Multi-Sub Optimizer - EQ For Multiple Subs

AdamAttewell

Prominent Member
Joined
Jan 21, 2015
Messages
1,379
Reaction score
1,077
Points
420
Location
UK
I will just echo what others have said about the 2x4 HD & MSO that they are amazing bits of kit.

Like Richard says getting the subs placed in the room in the correct position before any EQ is vital if you want a good result.

I took over 300 measurements with REW to find the best place for both my subs, I already thought I had found the best place for my subs which gave me this response.

Starting-Point.jpg


It was not bad but my subs starting rolling off at 100Hz & I wanted to make the crossover point smoother so I started moving a sub into every place a sub could be positioned & fired.

Then I used the trace arithmetic feature in REW to combine the best responses to see what it would look like when both subs were playing at the same time. With four contenders I settled on this one.

Front-Right-Fire-Forward-Rear-Left-Fire-Left.jpg


Then I loaded the measurements into MSO & with a lot of help & some trial and error I got this.

MSO-Predicted-Response-All-Four-Seats-Combined.png


Then to double check after loading all the filters into the MiniDSP I measured with REW to see how it looked at all the seating positions again.
Measured-Response-MSO-Applied-All-Four-Seats-Combined.jpg


This is what all the measured responses look like when separated & put next to the predicted response from MSO.
Predicted-vs-Measured.png


I am very happy with the result as being able to get this response over four seats spread over two rows I feel is pretty good.

Massive thanks to Richard for all his help & time as I would have never worked out half this stuff without his advice.

It took much longer than I wanted it to but no pain, no gain as they say.
 
I will just echo what others have said about the 2x4 HD & MSO that they are amazing bits of kit.

Like Richard says getting the subs placed in the room in the correct position before any EQ is vital if you want a good result.

I took over 300 measurements with REW to find the best place for both my subs, I already thought I had found the best place for my subs which gave me this response.


Nice work Adam. When you say 'four seats', are these inline with each other or two front and two behind?
 
Nice work Adam. When you say 'four seats', are these inline with each other or two front and two behind?

A front row & a rear row, one in front of the another.

They all have pop out foot rests so there is a large gap between to two rows to allow for this. Also the rear row is on a riser.
 
A front row & a rear row, one in front of the another.

They all have pop out foot rests so there is a large gap between to two rows to allow for this. Also the rear row is on a riser.

Cool. So has the 'main' seat SQ been compromised at all. Or would you say similar to before except now all the seats are equal?
 
Cool. So has the 'main' seat SQ been compromised at all. Or would you say similar to before except now all the seats are equal?

I will be honest I am now just sorting out Dirac so I am yet to fully test it out but I will be sure to let you know.

All I will say is everything is a compromise in HT :rotfl:
 
Last edited:
Then I loaded the measurements into MSO & with a lot of help & some trial and error I got this.

MSO-Predicted-Response-All-Four-Seats-Combined.png
Is the hump under 20Hz deliberate?

If you are happy then ignore but imv the shape of those curves is a bit u shaped. If you are so inclined then experiment with target curves.
 
Is the hump under 20Hz deliberate?

If you are happy then ignore but imv the shape of those curves is a bit u shaped. If you are so inclined then experiment with target curves.

That curve is from the tuning of the ports, I could of used MSO to flatten in out but I think you need all the output you can get below 20Hz. With them being ported they roll off heavy at 16Hz, with not much output after.

After speaking with Ed Mullen he recommended to not EQ it flat is it will take away from the output & ULF considerably.
 
That curve is from the tuning of the ports, I could of used MSO to flatten in out but I think you need all the output you can get below 20Hz. With them being ported they roll off heavy at 16Hz, with not much output after.

After speaking with Ed Mullen he recommended to not EQ it flat is it will take away from the output & ULF considerably.[/QUOTE

Is it not better to let MSO flatten the response, then run Dirac 'this way Dirac will not use to many filters', then use the mini to add a house curve after this?
 
