Near Coincident Mics - phase problems?

Something of a discussion has been underway between email associates here concerning recording devices equipped with stereo mics such as the Sony PCM-D1 and the Zoom H4.  The right-facing mic is  located very slightly to the left of the left-facing mic - so they are like a crossed pair but not actually crossed.  See http://www.samsontech.com/products/productpage.cfm?prodID=1901 for a picture.  Compare and contrast with the Rode NT4 which does have crossed capsules - http://www.rodemic.com/?pagename=Products&product=NT4

Some have said that the Zoom H4  arrangement is seriously flawed in terms of imaging and phase errors.  I'm not so sure that in practice it would actually matter that much.  Any of you good gurus care to comment?

Some have said that the Zoom H4  arrangement is seriously flawed in terms of imaging and phase errors.  I'm not so sure that in practice it would actually matter that much.  Any of you good gurus care to comment?

Well, it's a fixed, therefore correctable error - but I'm  not convinced that it will make that much difference either. The capsules on my Soundfield mic are about that far apart (and corrected...) perhaps they'd like to complain about that too?

I could work out how much error was being introduced quite easily - but I can't really see a lot of point in the exercise - and anyway, the prevailing conditions during recording will make far, far  more difference to the sound than the odd cm of difference in spacing between the zoom and the Rode. Even with the most coincident mics in the world, you can't get the capsules absoulutely adjacent anyway - and IMHO, the zoom arrangement is no more flawed from that POV than having the capsules on top of each other, as the Rode does.

Well, it's a fixed, therefore correctable error

It's not correctable, as it varies with direction (the case of the soundfield isn't directly comparable).  Michael Gerzon preferred, when recording with two cardioids (which he did a lot, never having a SoundField of his own), to have a slight overlap (an inch or so) which he found improved imaging precision.  Equally he was very clear that both theoretically and in practice, "underlapping" the capsules, as in the Zoom H4, had a bad effect on imaging.  Of course, some people don't place much importance on precise imaging, and so have no need to worry about things like this.

Paul

Well, it's a fixed, therefore correctable error

It's not correctable, as it varies with direction

That doesn't mean that it's not correctable - it just means that nobody's bothered to write a DSP algorithm to correct it. And you can't claim that its not possible, either, as Audition can extract directional signals at different angles - that's how the center channel extractor works, and I'm sure that the principle could be extended to 'correct' a stereo field. But as to why nobody's bothered...

Well, I'm not doubting for one moment that Gerzon's experience was as he stated - but as to how much it actually matters, or is audible in any given situation is far more debatable. And as far as I'm concerned there are times when too much of a precise image can actually hinder the sound being recorded, anyway, because it can be distracting. I used to take a lot of trouble to ensure that imaging was as good as I could get it, but over the last couple of years, based on some end-user feedback, I haven't worried so much on several occasions. And it's not that these end users aren't being critical, because they are - they are just saying that there are times when they wish to hear an amorphous mass of sound across a soundstage, without individual sounds being identifiable at all. Some venues obviously make this a lot easier to achieve than others, but there are some mic techniques that make it easier still!

I agree that for most people it's unimportant; but I would need more convincing (and here's probably not the place) that a decent correction could be done. 

But if there's any audible effect I would expect it to be deleterious.  Consider the way in which spacing and pattern work together effectively in ORTF, and then consider the effect of reversing the positions of the mics; it simply wouldn't work because the directional cues would be directly opposed - one mic would be "left" in timing but "right" in amplitude and the other the opposite.  The H4 arrangement is a modest step towards that situation.

Paul

Interesting.... I have the more recent Zoom H2 arriving in the next week or so, I hope, and that's said to have similar mic placement internally (except four mics) - but without ripping the top off, you can't be sure.   Anyway, I'll be testing the hell out of it and will report back.

Perhaps it's all partly a matter of personal preference - I totally hate the imaging from the Rode NT4 because it seems to provide no information about the room acoustic (and to me it has a hard and brittle sound too) - recording others have made with it I've found need to have 4dB of side gain added (using an inline MS VST) to produce at least some kind of acceptable sound from a classical recording.  That's compared to my own MS rig, which I've used for 15 years and therefore that's what I like!

But if the Zoom H4 configuration does lead to incorrect time-of-arrival info conflicting with the intensity info, wouldn't that mean that if you mounted two small omni mics right beside each other pointing forwards, so that only time-of-arrival info was obtained from the closely-spaced pair, you'd hear some kind of stereo?  I suspect you wouldn't.  Easily tested of course.  If you didn't hear that ToA info from such a configuration, why would you hear it (unwanted) from a similarly closely spaced cardioid pair?  Any further thoughts?

