for a two-CD set from each of the four pre-dawn / early morning recordings — again absolutely five-star and capturing everything I was after, with the microphone self-noise too low to be an issue.

Exploring possible workarounds — looking at Zoom H5 and H6

My further online researches highlighted the Zoom H6 as having some potential. Somewhat less expensive than the D100, and also having a similar type of microphone pair, though near-coincident XY and retaining their crossed near-coincident arrangement even when switched to 120° angle. Intriguingly, the H6 comes with a Mid-side (MS) mic as well, the two being quickly swappable. I fancied the idea of experimenting with MS recording, and sometimes being usefully able to further expand the soundstage of the recording. But plenty of online reports on that mic complained of its high self-noise, while the XY mic was apparently quite reasonable.

Still further research showed that the H6 had particular more or less show-stopper issues from my perspective, BUT the smaller and lighter (and less expensive) H5 was of the same vintage and had similar basic quality, and lacked a lot of H6 features that I'd never need and which cluttered its user interface. So I ended up taking on an experiment, ordering an H5, whose swappable on-board XY mic pair wouldn't be of any use to me because they were fixed at 90°, and ordering separately the XY and MS mics of the H6, which fit and work on the H5.

Once I got those I carried out comparative tests. To save space I'm giving here just a summary, not raw data.

PCM-D100 with its built-in mic pair
- Input level from mics: Rather low. A recording level setting of between 6 and 7 (out of 10) is required to produce a recording that in most cases plays back at a life-like level if playback level is already set at a sensible level.
- Self-noise: Low, despite my problems with it in very quiet soundscapes (i.e., the latter require still lower self-noise)
- Wind sensitivity: Insane! Very high, and a big problem.
- Frequency response at 24/48 resolution: no issues (reasonably consistent 20Hz to over 20KHz)
- Overall sound except in very quiet soundscapes: outstanding: disconcertingly real-sounding details, precisely located and focused, yet also breathtaking 3-dimensionality and width of soundstage (with 120° setting), which makes my recordings sound as though extending laterally beyond the speakers. The 90° mic setting is no good for natural soundscapes, but excellent for capturing, e.g., solo birds — but only if you want a narrow soundstage without significant ambience at all — a sort-of audio 'tunnel vision'.
Zoom H5 with its own XY mic pair (90° fixed angle)
- Input level from mics: Very low — some 23dB lower than the D100! And that's without any low-sensitivity setting (I checked carefully). With that silly low level, even recording with full gain (10 out of 10 on the recording level knob) couldn't get me a recording that would play back at a life-like volume (i.e., with a sensible playback volume level already set).
- Self-noise: Relatively high, c.11dB higher than D100
- Wind sensitivity: Insane! — much as D100
- Frequency response at 24/48 resolution: no issues (reasonably consistent 20Hz to over 20KHz)
- Overall sound: Useless for natural soundscapes, because of the narrow soundstage. Despite good clarity of sound, the high self-noise and narrow angle makes this really a no-no for natural soundscapes, and generally a rather poor mic for an otherwise such good recorder.
Zoom H5, with H6 XY mic pair (both 90° and 120° setting)
- Input level from mics: Very low — some 23dB lower than the D100
- Self-noise: Low — no more than 2dB higher than D100
- Wind sensitivity: Insane! — much as D100
- Frequency response at 24/48 resolution: More uneven than the D100, with two broad rises in the treble region with a dip in between; a rather confused-looking real-time bass curve, falling off more strongly than the D100.
- Overall sound: apart from slightly higher noise level, which would affect only very quiet soundscapes, this appeared at first glance to match the D100 pretty closely, though an initial impression from one full-length recording does hint at not quite so three-dimensional a quality to the soundstage. If that is borne out by further recordings, I'd put that down to the coincident crossed configuration of the mics, so removing virtually all phase (time-difference) cues for location of details.
  
