Page 1 of 1

stereo + "mop up"

PostPosted: Sat Oct 09, 2010 6:43 am
by Stephen Phillips
Hello all,

There was a long thread....

I think we all learned at least something about IMD, air-mixing vs. electronic signal-mixing, advantages of bi- or tri-amplification etc. On the whole, a far from useless exercise. I admit to an 'incomplete' digestion of the thread's contents, but that does not detract from my above statement.

"The best lies have some truth"

I attended two hectic International acoustics conferences recently, and found much to fascinate and humble. One particularly striking research effort had been conducted by a trio of intrepid Finnish scientists who had gone to the trouble of recording, using a 'shroud' of omnidirectional microphones in an anechoic chamber, many of the instruments of the symphony orchestra (including soprano voice). Their purpose was two-fold: to investigate in high resolution the directivity characterisitcs of the various instruments, free of polluting reflections, and to prepare discrete orchestral-part recordings that could be synchronized and reproduced over a suitably complex multi-channel playback system in actual concert venues, with a view to comparing conventional 2-channel or surround-sound recordings made at these venues (made using these pre-recorded takes) from equivalent listening positions. I was treated to a demonstration (as in any conference venue, less than ideal) which proved to me the integrity of their painstaking work and the believability of the resultant 'live' playback. For anyone interested in the papers dealing with these efforts, please PM me.

I mention this because there was an unmistakable 'trend' in the results of the directivity research, suggesting that indeed, for the greater part and largely irrespective of instrument, directivity of real orchestral instruments is not so very far removed from the very familiar behaviour of standard monopole loudspeakers i.e. omnidirectional at sufficiently low frequencies, narrowing to a predictable beaming at some 'crossover point'. Interestingly, the woodwind instruments (as might be expected) exhibited the most complex and unpredictable behaviour, I expect on account of the many possible points of emanation for their sound output. This behaviour, naturally, is frequency dependent. Brass instruments display an expected almost constant directivity on the axis of the bell.

We are all now wondering: what would have been the result had various kinds of organ pipe been tested?. I will harrass the excellent scientists of my recent acquaintance towards this end; meanwhile the question remains: how do pipes of various pitch and nature project their sound (fixed frequency and amplitude, of course) into the hall or casework? We might aid in the ultimate realism of reproduction by an awareness of these physical properties.

The other point is the one raised by Martin: the various approaches to high-resolution sampling do capture in a very substantial way the 'character' of the pipe. This surely can apply to close-sampling as well as relatively distant. Of course pipes are not meant to be 'listened to' at point-blank range and, if recorded in this manner, had best be allowed to 'develop' in either a real or virtual acoustic. I'll add my two cents worth on the issue of signal mixing effects on dry samples — it (signal-mixing distortion) is obvious and objectionable using even very high-quality 2-channel systems. IMD as such is of course a quantifiable problem; if one is limited to a two channel source (or prefers to be limited to such) then we rely on the competence of the relevant reproduction equipment. The case of the chromatically sampled pipe organ is special, and we need not be so absolutely bound.

Conventional 2-channel playback (or, for that matter, it's extensions ito 5.1, 7.1 etc.) must be acknowledged to have its own problems. I have more than once on this forum mentioned the ravages of inter-aural crosstalk and state-of-the-art means to ameliorate this problem. On the other hand, the establishment of of a believable ambient/reverberant field is something that can (or should) command the attention of all enthusiasts, not just 'domestic' players or listeners. In some senses the home user has a certain advantage in this regard, at least potentially, against the predicament of the large-venue installation which even so may lack a really generous reverberant field.

What is best? Principles developed for hi-fidelity listening can be gainfully employed. One thing we must acknowledge is that, even when equipped with two additional 'surround' channels, we should not (ideally) attempt to push this audio data out into the room through a pair of widely-spaced loudspeakers. Our hearing mechanism cannot make much of this in terms of a believable soundfield — though the data is unquestionably of value and can be better deployed. How? Especially considering the relatively fixed seating position of the primary listener (in this case the player...) what we can establish, with the appropriate hardware/software is a rear 'stereo dipole' — that is, two loudspeakers positioned very close together and fed with crosstalk-cancelled signals. The rear soundfield will occupy a broad hemisphere with the directionality of the various later reflections intact (something that cannot be achieved through conventional reproduction). Those who might object that large venues exhibit statistically diffuse soundfields should do some critical listening in these spaces.

