MSO Newbie

[kirian]

Hello everybody! Happy but a liitle desperate newbie ahead!. Going straight to the issue: I just wanted to integrate 2 subs and 2 mains for music only (no AVR). After careful measurements, MSO was seemed as the land of hope. Thing is, after countless hours of testing myriads of configs, i decided to measure with REW some of the very best of them according to the performance metrics of the software. The end result is that i cannot replicate, or to put it in another tone, the REW images have little or not at all corelation with the MSO predictions, worse: the final result is totaly unpredictable given the MSO simulation. I checked step by step all the faq section of the online tutorial, but could not find something relevant. (wishful thinking?). To this end, i suspect whether the biquads that are generated as xo files and i then put them to the cros section of the plugin manually, could be the culprit, thinking some kind of mixing with the original biquads of the plugin. Is that possible? I use a 8m USB cable attached to the UMIK1-could that be of any concern? Please have a look at the pictures, is this possible? Any help, suggestions? Thank you!

 

[-Matt-]

Hi Kirian,
Welcome.

First off - an 8m USB cable is not a problem.
(Are you using an acoustic timing reference)?

...And you are on a good path with MSO but there can be some gotcha's.

What hardware are you using for DSP? You mention applying biquads to the crossover section so that sounds like it is probably a MiniDSP device, but maybe not, so please tell us which one?

Another possible catch is that you have to be absolutely certain that all filers are set flat before you make your measurements.

You also need to make sure that REW is outputting the filters in the right format for your DSP device. (For some MiniDSP devices the output format needed differs depending on which plugin is used).

How many positions did you measure for MSO?

Is the "after MSO" REW measurement made in pretty much the same position as the input measurements for MSO?

 

[kirian]

Thank you very much for your help! Yes I did all measurements with acoustic time reference. I use a Flex Digital miniDSP device and the output of the MSO is set for the HD plugin namely 96Khz. I did 3 measurements as per the N+1 hypothesis (2subs). The distances between the measurements positions were equal between the MLP (Pos 1) in the graphs (60 cm on either side of the MLP). I do not use an AVR. And yes the after measurements were done in the exact MLP, which was marked in the floor. My concern is that even though I do make use of 10 PEQs during the simulation MSO process, the output biquads are only 8! Is this correct? My working hypothesis for this disparate results is that there is a basic fault in the biquads part. Please take notice that when I attempted to further correct the graphs using REW and putting the REW correction biquads in the input section (as I have already exhausted the output resources of the flex), the actual post REW graphs are uncannily similar to the prediction!

 

[-Matt-]

To this end, i suspect whether the biquads that are generated as xo files and i then put them to the cros section of the plugin manually, could be the culprit, thinking some kind of mixing with the original biquads of the plugin. Is that possible?

My concern is that even though I do make use of 10 PEQs during the simulation MSO process, the output biquads are only 8! Is this correct? My working hypothesis for this disparate results is that there is a basic fault in the biquads part.

You should apply the PEQ filters (biquads) to the output section for each channel of the MiniDSP Flex (not the crossover part).

You should probably be using the crossover part to high pass your main channels and low pass your sub channels.

If you are using the correct plugin (2×4 HD1) you should get 10 biquads per output channel and also 10 shared input biquads (in addition to the crossovers). You can optionally configure MSO to also use the input (shared) filter.

It shouldn't be necessary to manually copy the filters. MSO should allow your to export the filters to a text file. Pretty sure you can copy all 10 of the output biquads in one go if you use the advanced tab in the MiniDSP plugin.

 

[Keened]

And yes the after measurements were done in the exact MLP, which was marked in the floor. My concern is that even though I do make use of 10 PEQs during the simulation MSO process, the output biquads are only 8! Is this correct? My working hypothesis for this disparate results is that there is a basic fault in the biquads part. Please take notice that when I attempted to further correct the graphs using REW and putting the REW correction biquads in the input section (as I have already exhausted the output resources of the flex), the actual post REW graphs are uncannily similar to the prediction!

You need to increase the number of allowed biquads in the application setting to 10 (and the cross over to 8). MSO doesn't have the newer models in it's default setting choices.

When you say the online tutorial, you mean this right?

Finally, when you did your original measurements in REW, you remembered to turn off/change the output from L+R right? The acoustic timing is mono signal and mono signal.

 

[kirian]

Thank you all for your comments. I do not use this tutorial for three reasons:: first it refers to an older version of the MSO-the most recent one uses the measurement input wizard which simplifies the work needed to be done especially for newbies as me. However, later down the road i had to deal with the various aspects of the filters building and parameter estimation, so this tutorial did made sense. The second, most serious reason, is that according to the latest ''official'' tutorial, leaving unlocked the low pass filter (as suggested by the minidsp tutorial), is not recomended at all, since the design of the software is unable to take care of any changes in this value if i understood correctly-so everything i do is,, with crossover points locked. The third reason is that as per the miniDSP tutorial, it is not recommended to add filters to the mains. If this recommendation is followed, you end up with a more or less corrected sub response up to the low pass point and the rest up to the 200Hz region looking as a total mess.
But....my case is even worse....I cannot replicate the MSO prediction especially in the low pass region! Is this due to the fact that i add PEQs in the mains?, is it a major, consistent fault on my side, since importing the generated biquads in the advanced mode of the plugin is NOT so straightforward as it looks-you have to insist on that and found many times, that the whole process is a bit tedious. Is it in my measurements? As a last resort i decided to make a whole buch of new measurements increasing the number per mic position. Even though this amazing software, to the best of my understanding, is designed for subs integration mainly in the home -theater settings, doing a subs + mains integration may offer serious advantages in the music only applications.