  Also my impression so far is that off-axis sound rejection is much stronger than with the D100. That has its advantages in minimizing various disturbances, but I'd expect it to reduce the spaciousness of the soundstage — in particular, to reduce any tendency for the soundstage subjectively to extend laterally beyond the speakers.
  
  A subsequent test with concurrent recordings by the D100 and H5 of synthesized white noise showed the H5's (H6) XY mics to produce a much thinner and seemingly more anaemic version of that white noise — which suggested that even this mic option wasn't a particularly high quality one. — Well, er, except that I soon came to realize that it was more likely that it was highlighting a slight coloration of the D100 sound, which was a little bit warmer and 'fatter' than would be an accurate rendering! (See further below…)
Zoom H5, with H6 MS mic
- Input level from mics:
  - For RAW MS setting: Very low — some 23dB lower than the D100
  - For stereo encoding options: a little better but not much.
- Self-noise: Relatively high; appeared to be c. 7db higher than D100 but sounded much louder than that.
- Wind sensitivity: Much less than the others. Fan test suggests c.12dB better than D100, but quick field test in breezy conditions suggests possible a still bigger difference in outdoor conditions.
- Frequency response at 24/48 resolution: Show-stopper — the treble goes only up to 12K!
- Overall sound: Many recordists love working with MS recordings and getting impressive-sounding stereo sound from them. However, my own experience with this shows its devotees to be listening with a considerable lack of mental awareness and discernment. If you're a serious recordist of natural soundscapes, surely you want accuracy in your recordings, NOT impressiveness in its own right — but you won't get accuracy with this mic, and indeed, as far as I can make out, won't truly get it with ANY MS setup.
  
  All stereo spread settings offered by the Zoom MS decoder sounded impressive, yes, but through a weirdly unreal sort of sound as compared with what I actually heard when making my brief test recording while walking by traffic on busy main road and then walking down to the quite busy Quay area of the River Exe, with a wide range of types of sound, near and far.
  
  Even at the most realistic setting (described as 120° angle), the ambience was an impressive wash of sound, but it was NOT what I heard, because what I heard was a mass of details with a much quieter background 'wash'. The recording's 'ambience' (the 'mid' element) sounded remarkably lacking in a sense of near and far. House sparrows chirruping away in some trees around me sounded not only not well focused and thus rather vague in their lateral positions, but didn't sound near and far, but just louder and quieter. Absolute rubbish!
  
  That no doubt partly results from the lack of really high frequencies, which provide the cues that I need for perceiving depth in a soundstage. I've experimented by applying a quite modest attenuation to frequencies above what I can hear (when presented as sinewave) in one of my D100 natural soundscape recordings, and that considerably reduced the clarity of detail and sense of depth.

On the basis of my experience up to this point, then, I'd just about given up on the Zoom H5 (or H6), but a subsequent concurrent pair of side-by-side test recordings (with identical two Windcut furry windshields on each) in the River Exe flood meadow, with some breeze and a mass of background and ambient sound, threw my ideas into the melting-pot again.

What these new recordings showed was that in fact the H5 with h6 XY mics at their 120° setting registered significantly stronger mic wind noise than the D100 (i.e., as shown on the waveform) — but, surprisingly, it sounded quieter, smoother and less intrusive. Examination of frequency spectrum showed that the microphone wind noise, as it got stronger, developed a massive excess of very low frequencies, the peak region being 10–20Hz.

Theoretically this could be filtered out, but because of the way actual curves in Audacity's Equalization function don't precisely follow the nodes that one specifies, this could not be done without somewhat impacting frequencies from about 30Hz downwards. In any case, this issue related only to the mic wind noise, and actually apart from that I saw no excess of those very low frequencies in the H5's recording.

