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

[scotchrocks]

Those subs would have a more basic functionality, with just filters, and usually only a few of them. Would that work? I don't know. I have a somewhat similar question. I am using MSO and a miniDSP 2x4HD for my three or perhaps later four subs, but I would like to equalize the 80-200 Hz range of my highpassed main speakers as well. I can use the filters on my RME ADI-2 for that, but first, those would also affect the response of the already equalized subwoofers, and they would also be more basic. Or would I do better to simply equalize the subs with MSO, and subsequently use REW to design a few filters for the range of about 80-200 Hz, and implement those on the ADI-2, and separately for each of the two main speakers?

I think the best approach would be to optimize the sub response with MSO independently first. Then, run the REW measurements with your mains + MSO subs together and apply the resulting filters to the entire system from your ADI-2. Since the measurements take into account the total system response, it should be fine that the filters apply to the mains and subs. This kind of mirrors the workflow with AVR setups, where you optimize MSO independently to get the best "single sub" response, and then you can use the AVR to run Audyssey or other EQs with the MSO sub response.

However, one thing I'm not sure of is how you're applying the crossover to the mains and subs. It sounds like you have a high pass for the mains, but is there another low pass filter for the subs somewhere (maybe in the miniDSP or MSO config)? Just wondering, since if you're applying the filters to the whole system response from the ADI-2, you'd want to make sure that the mains/subs crossover is already set up before doing so.

 

[Willem]

I think the best approach would be to optimize the sub response with MSO independently first. Then, run the REW measurements with your mains + MSO subs together and apply the resulting filters to the entire system from your ADI-2. Since the measurements take into account the total system response, it should be fine that the filters apply to the mains and subs. This kind of mirrors the workflow with AVR setups, where you optimize MSO independently to get the best "single sub" response, and then you can use the AVR to run Audyssey or other EQs with the MSO sub response.

However, one thing I'm not sure of is how you're applying the crossover to the mains and subs. It sounds like you have a high pass for the mains, but is there another low pass filter for the subs somewhere (maybe in the miniDSP or MSO config)? Just wondering, since if you're applying the filters to the whole system response from the ADI-2, you'd want to make sure that the mains/subs crossover is already set up before doing so.

Thanks, this is clear to me. Thus far I have only equalized the three (quite different) subs using MSO/2x4HD, and with very good results, but I wanted to go beyond that and also do a bit of equalization on the main speakers. So I will do that with REW/ADI-2 in a second stage, as if with an AVR with inbuilt eq. As for the crossover, this is quite basic. The power amp has an 80Hz passive high pass filter in the cable between the ADI-2 and the power amp. Speakers are Quad 2805 electrostats, so with somewhat limited headroom, hence the high passing. The subwoofers are low passed using their own inbuilt low pass filters. That works well, even though it does not have the flexibility of an adjustable crossover. I had to pick a fixed frequency for the passive crossover into the Quad 606-2 power amp, and adjust the subwoofer crossovers as best I could. But I think it works. Personally I think it is disappointing that consumer power amplifiers do not have inbuilt and adjustable high pass filters such as many pro audio power amps do, but it is what it is.

 

[Willem]

The only disadvantage of equalizing with REW is that it will only be for one listening position. I guess I shall have to measure somewhere in the middle of the room because there is no main listening position.

 

[Keith_W]

The only disadvantage of equalizing with REW is that it will only be for one listening position. I guess I shall have to measure somewhere in the middle of the room because there is no main listening position.

Not really, REW does what you tell it to do. There is nothing stopping you from measuring from multiple positions and averaging the responses. Or better still, use MSO.

 

[Willem]

This is to equalize the whole system, including the main speakers, after using MSO to equalize the subs, see above. But yes, I can of course average multiple positions.

 

[jaakkopetteri]

I would use MMM instead of measuring discrete positions

 

[scotchrocks]

This is to equalize the whole system, including the main speakers, after using MSO to equalize the subs, see above. But yes, I can of course average multiple positions.

I think that should be alright, since MSO would be taking care of the most troublesome lower frequencies. The rest of the frequency response shouldn't be quite as position-sensitive, based on the dispersion of your speakers. I think a few REW measurements should get you in the right ballpark.

 

[mmares056]

I ran a new optimization using the new Mult-Stage option and this is what I got. As you can see based on the results my seating position is not optimal but this is the best I can do in my room as far as room treatments, seating position and speaker placement go. I think these results are a good compromise for me since I'm the only one that cares about home theater sound in my home. The double graph is a comparison of my seat to seat variation latest MSO optimization vs current REW spatial averaged measurements at the MLP(6 measurements, between 6-8 inches from each other. Comments welcome.

