MSO- Multi Sub Optimizer Beginners Thread

[NorCalRP]

Looking forward to the next steps towards getting everything corrected.

 

[NorCalRP]

Not what I was expecting. I'm guessing I did something wrong because this is nowhere near as flat as what MSO led me to believe it would be

 

[NorCalRP]

Before and after- I didn't set the level to 75dB on the before. I tell ya- this looks like nothing happened.

 

[Pogre]

View attachment 31478

These are the initial measurements of each sub, all three positions averaged.

View attachment 31479

This is all subs averaged (the average of the three above averages)

View attachment 31480

The red line is the average of all 9 individual measurements.

I have a pile of charts, lol.

 

[JStewart]

View attachment 31485

Before and after- I didn't set the level to 75dB on the before. I tell ya- this looks like nothing happened.

Houston we have a problem. Def not my mso experience.
Btw your before graph in post 41 looks awfully similar for the 3 positions. Possible but are you sure the data for each sub and position is correct?

 

[NorCalRP]

Houston we have a problem. Def not my mso experience.
Btw your before graph in post 41 looks awfully similar for the 3 positions. Possible but are you sure the data for each sub and position is correct?

The traces of the three measurement groups (left seat is the response of all three subs combined at that seat, so on and so forth? Of that graph are offset from one another by 15dB to be able to watch the optimization of each response curve in real time while it's running.

 

[NorCalRP]

Last night by the time I got my bi-quad filters into miniDSP I had been at it for almost 8 hours and was 6 hours away from needing to wake up and go to work. I'll be back at it tonight to see if I can determine what's going on. Basically, the PEQ produced by the bi-quad values apparently did nothing to the resulting measurements, so I'm fairly certain I did something wrong between running the optimization and importing the bq info into the PEQ of the miniDSP

 

[andy_c]

Before and after- I didn't set the level to 75dB on the before. I tell ya- this looks like nothing happened.

Yes, the before and after data look suspiciously similar. This suggests that the miniDSP filters might be in an unexpected state. One thing you can try is in MSO, create a new graph and add the three sub channel frequency responses to it. To do this, you choose "Filter Channels" under "Data" in the Graph Properties dialog that appears when you first create a graph. Check the three sub channels to add their responses to the graph. Then connect your computer to the miniDSP and sync them in the miniDSP plugin software. When you observe the response of each output channel in the miniDSP software, it should match up with the corresponding response in the MSO plot. The exception is that MSO includes the effect of any attenuation added to the channel, while the miniDSP software shows only the effects of the biquads, not any gain/attenuation effects.

 

[NorCalRP]

Yes, the before and after data look suspiciously similar. This suggests that the miniDSP filters might be in an unexpected state. One thing you can try is in MSO, create a new graph and add the three sub channel frequency responses to it. To do this, you choose "Filter Channels" under "Data" in the Graph Properties dialog that appears when you first create a graph. Check the three sub channels to add their responses to the graph. Then connect your computer to the miniDSP and sync them in the miniDSP plugin software. When you observe the response of each output channel in the miniDSP software, it should match up with the corresponding response in the MSO plot. The exception is that MSO includes the effect of any attenuation added to the channel, while the miniDSP software shows only the effects of the biquads, not any gain/attenuation effects.

I'll check that out tonight and report back

 

[andy_c]

Also, in in MSO's Tools, Application Options, choose "Hardware" and ensure that the sample rate is what the 2x4 HD uses, which is 96 kHz. If this was incorrect, you'll need to re-export the biquad files from MSO and re-import them into the miniDSP software. The maximum biquad count for input and output biquads should be set to 10.

 

[NorCalRP]

Also, in in MSO's Tools, Application Options, choose "Hardware" and ensure that the sample rate is what the 2x4 HD uses, which is 96 kHz. If this was incorrect, you'll need to re-export the biquad files from MSO and re-import them into the miniDSP software. The maximum biquad count for input and output biquads should be set to 10.

Ahh... didn't do that... cool. Is the max count set to 10 as the miniDSP has a 10 band PEQ? Also, would I correspondingly assign 10 PEQ filters per channel?

 

[andy_c]

Ahh... didn't do that... cool.

If the default sample rate value of 48 kHz was in effect when the biquad files were exported, the filters will be bonkers for the HD (but okay for the non-HD 2x4). To maintain a consistent filter frequency response with changes in sample rate, any system using digital filters (be they biquads or other structures) would need to change its internal filter coefficients. That's why DSP boxes having digital inputs use sample rate conversion to maintain a fixed internal sample rate regardless of the sample rate of the digital signal being fed to it. It allows them to not have to deal with the problem of changing filter coefficients when the sample rate is changed.

Is the max count set to 10 as the miniDSP has a 10 band PEQ? Also, would I correspondingly assign 10 PEQ filters per channel?

10 biquads would be overkill. Whatever you used looks to be plenty, as you got some really good predicted results. For now, we need to concentrate on getting the measured results to match the predicted ones. When done properly, the differences after level shifts are taken into account should be on the order of less than a dB, except maybe near 200 Hz, where, depending on the room, the response starts to look like "hash".