Perhaps it's all partly a matter of personal preference - I totally hate the imaging from the Rode NT4 because it seems to provide no information about the room acoustic (and to me it has a hard and brittle sound too) - recording others have made with it I've found need to have 4dB of side gain added (using an inline MS VST) to produce at least some kind of acceptable sound from a classical recording.  That's compared to my own MS rig, which I've used for 15 years and therefore that's what I like

That's because 90 degrees isn't the right angle for a pair of cardioids in this usage.  See this paper by Gerzon (again - sorry!) which shows the distribution of reverberant information across the stereo soundfield (quaintly represented in terms of needle movement on an LP in this early paper!) for several coincident techniques; 120 degrees is about best for a uniform spread using cardioids - anything less and the acoustic piles up in the centre of the image (but as it's not MS, widening the image after recording won't be as effective as widening the mic angle before).

Paul

That's because 90 degrees isn't the right angle.....

....But Paul, 90° IS a Right Angle Sorry! only kidding, carry on the discussion. Try "correct" next time

That's because 90 degrees isn't the right angle.....

....But Paul, 90° IS a Right Angle Sorry! only kidding, carry on the discussion. Try "correct" next time

   Quite right ; but in retrospect, "best" would have been even better in this case.

Paul

That's because 90 degrees isn't the right angle for a pair of cardioids in this usage.  See this paper by Gerzon (again - sorry!) which shows the distribution of reverberant information across the stereo soundfield (quaintly represented in terms of needle movement on an LP in this early paper!) for several coincident techniques; 120 degrees is about best for a uniform spread using cardioids - anything less and the acoustic piles up in the centre of the image (but as it's not MS, widening the image after recording won't be as effective as widening the mic angle before).

I don't think that it's anywhere near as simple as that, although I agree that 90 degrees generally isn't a good angle. It's not too difficult to calculate that something nearer to the 'correct' angle for a pair of accurate cardioids is about 115 degrees, but that doesn't really matter, because it's only a theoretically correct answer; real mics don't exhibit accurate cardioid characteristics over the full frequency range anyway unless they cost a small fortune...

The problem with the Rodes, and any other mics like this is that the angle is fixed. By Gerzon's own admission, his recordings were limited, and perhaps he didn't try recording in some venues where 120 degrees would have been completely inappropriate, and left a large hole in the centre of the stereo spread, but I have - notably on two occasions. On both of them, 90 degrees produced a much better result. The first time I noticed this significantly was years ago, on a Cathedral organ that was (still is, actually) laid out so that the disposition of the instrument left a large hole in the centre of the casework where nothing effectively speaks from, and it has all the reverb not in a diffuse field, but predominantly from the far left and right when you get microphones in a sensible position. On the occasion that I recorded it, 90 degrees worked out to be a very good compromise in terms of the stereo field.

The other time was the recording I had to take a guess at in a hurry, and I think that there's a link to it somewhere on the site still, although I haven't looked to see if the file is still there. This involved a wide, relatively shallow room which has hard walls with a choir in it, taking up too much room for my liking. More of them were on the sides than in the centre, and this worked out pretty much the way I expected it to - and sounded well-spaced with, once again, 90 degree mics.

So I'd say that you have to judge what's going to work best in any situation based on more than just the correct theoretical angle for a crossed pair - because that won't always get you the best result.

But if the Zoom H4 configuration does lead to incorrect time-of-arrival info conflicting with the intensity info, wouldn't that mean that if you mounted two small omni mics right beside each other pointing forwards, so that only time-of-arrival info was obtained from the closely-spaced pair, you'd hear some kind of stereo?  I suspect you wouldn't.  Easily tested of course.  If you didn't hear that ToA info from such a configuration, why would you hear it (unwanted) from a similarly closely spaced cardioid pair?  Any further thoughts?

Okay, returning to the original question about the spacing - well, if you look at the theoretical basis for ORTF, and the acoustics behind it, you discover that the ear relies only on phase information for location up to about 2kHz, and that this process starts to fall off at about 700Hz. - ORTF at higher frequencies relies on amplitude differences between the mics. This isn't unreasonable - at 20kHz an entire cycle of a wave only occupies 16.5 mm - so moving a mic that distance is going to give you a 360 degree phase shift. Ergo, all mic pairs, however carefully you align them, are going to result in phase shift errors at high frequencies, even if it's only a few degrees. And I would contend, on the  basis of what's known about ITD and IID, that this really isn't significant at high frequencies, and only makes a very, very small difference at low ones, where the wavelengths extend to more like the length of a bus.

As an aside, A.D Blumlein knew this (the ITD/IID tradeoff) in 1931 - as anybody reading Patent 394,325 would realise. Here you will find the first drawing of what has subsequently become known as the Jecklin disc, although I'm not sure how much he said about the baffle. The drawing shows something suspiciously similar, though. Yes, he was using omnis - he specifically refers to 'pressure' mics. And he says specifically "the microphones are sufficiently close together as to render phase differences of the incident sound negligable, and the output amplitudes therefore differ approximately proportionally to the obliquity (lovely word!) of the incident sound". He used the original 'shuffler' on these signals - and it has been sugested that if you increase the amount of shuffle quite a bit, you can still get distinct stereo from two relatively closely-spaced omni mics.

