Multi Subs- How to time align with different distances - Page 2

I delay the closer units to provide the best blend with the furthest. Once they are behaving well together, then treat them as a single entity combining with the center, followed by the L/R.

Interesting that you chose to optimize crossover with the center first. It makes perfect sense though. I have to figure out how to do HDMI/ASIO4ALL setup in REW so that I could route signal to any channel.

When combining subs with the center, do you adjust delay for the sub group or only change the crossover point? Does changing "global" delay for the sub group breaks the blend of the individual subs in any way?


Thank you

Arny Krueger posted this presentation by Dr. Geddes on another thread:

http://www.ustream.tv/recorded/28637776

In the presentation, Dr. Geddes addresses many of the issues discussed here.

Dang it! Now I have to add constraind layer damping, an additional sub or two and alter my approach........

But, thanks for the link

I must admit that I am new to some of these terms, so can anyone tell me what is impulse responce? When Mark was talking about group delay and distance, how can those be adjusted in a normal receiver or pre/pro? (note-I have a Denon AVR-3312). Or is a seperate EQ devise needed, ie the Minidsp, SMS1, ect..?

The HDMI connection in REW will be very cool, but isn't required to test the L/C/R. Put your preamp in PL2 Movie (not Music), and you can then send analog input to the left and right individually, with a mono signal to both channels being routed to the center at +3dB vs. the individual L/R.

The blending of the subwoofers is only dependent on the relative differences. You can delay the set of subs as a whole in the preamp's sub channel, or at the input of your DSP. If they are all delayed the same amount, nothing changed other than how they interact with the speakers. You also want to get the distances set for the main speakers before you start tinkering with the subwoofer, as you can then compare and find a best compromise for the center and left/right speakers. Sometimes you get lucky, other times it's a balancing act, especially if the center is very different from the left/right.

Thank you very much, Mark. It is a neat trick, using PL2 for feeding the center channel.

thank you for the pointer.

so what's the difference between what Harman people do with SFM and Geddes with his proprietary software? I do not see any.

I think "convergence to best practices" is an apt phrase here. Both SFM and the Geddes approach, in their current versions, seek to BOTH reduce mean spatial variation AND to flatten FR by placing multiple subwoofers randomly in a room and varying the subs' individual level and parametric EQ. Harman seems to have moved a long way forward from the original Welti stuff, and Geddes seems to have simplified his recommendations quite a bit (less emphasis on placing a sub above the room centerline, though I've found that making the subs non-coplanar is useful every time I've done it, "useful" defined as "reducing the need for EQ."

Now, if only someone would release software (preferably OSX-native) that did the calculations, and avoided the tedium of iterative measure-manual EQ-measure loops. I'll buy it!

If you read Welti's paper on subwoofer positioning and his paper on SFM, you'll notice no mention of improving frequency response. The entire goal is to minimize seat to seat variation. To that end, SFM uses three parameters: level, delay, and a single band PEQ (again, not to improve bass response, just make it as similar as possible across the seating area).

If you watched the video linked above, you'll notice that the final result for Geddes was smoother/flatter frequency response. Didn't matter to him whether it was for a single seat or multiple seats, his goal was to improve the bass. By comparison, SFM (as currently used by Harman) is pointless for a single seat, since its only goal is minimizing spatial variance.

It might or might not. I am assuming it was a mic at the LP, both speakers running. Move the mic 1/4 inch, remeasure and watch a vastly different top end FR appear. Bass will hardly change, 10 k and above assuredly will.

Measure left and right alone to get an inkling of what happens up there. They are interfering with each other and the room and and and. At those wavelengths the slightest movement of the mic will make a huge difference.

He doesn't have much concern about group delay below 80Hz either. Earl and I don't agree on everything, but on these two points we do.

I understand the distinction you are trying to draw, but i think it may be without a difference. Both parties are clearly convinced that spatial variation must be minimized for the EQ to be effective. Geddes spent a big chunk of the presentation explaining this. Welti assumes post EQ, Geddes does not (that's the biggest problem I have with his approach, but that's another topic) , that's why his software also equalizes for the best frequency response. I guess it is a difference but it's not a fundamental one. In the example he used, the room was EQ'd for a single location but he stated that the software can do this for multiple.
Edited by zheka - 1/24/13 at 6:21pm

Sure, because the number of seats (single or multiple) isn't the main concern of the Geddes software, since the purpose is best frequency response. By comparison, frequency response isn't a concern for the SFM algorithm, since its only purpose is consistency over multiple seats. Looks like opposite approaches to me; but if you feel they are "without a difference", then I won't argue the distinction any further.

Design of digital filters that don't fit the pattern of "constant in the passband, close to zero in the stopband" rely on numerical optimization techniques to minimize an objective function. The objective function takes a collection of error values (actual response minus desired response) and computes a single non-negative scalar that's some measure of the error on each pass of the optimization. What Earl is saying in the video is that his algorithm can be used to optimize multiple seats. This just means that error values from multiple seats are all folded into the objective function evaluation, so his algorithm is not limited to optimizing a single seat only.