 

[scotchrocks]

I ran a new optimization using the new Mult-Stage option and this is what I got. As you can see based on the results my seating position is not optimal but this is the best I can do in my room as far as room treatments, seating position and speaker placement is concerned. I think these results are a good compromise for me since I'm the only one that cares about home theater sound in my home. The double graph is a comparison of my seat to seat variation latest MSO optimization vs current REW spatial averaged measurements at the MLP(6 measurements, between 6-8 inches from each other. Comments welcome.

Looks like a pretty solid result for the optimization range. It's great that you confirmed the response with REW versus the modeled result from MSO. They match incredibly closely.

One question is whether you tried expanding the optimization range on the upper end. I optimized up to 200 Hz, but I've heard you generally want to optimize at least to 2x your crossover frequency. So, 160 Hz if you're running an 80 Hz crossover.

 

[mmares056]

Looks like a pretty solid result for the optimization range. It's great that you confirmed the response with REW versus the modeled result from MSO. They match incredibly closely.

One question is whether you tried expanding the optimization range on the upper end. I optimized up to 200 Hz, but I've heard you generally want to optimize at least to 2x your crossover frequency. So, 160 Hz if you're running an 80 Hz crossover.

Yeah others have commented and recommend optimizing to the upper 100Hz range for smoother subs to mains integration. I need to do more raw measurements because my subs are rolling off significantly above 120Hz. Someone on the forum said the roll off is a result of my raw sub measurements being Low Passed in your AVR / Processor.

 

[JStewart]

Yeah others have commented and recommend optimizing to the upper 100Hz range for smoother subs to mains integration. I need to do more raw measurements because my subs are rolling off significantly above 120Hz. Someone on the forum said the roll off is a result of my raw sub measurements being Low Passed in your AVR / Processor.

It’s true there is typically a low pass filter at 120Hz on the LFE channel (LPF for LFE). With some devices it can be turned off, some not.
A way around it is to set the XO for your fronts as high as possible, then disconnect one of the front speakers so when you send your sweep to it the bass management will redirect to the sub you are measuring.
And while it shouldn’t matter given the frequencies in REW’s timing chirp, I would still set whatever speaker that will be used for the timing chirp to Large.
Another option, but a bigger hassle if you’re using a processor or AVR with pre-outs, you can for example, set the center channel to Large, physically disconnect the speaker and then connect the miniDSP to the center channel pre-out. Now you can send a full range signal to the sub being measured that will roll off at the subs limit instead of a lower XO.

 

[Paffi]

After the release of the multistage optimization: is there a consensus which method is the "best". As for now I'm only using multistage and I'm wondering if I could achieve better results with the other methods.

 

[DaveBoswell]

The multi stage is your best option if you care about keeping the maximum output and dynamic range of your subs. The other methods were created as MSO evolved over time but now IMO are mostly useful if you want to experiment and force MSO to do unusual things.

If you are looking to improve seat to seat variation while using the multi stage approach, just increase the SPL Penalty allowed parameter to a higher number (see below). Personally I don't like setting any higher than 0.2 to 0.5dB such that I force MSO to find a "good enough" solution across multiple seats without sacrificing output (MSO default is 1dB which for my setup where I only care about 2 or 3 seats is too high).

If you allow for a very high SPL Penalty (say over 5 or 10dB), then the multi stage with perform much like the "classic" method - it will definitely result in better seat to seat variation, but will not improve the MLP response, and could very possibly result in very significant reduction in output capacity of the subwoofer system as it may use phase cancellation as a method to refine seat to seat which can really penalize efficiency.

 

[Paffi]

Thank you - Yes, I already increased the Penalty accordingly to the instructions of MSO online. But It´s a good idea, to increase it further - I´m trying to force MSO to achieve the lowest reverberation time (with phase canceling).

 

[DaveBoswell]

If you are thinking using phase cancelling with randomly placed subs then I would suggest that the multi stage approach using delays/AP/PEQ will give you equally good if not better results with better (potentially much better) efficiency and output so would recommend against it.

If you are thinking of a dual bass array approach mimicking Trinnov, etc then MSO might be helpful to get the front arrays and rear arrays (analyzed as separate sub groups in different MSO files) time aligned and in that case I would not recommend using multi stage but a combination of separate runs of "align front array", "align rear array" then treat front array as if it is a single sub and rear array as it is is single sub and doing a 3rd MSO run to get the final delays and shared PEQ between the front a rear arrays for the DBA.

