Multi-sub and Dirac. Do you need a minidsp? (beginners REW)

[nickwin]

Mar 13, 2015 15:58:17 GMT -5 ansat said:

Mar 13, 2015 15:53:45 GMT -5 nickwin said:

I finally broke down and decided to try out REW again. I had trouble getting it to work 10 years ago on a macbook, Its not working out as easy as I hoped this time either, but I did manage to get some measurements of my subs during the brief time that It was working for me.

The first picture is each sub playing on its own over laid with both together. The combined graph is how I had previously set up the position and phase by ear, I think it sounds significantly better than either on there own.

The 2nd graph is both subs playing together with different phase settings on the MFW15. I was kind of surprised how these looked. I playing with this a little by ear and the 7 o'clock (green) setting sounded the best to me, yet just looking at them the the other ones appear to be more in phase and flatter. Looking at the 3rd graph explains a little bit because with no smoothing my original setting produces a much smoother response. I think I must have been picking up on that peak at 75hz which I don't get with the setting I had.

What you guys think? Which setting should I use before I run Dirac? Do I need a miniDSP?

I think your good with no minidsp. Try the 3rd, for max output and allow dirac do its thing over it.

Tony

Sorry about the repeat colors on the graphs, still figuring out how to format those. Do you mean the second graph? The middle green or red line? Thanks! Also, is it best to show subwoofer graphs with no smoothing?

[ansat]

Mar 13, 2015 16:06:49 GMT -5 nickwin said:

Mar 13, 2015 15:58:17 GMT -5 ansat said:

I think your good with no minidsp. Try the 3rd, for max output and allow dirac do its thing over it.

Tony

Sorry about the repeat colors on the graphs, still figuring out how to format those. Do you mean the second graph? The middle green or red line? Thanks! Also, is it best to show subwoofer graphs with no smoothing?

Both phase 3avg on the second graph.

Tony

[pmd918]

Tony,

I would like to add a comment on your approach. Thanks, BTW for putting this together. Should be helpful to a lot of people.

My comment is wrt the final EQ after the subs are time aligned. Your tutorial has the user doing a separate EQ to flatten each of the subs independently.

Based upon my reading, it is better to EQ the subs as one unit after they are time aligned, applying the same filters to both subs using the combined output to determine the filters. So the left-most PEQ block would be used instead of the two right PEQ blocks.

I believe the issue relates to phase again. Any PEQ filters that are applied will impact the phase behavior of the output. So if you EQ independently, it may put the subs out of phase again. By using the same filters on both subs, any phase shift will occur to both subs, so the initial work to get them in phase will not be lost.

Of course, anyone can chime in on this one if I am technically wrong. But this is the way I have learned to do it.

[ansat]

Mar 14, 2015 18:15:43 GMT -5 pmd918 said:

Tony,

I would like to add a comment on your approach. Thanks, BTW for putting this together. Should be helpful to a lot of people.

My comment is wrt the final EQ after the subs are time aligned. Your tutorial has the user doing a separate EQ to flatten each of the subs independently.

Based upon my reading, it is better to EQ the subs as one unit after they are time aligned, applying the same filters to both subs using the combined output to determine the filters. So the left-most PEQ block would be used instead of the two right PEQ blocks.

I believe the issue relates to phase again. Any PEQ filters that are applied will impact the phase behavior of the output. So if you EQ independently, it may put the subs out of phase again. By using the same filters on both subs, any phase shift will occur to both subs, so the initial work to get them in phase will not be lost.

Of course, anyone can chime in on this one if I am technically wrong. But this is the way I have learned to do it.

I can agree with the summed eq for all the subs, as long as the subs are symmetrical and either in front or the rear of the listening position. But if you have subs in the front & rear, eq'ing the summed front and the summed rear separately would offer better control over a wider listening area. I make these statements "generally speaking" as in my room the right front subwoofer causes some nasty issues in 1 of the 7 listening positions (which happens to be my wife's preferred seat), So my front 2 subs have some are separated as well as front vs rear.

So many methods and thoughts out there on this stuff, it crazy.

Tony

[pmd918]

Mar 14, 2015 19:24:59 GMT -5 ansat said:

So many methods and thoughts out there on this stuff, it crazy.

Tony

I know. And it does definitely get confusing. There is no one size fits all solution. Compromises everywhere.

[nickwin]

Here is another graph showing each individual sub on there own overlaid with both together. This is after using REW to help me adjust the phase on one of the subs. Its easy to see that his is a big improvement over either sub on there own as well as both together before I made this phase adjustment. REW is a awesome tool (once you get it runnig), Im really pumped to play around with it more. IMO It can be a pita getting set up on a MAC so if anyone is having trouble you can PM Tony or I about it.

