Another multi-sub setup

[Conrad.]

I was reading this thread and it got me thinking that I could probably do with some similar help and/or confirmation.
I have a four sub set-up in a significantly smaller room. This is both good and bad as I get great room gain, but I have very limited positioning options, and the different positions I do have make very little difference to the response.

I'm running a PSA S3000 and an S3600 at the front, inside my main speakers.
Behind the couch on the back wall I have another S3000 and in the rear right corner I have a Velodyne DD15.
I have all subs connected to a miniDSP 2x4HD which is coming from a single sub out on my processor. This gives me a shared set of filters (5 bank PEAK only) for all subs, plus the full DSP in the miniDSP.
I have no Auto EQ.
Crossover is currently set to 100Hz with a 24dB/Octave slope. Main speakers are set to small.

I've attached a room sim showing my layout. As three of the subs are dual opposed I've added them as individual subs, one per driver.

Looking at the good feedback and analysis @Sonnie got, especially from @jtalden, I was hoping to achieve similar results.
Previously I had always measured each sub, EQd to flat (well, with a house curve) and then measured together. As I measured and added each sub to the combined response I would adjust the delay on the current sub until I got the smoothest response. So my process would be:

- Measure S3000 front, EQ to flat
- Measure S3600, EQ to flat
- Measure both S3000 front and S3600, adjusting the delay on the S3600 until I got positive summing across the whole range
- Repeat for the remaining subs
- The use the sub distance control in the processor to move the subs as a unit to integrate with the mains.

That had previously worked ok, but never gave me results I was delighted with.
Having read Sonnie's thread I noticed the advice was different. Measure all subs combined, then EQ the combined output, I'd like to give that a go.

I did read the longer steps which involve measuring flat before applying any kind of crossover, but my processor only allows up to 140Hz and the subs only allow up to 150Hz so I don't know it's going to tell me much different? I did run my measurements through Multi-Sub Optimizer, but that wasn't able to do much.

This morning I measured my subs with no EQ and no time alignment, individually, and a set. The result of all four with no EQ is actually pretty good. I have a couple of deep cancellations, but it's fairly smooth other than that. My waterfall isn't great below 20Hz, but I don't know I can do much about that without some serious bass traps that I just don't have the room for.

I've attached a couple of MDAT files. One is just the subs this morning, no EQ, one is from yesterday which is full range mains, as full range as I can get on the subs (140Hz) and multiple positions.

So, my questions:

- is there anything I can do to remove those dips in the un-EQ response?
- Anything else I should be measuring or looking at?

I'll be honest, I've never had much success understanding phase. I get it in theory, but I don't know what I'm doing in practice and everything I change either doesn't work as I expect or makes thing better in one place and worse somewhere else.

Thanks, apologies for the long, rambling post.

 

[Sonnie Parker]

Hi Conrad... yeah... I think you will definitely get better results measuring the subs as a whole... as one sub. If you EQ them prior to combining them, you don't get the same response and you are taking into consideration the interaction between all the subs. If you are limited to the position of the subs, about the only other option is to try adjusting the distance on one or two of the subs. It would no doubt be a trial and error type method, but you might get better results. Try taking one sub at a time and changing the distance by adding 4-5ft, measure and see how it effects the change. Then set that one back and perhaps try one in the rear and see how it measures. Then you might try extending two at the same time and see how those combined effect the measurements. You can also experiment with the volume levels. Try reducing one sub at a time by 5dB and see if that effects the response... but raise that volume back up before reducing the next.

The only downside in your setup is multiple different subs. It's typically better to have identical subs. Some of the issue can be will each sub keep up with the other. Once you EQ them at a certain volume level... if one gives out of steam at a higher volume, the overall response changes.

 

[Conrad.]

Thanks Sonnie.

Having had a play today I agree, EQ across the subs as a whole is definitely the way to go. Now my question is, what's the best way to do that? If I use REW to EQ the combined response, how do I actually apply that to the combined output? Would that be the EQ in the processor? I keep the inputs of the MiniDSP free for loading BassEQ filters for movies.

