My Opinions of Three Different Multiple Speaker Groupings

What I am reporting about here involves the Hereford Cathedral Organ used with reverb tails cut to 250 ms to give a dry organ sound.

Having a lot of speakers to use for my HW organs has led me to a number of experiments in speaker groupings. I find they fall into three major types: (1) The all stereo group (42 stereo groups); (2) the mostly mono/cycling groups of 12 speakers per group; and (3) a new one where there are cycled groups of stereo pairs.

Here's what I have found:

(1) With 42 separate stereo groups one can really separate out the organs' ranks. The sound goes from very quiet to very big when using the Crescendo Pedal. As new ranks are added, new speakers come into play. A nice sound. The big advantage to using stereo for a wet (but simulated dry) stereo set is that the individual pipe sounds are more natural---more real sounding in stereo (IMO).

HOWEVER, I hear a big problem with this sound. When using the original temperament on the Hereford, when playing a single rank, when notes are out of tune a bit, one hears beats that are rather ugly I think. For example, playing octaves on a 2' Principal gives a sort of vibrato effect. I don't like that.

My solution was to go to equal temperament and set HW's pitch randomizing to 0%. That eliminates those out-of-tune beats since the pitches are now exactly in tune. But, of course, the sound then becomes more sterile. The partial solution I found was to hand de-tune various ranks from others. Thus, the 'sterile' sound was there only if just one rank was playing, and the more ranks, the bigger and better the sound. All the beating was then occurring in the air of the room, rather than in each stereo speaker pair.

I used this grouping for some time and thought it was really quite nice. Then. . .

(2) Last week I thought I would try the set-up again where I used groups of 12 speakers just to see if I could hear a difference. THE DIFFERENCE WAS HUGE!!!

The Hereford with original tuning was transformed into a big, beautiful sounding instrument, whether one rank was playing, or whether many were playing. This grouping loads the ranks in monaural and cycles the pipes through the 12 speakers per group. Each group's speakers were spread horizontally across the room's wall. So, now all the out-of-tune pipes were celesting in the room air, not just rank against rank, but pipe against pipe too. I was amazed that there was that much improvement in the sound.

I then selected the equal temperament and set HW's randomizing to 200%. Ahhh . . . there comes that big old cathedral sound where you find organs that are in need of some re-tuning. This is really a grand sound, especially with an instrument like the Hereford which has so many ranks (67).

HOWEVER, there is a problem with this sound too. To load the organ's ranks in mono, HW converts the stereo wave files to monaural. To me there is a noticeable loss of fidelity in this for some pipes. Especially with reed ranks. Some pipes take on a sound that varies too much from the others in the rank. This is mostly a problem if a rank is played solo. In big registrations, the sound is really great.

That brings me to:

(3) A few days ago I got to wondering about something Brett and Martin had suggested some time back. That was to put together stereo sub-groups into larger cycling groups. So I worked out a plan that required me to buy two more of the Rokit KRK-5 speakers. The result is two groups of 12 stereo pairs (48 speakers), and three groups of 6 stereo pairs (36 speakers). (The low Pedal ranks go to two large full-range speakers in stereo separate from these new groups.) I have found that groups of 12 or 6 make the most efficient cycling use of the speakers in a group.

My hope with this arrangement was that I could get the best of the other two groupings. AND IT WORKS!!!

Loading the Hereford ranks in stereo gives a much more pure and consistent sound to pipes within a rank (really nice reeds now, for example). And, having the pipes cycle through 12 or 6 speaker pairs gives that big room-celesting sound. Again, with equal temperament and randomizing set to 200% the sound is really awesome!

NOTE: With many ranks stuffed into these five groups, there is bound to be some of the celesting within stereo sets. but it is not a noticeable problem now.

KEEP IN MIND:

What I refer to here is playing a WET organ as simulated DRY in one's living room using lots of speakers. Actual recorded dry sample sets do very well with the regular mono cycling groups. . . stereo pairs would be a waste of speakers with a dry set that was recorded as mono samples, such as the dry Sonus Paradisi organs. However, the wet ones from Sonus Paradisi ought to do very well using my #3 grouping. I will be loading the Caen and Zwolle surround sets this way next and I expect them to be spectacular too.

This whole HW thing has grown into something of a monster by now. All 96 of the MOTU interface outputs are now in use. Boy, do those little green diodes light up all over the front of the MOTU units when the Hereford's Crescendo Pedal is opened up!!!

