Rational Acoustics

August 22nd, 2008, 05:38 PM
Hey all,

Calvert just passed me this document. It's a "classic thread" on setting delays from the old SIA forum. At some time in the not-too-distant past, Calvert embarked on a mission to collect and archive classic threads from the support forum - while the project was waylayed by some occurance (now forgotten), he did get this one put together. It's interesting reading both for the subject matter and for the cast of characters that chime in (Dave Gunness, Pat Brown, Thorny, Stephen Siegle, . . .).

I submit it for reading and comment.

Note: I do not necessarily agree with all of the methodology espoused herein - I am a big fan of setting delay exactly at at least some reference point in the delay speaker coverage. Extra delay = improved imaging at the expense of reduced sonic quality.

Setting delays is an interesting mixture of Art and Science. So many factors to weigh in deciding a simple question of delay and level . . .



Harry Brill Jr.
August 23rd, 2008, 01:10 AM
Well what can I add to that?

I also lean toward no error in my delay setting (notice my wording ha ha). The key to this is in the design, then in the compromise made during optimization.

In a perfect world, the design would have relatively narrow coverage delay speakers, spaced evenly in a radius with the center being a well designed cluster. As you walk out of the coverage of one delay speaker, you would walk into the coverage of another. The delays would be equal in level and time at the spatial crossover between them. The delays would be equal in level and time with the main system on axis to each delay. As you move toward the main system, under the delays, you would move off axis to the delay, lose level, and the combing would be dramatically reduced, while the image was maintained at the main system.

My world is full of compromise yet I still prefer the method which gives no error. The key for me is choosing the spot where the delay will be equal in level with the main, then aligning the delay to that point in space. This choice is critical and I often fail on the first attempt. Ultimately my ears tell me if I've succeeded at finding the right compromise.

I'm usually working in corporate theater (talking heads), so delays are generally on a straight line (no arc), and there is a left and right main cluster, not a single center cluster. Since the delays are often taking over, and I'm likely to be asked to turn them up if I measure on axis, I will tend to walk under the delays, not completely off axis, but close to the -6dB edge. Since they are flown in this case, the loss in level at the edge of coverage is offset by a smaller distance compared to the on axis placement. A well behaved loudspeaker will still have a reasonably consistent response off axis. The floors are flat in most ballrooms. In this case the error under the speaker is far less noticeable since it is removed, and the error on axis is not much of an issue because the delay speaker is losing level faster due to inverse square law.

This is just one example of a compromise based on the ideal scenario. I've found with a correct delay setting the coherence increases indicating an increase in intelligibility. If the intelligibility is not going to increase for the sake of imaging to the stage, then we would be better off losing the delay speakers altogether. I never said I wasn't opinionated.

I have printed and occasionally hand out a post made by 6o6 at PSW which I'd like to post here. I feel like I can take the liberty to re-post these since I dragged him into that thread in the first place and I don't think anyone is going to come get me. I especially like where he points out that by intentional misalignment you are actually aligning the speakers SOMEWHERE ELSE. DUH, I wish I had thought of that.

Re: Delay Timing??? Thu, 11 October 2007 10:05
Bob McCarthy
Location: St. Louis

There are many aspects to data interpretation of acoustic analyzers that require complicated interpretation. Delay setting is not one them. It is clear cut and extremely accurate. Time arrival of A vs time arrival of B.

Therefore it only makes sense that we complicate the issue by adding layers of complexity by miscellaneous interpretations of the precedence effect Razz . Having set delays with an analyzer for over 25 years I can tell you that I have never once added extra delay for the "Haas effect". Careful calibration of the system can achieve the desired image placement without damaging the intelligibility/coherence, which is IMO "Job 1" of the delayed system.

To fully grasp the Haas research one must consider the following factors:
1) Precedence localization occurs only in the horizontal plane - as it requires binaural transducers for its detection.
2) Time is only part of the image placement formula - the other factor is level - a glimpse at the precedence effect graph will remind one that that are two axes here.
3) Vertical localization is driven by relative level.
4) It is not required in most cases (massive delay/relay towers excepted) for the delayed speakers to be level dominant over the mains, but rather to be at level parity where the maximum coherent addition (when accurately timed) can occur.
5) A 5th neglected factor is that of frequency response - which was matched in the Haas research and is unlikely to be in the field. Efforts made to provide some matching of the response of the delay to mains will help keep the image toward the mains. (this involves reducing the very top end to prevent it from standing too far above the mains)
6) The most common side effect of this appropriation of the Haas effect is that the intelligibility is reduced due to comb filtering,(the certain consequence of delay misalignment). This is then followed by an upward level adjustment of the delays,(and resulting image shift) which would not have had to occur if the alignment was precise.

Note: The case of high power, wide angle delay towers: these are much more RELAYS, than DELAYS, since they provide substantial power boost (in contrast to the reason we need UBALC delays). In such cases the leakage in all directions is severe and there is no setting that can work for more than a small area. In short - the more dominant the mains are, the larger the percentage of area that maintains reasonable timing. As the level of the delay increases, the leakage widens the areas of destructive interference, no matter what the timing is.

Harry said this all much more briefly......... Smile



Re: Delay Timing???
Thu, 11 October 2007
Bob McCarthy
Location: St. Louis


Agreed as to the inherent presence of errors, agreed as to the motivation towards - when error, error in favor of best image.

What one must consider when they are purposefully adding delay time errors is that they are making the systems in sync SOMEWHERE - just NOT where they are referencing to. The question is where, and for my money, I would rather know precisely where synchronization occurs so I know where the errors will begin their roll call.

If the delay is set to be sync'd where the mains and delays are in a straight line, this will yield the LONGEST delay time calculation, compared with positions that triangulate (main/dly/mic). In practice, of course we have both H and V planes at work and they can play for, or against each other in maintaining a constant time difference. If one does their alignment from a triangulated position, and adds extra delay, they are making the equi-temporal point move closer toward the straight line position. In other words, moving to the side and offsetting the time, may end up with the same time as referencing from the center..... but might not.

The vital characteristic for system intelligibility/coherence is that the synchronization occur at the position where the two speakers are equal in level. This is where the comb filter stakes are the highest. If one system has level dominance, errors in timing cause less damage.

Re freq response: It is common to think of imaging issues between two speakers as digital (1 or 0, all or nothing) but this is often not the case. The precedence game plays with BOTH level and time and can play out differently over frequency. The mains LF and MF can be ahead in time and matched in level (image to mains). But if the HF extension of the delay is well beyond the mains, then for THAT range the level dominance moves the image to the delays - overriding the timing (look at your Haas chart again). This is recognizable when sibilance or high hats come through the system and the image splits and shifts around momentarily. Needless to say, this is highly detectable. Therefore, care is warranted in the HF extension of delays, if there is to be hope of running them submerged and undetected.

This is a particularly important issue for small format speakers with ultra-light tweeters. These systems may be the only ones that the dogs can hear, and need to get brought into spectral conformance with the rest of the system (sizzling Ubalcs and Frontfills are symptoms of this).


Here is a picture for anyone that wants to know what 10ms of delay error looks like. Notice the coherence hit.