Attachments:

[wilburthegoose]

Just an opinion from a novice. Use the sound you hear over the graphs. Chasing perfection with numerical graphical analysis is a guaranteed losing proposition in the long run to me.

[socketman]

Tony , would you not want to phase align your mains to your subs and then add the second sub and phase align it to the first sub or does Dirac handle that part itself.

[ansat]

Mar 15, 2015 13:28:45 GMT -5 socketman said:

Tony , would you not want to phase align your mains to your subs and then add the second sub and phase align it to the first sub or does Dirac handle that part itself.

I phase align the subs to each other then adjust the phase of the summed response to the xover point of whatever speaker your focused on. (currently the front 3 are full range speakers in my system, so I just phase align using enhanced bass so I avoid as much cancellation from the mains output)


Tony

[socketman]

I have my minidsp sitting at the local PO waiting for delivery tomorrow. Of course I had 10 days off and it arrives the day I go back to work. LOL . I have run a bunch of tests with REW and I discovered that running my 2 subs out of phase by 180 was a big mistake and was casueing all kind of phase shifts and nulls. I will set up the minidsp and then hopefully the latest firmware will be here and then I can rerun dirac.

[ansat]

Corrected the images.

Tony

[jpwmeyer]

Thanks very much for this post/tutorial... it has really helped me get up the REW and miniDSP learning curves.  I have applied your logic to my situation/system, and have some questions for you and/or others if you will bear with me.

I have an XMC-1 that was initially setup as a 7.2 configuration with the subs setup as stereo; I  have implemented the full version of Dirac with a calibrated UMIK.-1.  I was pleased with the improvement that Dirac did to the clarity of the bass, and have recently purchased a balanced 2x4 miniDSP to see if I can further improve upon the performance of the subs. The subs are dual-opposed, floor-mounted, infinite baffle subs with 18” drivers (4 drivers total) driven by a Behringer EP4000 amplifier using the balanced inputs from the XMC-1 and miniDSP. The room that the system is in is large (30’L x 25’W x 18’H) and this room opens into 2 other rooms. The subs are installed near the front wall about 3 feet from the mains, and equidistant to the main listening position.  As recommended, using REW and my cross spectrum calibrated UMIK-1, and using only the left sub output from the XMC-1 to the miniDSP and switching to mono, I measured the response of each sub, then both subs, and concluded that I could benefit from a miniDSP. The following plot illustrates these initial REW readings at the main listening position:



I then purchased and added the balanced 2x4 miniDSP into the signal path, then very briefly played around with EQing the subs with the following REW results:



Based on the results, I have the following preliminary comments/observations:

• I was unable to eliminate the null at 30 Hz by simply adjusting the output delay and/or polarity using the miniDSP (using the recommended 1 ms steps with the polarity toggled both ways).  The null can be improved upon by using these adjustments; however, other anomalies/nulls in the frequency response curve are introduced that seem to create more problems than are being solved... a bit like whack-a-mole...
• I removed the cover from the miniDSP and it is setup with the jumpers on the 2.0V RMS input sensitivity, and (based on research and posts from the miniDSP community) the jumpers in this configuration are resulting in a loss of 7 dB in gain, which I am not thrilled about.
• To resolve the null at 30 Hz I introduced a -16 dB cut at 30 Hz on one of the subs with a +5 dB boost to the other sub, which seems to have substantially flattened the frequency response curve.  I added a couple of other minor peak filter adjustments but before I got too far I thought I would seek guidance from you gents.

So... here are my questions:

• Am I on the right track with the -16 dB cut at 30 Hz or is my logic/approach flawed? Given the frequency should be non-directional at 30 Hz I thought this was a feasible solution but wasn't certain if this will cause issues that I am unaware of.
  
• Can/should I adjust the input sensitivity jumpers to reduce the miniDSP input sensitivity to 0.9V RMS to eliminate the 7dB drop in gain when adding the miniDSP or will this cause other problems/damage that I should be leery of? I considered whether I should simply bypass the miniDSP with one of the subs in order to avoid the 7 dB drop in gain (for one of the subs) and only EQ one sub, but my preference would be to move the jumpers, but thought I would enquire here...
• Is there a more automated way to EQ the subs with miniDSP/REW (before Dirac) or is brute force/trial and error the only way to go?
  

• Should I be worried about time aligning the subs (and looking at other outputs from REW) given that the subs are equidistant from the main listening position?