And the room shakes . . .

Leo Chris.

Hi Leo,

Thanks for your always interesting work in the multi-channel area. I thought that it was about time that you realised that some is reading these particular posts of yours!!

Your views on 12 channel groups concur with my experiments, and one 12 channel group is now a key element of my VTPO's audio configuration. Your audio distribution spreadsheets, which have been linked to, or included on my website for years now, have always proved to be an invaluable resource to me. Thank you once again.

When I set out on the HW multi-channel journey in 2006/7, it was by deploying 2 channel C C# groups which I had used successfully in my GS3 VTPO. I simply brought along to Hauptwerk that which had worked for me previously. However, once up and running, I then commenced experimentation with the different HW distribution algorithms. After toying with Tone Matching configurations where I attempted to offset 3rds and 5ths problems which often sounded like IM type distortion, C/C# ranks and octaves cycled became my preference.

I continued with my preferred C/C# ranks and octaves cycled as the audio grew from 24 channels to 39. Well, with the exception of two tone matching audio groups which contained the larger reed ranks where the 3rds and 5ths issues were unacceptable to my ears.

Once my custom ODF started to progress, and my confidence in my project grew, I was able to eliminate the 3rds and 5ths issues altogether without having to use any ensemble reducing Tone Matching algorithms. As your spreadsheets demonstrate, Tone Matching algorithms cause the notes from each rank to follow each other from speaker to speaker, rather than rank 1 in one channel and rank 2 in the opposite channel. TM is therefore taking one further away from a more authentic pipe ensemble and dynamic build up which C/C# ranks and octaves cycled does deliver. But there was more to learn!

I had always employed bass splits to increase ensemble, transparency, and to better match samples with audio capacity. However, I also used Bass Split to spread the two major Vox Humanas over two 2ch groups in each chamber. So, four identical speakers then handled each Vox. While this was certainly an improvement over only one 2ch group in each chamber, I still did not know what I did not know (thank you Donald Rumsfeld!). Thankfully, I have enjoyed detailed feedback from organist, Simon Gledhill. Simon took a very keen interest in what was happening my 'Ruby' project, and through his skills and expertise my eyes and ears were opened.

Simon, through his work with larger Allen VTPOs, had worked through the implications represented by musical intervals created by speaker groups of differing numbers. You can read about some of this here: http://theatreorgans.com/ianmclean/Conf ... 20Blog.htm . What is specific to this part of this post, is "Config 5" from 2011. The change wrought on the Vox Humanas (and also the Brass Saxophone) by the change from two 2ch groups to one 4ch group was incredible to me.

I then spent a great deal of time reconfiguring the, then 39 channel config to a mixture of 4ch and 2ch groups. Once again the improvement was astonishing.

But! With tremulants, there remained issues even with 4ch groups (also mentioned on my Configuration Blog). So, there came into existence one 8ch group to limit tremulants of different ranks beating against themselves within a single channel when some triads were being played when using C C# Ranks and Octaves cycled.

To make the 8ch group work effectively, I had to undertake some ODF work to alter the RankIDs which determine the channel sequence number for a multi-channel HW VTPO. To that end, I shall be forever grateful to Joe Hardy of Paramount Organ Works, who was not only prepared to make public his own audio sequencing work, but also assist me with comprehending how it all worked.

Without Joe's discoveries, simply creating an 8 channel group might not be enough. Except by pure luck if perchance, the RankIDs and their resulting sequence deliver the desired outcome for a particular sized audio group. However the reality is that the larger in channel numbers an audio group becomes, the more likely an organ is to end up with outcomes similar to the Tone Matching series. That is, ranks following each other around in the same channels, resulting in a diminished and.or unrealised ensemble, and for VTPOs, compromised tremulants. So, without ODF changes, sometimes numerically larger audio groups may not be the answer. Well, it could be a salve to one problem, only to create another.

Then, it occurred to me that the only way to address many more of the beating tremulant issues was to move almost completely away from smaller audio groups each fed with specific ranks. The smaller 2ch to 4ch groupings permitted me to 'terrace' groups of ranks from front to back, as they might be in a real pipe chamber, and this also delivered an almost 3D outcome in room. While although this could be very impressive, it wasn't actually as authentic to the chamber mix reality as it could be, not to mention the other issues already mentioned . So, I once more revisited your work, reconsidered Simon's input and the in room realities, and then decided that 12ch groups would be the best solution. It also occurred to me that 12ch groups could also better address another of my authenticity conundrums.