By definition, these sorts of recordings are 'wet'. The same sort of subjective result can be had through the use of appropriate convolution engines and a multi-channel output and side-rear loudpeaker array. Of course, the convolution data, even if from the exact same organ-installation venue, is uncorrupted by the inevitable presence of 'frontal sound' in the rear-hemisphere 'channels' present in conventional surround-sound recordings. If opting for 2-channel frontal reproduction, the closely-spaced speaker + crosstalk cancellation approach will deliver huge dividends as well.

What of dry sample reproduction in domestic, relatively damped acoustical conditions? Here we can easily differentiate between the frontal and rear soundfields, with one important included observation. The 'dry' pipe sound, naturally would be expected to emanate from the front (except for those occasional remote-rank situations such as I encountered recently in Melbourne), though even this is necessarily complemented with a significant contribution of case and frontal room ambience/reverb. This must be provided, if not in the recording, then by subsequent processing (appropriate impulse responses will do nicely). The 'greater' soundfield can be similarly and authentically presented, preferably with a bank af angularly displaced loudspeakers fed with discrete convolved signals. In a properly damped room, the results can be very believable and powerful. In this there is a 'crossover' between the requirements of the wet- and dry sample user. The fact that rear reverberation is unavoidably captured in either 2-channel or 4-channel (front L-R) 'wet' sample libraries, while regrettable, is not terminally destructive to the beauty or usefulness of these virtual instruments, anymore than would be the case for a conventional stereo recording of the same instrument and venue. We might, however, consider the advantages in addressing the binaural behaviour of our hearing mechanism when recordings (CD-type or sample) are made. Many issues with signal-to-noise ratio and directional issues can be eliminated effectively through recording techniques that focus on the frontal sound and suppress the rear, preferably through acoustic baffling (as directional microphones have many anomalous behaviours in addition to their 'hesitant' bass response) with later addition of convolved ambience/reverberation. A body of empirical evidence should encourage us in this direction. A believable ambience is indeed the linchpin of our efforts and hopes in this regard. Having this successfully established, much else can be less than optimal to little negative effect.

I am conscious of the length of this contribution, and promise not to make a habit of it..... (I await HW 4 as eagerly as the next man),

Cheers to all,

Stephen.

Re: stereo + "mop up"

PostPosted: Sat Oct 09, 2010 8:05 am
by pwhodges
I would suggest further reading about Ambisonics (for reproduction that aims for realism with no preferred directions (in full 3D), and Ambiophonics (for crosstalk-cancelled reproduction through close speakers). Some people even use ambiophonics for (2D) reproduction of ambisonic recordings.

Paul

Re: stereo + "mop up"

PostPosted: Sat Oct 09, 2010 9:14 am
by engrssc
stephenphillips wrote:I am conscious of the length of this contribution, and promise not to make a habit of it.


Most interesting and even enlightening. I needed to read thru it several times. Sort of like searching for the Holy Grail.

Rgds,
Ed

Re: stereo + "mop up"

PostPosted: Sat Oct 09, 2010 9:31 am
by toplayer2
Stephen and Paul,

Oh how wonderful it is to read discussions of this sort from well informed contributors such as you!

The two schools, ambiophonics and ambisonics (similar sounding names for two very different approaches) both have merit. Glasgal's methods seem to provide convincing results from 2D pre-recorded material whereas ambiosonics relies on special mic arrangements to produce the 3D effect that is also quite convincing. Please correct me if either assertion is off the mark, I am working from memory.

Virtual organs provide an opportunity to explore yet another avenue in some ways akin to the Finnish explorations. We are not limited to two channels of pre-recorded material. Taken to extreme, each pipe could have its own speaker. A pet theory of mine is that a multiplicity of directional queues helps with the perception of individual pipes. Further, it seems that the comb filtering effect as frequencies mix in the air also fosters, almost paradoxically, a perception of clarity and tangibility to the soundfield.

