ProSoundWeb Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: [1] 2 3  All   Go Down

Author Topic: Trace Averaging Separation  (Read 820 times)

Russell Ault

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 334
  • Edmonton, AB
Trace Averaging Separation
« on: March 05, 2019, 09:49:53 pm »

I can't seem to remember the rule of thumb for how far apart microphones should be when doing TF trace averaging to maximize data quality while minmizing the number of measurements to be taken. I feel like it was something like 1/3 wavelength at 100 Hz (i.e. roughly a metre), but I'm not sure now.

Can anyone enlighten me?

Thanks!

-Russ
Logged

Michael Lawrence

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 72
Re: Trace Averaging Separation
« Reply #1 on: March 06, 2019, 06:43:22 am »

I can't seem to remember the rule of thumb for how far apart microphones should be when doing TF trace averaging to maximize data quality while minmizing the number of measurements to be taken. I feel like it was something like 1/3 wavelength at 100 Hz (i.e. roughly a metre), but I'm not sure now.

Can anyone enlighten me?

Thanks!

-Russ

As part of the Smaart training class, Jamie takes three traces within a couple of inches of each other and shows how that creates a sufficiently reliable average. I think Roger Gans's star pattern used a foot separation.

For what it's worth there is a bit of a divide in the system optimization community about the validity of trace averaging that reaches back to the Don Davis days and it is fairly heated, so you will get different answers depending on who you ask.
Personally I tend to use optical averaging for most applications but if I am going to do a trace average I'll move the mic by a few feet. Moving vertically a bit is also usually just as effective.

M
Logged

Russell Ault

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 334
  • Edmonton, AB
Re: Trace Averaging Separation
« Reply #2 on: March 06, 2019, 01:24:06 pm »

As part of the Smaart training class, Jamie takes three traces within a couple of inches of each other and shows how that creates a sufficiently reliable average. I think Roger Gans's star pattern used a foot separation.

That's what I was trying to remember. That class was a couple years ago now, and the only thing I could find about the star pattern was that the size of the pattern needed to fit within a speaker's isolation zone (which makes sense, but doesn't tell me anything about minimums).

Personally I tend to use optical averaging for most applications but if I am going to do a trace average I'll move the mic by a few feet. Moving vertically a bit is also usually just as effective.

A few feet is what I've been doing and it seemed to have been doing the job. What I've been debating is whether putting three measurement mics on a single stand using a wide bar might be an effective approach (regardless of how the traces are averaged).

For what it's worth there is a bit of a divide in the system optimization community about the validity of trace averaging that reaches back to the Don Davis days and it is fairly heated, so you will get different answers depending on who you ask.

I always forget about this. I know Jamie Anderson is strongly in favour, and I seem to recall that 6o6 is strongly against, which I feel like is enough of a divide for anyone. (For what it's worth, if I'm going to use trace averaging I'll always compare it to the optical average to make sure I'm getting usable data. Please don't flame me!)  :)

-Russ
Logged

Michael Lawrence

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 72
Re: Trace Averaging Separation
« Reply #3 on: March 06, 2019, 05:29:10 pm »

What I've been debating is whether putting three measurement mics on a single stand using a wide bar might be an effective approach (regardless of how the traces are averaged).

Maybe two on a bar, and one a foot or so lower just to make it less ridiculous to move around and to add another plane of randomization to it.

I know Jamie Anderson is strongly in favour, and I seem to recall that 6o6 is strongly against, which I feel like is enough of a divide for anyone. (For what it's worth, if I'm going to use trace averaging I'll always compare it to the optical average to make sure I'm getting usable data. Please don't flame me!)

I recently discussed this with Jamie, and he said although he was trained to be distrustful of averaging (I'm paraphrasing), he realized through experiment that you had to do some truly contrived situations in order to create an example where the averaging wasn't representative.
I also understand Bob's reservations based on the mathematical difficulties with averaging, and so I see legitimacy in both sides. I have the highest respect for all involved parties and so my strategy is to take it all into consideration and make my own judgement call.
Logged

Riley Casey

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1275
Re: Trace Averaging Separation
« Reply #4 on: March 06, 2019, 08:56:18 pm »

Mild topic swerve. Anyone have an idea for what would constitute best practice for mic placement for getting nearfield curves to average for setting up DSP settings for stage monitors?  Anything not obvious to consider?