In general, the number of biquads is one-half the order of the filter. A PEQ is second-order, so that's one biquad. Odd-order filters waste a half a biquad. A first-order low-pass filter, for example, would use one biquad. The 24 dB/oct low-pass used in a typical AVR crossover is 4th-order, so two biquads are needed to implement it.

 

[andy_c]

A few random, unrelated thoughts came to mind:

1) The last time someone had a problem in the big MSO thread with the final measurements in REW not matching MSO predictions, it was caused by a failure to use the acoustic timing reference when making the original REW measurements. I thought I'd mention that just in case.

2) To save a graph in MSO, just right-click on it and choose "Save as PNG File..." from the context menu.

3) This is nitpicking I know, but it's become a convention of sorts on the forums to adjust the y-axis range of graphs to get 5 dB/div. Using 10 dB/div or larger tends to hide problems.

 

[NorCalRP]

A few random, unrelated thoughts came to mind:

1) The last time someone had a problem in the big MSO thread with the final measurements in REW not matching MSO predictions, it was caused by a failure to use the acoustic timing reference when making the original REW measurements. I thought I'd mention that just in case.

2) To save a graph in MSO, just right-click on it and choose "Save as PNG File..." from the context menu.

3) This is nitpicking I know, but it's become a convention of sorts on the forums to adjust the y-axis range of graphs to get 5 dB/div. Using 10 dB/div or larger tends to hide problems.

Not sure what #3 means. Regarding the timing reference, I'm pretty sure I did. It was my first time using them, and hadn't heard a high pitch squeak at the beginning of the sweep. Is that what the timing reference is?

 

[andy_c]

Not sure what #3 means

If you make the difference between the maximum and minimum y-axis values (the y-axis span) quite large, both REW and MSO will set the dB per vertical division (which I've abbreviated dB/div) to 10 or even more. But when it's that large, it can tend to make a response look better than it really is. Here's a graph for which the y-axis is at 5 dB/div (horizontal lines at ...75 dB, 80 dB, 85 dB... etc., all in 5 dB increments.

You can usually achieve this by reducing the y-axis span, but sometimes you may need to maximize the app to get the graph display software to cooperate. Both REW and MSO behave this way.

Regarding the timing reference, I'm pretty sure I did. It was my first time using them, and hadn't heard a high pitch squeak at the beginning of the sweep. Is that what the timing reference is?

Yes. The timing reference is a chirp that starts at 5 kHz and goes up. So if you heard that during the measurement, you should be fine.

 

[NorCalRP]

I have a pile of charts, lol.
View attachment 31489

That looks delectable! What subs do you run?

 

[NorCalRP]

Did some investigating. I think I may have made some mistakes labelling my measurements. For starters, when I imported the BQ info sets into the miniDSP, the SVS and HSU were swapped. Also, looking at one of my MLP measurements in REW, then graphing that .frd file in MSO, they definitely didn't look the same- almost inverted.

I'm going to run a new set of measurements, be more careful about properly labeling and saving each one, then go from there. Having spent quite some time going through the tutorial yesterday, I feel good that it's going to be a much quicker process

 

[andy_c]

Did some investigating. I think I may have made some mistakes labelling my measurements. For starters, when I imported the BQ info sets into the miniDSP, the SVS and HSU were swapped.

If need be, you can rename the MSO sub channels (to e.g. match the channel names you picked in the miniDSP software) by selecting the channel node in MSO, then pressing F2. This opens up an in-place edit control, much like editing folder and file names in Windows File Explorer. Here's an example.

 

[andy_c]

Also, looking at one of my MLP measurements in REW, then graphing that .frd file in MSO, they definitely didn't look the same- almost inverted.

I missed this the first time. If you do a measurement in REW, then export it as an FRD file, then import it into MSO using File, Import Plot-Only Measurement, there is absolutely, positively, no f'ing way that MSO will show a different result than REW. Anyone making such a claim is wrong.

 

[NorCalRP]

Boy, This is really frustrating... I ran all the measurements again, and ran a new configuration. The optimized results looked pretty much identical to the first set of optimized results. I compared the filter channel response curves with what I imported into MiniDSP, and the peq curves that generated. Using the export feature from MSO, along with the import feature from MiniDSP, my HSU information and SVS information seemed to have gotten swapped somehow, as the curves did not correspond correctly. I then manually copied and pasted the biquad information, and got the curves to all line up and be looking the same, minus the attenuation mentioned. Here's what I ended up with:

This is a measurement of all 3 subs, at the main listening position, before running MSO

This is the same measurement, post optimization, with biquad info having been input into the MiniDSP plugin using the import feature

This is the same measurement, post optimization, having manually cut and paste each biquad into the corresponding PEQ slot (bq 1 into peq1, etc.)

@andy_c I did rename the sub channels. One thing I found very interesting was that when I entered all the information in the config view (adding channel, filters, associated measurements) I entered them in this order SVS(sub ch1), HSU(sub ch2), Ultimax(sub ch3). After the optimization finished, the SVS and the HSU had flip flop Ed on the tree. That's what clued me in to taking a look at the biquad values displayed in MSO to see how they translated into MiniDSP- they didn't go in as they should have.