I made a further test to see if the problem was the way the furry windshields were interacting with the H5's crossed mic configuration: I made a short side-by-side pair of recordings in my living room (no windshields) of a playback through my notionally monitor-quality stereo system of the very deep cave rumbles and booms on the clifftop of Beeny Cliff (near Boscastle). Both recorders came out similarly then, with a general steep roll-off below 20Hz — though the H5 lower bass actually sounded a bit cleaner and 'tighter' (probably good!). So, the very low bass excess in mic wind noise was clearly nothing intrinsic to the H5.

I would clearly need to change the way any furry windshields fit over the H5's mics, though. Most likely the way forward there would be for the inner windshield to be a reasonably thick foam one, on which the outer, furry, windshield would then rest, for that ought to prevent very low frequency vibrations from occurring.

Also, however, the latter two test recording sessions revealed a distinct show-UNstopper with regard to the H5's sound quality. The soundstage was much clearer, more sharply focused and three-dimensional, sounding altogether still more 'real'. Not just 'realistic', but sounding to be exactly what I heard while making the recording. After this, listening again to the D100 concurrent recordings was like listening through an obfuscating veil or fog. The main factor here is the H6 XY mics being significantly more directional than those of the D100. That makes for still clearer stereo separation. Differences in the treble frequency response most likely also to contribute to this clarity.

…But then a later pair of test side-by-side recording sessions at Boscastle Harbour and beside the southernmost cave of Beeny Cliff, near Boscastle threw my ideas into the melting-pot again. Yes, the Zoom's greater stereo separation indeed made for particularly sharply-focused and located details, but, however I sought to tame its brash and harsh sea sound, it was no match for the subtlety and three-dimensionality of the D100. It was also clear that at the Beeny Cliff cave location some of the cave sounds (right, but not really hard-right) were too precisely located — i.e., were more so than as I head them, and thus sounding thus smaller and less thrilling.

But actually, prior to that session I had my first 'for real' recording session with the Zoom, three hours of quite beefy wind gusts chasing around in the trees below Cranbrook Down, on the upper slope of the Teign Gorge. I was able to improve that recording, not by taming the treble, which sounded fine in that situation, but by applying a lower strength of the bass boost curve that I worked out for the Beeny Cliff cave. However, I found that any prominent wind-in trees sound that was hard-left or hard-right sounded as though it was coming solely from the respective speaker and was mono and not a part of any overall soundscape. Indeed, more surprisingly, occasionally a gust only a little off centre would sound as though being mono.

It's thus become clear that the Zoom's mics, as they come, are just too directional for decent natural soundscape recordings, being prone to producing unnatural effects through being too directional. You hear a mass of clearly-defined details, yes, but they aren't creating a really coherent three-dimensional soundstage.

Actually the aforementioned sea recordings showed clearly that NEITHER recorder reproduced the sea sound fully accurately. At Boscastle Harbour I made a particular point of examining and memorizing very carefully the sea sound there, and I noticed a particular characteristic of the sound of small waves breaking there, with a lovely feather-light and airy quality to the treble aspect of that sound, which latter wasn't being reproduced significantly by either recorder.

However, I soon got encouragingly closer to getting both models reproducing the sea sound much closer to how I did really hear it — through two measures:

Using gentle frequency-range tilts, with carefully chosen start and end points (i.e., the tilt not being over the whole frequency range). For the Zoom the tilt is descending towards the treble, and in the D100 the tilt is ascending. That's easily achieved in Audacity, using its 'Draw' graphic equalizer. It's best to create any curve as a descent (in either direction) from the 0dB (no change) line. That way you don't risk any of the waveform clipping. The aforementioned wind-in-trees recording benefited quite spectacularly with a mere -1dB tilt towards 8K, making the whole soundstage much more three-dimensional and the loudest wind gusts through the trees much less abrasive and more powerful.

Similarly I worked out that for most D100 recordings a quite dramatic increase in clarity and general realism was achieved with a +2dB tilt towards 8K (again, configured as a -2dB tilt in the opposite direction), with subsequent amplification of the recording, up to 3dB maximum, to get the most lifelike rendering — the degree of that amplification having to be worked out on an individual recording basis, so it cannot be incorporated in any automated action. Both tilts extend between 100Hz and 8K (only).