I'll keep playing with the delays in the minidsp and see where that gets me. What I was hoping was that someone would be able to suggest some delays from looking at the phase of the measurements. If that can't be done I'm happy to adjust, measure, adjust and see where I end up.

As for different subs, three of them are all PSA from the same series (2x S3000 and an S3600) so they all use the same drivers. They should all be able to keep up with each other. The 3600 (dual 18") can't get as low as the S3000s (each dual 15") but that's due to the different DSP in the amp. The outlier is the Velodyne and is definitely the weak link. I might remove that and see if it makes any difference.

 

[jtalden]

I time aligned the 4 SWs in the "Subs Only" file and attached an SPL overlay chart showing the result as compared to the current "All" measurement you provided. I have not yet gone back and detailed the new settings, but can do that if you like.

There are many ways to setup 4 SW with good results. I think from a measurement perspective this result is a little better EQ starting place than the current one, but in the end it may not necessarily translate into a sound quality improvement.

General Comments:
My understanding is that there is not suppose to be any content in the LFE channel above 80 Hz. It would thus not seem necessary to bother to measure the LFE channel when using a 100 Hz XO. A setup with a much lower XO may want to do that. I have seen comments that sometimes there is some content in the LFE as high as 120 Hz, but have chosen not to worry about exceptions instead feeling it is more important to prioritize the proper XO timing and EQ using the redirected bass.

Current Timing:

New Timing:

 

[Conrad.]

Thank you for that, that's very interesting. I didn't know about the time alignment tool, that's really cool. And I need to read up on the Vector Average function as well. Is that different from the Average the Responses button on the main window?

I agree about the not measuring to flat with the LFE range, but I've heard lots of people tell me it's the way to do it so I thought i'd get as close as possible. It didn't seem to help.

Looking at your file, it makes sense. What's really interesting is that I saw a similar result with the inversion of the DD15 when I looked in multi-sub optimiser. I'll try and that measure and see if that gives me the expected outcome. Adding the delays and the inversion in the MiniDSP is easy so I'll do that there.

The two humps in the vector average can be easily fixed with a couple of fairly broad filters in my processor, so that's very possible too. They make sense as well, most rooms in the UK seem to have a mode in the 40Hz area, they're all about the same size so it makes sense to me that they're so common.

I'll follow the process through and see how it goes, and report back.

There is something I quite like about not having any EQ in the chain though, so I'll save this output as a separate config so I can jump between no EQ and the set of filters this config requires and see which I prefer.

 

[JStewart]

They make sense as well, most rooms in the UK seem to have a mode in the 40Hz area, they're all about the same size so it makes sense to me that they're so common.

That's the axial mode for your room length. You can see it in the REW room sim you posted above. You would have the same around 75Hz or so where the axial modes for the room's width and height are, but there's a phase shift you can see in the vector average for 1,3 there for some reason that's causing a cancellation.

 

[Conrad.]

How did you manage to invert the DD15 response in the vector average? I can average the responses and match yours for all but the one with the DD in it.
I've googled but I couldn't find much. Is there a guide you can point me to? I read this (https://www.roomeqwizard.com/help/help_en-GB/html/graph_allspl.html) but I'm not sure it's what I need.

I presume the odd phase issues are to do with the dual opposed nature of the subs and the fact I'm getting reflections that are close to the direct sound from the driver that's facing me? But I'm guessing.

 

[JStewart]

I presume the odd phase issues are to do with the dual opposed nature of the subs and the fact I'm getting reflections that are close to the direct sound from the driver that's facing me? But I'm guessing.

I would guess not related to the dual opposed and to your point, a room caused anomaly of some sort. Firing directly at a wall won't cause a dip within the subs output range. I don't think its something to go chasing either, although others might disagree. Basically the only tool in the toolbox to combat these issues is multiple subwoofers set up as best as possible and/or movement of subs or the LP.

 

[jtalden]

For this purpose it can be selected most directly within the alignment tool.