My humble experiment will be to set up two twelve channel arrays using identical speakers (Mackie HR824), one array for the left stereo signal and the other for the right. The dry sampled ranks will be routed to these arrays using "Cyclic with octave, octaves and ranks cycled". These arrays will therefore constitute the direct component. Reverberation will be generated by capturing the air mixed output from the twin arrays using stereo microphones. This signal will drive Gigapulse convolver running in a separate computer. There will be 9.1 channels exclusively for the convolver. Larry Seyer has recorded some beautiful IRs for Gigapulse that include up to fifteen unique mic positions. It is hoped that the use of nine channels will provide enough directional queues to create a good illusion of a solid reverberant soundfield.

To build on Stephen's point about recording pipes too close, I couldn't agree more. To me, such recordings sound strident and unnaturally bright and lack any sense of space that is provided by a somewhat more distant mic placement. For the Paramount, I use recordings made with the microphones placed just outside of the pipe chamber. This gives a less strident sound and captures the air mixing (from reflections) that takes place inside the chamber.

Joe Hardy
Paramount Organ Works

Re: stereo + "mop up"

PostPosted: Sat Oct 09, 2010 4:51 pm
by pwhodges
toplayer2 wrote:Glasgal's methods seem to provide convincing results from 2D pre-recorded material whereas ambisonics relies on special mic arrangements to produce the 3D effect that is also quite convincing.

Well, I wouldn't say "special" mic arrangements, except to the extent that any specific technique is special when viewed from elsewhere! Yes, for ambisonics you need several microphone capsules closely coincident - but I use coincident mics for stereo anyway. You can see an arrangement of conventional mics for ambisonic recording (without height) on my website here, and the mic I currently use for (3D) ambisonic recording here; of course, I use this when I require stereo as the end product as well. Some major ambisonics researchers use ambiophonic techniques for playback, in 2D at least; but I find it philosophically at odds with the complete uniformity of ambisonics regardless of direction.

To build on Stephen's point about recording pipes too close, I couldn't agree more. To me, such recordings sound strident and unnaturally bright and lack any sense of space that is provided by a somewhat more distant mic placement. For the Paramount, I use recordings made with the microphones placed just outside of the pipe chamber. This gives a less strident sound and captures the air mixing (from reflections) that takes place inside the chamber.

Two comments here. Firstly, the increased HF content of close-miked stereo recordings has often been noted, but seems to have become the standard hi-fi sound, any departure from which is cause for criticism - this is a Bad Thing™. Secondly, in principle, if closely miked dry samples are played back in an installation, the distance of the speakers from the listener should restore the HF loss appropriately. Of course there are then other issues, such as avoiding electronic signal mixing by careful use of multiple speakers. On another forum (!) a contributor has made the point that "nearly dry" samples (as you describe doing) do not suffer from the signal mixing problem nearly as much as dry ones - this is a good observation, I think. He went further to say that he had taken a set of very dry samples, and had made them "nearly dry" by convolving them individually with impulses representing slightly different positions, as they would have in a real organ, and that these had then been playable without signal mixing problems as well.

Paul

Re: stereo + "mop up"

PostPosted: Sat Oct 09, 2010 6:01 pm
by toplayer2
pwhodges wrote:Two comments here. Firstly, the increased HF content of close-miked stereo recordings has often been noted, but seems to have become the standard hi-fi sound, any departure from which is cause for criticism - this is a Bad Thing™. Secondly, in principle, if closely miked dry samples are played back in an installation, the distance of the speakers from the listener should restore the HF loss appropriately.

Just yesterday I happened to listen to a recording made with my own samples that were produced ten or fifteen years ago. The samples were recorded as close to the speaking pipe as I could get without bombarding the microphone (mono) with wind blast. I mused to myself as I listened just how bloody awful the sound was. Terribly screechy. The sound one hears out in the auditorium is never like that. I too have used the term "slightly wet" (1) to describe samples that are mic'd with a bit more distance. Trial and error has led me to the current practice of positioning the stereo mic pair just outside of the shades. I'm still not totally happy with the ensemble with large registrations, but I think signal mixing is the main villain causing the perceived harshness.