(2021 note)
Watch out! Please don't make my mistake here with regard to the D100 recordings! See my relevant Blog entry.
(Zoom only) Mixing into the stereo a much level-reduced mono version of that stereo track, to moderate the over-separation of the stereo channels' contents. That is easy to do in Audacity, though quite long-winded. It worked very well with the aforementioned wind-in-trees recording — the added mono having been reduced by 15dB first. That was quite enough to get rid of the impression of mono sounds coming directly from the speakers, without spoiling the subjective width of soundstage. It's a bloody nuisance that this would be needed for every Zoom recording, because that would mean a lot of extra work in the processing of each new batch of recordings.

Update, February 2019

I have retrospectively applied a mostly 2dB tilt to most of my D100 recordings, with great improvement in clarity, detail and overall realism, in each case sounding as though a murky midrange fog has been removed.

As for the Zoom, I was delighted to succeed in working out a macro in Audacity that automates the process of extracting a mono version from a recording, reducing its level by a specified amount, and then mixing it back into the stereo version. Once I've been able to test different levels of that mono component with suitable recordings I'll add into the 'New Files' standard initial processing macro a line that calls the 'Add Mono' macro, configured to add the optimal level of that mono component.

…But, Update March 2019 — the Zoom not so good

Having at last had two extended recording sessions in which I made long side-by-side Zoom and D100 recordings of sea dramatics, I was able to experiment with a variety of processing parameters, and came to the rather disappointing conclusion that, on balance, I'd use only the D100 in future, at least for most outings, and I'm unlikely to find much use for the Zoom after all. Unfortunately the Zoom's more neutral treble and initially compellingly detailed stereo image are outweighed by a major coverage and soundstage limitation that cannot be compensated for by any processing.

I've already remarked about the Zoom's mics as being over-directional, so that a certain proportion of mono has to be added to each recording to properly join up the sound from left and right channels — but the problem is that that over-directionality cannot be compensated for in any direction that doesn't lie between the forward axes of those mics. So, as a sound moves in the rearward direction beyond a microphone axis, it rapidly loses volume and higher frequencies, sounding very distant, and all but completely disappears within a fair area behind the recorder.

That means that not only is there an extensive deaf area behind the recorder, but all sounds round the edge of the soundstage sound wrong, especially through headphones, because they sound far too distant compared with how they'd actually sounded, and sound to be still too restricted to their particular side. This would be the case also for sounds above or below the recorder; anything significantly more rearward than the mic axis will sound over-distant or be more or less inaudible.

That would presumably suit some people, but I think normally not really dedicated natural soundscape recordists. Speaking for myself, I generally want the widest-possible soundstage, even for playing back through just two speakers. Indeed, the D100's soundstage usually seems to extend significantly beyond the speakers on either side, as well as altogether being more three-dimensional than the Zoom's rendering. In one of my comparison recordings there was a little trickling streamlet running over the rock slabs immediately behind the recorders. The D100 heard it clearly, and it sounded so clear as to seem to be immediately in front of the recorder. But the Zoom didn't hear it at all through the sea sound ahead, which was the main subject of the recording.

I eventually sold those Zooms off.

A new twist — late March and April 2019 — return of the PCM-M10?

I got put onto the remarkable audio processing VST plugin called TDR Nova, which (particularly in its paid-for Gentleman's Edition) enables me now to minimize and in some cases to more or less resolve certain issues with my recordings — such as more effective mic wind noise reduction and (in the case of recordings of my music compositions) increasing dynamic range. Another thing that the paid-for Nova does is enable one to adjust stereo width and sharpen up the soundstage — which is something I'd been wanting to do for all my PCM-M10 recordings, to upgrade their stereo imaging and soundstage width to be something more like that of the D100.