 

[JStewart]

How did you manage to invert the DD15 response in the vector average? I can average the responses and match yours for all but the one with the DD in it.
I've googled but I couldn't find much. Is there a guide you can point me to? I read this (https://www.roomeqwizard.com/help/help_en-GB/html/graph_allspl.html) but I'm not sure it's what I need.

For this purpose it can be selected most directly within the alignment tool.

@ConradNash , the link in your post is either not for the newest versions of REW or John M. hasn't had the chance to update the help files yet.
The alignment tool is in the All SPL graph settings and looks like this...

As you can see I'm using REW 5.20 Beta 33.

 

[jtalden]

Hmm, I did the alignment manually out of habit rather than using the alignment tool. I just tried to use the alignment tool to recreate my result without success. It appears the "Alignment copy" is not placing the impulse at the designated location and with the appropriate ref point. Have either of you had success recreating my posted SPL response using the tool?
I will investigate further to see if I can resolve this. I will reconfirm my manual actual Impulse timings and post those if you want to implement them before this is sorted out. The actual Impulses location of the individual SWs in my posted file may not exactly reflect what was used because several adjustments were made and their locations may have been changed a little for a subsequent trial. I will check.

 

[jtalden]

I did another timing alignment, this time using only the REW alignment tool. It worked just fine. No FDW was applied during the process. I had to try a lot of different options to find this one. It will be hard to find a more favorable setup. The alignment was done by:

1. First the 4 SW impulses were aligning at 0 ms.
2. The alignment tool was used on 1 and 3 measurements.
3. The alignment tool was then used on 2 and 4 measurements.
4. The alignment tool was then used on the results of those 2 efforts (measurements 8 and 11).
5. The result is measurement 14 It will be easy to EQ to a very smooth SPL response.

The relative timing if the 4 SWs is not easy to pull out of the attached mdat because of the way this was done, but all the needed info is there. I may do this tomorrow after I confirm that I did it correctly.

The SPL chart below shows the comparison of this new alignment (blue) vs my original suggested alignment (green).

Measurement 14 EQ Example:

 

[Conrad.]

Thank you @JStewart. I'm on the latest beta, 5.20.34 and I found the alignment tool. My challenge was that I wasn't getting the same results as jtalden.

As for the new files, I'm able to follow along a little.
Combining the time aligned copies gives me an output similar to the pairs of subs combined. Combining those pairs doesn't though. I also see that you've adjusted the timing on the DD15 by a different number than REW said it was off in its initial sweep. Presumably this is to better align the impulse response and/or phase. Then you inverted the phase and I made further adjustments and I got lost.

If you're able to provide the net delays for each sub, that would be awesome. I can then plug them in to the minidsp and measure the pairs, and the quad combined. I'm pretty comfortable taking that and applying EQ on top.

I'll keep looking at the timings to see if I can mirror your thinking.

 

[jtalden]

Combining the time aligned copies gives me an output similar to the pairs of subs combined.

Well done.

Combining those pairs doesn't though.

Note that pair 2,4 was inverted when the 2 pairs were aligned.

[/QUOTE]I also see that you've adjusted the timing on the DD15 by a different number than REW said it was off in its initial sweep. Presumably this is to better align the impulse response and/or phase.[/QUOTE]

??, The initial step I made was to time align all 4 SW to start their impulse rise at 0 ms. That is, I abandoned the initial acoustic timing for this setup. The resulting relative positions of the impulses (not yet pulled from the data) will still be correct for the 4 SWs. I will need to recover a reference to initial timing in order to suggest specific timing changes for each SW.

The initial step to move all SW to 0 ms was done to help assure me that I was not moving the timing too far from the theoretical idealized acoustic timing. Departing more than 1/2 wavelength from that position should be avoided if possible. In this case for 1/3 alignment I only moved 3 as needed until a good combination of SPL and phase tracking occurred. I kept that same pattern for the other 2 alignments as well. For the 2/4 alignment I only moved 4 as needed until a good combination of SPL and phase tracking occurred. For the 1,3/2,4 alignment I only moved 2,4 as needed until a good combination of SPL and phase tracking occurred. I may have go a little further than 1/2 wavelength in one of the 3 alignment steps.

