Multi-Sub Optimizer (MSO): Lessons Learned, Tips & Tricks

[Keith_W]

Is this just something i'll have to deal with or is there a way to eq it out for my mlp?

It looks like it was in the original measurement before you ran it through MSO. It's likely not equalisable since it's probably not minimum-phase. Can't beat physics with DSP i'm afraid. It doesn't look too bad, I doubt if it would be very audible. If it bothers you, consider finding a better position for your subwoofer or getting another one.

 

[win]

hi guys. just watched the 2.0 video and i still have a noticeable dip at about 66hz. right now, I'm using a mini-dsp 2x4hd and for my shared filters, i have 10 peq's. For my per sub filters, i have 2xall pass second order filters, and 8 peq filters w/the settings recommended by the 2.0 video.

Is this just something i'll have to deal with or is there a way to eq it out for my mlp?

View attachment 457377

1. I think you are giving up a lot of headroom and probably increasing distortion substantially by boosting so much outside of the subwoofer effective range. Consider setting your curtains at around 25hz to 170hz.

This will give you more headroom to play with to fill in that dip at around 66hz and should produce a much better overall result.

2. With that said, I highly doubt you'd notice that narrow of a dip.

 

[Omartian]

awesome, thanks, i'll leave it alone. thanks guys

 

[Omartian]

Also, just want to confirm the steps with regards to importing my settings for my minidsp 2x4 hd:

1. for Input PEQ on my minidsp, use my shared filters biquad
2. for output individual sub peq on minidsp, use individual sub filter biquad
3. for delay on minidsp, use individual delay blocks for each sub
4. Leave inversion on minidsp alone
5. For the gain on my minidsp, do i take the gain block on my shared filter and apply it to my minidsp's input gain?

 

[Omartian]

ok, just took some measurements post MSO and by my eye, i'm pretty happy with the results. Attached are my measurements w/a few different hpf attached. I have the lavoce saf 184.03 so thinking of doing the 12hz hpf. Thoughts?

 

[DaveBoswell]

Also, just want to confirm the steps with regards to importing my settings for my minidsp 2x4 hd:

1. for Input PEQ on my minidsp, use my shared filters biquad
2. for output individual sub peq on minidsp, use individual sub filter biquad
3. for delay on minidsp, use individual delay blocks for each sub
4. Leave inversion on minidsp alone
5. For the gain on my minidsp, do i take the gain block on my shared filter and apply it to my minidsp's input gain?

1 to 4 are correct. No need to bother with 5 - you can ignore the shared filter gain setting in the transfer to minidsp.

 

[Omartian]

1 to 4 are correct. No need to bother with 5 - you can ignore the shared filter gain setting in the transfer to minidsp.

it's weird, my shared gain block value was 5. when i ran my input gain at 0, my curve was below my goal, but 5 gets me there perfectly.

 

[Omartian]

have a huge null at about 140 hz. is it ok to have my optimization range from 15-137 to avoid overeq'ing something that is above my xover point? i noticed in dave's video he went all the way up to 250hz or something like that but just wondering if that's necessary

i do notice that the response jumps outside of 137 but will that be problematic?

 

[Paffi]

I have the issue, that the individual filtered sub measurements won´t align well, using delay blocks and all pass filter. Even after the complete multistage optimization, the individual filtered subs have high discrepancies (yello circled at the screenshot). According to the (really great) help video, they should align much better after the filter setting. It´s four subs but I can´t explain why.

Ah one further question: after the update of mso I can´t find the option to display the filter channels, only the filtered measurements - how do I get that back?

 

[kiwifi]

I have the issue, that the individual filtered sub measurements won´t align well, using delay blocks and all pass filter. Even after the complete multistage optimization, the individual filtered subs have high discrepancies (yello circled at the screenshot). According to the (really great) help video, they should align much better after the filter setting. It´s four subs but I can´t explain why.

Ah one further question: after the update of mso I can´t find the option to display the filter channels, only the filtered measurements - how do I get that back?

I find it best to impose constraints on what MSO can do, in order to focus it towards a solution that makes sense. For instance you can start by only allowing it to alter the delays, then lock them down and only allow shared PEQ's to be modified. Lock those down as well and then allow the individual PEQ's to vary.

I haven't used MSO in a while, so I can't tell you how to display the filter responses in the new version, but it is definitely worth doing, to make sure that they look sensible.

 

[andyc56]

Ah one further question: after the update of mso I can´t find the option to display the filter channels, only the filtered measurements - how do I get that back?

In the Graph Properties dialog, expand the Trace Data Types folder as shown. Select Filter Channels. Now check the checkboxes for the traces you wish to add as shown below. This will enable the Apply button. Press it. You probably have manual y-axis scaling on your graph. In this case, you won't be able to see the added traces, as the nominal gains for the output channels are 0 dB, and thus out of the y-axis range.. That means you'll need to add trace offsets to them.

To add trace offsets, expand the Trace Properties node. This will show the added trace(s) (four in the example above and below). I have added an offset of 80 dB to the traces. Do this for each of the traces you have added (four in my case). This change will activate the Apply button. Press it to apply the changes.

The result for my example is shown below. You can see the traces have been shifted to 80 dB, so they are now visible.