However, I found that the same source also offers a free plugin called Proximity, which enabled me to make adjustments to stereo width and sharpness much more simply and quickly, so I got using this to transform my M10 recordings. This also meant that I now regretted that I'd sold off (in 2016) one of my two M10 recorders, and I'd now be bringing the one I still do have into active use once more, along with the D100s. After all, the M10 is significantly less wind-sensitive, and its somewhat narrower angle can sometimes be a considerable advantage if the soundstage can easily and quickly be widened and sharpened-up afterwards.

Then, within a further week or two, I found that I could bring about a further improvement in the quality of the stereo imaging in my M10 recordings. Fantastic! At that point of improvement, any sense of listening to a recording is usually dissolved, and it sounds as though I'm back there in the recording session.

This further improvement has been facilitated by my switching from the Proximity plugin to the excellent A1 Stereo Control, and using two passes, at maximum width (200%) and then at 110%, but first having applied a compensatory EQ tilt falling from no change at 400Hz to -9dB at 8K.

Actually, further side-by side field recordings by M10 and D100 recorders beside Bellever Tor, Dartmoor, revealed significant limitations on the stereo imaging improvements obtainable by such stereo widening software. Although the transformed M10 recordings sounded convincing in themselves, when compared with the respective side-by-side D100 recording, each of the stereo-enhanced M10 recordings showed remarkable deficiencies:

Despite the M10 capturing sounds from further round behind the recorder, bird sounds were almost all clustered at or closely around the centre, whereas many of them had really been in a very wide range of positions (and were correctly positioned in the D100 recordings), including ones at least to some extent behind the recorders. Various other sounds, however, particularly bees and flies, were more correctly located and could be anywhere within the wide soundstage.
Although the spatial location of details initially seemed precisely defined, just as with the D100 recordings, when the same sound events were compared in the M10 and D100 recordings, the super-enhanced M10 results sounded a lot less convincing and 'solid'. A real give-away, which pointed to the root of the issue, was an event captured by both recorders, where a bumblebee got interested in the recorders (probably their furry windshields) and for a short while flew very closely around and about the recorders.

In the M10 recording, although the bee seemed convincingly to keep moving side-to-side, it was a confusing sound, just moving side to side, with something about the sound giving a strong feeling of it not being quite right, and indeed being an artificial sound effect ('moronic' and 'aggressive' come to mind to describe it). By contrast, the D100 rendition of the same event kept the bumblebee sounding solidly real (its sound seemed to radiate its furriness and colours, so authentic did it sound), and it gave an uncanny sense of reality about its flying around the recorder (including going transiently behind it).

Further, a look at the waveforms of the two versions of that bumblebee event drew my attention to something I'd been sort-of noticing all along about the transformed M10 recordings' waveforms. In fact the M10 waveform for that event still looked as though it was almost mono, with almost identical waveform in both channels, whereas in the D100 version the bumblebee could be seen to shift strongly between left and right channels.

That nicely explains what's been going on. These stereo widening / sharpening programs are NOT significantly (or at all) increasing the genuine directional information in the respective channels, but only using a bit of wizardry to increase the phase / time-delay differences of each sound to give the illusion of greater stereo separation! This is nothing peculiar to A1 Stereo Control, but is definitely the case with the Proximity plugin that I used, and, as far as I can see, ALL programs that do stereo widening. They all convert standard stereo recordings into mid / side (M/S) format and then increase the volume (and, I think, treble) of the 'side' channel, then re-converting that to standard left / right channels again.

It's no coincidence that I'm very lukewarm towards recordings made in M/S mode, because to me, although they can sound very impressive, they always sound at least slightly unnatural, because the exaggerated phase / time-delay information conflicts with the 'pan' directional information (i.e. the relative strength of any particular sound in each channel). Most listeners would probably not notice anything was amiss, because they don't listen sufficiently critically to sounds in their everyday life.

