Multiple Small Subs - Geddes Approach

Re: Re: sims

auplater said:


Hi Todd..

Not trying to be difficult or contrary (and I've been trying to follow the thread)... just trying to figure if the math can be manipulated and or massaged to help with improved music quality in my setup.

I currently have 3 12" sonosubs in various ported configurations, a dual 10" isobaric BR, plus dual 6" mid subs in main 6' dipoles, plus 10" woofers in old Advent 6002's for sides in 7.1 configuration.
John L.:

Wow, you have a lot o' subwoofers! Geddes method does not really address different resposnes in different locations. He is taking an average over a given area and optimizing that to be smooth. you could modify his approach so as to look at two separate areas and try to make them more similar. If your room is rectangular and reasonably symmetric, you could try a midwall, or 4 corner setup. Though in that case the subs should be similar if not identical.
 
cap'n todd said:



Oh, I was thinking the entire modal repsonse, not a single mode. Separating out one mode like that is even more theoretical than some of the other stuff we have been discussing. No physical standing (even just two walls) wave would ever have that response, since there would always be standing waves at harmonics of the first on (95 hz in your plot). but anyway I see what you are saying.

I think there is some confusion between a mode and a resonance (even to me). I hold that a mode refers to the spatial standing wave, which will never look like the plot you showed. That would more properly be called a resonance.


Yes, agreed. But if we use the modal expansion to perform the simulation then we (I) need (want) to understand how each term (mode) contributes to the over all response. After all, if the source are positioned so that certain modes are not excited, even though the harmonics are, then the contribution of that mode to the net response is not just at the modal frequency.

Anyway, I have performed a number of sim. I completely rewrote one of my codes so that there was no garbage hanging around that might interfere with the result.

Here is a link to some of the results.
 
john k... said:



Yes, agreed. But if we use the modal expansion to perform the simulation then we (I) need (want) to understand how each term (mode) contributes to the over all response. After all, if the source are positioned so that certain modes are not excited, even though the harmonics are, then the contribution of that mode to the net response is not just at the modal frequency.

Anyway, I have performed a number of sim. I completely rewrote one of my codes so that there was no garbage hanging around that might interfere with the result.

Here is a link to some of the results.

OK. I looked at the sims. It seems you are showing more difference between the N=5 and N=30 cases than I remember seeing when i tried it. Also I should expect to see much more difference between N=5 and N=30 above around 100 Hz. N=5 for that room should get modes up to around 100 Hz, the error being the loss of the "skirts" of the higher order modes impingin on the <100 range. So, it looks a bit strange. Also your plots showing addition of direct term are higher than mine.

BTW did you include the scaling factors for ex ey ez?

I'll have to look at this more. Unfortunantly, it may be a few days....(deadline approaches!)
 
cap'n todd said:


OK. I looked at the sims. It seems you are showing more difference between the N=5 and N=30 than I remember seeing when i tried it. Also I should expect to see much more error above around 100 Hz (N=5 for that room should get modes up to that frequency, the error being the loss of the "skirts" of the higher order modes impinging on the <100 range). so, it looks a bit strange. Also your plots showing addition of direct term are higher than mine. I'll have to look at this more. Unfortunately, it may be a few days....(deadline approaches!)

In case you missed it, the N = 5 compared to N = 30 are for what would be a near field simulation with the mic position directly in front of one source. As has been discussed, when the mic is close to the source the series converges much slower. When the mic is at one of the listening positions given by your matrix there is very little difference between 5 and 30 modes below 100 Hz.
.

I don't understand you comment about the "Also your plots showing addition of direct term are higher than mine. " All I see is that there is a significant difference in the response w and w/o the direct term. But the variation with position is still fairly minor.

I also looked at different values of alpha. All it really did was, as expected, change the amplitude of the response near the modal frequencies when the direct term is included.

And remember I also have the woofer response LP filtered at 100 Hz, so you may see some differences in the details as F approaches 100 Hz.

The only thing that is different here, since I rewrote the code, is the contribution from the direct term. I initially coded it from memory and didn't have the phase correct.