Joe

(1) http://forum.hauptwerk.com/viewtopic.php?f=6&t=2602&p=16235&hilit=+slightly+wet+#p16235

The Barton samples were later discarded, they were mic'd too distantly.

Re: stereo + "mop up"

PostPosted: Sun Oct 10, 2010 3:42 am
by Stephen Phillips
Dear Paul,

Thank you for again making those links available to the Hauptwerk community. The information is important and much of it can be directly applied to virtual organs. Your own very well organised web pages are full of interesting and relevant facts and opinion too.

Ed,

Ha! I like your 'Grail' imagery. One of our broadcasters here in Australia recently aired "The Da Vinci Code" which I finally watched (but I dare not go there...) Actually there is a piano sample set, released quite a few years back, that 'dared' to call itself "The Holy Grail" which I found, based as it was on a Kawai of all things, to be more than a little presumptuous. We will in any case continue our quest, with Martin's deeply engineered software and a genuine wealth of internet-and-other-sourced information — at least while our bank accounts hold up!

I think it valuable, at this stage of the thread, to make a couple of additional observations, especially for those possibly a little lost in the woods. Ambiophonics and Ambisonics are both theoretical approaches intended to construct, for a single listener for best results, a soundfield that bears no resemblance to that native to the listening room. I am trying to find my words carefully here because, with Ambisonics the resultant desired field is that of the recording venue, with scope for substantial manipulation of this coded data. In the case of Ambiophonics, the primary stress is upon realism, and therefore, as a fully-realised discipline goes far beyond a record-replay strategy (though it does dwell on optimised microphone designs and practice), encompassing such things as speaker and room correction, room treatment etc. How much of all this is directly relevant to a particular situation depends on the goals of the user. It is also surprisingly (and thankfully) extremely backward-compatible with zillions of already existent recordings, be they two- or multi-channel or, for that matter, pipe organ samples. Since the dedicated hi-fi enthusiast and the virtual organist (like the real) are both obliged to stay relatively stationary in pursuit of their (in many ways) essentially individual passion, the benefits of the philosophy are there to be taken advantage of. This, of course, is also true of Ambisonics.

An important point of distinction is that Ambiophonics is at heart a binaural approach, achieving its great success (in part) by permitting the pinnae of only the end-listener to 'sample' the resultant soundfield (as opposed to classic dummy-head type arrangements). It is also not particularly interested in any one acoustical field in particular, only that the one generated be believable. This is the part of it that can be gainfully exploited by dry and 'drier' sample set users, or those wishing to experiment with truncated release tails. There is plenty of information on the website that can assist in realising working systems, and I have nothing that I can really add to that, except perhaps to assist in a thought process that would translate some of these concepts into working virtual organ systems. Taking on Martin's assertion that the replay of a single-position stereo wet sample set is an equivalent audio event to the replay of a CD made at the same location, then, for those players wishing to exploit this realism, Ambiophonics will vastly increase the spatial accuracy and dimensionality of the experience. As my earlier post mentions, this is also true of those organs made available in surround — they have a glory burned into them that even the sample developers may not have guessed at themselves.

Cheers,

Stephen.

Re: stereo + "mop up"

PostPosted: Sun Oct 10, 2010 4:18 am
by Stephen Phillips
And Joe,

24 Mackie HR824s is in no way humble!

If I understand your proposed experiment correctly, there is the risk of provoking feedback in the system, as the sounds of the reverberation loudspeakers stand a good chance of 're-presenting' at the microphones which, of course, will not differentiate this from the input from the 'pipe' speakers. Would still be intriguing to hear the results and, as you say, the work Larry has done is top-notch and extremely versatile.