Any and all comments/suggestions would be appreciated.
John

[ansat]

Jan 20, 2016 0:48:58 GMT -5 jpwmeyer said:

Thanks very much for this post/tutorial... it has really helped me get up the REW and miniDSP learning curves.  I have applied your logic to my situation/system, and have some questions for you and/or others if you will bear with me.

I have an XMC-1 that was initially setup as a 7.2 configuration with the subs setup as stereo; I  have implemented the full version of Dirac with a calibrated UMIK.-1.  I was pleased with the improvement that Dirac did to the clarity of the bass, and have recently purchased a balanced 2x4 miniDSP to see if I can further improve upon the performance of the subs. The subs are dual-opposed, floor-mounted, infinite baffle subs with 18” drivers (4 drivers total) driven by a Behringer EP4000 amplifier using the balanced inputs from the XMC-1 and miniDSP. The room that the system is in is large (30’L x 25’W x 18’H) and this room opens into 2 other rooms. The subs are installed near the front wall about 3 feet from the mains, and equidistant to the main listening position.  As recommended, using REW and my cross spectrum calibrated UMIK-1, and using only the left sub output from the XMC-1 to the miniDSP and switching to mono, I measured the response of each sub, then both subs, and concluded that I could benefit from a miniDSP. The following plot illustrates these initial REW readings at the main listening position:

View Attachment

I then purchased and added the balanced 2x4 miniDSP into the signal path, then very briefly played around with EQing the subs with the following REW results:

View Attachment

Based on the results, I have the following preliminary comments/observations:

• I was unable to eliminate the null at 30 Hz by simply adjusting the output delay and/or polarity using the miniDSP (using the recommended 1 ms steps with the polarity toggled both ways).  The null can be improved upon by using these adjustments; however, other anomalies/nulls in the frequency response curve are introduced that seem to create more problems than are being solved... a bit like whack-a-mole...
• I removed the cover from the miniDSP and it is setup with the jumpers on the 2.0V RMS input sensitivity, and (based on research and posts from the miniDSP community) the jumpers in this configuration are resulting in a loss of 7 dB in gain, which I am not thrilled about.
• To resolve the null at 30 Hz I introduced a -16 dB cut at 30 Hz on one of the subs with a +5 dB boost to the other sub, which seems to have substantially flattened the frequency response curve.  I added a couple of other minor peak filter adjustments but before I got too far I thought I would seek guidance from you gents.

So... hear are my questions:

• Am I on the right track with the -16 dB cut at 30 Hz or is my logic/approach flawed? Given the frequency should be non-directional at 30 Hz I thought this was a feasible solution but wasn't certain if this will cause issues that I am unaware of.
  
• Can/should I adjust the input sensitivity jumpers to reduce the miniDSP input sensitivity to 0.9V RMS to eliminate the 7dB drop in gain when adding the miniDSP or will this cause other problems/damage that I should be leery of? I considered whether I should simply bypass the miniDSP with one of the subs in order to avoid the 7 dB drop in gain (for one of the subs) and only EQ one sub, but my preference would be to move the jumpers, but thought I would enquire here...
• Is there a more automated way to EQ the subs with miniDSP/REW (before Dirac) or is brute force/trial and error the only way to go?
  

• Should I be worried about time aligning the subs (and looking at other outputs from REW) given that the subs are equidistant from the main listening position?

Any and all comments/suggestions would be appreciated.
John

Your on the right track, if you do not have a gain matching device inline, go back to the 0.9. Your method is exactly what i would of done to fix the null.

For automation, there is nothing that can be done about the delay part, but you can automate eq filters and load them into the minidsp. But given the challanges that you are already having, i font think it would yeild better results then what you already discovered.

Tony

[jpwmeyer]

Thanks Tony.
Do you recommend that I use both inputs on the miniDSP for my setup (i.e. use an XLR wye from the left sub out on the XMC-1) to double the number of potential usable filters, or should I just use one to reduce the potential for distortion and/or clipping?

[ansat]

John,
Single input will be fine. Make sure that you don't go overboard with the filters since Dirac will still want to apply its own. And be careful with what dirac tells you it wants to correct. You might need to create a custom filter so that it doesn't try to apply boost where it will have no advantage.

Tony

[jpwmeyer]

Will do Tony.
Thanks again.

[KeithL]

A general comment on Whack-a-Mole......