[/QUOTE]If you're able to provide the net delays for each sub, that would be awesome. I can then plug them in to the minidsp and measure the pairs, and the quad combined.[/QUOTE]

I will work on the relative delays now. This is the first time I have done it this way so I need to pull out the relative offsets from the 3 steps and then manually confirm I did it correctly. Then they need to be adjusted by the initial acoustic timing position so I can provide specific timing changes. This could have been handled more easily, but I did not plan ahead well. Next time I keep one of the SW initial position in the file for reference.

 

[jtalden]

Okay, I confirmed the timing changes needed. It was more difficult than necessary because of some unnecessary steps I took. I should not have adjusted the initial timings in the way I chose. I was wrong in the value of doing that as with different model of SWs that have different bandpass ranges means that aligning the Impulses is not really an effective starting point. The table below contains the changes needed to your initial delay setting in order to get the responses shown in Post 12. I found some minor errors in my notations in that mdat file and I am sure it did not help your understanding. I think it will be more confusing to go back and explain those so I suggest you make the SW delay adjustment suggested below and remeasure each individually and together to confirm the result. Note that depending on you current MiniDSP settings you may need to shift all settings up or down in order to stay within the total delay range allowed.

After confirmation of the overall response of the SW group given the new timings, the next step is determine the timing of the SW group as a whole to the mains.

The last columns in the top table is the additional delay and polarity change needed to your current SW settings.

 

[Conrad.]

Thank you for the reply. I was just about to sit down and run through your steps and ask some follow up questions. While I really appreciate you giving me the results, I'm also keep to learn the thinking behind how and why you chose what you did. However, in this instance it doesn't seem like learning it would be particularly advantageous.

You've answered a number of my questions above anyway, like the fact that you inverted, and then re-inverted the DD15 meaning that it's phase ended up at zero. The inversion of the S3600 is interesting though, and unexpected.

One question I have, you paired the subs and then paired the pairs. Is that beneficial over having a single group and then then adding to it? So, aligning 1 and 2, then 1 and 2 and 3, and then adding the fourth to that collection?
Also, you paired them as 1,2/2,4. Presumably that's because 1 and 3 are the same make and model of sub, or is there another reason?

I'm going to play with the impulse response and time alignment using the numbers above and see if I get the same result. I'll also, of course, apply these measurements, measure, and confirm. I'll report back.
Once that's done and I've EDd flat, would you be ok to take a look at the mains and sub integrations with me as well?

Lastly, I presume I can add 2.35 to each measurement above so that my minimum delay is 0. I think the minidsp will handle 12.45 without any trouble.

Thanks again. I'll try and get these loaded and tested tomorrow evening.

 

[jtalden]

One question I have, you paired the subs and then paired the pairs. Is that beneficial over having a single group and then then adding to it? So, aligning 1 and 2, then 1 and 2 and 3, and then adding the fourth to that collection?

I felt it was advantageous to pair the SWs that had similar low end responses. I don't know what impact chosing a different pairings would make.

Also, you paired them as 1,2/2,4. Presumably that's because 1 and 3 are the same make and model of sub, or is there another reason?

That was notation error on my part; 1,3 and 2,4 were the pairings. The measurement notes confirms the actual SWs used in the pairings.

I'm going to play with the impulse response and time alignment using the numbers above and see if I get the same result. I'll also, of course, apply these measurements, measure, and confirm. I'll report back.
Once that's done and I've EDd flat, would you be ok to take a look at the mains and sub integrations with me as well?

Yes, I will check your SW to mains timing. Only 3 measurements are needed; L, R and SWgroup.

Lastly, I presume I can add 2.35 to each measurement above so that my minimum delay is 0. I think the minidsp will handle 12.45 without any trouble.

Yes

 

[jtalden]

I forgot to explain that paring was done so each SW was represented exactly twice so all are equally represented. If the tool is used sequentially as: 1/2, 1,2/3 and then 1,2,3/4 then in the result 1 is represented 4 times, 2 represented 3 times, 3 represented 2 times and 4 only once.