Edit: There's a potential point of confusion here. The Geddes multi-sub algorithm seems to have at least two incarnations: the one he does and the one he tells everybody else to do. The latest version of the one he tells everybody else to do indeed only optimizes one seating position (it used to optimize the spatial average, but he dropped that requirement). The one he does uses his proprietary software and is capable of optimizing multiple seats (according to the video).
Edited by rock_bottom - 1/24/13 at 10:21pm

it did come out rather arrogant on my part, did not it? I am sorry

Not at all. Guess I just see more differences than similarities between the Geddes software vs the SFM algorithm when it comes to both approach and goal, whereas others feel the opposite (remember Wayne's recent posts in another thread). Not worth arguing over.

You sir owe me a keyboard. Your line above made me spray my morning coffee out my nose. It's only funny because it's so true.

I hear you!

Still, I consider Earl to be the most interesting and thought-provoking person in audio today, not just because he's a very bright guy, but also because he's an unconventional and original thinker. I like the way he's trying to make the most of low-cost, readily available EQ solutions using IIR filters, as opposed to unobtanium or super-expensive alternatives. Still, sometimes some of the things he says leave me scratching my head.

Those filters he shows at 32:21 into the presentation are pretty interesting and non-obvious. I noticed also that he's not using individual delays for each sub, only IIR filters, but he has the advantage of his proprietary software to make that happen.

Like in the presentation, where he says that room gain doesn't exist or corner placement doesn't make subs louder. He also switches opinions on certain things, which isn't so bad, but he seems to go from one extreme to another. From what I remember, at first he was absolutely against using equalization, compared to the presentation where he states that you absolutely will never get proper response without EQ. Middle ground is for wimps!

He is a piece of work, ... that said, I do try to read everything I can that he puts out.

When I was searching for mains, the Summas/Abbeys were on a short list to hear. We exchanged several e-mails attempting to hook up, and discussing various options unique to my room/system, etc. Never got to experience them. There are few things like a well executed two-way,..and those dense, inert, composite, low diffraction offerings of his, seem to hit on so many levels. But they are expensive.

Anyone know the basis of his room gain, boundary loading statements?

http://www.avsforum.com/t/1329334/no-room-gain-in-most-rooms

thank you for the pointer. fun read.
it seems the most charitable explanation is that his subs being tuned to 45Hz do not go low enough to benefit from PVG.

Earl has a way of bending the laws of acoustics to better suit whatever he's selling. Yes, you might not get 12dB/octave below where the longest room dimension is a half wavelength, and you might not get 12dB more in the corner than in the center of the room, but you won't get nothing either. Except maybe in Earl's World.

This, along with the above keyboard soaking remark, made me laugh out loud as well...

JSS

I'm not sure if this presentation is new or old, but the PowerPoint slide show that Earl used in the video is available at his web site (pptx file).

So you can now see the dB/division and frequency scale of the graphs etc.

Can I put in a dissenting opinion?

I am not up to arguing audibility (I think it is but not that certain to argue the point) so instead will argue from a reverse position.

Practicality. (I am assuming that what all would agree on is a final 'in room' response)

Firstly, if 'poor time aligning' with subs is inaudible, it follows that 'perfect time aligning' will at least not introduce undesirable results.

So on to the practicality. IF we are to eq to get the final response, then it is best to start any equing stage from the best position, or baseline.

That is when the time alignment is optimised. THEN do the needed eq.

Saves amp power, you are eqing when the subs are not fighting themselves.

IF I could remember the site where I have parked all my 'photos' there might be some graphs there that show the differences in FR down only to differences in sub impulse arrivals. The difference is not trivial. Basically, having them time aligned is 'free gain', so why waste it? (*if you have the ability to do so)

Once you have them time aligned, which automatically give the 'best' eq free response, THEN we do needed eq.

I think it's the integration of the mains and sub(s) in the crossover region that matters most for the delay (although Earl would argue there's no "crossover" per se in his method). Just for the sake of round numbers, assume the crossover frequency were set to 100 Hz and the relative delay between mains and sub were varied by 5 msec. That's a 180 degree variation in relative phase between mains and sub at the crossover frequency, which will surely affect their integration.
Edited by rock_bottom - 2/7/13 at 10:38am

For sure. Which makes it all a bit more nuanced than 'we can't hear phase/time whatever in the bass so don't worry too much'.

As I am trying to say, let's ignore any and all points about 'audibility' for now, just move on to the 'practicality'.

At the end of that day we do want a smooth even bass response of whatever shape best suits our taste.

WE MAY be able to eq that dip out of the response by throwing power at it. However, if that dip we are eqing is due to phase (time) differences which can be handled, handle it FIRST with the appropriate tool (fix time caused problems with time) then do your eq. I can pretty well guarantee you that you will need less eq after fixing time than if you left it alone.