I don't mean this as a criticism of the stereo widening programs I've used, because clearly they can do a fantastic job considering — but in my case we're dealing with recordings that have pretty weird and diffuse stereo to start with. I checked the original unprocessed recording containing the close encounter with a bumblebee, and with robins far round on the right to the rear, and, sure enough, the bumblebee had the same poor sound, so the stereo widening wasn't responsible for the issue — it was simply trying to make the best of a bad job.

And similarly, the robins were just a tad right of centre in the original, so the stereo widening hadn't succeeded in placing them further to the right, though it had sharpened-up the sounds, so that the robins sounded all the more as though they really were just right of centre (front).

Until I'd made that last series of side-by-side recordings, I'd been keen to get using the M10 again alongside the D100, and even look out for further M10s to buy on the second-hand market, as they have the advantage of being less wind-sensitive and of course smaller and lighter. But now I'm not so keen after all, despite all the advantages that use of the M10 could bring. On balance, I do intend to take the M10 out with me with my D100s on some occasions (and would presumably buy an extra one, second-hand of course, if I could find one at a fair price), but only to use as a 'second string', in order to be able to record something special where the wind was likely to ruin a D100 recording, or of course to take with me when I haven't carrying capacity for a D100 (or an additional one).

Stereo 'enhancement' no good for most sea recordings!

Belatedly, further careful listening to M10 sea recordings with the super-duper stereo enhancement brought me an unwelcome surprise. Although those recordings all sounded excellent through my headphones, when they were listened to through speakers unhelpful phase cancellation effects ruined the stereo imaging.

Only a few of my M10 sea recordings still had sufficient stereo image integrity (as heard through speakers) for them to be worth keeping. I therefore have withdrawn all the relevant CDs, and intend as soon as possible to delete the particular recordings themselves — even the archived originals of those. That would have a very positive effect, in bringing about long-overdue pruning-down of the collection to a more manageable number, which would then be of more consistent quality, and it would leave me more motivated to capture equivalent soundscapes with my current recorders.

On this basis, it's now unlikely that I'd use the M10 again, so would actually come round full circle and put my original one up for sale once more!

February 2021 update — just why the M10 has that problem, and why I'll use it more in future — except that I won't!

As a result of a side-by-side test recording session in a local park last May, I was able to work out why the M10 was problematical for stereo widening in some recording situations — and I'm pretty sure that there would be no significant problem with stereo widening for D100 recordings made with the 90-degree mic angle, so the problem lies with the M10 and not the A1 Stereo Control software.

The problem with the M10 is its use of very closely-spaced omnidirectional mics without an acoustic barrier between them. That straightaway is the cause of the poor stereo imaging, because the two mics are hearing almost the same thing. Only the overlap of the two mics gives you your impression of a soundstage, with sounds heard further right or left sounding uneasily off-beam and rather cramped-in. Then, when stereo widening is applied (i.e., strongly enough to fully compensate for that problem), all the peripheral sounds, including those from right behind the recorder, are superimposed upon the main soundstage, many of them from hard-left or hard-right actually coming out nearly central.

That in itself doesn't necessarily produce a poor-sounding recording, even though many sounds may be heard in wrong positions. Indeed, I have many widened M10 recordings that sound absolutely excellent despite various individual sounds presumably being incorrectly positioned. The sound quality problem kicks in where there is generalized (mostly background) sound, a good proportion of which is right round at the sides or indeed in the rear half of the recorder's surroundings. All that gets superimposed upon the roughly same sound recorded in front, and has at least the potential to cause problems of addition / subtraction effects, precisely as I've observed with panoramic generalized and especially hissy sea sound. Then, as part of that phenomenon, you can hear phasiness effects too.

It's therefore best to restrict use of the M10 to situations where there isn't a relatively homogeneous generalized sea or other water (e.g. waterfall / weir) sound widely spread (or even occasionally wind in trees), which can make a pig's ear of any fully widened version of the recording as heard through speakers.