The recent changes for the new Graph Properties dialog were made so you no longer have to keep going back and forth between the Graph Properties dialog and the since-removed Trace Properties dialog. As you add or remove traces, they will show up or be removed under the Trace Properties node as above. To rename a trace, select its node and either click it a second time or press F2 (same as Windows File Explorer). Edit the name in place, and the new trace name will show up in the legend right away. You can also delete a trace by selecting its node and hitting Delete. This is easier then finding it on the property page from which it was added, then unchecking it and hitting Apply, although you an still do that too if you want.

For help on the dialog as a whole, put the cursor in the tree on the left and hit F1 or click the Help button. For help on a particular page, activate that page, put the cursor in a control on that page, and press F1 or click the Help button.

 

[Paffi]

You probably have manual y-axis scaling on your graph. In this case, you won't be able to see the added traces, as the nominal gains for the output channels are 0 dB, and thus out of the y-axis range.. That means you'll need to add trace offsets to them.

That will be the issue, as I added the traces before but couldn´t see them. Thank you!!! Now I only have to figure out, why the individual filtered measurements don´t align well, after delay and allpass was set.

 

[DaveBoswell]

That will be the issue, as I added the traces before but couldn´t see them. Thank you!!! Now I only have to figure out, why the individual filtered measurements don´t align well, after delay and allpass was set.

The individual filtered measurements of Frequency Response for the subs on their will not align well to each other in the optimized result - the only thing you should concern yourself with is the combined Frequency Response of all subs on at the same time as that is what you hear. The individual Phase response however is very useful as I noted in my video to ensure that your subs are reasonably well aligned to each other in the Phase domain as that will result in the best total output of your system.

IMO Your phase plots look pretty good overall with good overlap between the subs so you should be good to go. If you'd like, post your updated plots here once you have activated the Filter channels, and turned off the Magnitude response of the individual filtered measurements (keep the Phase response on) and we can have a look.

 

[Paffi]

The individual filtered measurements of Frequency Response for the subs on their will not align well to each other in the optimized result - the only thing you should concern yourself with is the combined Frequency Response of all subs on at the same time as that is what you hear. The individual Phase response however is very useful as I noted in my video to ensure that your subs are reasonably well aligned to each other in the Phase domain as that will result in the best total output of your system.

IMO Your phase plots look pretty good overall with good overlap between the subs so you should be good to go. If you'd like, post your updated plots here once you have activated the Filter channels, and turned off the Magnitude response of the individual filtered measurements (keep the Phase response on) and we can have a look.

the combined frequency was / is fine - I was just curious because in your great video, the individual subs did align very well after time and phase alignment and you explicitly mentioned it, so I thought it should be the same for me.

With the adjusments in the trace properties, I get the correct view now and I think the individual filters makes much sense, when compared to the "natural response" (light grey) and the necessary adjustments.

The result, measured with rew, is also fine. I know, it could be even more linear, I had that in the past, then its a perfect straight line (It´s not that hard to get that for showing purpose). With this version I tried kind of "chaotic" measurement points (12) with different hights and position instead of a structured approach with everything at ear level and defined distance. Hence the more uneven result, but I´m not doing this for text book perfect frequency responses at the main spot but a more even distribution in a wider area and more focus at the decay.

What I´m quite unhappy yet is the step response. with 3 peaks nearly the same volume. Thought I would be correct after mso, but in fact it isn´t.

Furthermore I want to try mso to find the perfect phase cancellation delay, instead of phase alignment, for the rear subs. While I know it in general, there may be a more frequency specific phase cancellation with allpass possible. Until now I can´t figure out to force mso to search for that - will be the next project.

 

[andyc56]

Furthermore I want to try mso to find the perfect phase cancellation delay, instead of phase alignment, for the rear subs. While I know it in general, there may be a more frequency specific phase cancellation with allpass possible. Until now I can´t figure out to force mso to search for that - will be the next project.

I looked through some of your previous posts in this thread and noticed you have a DBA. I didn't realize that before. A DBA will give better results than MSO will in all respects except for SPL capability. MSO is not intended for use with a DBA. It is for use in situations where domestic considerations prevent the use of a DBA. The documentation warns against using SPL optimization with a DBA. SPL optimization is part of the multi-stage optimization. In a section titled Some Notes About SPL Maximization, it says:

You should not run an SPL optimization on a Double Bass Array (DBA) system. A DBA system depends on having the rear subs in opposite polarity with the front ones. It deliberately sacrifices SPL capability in order to optimize the uniformity of the sound field over position in the modal frequency region (which it does very effectively). MSO's SPL optimization will likely interfere with the proper operation of a DBA.

 

[Paffi]

I looked through some of your previous posts in this thread and noticed you have a DBA. I didn't realize that before. A DBA will give better results than MSO will in all respects except for SPL capability. MSO is not intended for use with a DBA. It is for use in situations where domestic considerations prevent the use of a DBA. The documentation warns against using SPL optimization with a DBA. SPL optimization is part of the multi-stage optimization. In a section titled Some Notes About SPL Maximization, it says:

The measurements postet are not a DBA, just 4 subs. As my room is not textbook perfect for dba, im trying with different configurations, DBA is included. But it´s not an real dba, just 2 subs on each the front and back wall. Currently I´m also using 2 active bass traps in additon (PSI Avaa C214). So I try what is possible with multi sub and mso and I´m also experimenting with phase cancellation.