 

[-Matt-]

I caveat the following with the fact that I've only done sub EQ - not sub-main integration so far...

Within MSO you can check your hardware configuration by clicking Tools > Application Options...

Obviously you won't select the AVR or crossover checkboxes - but make sure the miniDSP specific biquad one is selected and set the sample rate to 96 kHz. To start with you probably just want to use the 10 output biquads. If you use the shared input biquads they may be unintentionally applied to the mains.

Once you have run MSO export the filters to a set of text files by clicking Config > Save Config to miniDSP Biquad Text Files...

Inside the text files for each sub there is a block of code (the biquad) for each filter. You can paste this text into the advanced tab for each filter in the miniDSP plugin...
Note that I'm copying biquad6 from the MSO file into the output EQ6 of the miniDSP plugin but that inside the miniDSP plugin the first line of text always says biquad1, regardless of which filter you are editing.

When you have pasted the biquad in, click process and you should see the filter graph update. You can also toggle the bypass to check that each biquad makes the expected change. Make sure that all biquads are pasted into the appropriate filter position, and that all 10 filters are correctly filled and aren't bypassed by mistake.

(Yes, I allow boosting as well as cuts. I don't care about headroom as I run my subs at half volume and can already make more bass than I'd ever want)!

I also run correction to as high frequencies as possible (250Hz) to try to keep the sub response as smooth as possible throughout the crossover region.

 

[kirian]

Thank you so much for the hassle to presenting the process! Really appreciate it! So far, what do you do with the xo files? On my side, these files contain 10 (!!!) biquads, which I copy/paste to the crossover section of the plugin. Is this totally wrong? Are these files embedded in the .txt files already? As per MSO tutorial the crossover section take priority in filling up by the generated biquads of a config that exceeds the output biquads capacity of the plugin.
Thank you so much for your time and suggestions.

 

[-Matt-]

In my setup I use the filters in the crossover section as additional filters for sub correction because my AVR already applies its own crossovers.

In the right of the second image in my post above you can see all of the text files that MSO generates for my configuration. The ones with the suffix "_xo.txt" contain the biquads that I paste into the miniDSP crossover sections.

I assume that you will need to paste the biquads that define your crossovers into the crossover section. (Although the same can likely be achieved using the basic tab and setting the crossover frequency etc).

However, I think the crossover sections should only have 8 biquads, whilst the output sections can have up to 10. Perhaps this wasn't configured correctly during MSO setup?

 

[Keened]

But....my case is even worse....I cannot replicate the MSO prediction especially in the low pass region! Is this due to the fact that i add PEQs in the mains?

Quite possibly actually. I would generate EQ filters for the mains using AutoEQ on REW after you've integrated the subs and remeasured the new in room response. And or if you have the Dirac version let Dirac handle it. On that note, you can't do any kind of IIR filtering/input filtering if you have the Dirac plugin. That is an either/or kind of situation, so try it with no input filters/all pass filters/etc.

I can't speak for leaving the cross over locked/unlocked, since it always wants to push the cross over as high as possible either way in my setup, so there is very little difference.

But I would ask how steep your cross over filters are; is the messiness coming from the interaction between the subs and the mains, or is the low end of the mains messy? One of the benefits of doing crossovers entirely in the digital realm is you can use super steep crossovers like the 4th order Linkwitz-Rileys.

Also don't forget to turn on the biquads you paste in when doing the cross overs. The PEQ biquads let you import and turn themselves on, but the crossover biquads require a few extra clicks.

 

[kirian]

Keened thank you very much for your comments!
Ok! I have done countless simulations during the last three weeks. What I found amazing with the software is that you can directly, in real time, eyeball the changes you induce to your basic measurement curves by adding crossovers or changing the cross points. This feature is unbelievable! My two best configs, out of trial and error attempts and pure luck, because I CANNOT predict the final REW response, have been accomplished using 48 Linkwitz in the subs and 24 Linkwitz in the mains. Whenever I tried to add all pass filters the REW result was even more horrid, and yes the most horrendous part of the spectrum is below 30Hz. Take notice please that with my two subs in the initial uncorrected measurements, 15 Hz is easily reached. I thought that the software could be ideal since you can apply all these different values in one go and see what happens. But for me, subs and main integration without AVR remains unsolved, as I am not able to configure the end response of the combination. Instead, doing my subs first and then putting crossover points in the miniDSP after, seems futile, as I would need the same or even better luck to accomplish the task. Dirac surely remains the last option. But, subs + mains integration remains an unmet need and hoped a lot in MSO!

 

[kirian]

However, as a second thought, your genius suggestion to try using REW correct above the cross points after doing the subs makes sense and I will definitely try it. This requires to leave mains without PEQs and correct the subs, which seems a worthwhile effort before Dirac.