The bass response you experience in a specific room, and at a specific listening location, is a combination of several factors: For example, the resonant modes of the room itself are related to the dimensions of the room; you can alter how much you excite those modes by moving speakers around, and with EQ, but you can't move them unless you change the room itself, or move your listening position in relationship to them. Likewise, you have multiple speakers, each producing a range of frequencies, which means that there WILL be interactions between them. By adjusting levels, or shifting phase and delay, you can shift where nulls and peaks occur, and you can arrange things so that multiple different peaks or nulls don't coincide at the same location, and you can often fix things so that no especially nasty ones occur at your listening position, but you cannot completely eliminate them. You simply cannot fix everything, everywhere in a room, by altering the signal with a processor..... and, if you plan to try, then you need to pick your battles.

The other thing is that some problems are mechanical - and so require mechanical solutions. For example, if you have a room that is too live (which can happen at high frequencies - or at a specific bass frequency), then sound at the particular problem frequency "bounces around and builds up" because it isn't absorbed. If it's a bass frequency, this will result in a peak, and an obvious boominess at that frequency. You can reduce the overall amount of sound at that frequency with an EQ, but, even though you can eliminate the frequency response peak, you can't entirely fix the boomines at that frequency. Assume I have a room which has some sort of mechanical issue that makes it much less absorbent at 50 Hz than at 40 Hz or 60 Hz. If I then play something that has energy at those three frequencies, the energy at 50 Hz will remain in the room longer, and build up more, resulting in a peak at 50 Hz.... in what we call the power response - which is the long term measure of how much energy is there. However, if I use an EQ to correct that peak by inserting a response notch at 50 Hz, and get the long term power response correct, as a result the short term response will then have a dip at 50 Hz, and the sound that is present at 50 Hz will still sound blurry because I haven't fixed the excessive reverberance time at that frequency. (And, if I do a T-60 graph of the reverberance time in the room vs frequency, it will still show the same peak in T-60 at 50 Hz.) And, if I try to cancel out that energy by running another sub out of phase at 50 Hz, then I will create another interaction at another point in the room.

The only way to fix the problem without creating a new one is to address the reason why the room fails to absorb energy at 50 Hz - perhaps by adding a trap tuned to that frequency - of by figuring out why the wall materials I've chosen have an odd peak at that frequency and changing them.  

To answer your question, though..... by introducing a major notch at 30 Hz in one sub, and a slight boost in the other, you've essentially eliminated the interaction between those subs at 30 Hz (the sub with the 16 dB notch is effectively not doing anything at all at 30 Hz, and the one with the 5 dB boost is doing all the work for both of them, therefore there is no more interaction between them at 30 hz - because, at 30 Hz, you effectively only have one sub.)  In general, assuming the subs themselves are identical, and they're the same distance from the listening position, and you haven't got any crossovers or processing affecting the timing of each to a different degree, then they should already be time aligned relative to that position. Most of the ways to go about figuring out the correct settings comprise processes rather than specific calculations..... choose a reasonable starting point, take readings, adjust things based on those readings, repeat until you're happy with the results. There are usually many combinations that will approximate a chosen result, none of them perfect, and all involving compromises - so it's rarely a matter of "finding the right formula that always gives you the right answer".

Your second to the last question is somewhat rhetorical - Dirac (or the equivalent) IS the automated way to make these corrections. However, as a general situation, really deep nulls can be difficult or impossible to correct using EQ, and allowing an automated correction process to attempt to do so can often lead to the creation of other problems. So, for the best overall results, when making your initial adjustments, and figuring out where to position your subs, try to eliminate sharp deep nulls (which may be difficult or impossible to remove with processing), then try to eliminate or reduce sharp high peaks. If you can eliminate, or at least reduce, most of those, then the automatic correction process (or manual EQ) will have a better chance to produce a really good result without causing more problems.

The general best order to proceed is:

1) make your room as good as it can be
2) find the best location for your subs (to avoid sharp peaks and especially deep nulls)
3) make any EQ or phase adjustments you have available to improve 2)
4) use your automatic room correction mechanism to fine tune the remaining response irregularities

Jan 20, 2016 0:48:58 GMT -5 jpwmeyer said:

Thanks very much for this post/tutorial... it has really helped me get up the REW and miniDSP learning curves.  I have applied your logic to my situation/system, and have some questions for you and/or others if you will bear with me.