If you do run up against instability, you could at least do a variation of the experiment by committing the two-track mic feed to 'tape' and then putting this through the same Gigapulse system. You will have the air-mixed signals and eliminated the possibility of feedback but, of course, you cannot play this as an instrument in a room. Or you could try baffling the microphones, something after the style of the Ambiophone, a description of which you will find on the Ambiophonics website. It is structured so as to deliberately suppress overhead, side and rear pickup, which seems to me to be very much in your interests. An entire session of the Melbourne conference was devoted to assisted acoustics systems, really for already-large venues, but these things are stratospherically expensive. Lexicon did at one time make available, on the now-discontinued MC12 controller, a 'domestic' version of their LARES system, called LIVE (Lexicon Variable Environment). Using permanently-mounted microphones, it is designed to avoid feedback by ingenious manipulation of the processed signal. This is not a convolver but an elaborate programmable algorithm base devised by David Griesinger, formerly Lexicon's chief engineer. As he was present at that very conference (indeed he seemed 'omnipresent' in certain respects) I took the opportunity to ask him about all this. He made the very relevant point that you still need a decent room in which to operate, that is, large enough not to have insurmountable problems with midbass standing waves and sufficiently treated to avoid nasty flutter echo etc. His work on the subjective effects of phase-shift and the fascinating (and non-intuitive) contribution direct sound makes to listener envelopment will, I think, be landmark contributions in that field. Ever the empiricist, he provided both public multi-channel loudspeaker and private binaural earpiece demonstrations of his wide-ranging international research, including recordings made at Leo Beranek's personal seat in Boston Symphony Hall. Beranek, at 96 (!) was also present, so David was not getting away with anything!

Anyway, please keep us informed of your progress. How do you anticipate arranging the geometry of these two dozen active speakers? Distance from microphones? Other details? Pictures will be a great assistance...

Cheers,

Stephen.

Re: stereo + "mop up"

PostPosted: Sun Oct 10, 2010 8:38 am
by toplayer2
Stephen,

Thank you for another very informative and thought provoking post.

The idea of using a baffle had occurred to me. The organ room ceiling uses 2x2 drop ceiling (overlaid with 3"rock wool insulation). I had built a ceiling mounted 2'x2'x4" box with the open end facing forward to house a video projector. Another like this will be constructed as a baffle to house a Zoom H4n.

Here is a crude diagram of the organ room:

Image

The purple shaded area is completely enveloped with Auralex wedge foam panels to create a "dead end" and has bass traps in the corners. The Mackies are arranged into two 4W x 3H arrays. They are connected to a MOTU 24 io PCIe audio system. The second PC with Gigapulse uses an M-Audio 1010LT to drive the 9.1 audio system (which also serves the home theatre system). The corner location for the Trinity seems to work well to minimize the build up of standing waves thanks to the non-parallel boundaries.

This would be over the top for most VPO users, but the cost is considerably less than many brand name digital organs.

Joe

Re: stereo + "mop up"

PostPosted: Sun Oct 10, 2010 4:18 pm
by Stephen Phillips
Hi Joe,

A very instructive plan of your impressively equipped room, thanks. If it were not doing 'double duty' as a home theatre, I would strongly urge you to consider 'livening up' the area around the Mackie banks, to assist in generating some of the early reflections an organ case would naturally provide. You might be able to get that part of the soundfield 'behaving properly' by doing an electronic impulse chain for the front speakers — it would create a different total signal for the room pick-up mics but that is also true of an actual organ where the room is reflecting sound which has emanated in a complex manner from the case.

I don't know from where in the room you view the video material, or for that matter how many people you regularly accommodate for that purpose. If you are really interested in maximizing the dimensionality of your experience for a favoured listening position (like the organ console...) you should seriously consider deploying your speakers (not the Mackies obviously) along Ambiophonic lines, with crosstalk cancellation (and it frees up one of the front speakers as well). Check out the website for detailed information on how to do this if it is applicable for your case.

I also do not yet know what the conditions are like out in the room, as far as acoustic control is concerned. If you are trying to establish a soundfield foreign to that space (which would also be true of film) then it should be pretty well damped — with that many speakers to play with there is little point in trying to bounce the sound around further.

Hope these few remarks are helpful. Please let us know how things progress.

Cheers,

Stephen.

P.S. If these samples are truly dry they will probably need some EQ, either in HW itself or subsequently. The distance to listener is dramatically shorter than in most situations with a large-scale instrument.