Over the years one of the issues which I had thought about, especially with a unified instrument like a TPO, was the balance of relative tone sources. Taking my 44 rank custom specification as an example. Let's say (for ease of maths) that there might be 3200 pipes (tone sources) in a pipe organ of the same size. Now I was attempting to simulate that virtually with only 32 channels. That's a 100 to one relative tone source ratio! Little wonder that so many VPOs sound like someone playing a recording of pipe organ, instead of the illusion of a pipe organ in the room!

Using smaller groups was not authentic in this context if one is to equate a tone source as a pipe, or a loudspeaker. For example: if the highly unified 'glue' of a TPO, the Tibia, was in a 4ch group, and let's say, a colour reed like an Orchestral Oboe was in another 4ch group (which for my Config 7 was the reality), then if one was to play a 4 note chord using 8' pitches of both ranks, then both the Orch Oboe and Tibia would be playing 4 channels each. Fine.

But let's consider next an ensemble where the Orch Oboe could be drawn at one pitch, and the Tibia at three. So, four notes of the Orch Oboe would play through 4 channels, and 12 notes of the Tibia would play through 4 channels. On a real pipe organ, four Orch Oboe pipes would be playing and 12 Tibia pipes would be playing. It seemed to me that using a 12 channel group would make for a more authentic outcome in any tone source consideration. And, it does.

Not only that but a 12ch group limits tremulants of the same rank from layering on top of themselves at too close a frequency. So nauseating tremulant 'motorboating' is next to eliminated. To my ear this more of an issue than differing rank tremulants being mixed. But, of course, even 12 channels will not eliminate a unified Tibia drawn at many pitches from ending up with itself. But, the greater the difference in pitch when this occurs in a channel, the less that tremulant motorboating occurs.

I then position the 12 channels Left to Right 1,2, then 3,4, and so on to 11,12 from front to back in each chamber in order to maximise acoustic mixing. This is closer to, but still a long way from, the assymetric layout of Wurlitzer manual pipe chests.

These days, as less has proved to more, especially due to the greater note allocation efficiency of larger audio groups, I now only use 32 channels so that my RME HDSPe/MADI - M-32DA combo is the only audio device used. Also, the difference in audio quality between the RME and my previous audio devices is not subtle. However, wanting the highest audio quality means that while I can use one 12 channel group in my 'Solo' chamber, for the 'Main' chamber, I am limited by available channels to a 10 channel group.

Currently, that's 22 channels of Behringer B2031A Truths which are then bass split to two pairs of large three ways (Klispch La Scalas and KRK13000Bs) which are in two channel groups, and the two single channel subs (Rodgers P7.5a and JBL S1SEX). I also continue to use two pairs of Rodgers M-10 speakers for parts of string and other celestes. The M-10's much larger scale (relative to the Behringer Truths) and out of phase di-polar response remain very useful in balancing celestes from two channels against ten or twelve.

Of course in a perfect world, four 12 channel groups along with another two pairs of large 3 ways would be the ideal answer. And indeed, replacing the two ways with 3 ways would also improve the outcome noticeably. But, I have neither the room, or wish to spend the money on that!

In summation, if it hadn't been your inquisitive intellect, and experimental determination regarding multi channel Hauptwerk installations, and then your willingness to share what you have learned, I doubt that my eyes would have been opened to the possibilities. That then Joe Hardy was prepared to also share in such an extraordinary way something which he could have quite rightly kept as a proprietorial marketing point of difference is what has enabled me to practically make work what was possible.

Cheers,

Fascinating! Thanks for posting this!

I have still been experimenting too. In fact I am about to post yet another one of my speaker grouping layouts. It seems to me to be the best so far. This one is just for dry organs sets, and specifically for the theatre organs, though all my other dry sets sound much better too.

This time I'm using groups of six speakers, which cycle the six just as efficiently as groups of twelve do. However, of course, ranks can pile up on themselves sooner. Anyway, I'm running 14 groups of six in the Mac Pro for the Paramount 3/32 TO, and 3 groups of six in my older PC for the Freedom Morton 3/17 TO.

The sound of these two TO's together is really a thrill to hear!