What I'd still like to know, though, is how you are supposed to operate the controls on a H4 when you've got its microphones into a suitable position for most recordings!

So I'd say that you have to judge what's going to work best in any situation based on more than just the correct theoretical angle for a crossed pair - because that won't always get you the best result.

Don't expect me to disagree!  Theoretical solutions assume theoretical situations, which we don't often have.  I spout a lot of theory because I believe that an understanding of what's going on and also the limitations of the theory is a good starting point for moving on to making pragmatic decisions.

Paul

Thanks for the further intriguing responses - I'm getting the feeling that if it works, don't worry about why, and if it doesn't, the reason is clear!

What I'd still like to know, though, is how you are supposed to operate the controls on a H4 when you've got its microphones into a suitable position for most recordings!

Indeed!  In contrast, the H2 is operated upright, which gives you more of a chance of operating it while pointing it - see http://www.zoom.co.jp/english/products/h2/index.php - and cunningly provides mics set at 90 degrees on one face and 110 degrees on the other.  And you can record from one pair, or the other pair, or both pairs to one stereo file, or each pair to its own stereo file simultaneously - which should give rise to some very interesting possibilities in post production.  For example, in the scenario of recording a classical performance in a decent acoustic, one could vary both the theoretical width of each pair, and their relative level, to best advantage - or perhaps even turn it into an MS pair of sorts (sum the forward facing pair to mono, widen the rear facing pair to kind of simulate fig of 8, and mix with appropriate phase manipulation).  I may have a chance to try mine out on Stephen Hough on solo piano at the end of the month in rehearsal in a 3000 seater hall... or might even have the nerve to use it for the concert (as backup to the proper recording).

One of the troubles with any crossed-pair mic in comparison to your MS Sennheisers is that when you convert to MS, you are creating the M signal with the side response of two cardioids, rather than using the much better response you get from the Senn forward-facing cardioid and figure-8 combination. When I tried all the possible mic permutations as direct comparisons with my collection of AKGs, there was always a subtle difference in the M sound, even with good cardioids, from what you got with a dedicated M mic. It follows really that the more indifferent the side response of the mics you convert to MS, the greater the difference in the summed sound will be, in comparison.

I have to say that in the tests, the best sound I got was from a figure-8/omni combination - in other words, what amounts to a Blumlein pair, in terms of the response when converted to LR. And that was from a generally quite mismatched SD omni and LD fig-8. The other thing that worked quite well was a hyper-cardioid with the fig-8 - you still got some rear response, but not so much of it, and you have to muck about with the MS response a bit to get the most out of the sound. That's what a 'fatter' polar response gets you...

Some brave soul has now ripped apart a Zoom H2 and the photos of its innards can be seen at http://s243.photobucket.com/albums/ff112/aramri/Zoom_H2/ - the mic layout is intriguing.  They've gone for the same general arrangement as in the H4, namely with the mics in each of the stereo pairs facing across each other, but about as closely spaced as possible - which in the case of the 120 degree pair leads to one mic appearing to partly obscure the front face of the other.

Now there's no obvious reason why they could not have mounted the mics pointing outwards from each other - a square of mics in essence with their faces pointing outwards.  I can't believe that they didn't think of that and indeed I'd like to think that they'd gone so far as to test it.  But they've settled on the literally crossed pairs. 

Looking at it it's hard to take it seriously, but some of the recordings posted on line so far are pretty impressive, all things considered (eg price and size).

Well, I've been using - no, playing with - my Zoom H2 for the last few days, and as a point and shoot recorder it's pretty impressive for the price and size.  Imaging seems fine to me.  Most of the facilities (limiters, digital gain, etc) are best avoided, but if used as is for straightforward recording from either pair of mics or all four, it does the job it's supposed to do well.  Level setting is by three position switch but the mid position (used in the sample following) seems to cover everything from near silence to heavy traffic.  In fact, there's no noise penalty from raising the level in Audition of a recording made in "M" to the level that it would have had if recorded in "H" - in other words, the M setting is redundant really.  Makes life easy.

Here's a sample which helps judge the noise level.  This was recorded while strolling round the local park with the dogs - sounds of circulating panting dogs, Australian birds, distant helicopter, distant train and traffic rumble, and sounds of me getting the dogs back into the car and then climbing in myself, and starting the engine and the radio.  It was mixed down from the original two stereo 44.1/16 wave format files from the H2's 4 channel mode, using a method I've found gives good results, namely, treating the front pair as a Mid mic and the rear pair as a Side mic and mixing as if an MS pair.  Link is http://www.fileden.com/files/2007/9/22/1451533/Dogs%20in%20park%20for%20net%20MS.mp3 (3.9Mb 192kbps mp3 mix).

Not much going on here at present so no music samples to share yet.