In my experience those effects aren't heard through headphones, though all-round generalized background sound still gives a very confused and two-dimensional effect as compared with a similarly widened version of a D100 recording made with a 90-degree (i.e., narrow) angle, which comes out beautifully clear and three-dimensional.

So, why would I still want to use the M10 at all? — Because the D100 mics are so damned wind-sensitive, as previously mentioned ad nauseam. I've missed many, many wonderful recording opportunities because of that — a fair proportion of those quite needlessly because the wind had still been light enough to use an M10 with impunity!

But then again, I also would be much more aware of the shortcomings of further widened M10 recordings, and so always want to improve on them. …Eventually I came to the conclusion that it really was best for me to sell the M10 and concentrate on working with my three D100s. Then, as long as I could find sufficiently low-wind conditions, I could be sure of getting really high quality recordings, including occasional zoomed-in ones using the 90° setting and A1 Stereo Control for the post-recording widening.

Late 2022 update — a more balanced outlook on the M10, and some welcome reprieves

In late 2022 I revisited the M10 recordings that I'd notionally discarded, taking copies from an archive of pre-widening versions and this time applying a bit less widening — either 160% or the full 200% widening given by A1 Stereo Control (i.e., without a second pass in A1SC), according to which came out best with the particular content. While that couldn't rectify the M10's wrong placing of peripheral details in the original soundscape, it did produce less phasey effects than before.

Also, I found that I'd been blaming too much on the stereo widening, because I also got some phasey effects, including phase cancellation points, from some PCM-D100 sea recordings, which hadn't had any cause for me to try to widen them, when they were listened to with speakers, particularly my computer speakers (Audioengine A2+).

Consequently I replaced all my extant widened M10 recording versions with new copies widened that bit less, and also with a stronger subsequent EQ tilt away from the treble (a straight line from no change at 100 Hz to -7dB at 8K), followed by a level amplification to compensate for a reduction of level caused by the processing (the stereo widening presets used included a 2dB input level reduction to compensate for increased peaks that the widening can produce on some recordings).

That final tilt further reduced noticeable phasiness, and I found that I now had the best-sounding results from the M10 that I'd ever had, and restored some seemingly intractably unusable recordings to at least 4-star status (if we ignore the incorrect positioning of certain sounds).

Yes, I've restored some really thrilling recordings from the bin, and they'll be shared on Freesound in due course.

End of road for the PCM-D100 too!

No, I don't mean I'm stopping using mine, but rather, that an online check in August 2021 revealed that the PCM-D100 has followed the same fate as the M10 — i.e., Sony has discontinued it, without (so far) offering a viable replacement. They did release another model — the D10 —, which could be regarded as the replacement, but its noise level is distinctly higher than that of the D100, so for natural soundscape recording it really isn't a worthwhile replacement at all.

As yet, apparently no other 'hand-held' recorder of similar quality is available anywhere. The Tascam DR100 MkIII is reported to have the next-best s/n ratio, but in terms of practical usability for natural soundscapes, it would be a poor substitute. The Zoom H5 with H6 mics module would be a quite good contender, but battery life is relatively short, and, as noted further above, I found those H6 mics too directional, not being able to achieve a really well-integrated stereo soundstage — at least, in wide-angle setting.

So, I'd better not have any of my D100s fail or get lost. — And of course the second-hand price of that model will undoubtedly soar to real eye-watering prices that I most certainly would NOT be paying.

(More observations will be added here…)

My own recordings

You can explore my recordings on my Broad Horizon Natural Soundscapes page, and my Digital Downloads Store presents a listing of all my CD-quality downloads, together with preview excerpt links for each title.

'e-Store' sign with dramatic Cornish clifftop scene — The place to go for full-length CD-quality downloads of my natural soundscape recordings…