Russell Ault

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 334
  • Edmonton, AB
Re: Trace Averaging Separation
« Reply #5 on: March 06, 2019, 09:12:28 pm »

Maybe two on a bar, and one a foot or so lower just to make it less ridiculous to move around and to add another plane of randomization to it.

Talked me into it, although I'll probably go above instead of below (I spend most of my time in soft-seated venues). Now to find a wide enough bar.

I recently discussed this with Jamie, and he said although he was trained to be distrustful of averaging (I'm paraphrasing), he realized through experiment that you had to do some truly contrived situations in order to create an example where the averaging wasn't representative.
I also understand Bob's reservations based on the mathematical difficulties with averaging, and so I see legitimacy in both sides. I have the highest respect for all involved parties and so my strategy is to take it all into consideration and make my own judgement call.

I'm jealous you had the opportunity to speak to Jamie recently. I think he's been up in my part of the world exactly once (and I count myself lucky to have been there). Still haven't met Bob. I too have the highest respect for both (which is why I consider it to be a serious controversy whenever they disagree). I'm sure both would agree with making your own judgement call (insert your own quote about the real analyser being the one between your ears).

-Russ
Logged

Michael Lawrence

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 72
Re: Trace Averaging Separation
« Reply #6 on: March 06, 2019, 10:22:15 pm »

Mild topic swerve. Anyone have an idea for what would constitute best practice for mic placement for getting nearfield curves to average for setting up DSP settings for stage monitors?  Anything not obvious to consider?

Riley -
I have two comments for you. First of all, I submit for your consideration that the only mic placement that matters for stage monitors is artist's listening position, which is often not on axis. Hopefully it's close, but artists have a nasty and unreasonable tendency not to remain stationary on stage. So I'd start by tuning for the place where their head is actually likely to be, or maybe some representative positions through that area, averaged together or not.

The other one is more of a minority issue concerning phase alignment between drivers in a two-way monitor, which luckily is not something we deal with all that often these days but I find it interesting enough to mention: most people might think that onax to the HF driver is the best location to do the alignment. If you do that, the relative arrival times between the woofer and the tweeter don't hold over distance (unless it's a coaxial setup). If you use the point equidistant between the two, the phase alignment will hold as you move back because the relative time offsets don't change. Bob points it out in his 2nd edition. Probably not relevant but I certainly find it neat.

Oh, also - usually we'd not want boundary effects in a measurement of this type, but since the wedge will be used on the floor, I would say it makes sense to measure it as such.

I'm jealous you had the opportunity to speak to Jamie recently. I think he's been up in my part of the world exactly once (and I count myself lucky to have been there). Still haven't met Bob. I too have the highest respect for both (which is why I consider it to be a serious controversy whenever they disagree). I'm sure both would agree with making your own judgement call (insert your own quote about the real analyser being the one between your ears).

-Russ

Bob recently reviewed a draft of an upcoming article I'm working on and I didn't realize until after I sent it to him that the process I was explaining involved averaged data. He didn't mention anything to me about it but I'm sure he was silently judging :D
The system optimization community has many such 'controversies' because it's still in its infancy. There are things Bob and Jamie don't agree on, things Bob and Pat don't agree on (actually I just published an article that could be interpreted as a direct contradiction of some of Bob's work, although I don't view it as such), things Meyer and L-Acoustics don't agree on, even down to the fundamental mathematics used to describe the same reality - and though of course I have my own personal conclusions about who's 'closer to the truth' in each case, I think the larger lesson for me is that it's wonderful to be in the field during a time where I can learn directly from the pioneers themselves. All of this stuff is just a foundation for human judgement and intuition, so my approach is to learn as much as I can from as many people as possible, do my best to sort it out, and go from there.
Logged

Frank Koenig

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 662
Re: Trace Averaging Separation
« Reply #7 on: March 07, 2019, 01:21:38 pm »

This is an interesting and important topic and I'll throw in a couple of observations.