 

[Conrad.]

Ok, progress!

Here's the latest MDAT after applying the time alignment, and then applying the REW EQ.
I first time aligned all the subs, including inverting the S3600. I measured them all together and, hurrah, they looked like the predicted response in REW. I did have a deeper null at 75Hz than it predicted.

Then I ran that through the EQ, applied the filters as per REWs suggestion and it didn't quite work out. I guess there could be a number of things going on. Either:
- the processor's EQ might not be as accurate as it says it is, but I doubt that, and I hope it's wrong. I can test this by applying the EQ on the input of the minidsp as well. I'll do that later this week. It's also easier to change there, I should have done that earlier.
- the combined response of all subs isn't behaving in same way as a single sub would. This makes more sense. Cutting 44Hz on the input will affect all subs equally, but each sub is contributing a different amount at each frequency so it'll be causing additional modes and nulls. This, I'm guessing, could cause the response to act in a more extreme way than a single sub's response.
- some other unknown reason that I haven't though of.

I'll have a play with lowering the negative gains for each frequency and see how that behaves. I'm done for tonight.

Can I just say how much I detest the miniDSP software? I had to use the wireless bridge to get it to not inject noise through the USB connection and it's the least stable bit of software I've used in a long time. It completely refused to connect after a while, even after a complete factory reset. I think it might have something to do with me running a mesh network and it seeing each access point as a separate network? I don't know. But it has Alexa capability, so there's something pointless it does instead of "be stable". /rant.

So now I have to test which response I prefer. The no processing at all with the couple of deep nulls, or the aligned response after some tweaking. I'll give both some critical listening this weekend.

Thank you for all your help!

 

[jtalden]

The dip in SPL indicates the timing differs from what was intended. The EQ usually has no major impact to the Impulse locations so I will investigate the timing using this data to see how it differs.

Once the SW are timed properly EQ the group as a whole. The SW group should behave to EQ the same as a single SW would.

 

[jtalden]

I checked the SW timings in Post-19 vs the Post-12 target. The chart below indicates the timing delay changes needed relative to your current settings. These changes will provide the Post 12 timing and SPL. The polarities are all correct as they currently are set.

 

[Conrad.]

Thanks, I'll try those tonight.
Odd that they're out though. I guess the 0.6ms could be microphone placement issues?
Not sure about the 9.2 ms delay though. I'll apply and report back.

Thanks again!

 

[Conrad.]

Now we're cooking with gas.

I applied the adjustments as above and re-measured each sub individually and combined. Combined looked like the Before expected measurement. Then I EQd the combined sweep by (mistakenly) adding the REW filters to the input side of the MiniDSP. This gave a slight improvement. But then I re-read the posts and realised I'd misunderstood the instuctions to apply the REW filters to all subs. So I did that on the outputs - which is super easy as REW outputs for the MiniDSP 2x4HD - and I got the below.



That's a pretty sweet response. I might re-measure to flat rather than quite such an aggressive house curve, but it's nice to know I can do it. I also ran compression sweeps and I gave up long before the subs did.

I took the liberty of including the LCR measurements in the attached MDAT, as well as the individual, combined, and combined with EQ measurements.

Thank you for taking the time and helping me with this. I feel like I have a better understanding of the process, and I've achieved a better response than I could have imagined.
And it sounds great too. There's power there but it's not boomy or overpowering now, it feels much more controlled. The spectorgram looks good too, I think. Waterfall's a bit sloppy but I don't know I can do much about that.

 

[jtalden]

Should I still look at SWs to mains timing, or is that something you will do, or have you pre-pro handle?

 

[Conrad.]

Would I do more than time align them to a constant timing reference? Aligning the impulse responses is the same thing, right?
If you have a better or more comprehensive way of doing it I'd be very interested if you have the time.

Do you think the 100-127Hz dip in the L and R is an integration issue or could be fixed with timings? Given that that's above crossover and in the L only and R only responses I suppose it's a speaker/room interaction.

I'd love your thoughts.