Lastly, if I increase the damping to very high levels, I recover the direct response without including the direct term even when at a distance from the sources.
 
john k... said:


In case you missed it, the N = 5 compared to N = 30 are for what would be a near field simulation with the mic position directly in front of one source. As has been discussed, when the mic is close to the source the series converges much slower. When the mic is at one of the listening positions given by your matrix there is very little difference between 5 and 30 modes below 100 Hz.
.

I don't understand you comment about the "Also your plots showing addition of direct term are higher than mine. " All I see is that there is a significant difference in the response w and w/o the direct term. But the variation with position is still fairly minor.

I also looked at different values of alpha. All it really did was, as expected, change the amplitude of the response near the modal frequencies when the direct term is included.

And remember I also have the woofer response LP filtered at 100 Hz, so you may see some differences in the details as F approaches 100 Hz.

The only thing that is different here, since I rewrote the code, is the contribution from the direct term. I initially coded it from memory and didn't have the phase correct.

Lastly, if I increase the damping to very high levels, I recover the direct response without including the direct term even when at a distance from the sources.


OH, I thought you were just using the nearest seat as the "near" position. Still, when i did the N=5 vs N=35 comparison, i did it at 4". So, that doesn't explain it, unless you were much closer than that.

yes, but in my plots usually the direct only has an influence where the modal response is low.

if you put dampling at very high levels, the model is invalid anyway, so I'm not sure what it means to say you recovered the direct.

Again, did you include the scaling factors for ex ey ez (i.e. times 2 for N>1)??
 
capon todd said:



OH, I thought you were just using the nearest seat as the "near" position. Still, when i did the N=5 vs N=35 comparison, i did it at 4". So, that doesn't explain it, unless you were much closer than that.

yes, but in my plots usually the direct only has an influence where the modal response is low.

if you put dampling at very high levels, the model is invalid anyway, so I'm not sure what it means to say you recovered the direct.

Again, did you include the scaling factors for ex ey ez (i.e. times 2 for N>1)??

I get just about the same result for the mic 4" from one of the woofers. Di you remember to change the vertical position of the mic to match that of the woofer? I know you said you aren't concerned about the effect of vertical position but 6" vertically can make a big difference, particularly if it is 4.5' instead of 4'.


I think I noted this before. We have been throwing this around, "valid only for small damping". It really that the system isn't accurate as the damping get larger because some of the expansion are only to 1st order in admittance. But what constitutes small damping? Based on what I have read in Mechel, and in Morse an order of magnitude analysis indicates the alpha(n) << Pi x n. In any event, that the system recovers the free field response for large damping shows that the free field is contained in the series solution and indicated that the direct term should not be included. As I am sure you know, it is not uncommon in math and engineering to expand a solution in terms of some small parameter. While theoretically valid only if the assumption of the small parameter is enforced, many time the behavior of the solution when the assumption is violated still is indicative of the correct behavior, but in the absence of higher order effects which would change the character of the solution, but not the trend.


And of course I handled ex, ey... correctly. :)
 
Re: oops

auplater said:
Wondering what the practical goals of all these simulated responses are? What with the assumptions about, for instance, eigenmode amplitude approximations (seeing as the theory implies infinite impulse and zero width), room boundary condition compromises, room dimensional constraints, all conditioned by the model, what are the real world applications, or is this just an academic exercise.

Seems Dr. Geddes has been doing the math, as well as empirical work for over 2 decades, developed award winning speakers as well as an eminently usable semi-empirical method for subwoofer placement, and tirelessly worked to explain to others his rational based on theory.

So I'm not sure what the use of these sims in real home listening environments is, seeing that this is Geddes subwoofer approach thread.

John L.
From an engineering point of view, if it's possible to model a room and predict how the speaker+room system will respond within reasonable tolerance, then one can do a simulation to find the best room arrangement without having to do measurements and shifting things around on site. This is very common practice in almost any engineering project.
 
Only possible if you have accurate data on the damping properties of your furniture, curtains, walls etc.

Not very practical in my opinion. In the timeframe needed to obtain this data, you will be able to set up systems in several and very different rooms according to Geddes method.

No, just get a number of subwoofers, play around and measure, and EQ the last irregularities if needed.

Sims that need a lot of complex (and not very accurate) input data are mainly useful to gain insight and more understanding of what is happening. Also they are useful to compare systems properties when evaluating them. But for a practical end solution, stick with measurements and trial-and-error.