I have not done it, and I have heard a few others have tried but ended up going back to "classic" approach to increase efficiency and output, but conceptually I think it's possible but it would require a lot of experimentation and testing.

 

[Paffi]

Yes, I´m thinking about phase canceling - I used my setup for some time in dba-configuration. As the conditions are not the best, it´s no textbook installation (living room), but worked somehow. Now I try MSO to learn what It can do with a "regular" bundle of subs and further, if it can also optimize installations with phase cancellation. I´m already experimenting with locking some parameters (like delay, as I know it already). Will see whats possible. Will try your suggestions too! It´s just a rabbit hole...

EDIT: Do you know if I can export filtered measurements (the predicted frequency reponse after mso) out of MSO and import them in REW? Or at least exporting the filters, for importing in REW and apply there at the measurements. Just do avoid hundreds of measurements....

 

[DaveBoswell]

Yes, I´m thinking about phase canceling - I used my setup for some time in dba-configuration. As the conditions are not the best, it´s no textbook installation (living room), but worked somehow. Now I try MSO to learn what It can do with a "regular" bundle of subs and further, if it can also optimize installations with phase cancellation. I´m already experimenting with locking some parameters (like delay, as I know it already). Will see whats possible. Will try your suggestions too! It´s just a rabbit hole...

Interested to see what you can come up with so please share back here. I've given this some thought over the last while, but don't have the time nor the sub setup to give it a try but here's a bit more of my thinking:

Step 1: align front array
- goal of DBA is to use a 1 dimensional or 2 dimensional line array from the subs on both the front and rear walls that creates as close to a planar wave as possible
- if you have parallel front and rear walls for an array of say at least 3 subs, the length modes will be accentuated by room resonances more so relative to the other room modes and will show up as strong peaks in the Frequency Response plots
- as such, if you do a suite of measurements in the left-right plane at the room front-to-rear mid point (say a grid of measurements 3 to 5 wide by 1 or 2 high), then you should see strong peaks at the first, and to a lesser degree the third length (the second will be a null at the room mid-point) modes. So if room distance between front and rear walls is 6m, then the first modal frequency is at (343 m/s / 6m) / 2 = 28.6Hz, second at 57.2Hz, and the third will be at 85.8Hz.
- so you could use MSO to Maximize SPL between say 24Hz and 34Hz range (+- 5Hz from the first and highest modal frequency) but using only front subs and delays and some subtle relative gains (say +/- 2 or 3dB or range in MSO) between the subs using all measurement positions
- this optimization would have the result of phase/time aligning the subs along this measurement plane which IMO is a proxy for a planar wave - it will not be perfect the more the room is not rectangular or has significant furniture or other shapes that can disturb the sound wave, but it should be reasonably close at these very low frequencies

Step 2: align rear array:
Then you can do the same thing for the rear array (and turn off all front subs for this testing) but in reverse.

Step 3: determine best delay and gain between front and rear arrays:
Now you have the results for front and rear arrays that will result in planar waves, so you can lock these settings into the minidsp and remeasure the room but treat the the optimized front array as one "Front Array" sub and the rear array as a one "Rear Array" sub for MSO purposes. Then upload sub responses into a new MSO file and then reverse polarity and set a delay and gain setting to the rear array "sub" in MSO, set delay range to correspond to distance between front and rear arrays with plus or minus 5 or so ms, and then optimize in MSO to flatten response to a target (include a shared gain filter as well) with only the rear gain and delay filters as degrees of freedom. Lock these settings. Single stage optimization only! You do not want to use SPL Maximization

Step 4: fine tune with some PEQ for final response
If desired, you can add some shared PEQ to the MSO file in Step 3 for some fine tuning. I would recommend not using and sub level (i.e. Output) PEQ but you may want to experiment a bit, especially if there are some odd anomalies in the planar wave front caused by asymmetries in the room.

Just my 2 cents - I think conceptually this is a viable rational approach, and would be curious to know if MSO helps in optimizing, but have not validated by any trials on my end.

 

[andyc56]

EDIT: Do you know if I can export filtered measurements (the predicted frequency reponse after mso) out of MSO and import them in REW? Or at least exporting the filters, for importing in REW and apply there at the measurements. Just do avoid hundreds of measurements....

You can export the trace of any graph using the context menu shown below.

This will give you a text file with frequency in the left column, dB value in the middle one, and phase in the right column.

You can also use this method to export the trace as a target curve. When you import a target curve, the phase is ignored.