I'll get to the details in another post as soon as I have finished testing things out more.

Leo Chris.

Hello Chris,

A question of philistine: when you engage 2 stops of the same group in cycled outputs, do they both play the first C in the same speaker?

Another question: the user guide of HW doesn't explain anything about the different options in the sub-menu Output allocation within group. I'm sure you've tried each one meticulously. May I ask you a favor and tell me what each option does exactly? (I've not yet multi-channels and I cannot try it myself).

Your posts are always very interesting because few HW users have the opportunity to make so many experiences with a so huge material. What a pity I live in Europe! I should visit you very willingly...

Hello Luc,

Your first question: When using cycled groups, the first pipe of each next rank goes through the next speaker in the group. They then cycle around within that octave. Twelve speakers means a different speaker for each pipe in the octave. With six speakers, the pipes will cycle twice within the same octave. Also, since this mode cycles octaves, as well as within octaves, the next octave of a rank is shifted one speaker ahead in the group. That means that octaves of the same rank go through different speakers in the group. It's a real genius of a thing that Martin Dyde has done here! But, the more ranks per number of speakers in a group, the more likely that pipes will overlap on the same note.

It does take a lot of speakers to get the best results with a larger organ. I have tried varying the number of speakers to see what the result would be. When Martin D. showed what the algorithm for the cycling is (how HW decided which pipe goes where) I was able to put together an EXCEL file that lets you see what the results are. The file is on my webpage (the EXCEL file for download). Here is the name of the file on leochristopherson.com. Scroll down to this entry:

The EXCEL file for
'Cycle within octave,
octaves and ranks cycled'
ADDED: 4-25-2011

Have a look at that and it should explain a lot. As I recall, there is another file there somewhere that shows the results of using the different Tone Modes too.

And, if you do get a chance to visit the Northwest US, I'd be delighted to have you stop by and give this setup a try!

Leo Chris.

Hello Leo,

Many thanks for your kind answer. Your site is a veritable gold mine. Your contribution is significant for the HW community. Thanks again for all this.

Best wishes.

These days, as less has proved to more, especially due to the greater note allocation efficiency of larger audio groups, I now only use 32 channels so that my RME HDSPe/MADI - M-32DA combo is the only audio device used. Also, the difference in audio quality between the RME and my previous audio devices is not subtle. However, wanting the highest audio quality means that while I can use one 12 channel group in my 'Solo' chamber, for the 'Main' chamber, I am limited by available channels to a 10 channel group.

How do you feed the 32 channels from your computer? ADAT? I really like the look of this but cant figure out how I could get multiple ADAT paths out of a mac.

When I was using the RME gear (since replaced by Lynx Auroras for 14 months now), I used MADI optical fibre to pass the datastream using an RME HDSPe MADI card in the MacPro. From memory it will pass 64 channels at 48kHz and 32 channels at 96kHz.

For the Lynx gear I use the very bulky balanced AES connections. Two Lynx AES16e cards (in the MacPro) and four Lynx AES cables. For me, this is a superior solution. Some in the recording studio space consider AES superior to optical solutions. Maybe due to the extra overhead which the muxing and demuxing for the optical adds? It was not possible for me to do an optical vs AES A/B comparison with the RME gear as the RME M-32DA DAC did not have a provision for AES connections.

Cheers,

Ian

Multiple "Echo Audiofire 12s" I think you can loop through the Firewire 400 through multiple units, though I don't know the limit. I think the biggest is 5 units.units of 12 channels. You also must loop through the wordclock signal so that the timing stays consistent between all the units.

My latest setup is 3 divisions of 12 x 2031As; Main, Solo and Tuned Percussion; I have divided the pedal into main and solo, using the bass split thru 4 full range 15" speaker cabinets and the sub bass is setup as auxilliary in the audio assignments. I have 60 audio channels provided by five echo audiofire 12s. The 3 main groups are set to tone matching two.
The groups are as follows;
1: Main
2: Solo
3: Tuned Percussion
4: Main Pedal
5: Solo Pedal
6: Effects 1
7: Effects 2
8: Effects 3
9: Effects 4
10: Effects 5
11: Traps
12: Aux Main Sub
13: Aux Solo Sub
14: Strings thru the Conn Pipes C/C#
15: Aux Echo thru the Alesis Microverb.

This arrangement has been the most satisfying to my ears to date.