I have an XMC-1 that was initially setup as a 7.2 configuration with the subs setup as stereo; I  have implemented the full version of Dirac with a calibrated UMIK.-1.  I was pleased with the improvement that Dirac did to the clarity of the bass, and have recently purchased a balanced 2x4 miniDSP to see if I can further improve upon the performance of the subs. The subs are dual-opposed, floor-mounted, infinite baffle subs with 18” drivers (4 drivers total) driven by a Behringer EP4000 amplifier using the balanced inputs from the XMC-1 and miniDSP. The room that the system is in is large (30’L x 25’W x 18’H) and this room opens into 2 other rooms. The subs are installed near the front wall about 3 feet from the mains, and equidistant to the main listening position.  As recommended, using REW and my cross spectrum calibrated UMIK-1, and using only the left sub output from the XMC-1 to the miniDSP and switching to mono, I measured the response of each sub, then both subs, and concluded that I could benefit from a miniDSP. The following plot illustrates these initial REW readings at the main listening position:



I then purchased and added the balanced 2x4 miniDSP into the signal path, then very briefly played around with EQing the subs with the following REW results:



Based on the results, I have the following preliminary comments/observations:

• I was unable to eliminate the null at 30 Hz by simply adjusting the output delay and/or polarity using the miniDSP (using the recommended 1 ms steps with the polarity toggled both ways).  The null can be improved upon by using these adjustments; however, other anomalies/nulls in the frequency response curve are introduced that seem to create more problems than are being solved... a bit like whack-a-mole...
• I removed the cover from the miniDSP and it is setup with the jumpers on the 2.0V RMS input sensitivity, and (based on research and posts from the miniDSP community) the jumpers in this configuration are resulting in a loss of 7 dB in gain, which I am not thrilled about.
• To resolve the null at 30 Hz I introduced a -16 dB cut at 30 Hz on one of the subs with a +5 dB boost to the other sub, which seems to have substantially flattened the frequency response curve.  I added a couple of other minor peak filter adjustments but before I got too far I thought I would seek guidance from you gents.

So... hear are my questions:

• Am I on the right track with the -16 dB cut at 30 Hz or is my logic/approach flawed? Given the frequency should be non-directional at 30 Hz I thought this was a feasible solution but wasn't certain if this will cause issues that I am unaware of.
  
• Can/should I adjust the input sensitivity jumpers to reduce the miniDSP input sensitivity to 0.9V RMS to eliminate the 7dB drop in gain when adding the miniDSP or will this cause other problems/damage that I should be leery of? I considered whether I should simply bypass the miniDSP with one of the subs in order to avoid the 7 dB drop in gain (for one of the subs) and only EQ one sub, but my preference would be to move the jumpers, but thought I would enquire here...
• Is there a more automated way to EQ the subs with miniDSP/REW (before Dirac) or is brute force/trial and error the only way to go?
  

• Should I be worried about time aligning the subs (and looking at other outputs from REW) given that the subs are equidistant from the main listening position?

Any and all comments/suggestions would be appreciated.
John

[jpwmeyer]

Thanks for the comprehensive response Keith.
In general. I think I have followed your (and Tony's) recommended approach to bass and subwoofer management, but there's always room for improvement... More than once, I have second-guessed my decision to go with IB subs for the room as they can introduce difficult-to-solve problems (like my null at 30 Hz!) but I did have the foresight to use two conventional subs to test about 10 different pair configurations to arrive at their current/final location; however, the subs I used for that analysis were not particularly large (and my room is) and I did not find the 30 Hz null at that time. Regardless, I am pretty pleased with the results and I really enjoyed the project!
I have used 4" thick bass traps in the front corners of the room, and 2" OC 703 acoustic panels on the front wall behind the Maggies, and both have had a positive impact on sound quality, as has the XMC-1.
Maybe the next house/HT room will be benefit from this experience!
John

[goozoo]

Thanks for the always thorough and useful tutorial Ansat. A couple of simple suggestions to this would be that to properly implement multiple subs with DIRAC, a MiniDSP is a must, and that to gain or level match the subs to each other so the summed response is at your reference level. A simple way to calculate inter-sub time delay would be to use the furthest sub distance as zero and take the difference between it and the next sub and divide by the speed of sound. This then should give a fairly accurate measure of what the delay needs to be input into the MiniDSP. For a better understanding with sub placement, interactions, etc, readers might want to refer the link below.

forum.blu-ray.com/showthread.php?t=48286

There are also some very useful articles and tutorials found on the MiniDSP website that deal with this issue. At the end of the day, room acoustics and sub placement trump all the digital manipulation that one can do to correct for sound anomalies and the smart money is better spent dealing with those issues head on before putting too much faith into what room correction can accomplish.

[JKCashin]

Maybe it's been answered in the thread already, but I did not see it (searched on "balance") but can one use the recommended MiniDSP 2x4, the balanced model, and run the inputs in as balanced, but use just one "side" of the balanced outs to run subs with unbalanced inputs? $US125 for dedicated bass management is too hard to pass up!