Sims can only point you towards the right direction when the topic of this thread is concerned. In that respect I regard the sims presented here to be very useful, but remember that you will never get a close to optimum placement of subs, their number, their relative amplitude level and phase shift and crossover frequency from a sim.
 
I have added complex admittance to my sims which include complex eigenfunctions, complex modes and both wall absorption and wall flexing. Repeating the sims for Todd's 4 woofer array shows that adding the wall reactance adds some additional irregularities to the response but spatial variation over the listening area is still pretty much as before.
 
Using Multiple subs in small rooms via Geddes

Since the other thread has been confiscated by another subject that is not really germain to how I do the multiple sub layout, I thought that it would be appropriate to start a new thread to get the topic back on track.

Any questions about my approach should be posted here and I will deal with them hopefully without getting sidelined to other topics.

It was commented that my technique does not yield a spatially smooth result, but this was incorrect. The result is smooth over as large a space as one cares to take measurements in. Obviously outside of that area where one takes the measurements for the optimization one cannot be assured of an optimum result. But this is going to be true of any technique.
 
Re: Using Multiple subs in small rooms via Geddes

gedlee said:
It was commented that my technique does not yield a spatially smooth result, but this was incorrect.

You're referring to the following statement from Todd Welti?

Geddes method does not really address different resposnes in different locations. He is taking an average over a given area and optimizing that to be smooth

How does your method ensure that it does not only produce a smooth average of otherwise very uneven frequency responses?

Best, Markus
 
Re: Using Multiple subs in small rooms via Geddes

gedlee said:
Since the other thread has been confiscated by another subject that is not really germain to how I do the multiple sub layout, I thought that it would be appropriate to start a new thread to get the topic back on track.

Any questions about my approach should be posted here and I will deal with them hopefully without getting sidelined to other topics.

It was commented that my technique does not yield a spatially smooth result, but this was incorrect. The result is smooth over as large a space as one cares to take measurements in. Obviously outside of that area where one takes the measurements for the optimization one cannot be assured of an optimum result. But this is going to be true of any technique.

The devil :devilr: sitting on my left shoulder is prompting me to comment. The angle on my right is just shaking her head :no: . I'm going with the right for now, just shaking my head.
 
Re: Re: Using Multiple subs in small rooms via Geddes

markus76 said:


You're referring to the following statement from Todd Welti?



How does your method ensure that it does not only produce a smooth average of otherwise very uneven frequency responses?

Best, Markus
I think that gedlee is looking at the system from pro audio point of view, in such case of which requires technical experience to optimize layout compromising among many factors that are hard to describe.

Methods like DBA temps to eliminate first reflection is a more simpler way of adjusting layout to reduce the experience level necessary.
 
Re: Re: Using Multiple subs in small rooms via Geddes

markus76 said:

How does your method ensure that it does not only produce a smooth average of otherwise very uneven frequency responses?

Best, Markus

"smooth average" is relative. Given the number of subs that you intend to use it will always be smoother than a fewer number of subs. When the responses for various room locations are taken, the fitting procedure attempts to maximally smooth those responses. Since there are now a number of degrees of freedom which can be manipulated (like location, LP point, gain and phase) it is assured that the final response will be smoother in both space and frequency over that range of both variables that is being considered. Is this "smooth enough"? Well that is a serious question that no one has an answer to. Will it be "smoother" in BOTH respects? Absolutely.

The response can be further smoothed with a few bands of EQ or the use of more subs (as long as they are "independent" subs). In the end any desired or necessary smoothness in space and spectrum can be obtained.

In my experience three subs seems to be the most effective. Adding more usually only results in a small improvement. Again, in my experince, one sub should be in the corner, another along the longer opposing wall and a third is almost arbitrary, but studies have shown that placing one (but not the corner one) above the rooms centerline helps out a great deal.

Are there "better" ways to get smooth bass response - of course, there is always "better", but what is necessary, is, as I have said, unclear. The subjective responses from my rooms seems to indicate that what I do is "sufficient" as those responses are unanimously positive with many claiming "the best bass that I have heard". Hence, I must conclude that multiple subs is "neceesary", that three well placed and setup subs is "sufficient" and that virtually any listener will be "pleased" with the results.

I'll leave the academic "proof" of the ideal to others.