First, this is a sampling problem and how we think about the spatial sampling rate depends on what we're trying to achieve. If we're combining far-field measurements in a (real or pseudo) anechoic environment we should be thinking in terms of angle, not distance, as we would expect each response to be substantially the same (except for level) if measured at a different distance from the source but at the same angle. If, on the other hand, we're trying to wash out the effects of room modes at low frequencies, as I suspect we're talking about here, then distance is appropriate. I don't know much about the latter and am interested in the guidance given here.

In any case, it comes down to Nyquist, except that it's kind of tricky in that we have to look at the "bandwidth" of the variation in the response at each of a number of frequencies as we vary the distance or angle between the spatial samples. Given this complexity, I think a more practical approach is a statistical one where we try to find the spatial sampling rate that causes the combined measurement to converge sufficiently as the number of samples increases. In other words, if sampling closer together does not make a change we care about in the result then we're close enough. I'm assuming here that the statistic we're using (such as averaging) is a consistent estimator in that the result converges as the number of samples increases.

Second, while "optical averaging", in other words eyeballing a bunch of overlapping magnitude curves, may work better than expected it's not of much use for any sort of automated alignment. We need an algorithmic way of combining measurements. Intuitively we know we'd like to throw out certain outliers and perhaps put more weight on peaks than dips as generally they are more audible. There is also the question of how much smoothing to apply at each frequency and whether some form of non-linear smoothing could be used to achieve the peak/dip bias. Further, we may want to weight measurements differently -- screw the folks at 30 deg a little to get a flatter response onax, for example.

It's pretty hard to argue with magnitude averaging IF a large enough number of samples is available. (Phase averaging is more subtle but similar.) But measurement is expensive and we often would like to get a good enough result with just a handful of measurements. In any case, as above, when more fooling around doesn't result in a material difference in how we set the EQ or what our FIR filter response looks like, we're done.

Personally, I've been using 5 deg in most of my speaker measurements but admit that it's pretty much a wild-ass guess. I've also fooled with the trimmed mean as a way of automatically throwing out outliers. Often I don't find it necessary. Deriving minimum phase from averaged magnitude and averaging excess group delay gets around some phase averaging pitfalls. Too much to write about here but if anyone wants to visit and chat on these subjects I'd be very happy.

--Frank
Logged
Yes, it is a giant stereo system!

Michael Lawrence

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 72
Re: Trace Averaging Separation
« Reply #8 on: March 07, 2019, 04:58:20 pm »

Too much to write about here but if anyone wants to visit and chat on these subjects I'd be very happy.

--Frank

Count me in! Sounds fun... email address is on my profile.

M
Logged

Mark Wilkinson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 849
Re: Trace Averaging Separation
« Reply #9 on: March 07, 2019, 05:39:53 pm »

Interesting thread...with issues that often drive me crazy !

In my self-taught, measure-in-my-driveway/backyard type experience, I've come to a couple of simple preferences regarding averaging.

I try to average magnitude.
But only when there is high coherence, and response variations tend to be generally the same on and off axis.
So to keep it automated, I smooth measurements to taste before averaging, and the correct the average.
I do this driver by driver, assuming needed corrections are all minimum phase.
If corrections don't respond rather fully, I ditch them as non-minimum phase.
Hopefully I get smoother response optimized for a combination of on and off axis.. 

For phase ...no averaging...it seems to me phase is "inevitably to a spot"
So I use the previously corrected drivers' magnitude response, and phase align those responses to one point, at some distance that is close to equidistant to acoustic centers.

It's never measures as perfect as doing a mag and phase alignment to one spot,
but in today's world of easy to make damn near perfect response at a spot,
it seems like doing so may be its own new can of worms.  (at the expense of off axis)

I hope that made sense...please pass on any critique/observations/reservations on this methodology.


Thanks guys,  Mark
Logged
Pages: [1] 2 3  All   Go Up
 


Page created in 0.09 seconds with 22 queries.