That curve is from the tuning of the ports, I could of used MSO to flatten in out but I think you need all the output you can get below 20Hz. With them being ported they roll off heavy at 16Hz, with not much output after.

After speaking with Ed Mullen he recommended to not EQ it flat is it will take away from the output & ULF considerably.
Having an enormous hump in the response is v unusual, a smooth curve is generally preferred. The native response of the sub certainly will not look like that so your filters must be doing something a bit aggressive here. Can you post the filters? If it were me, I would definitely revisit this
 
Is it not better to let MSO flatten the response, then run Dirac 'this way Dirac will not use to many filters', then use the mini to add a house curve after this?

That is what I said, to let MSO EQ it flat if I wanted to.

Having an enormous hump in the response is v unusual, a smooth curve is generally preferred. The native response of the sub certainly will not look like that so your filters must be doing something a bit aggressive here. Can you post the filters? If it were me, I would definitely revisit this

You can see the slight hump in the before measurements of around 4dB created by the port tuning.

This has been exacerbated by only correcting to 20Hz with MSO & setting the MLP volume level to 80dB. No boosting within MSO.

Crazy amount of work now to redo all my Dirac projects for various seating positions, could just let Dirac flatten it out & see how it sounds but I am pretty happy with it.

I mean the hump is all below 20Hz which technically we cant "hear" only "feel" so its only coming into play when ULF active which is when you want that output in my opinion (explosions etc) as the way the human hearing works we are far less sensitive to ULF.
 
Thanks Adam. Please understand I’m not trying to be critical of your work flow. I’m just trying to figure out how to perform MSO :nono:
 
Thanks Adam. Please understand I’m not trying to be critical of your work flow. I’m just trying to figure out how to perform MSO :nono:

No worries, sorry if it came over that way. Thing do get a little lost in translation in text:smashin:
 
which is when you want that output in my opinion
The response shown is no different to a modal resonance really which typically means noticeably inferior to a smoother response. This is as true in the ulf range as it is higher up, there is just less content in that range so it won't be noticeable as often. If you are happy though then that is all matters in the end.
 
The response shown is no different to a modal resonance really which typically means noticeably inferior to a smoother response. This is as true in the ulf range as it is higher up, there is just less content in that range so it won't be noticeable as often. If you are happy though then that is all matters in the end.

Are you saying that there will be increase in modal ringing because of this curve?
 
I am saying it is no different to a modal resonance (which you would normally squash)

Ran this by Ed Mullen at SVS & he says this is room gain.

Ed Mullen said:
This is ‘room gain’ which is adding output at the deepest frequencies.

This is occurring at such a deep frequency that it will have a negligible effect on the subjective presentation of the subwoofer on most source material. You could certainly notch it out with a PEQ (if the software goes that deep), but I wouldn’t make it a high priority.

I then asked him if MSO was causing this hump below 20Hz & showed him the before EQ FR.

Ed Mullen said:
This one shows a rising response before 20 Hz which makes it appear the 15 Hz hump is smaller – but it’s about the same as the other FR graph relative to the 30-80 Hz drive level.

All well over my head but it seems there is more than one way to skin a cat.

Think I will stick with what I have for now as it sounds great.
 
Ran this by Ed Mullen at SVS & he says this is room gain.
there is no way that is room gain (and if it were, you'd see it in all measurements), it's simply a result of stopping MSO at 20Hz.

do you intend to use BEQ? if you do, just bear in mind that the amount of content you'll encounter in that range will go up *alot*. Depending on the filter shape and the levels you listen at, this could be quite dangerous for the sub (though it probably has a decent limiter to guard against this)
 
Last edited:
there is no way that is room gain (and if it were, you'd see it in all measurements), it's simply a result of stopping MSO at 20Hz.

do you intend to use BEQ? if you do, just bear in mind that the amount of content you'll encounter in that range will go up *alot*. Depending on the filter shape and the levels you listen at, this could be quite dangerous for the sub (though it probably has a decent limiter to guard against this)

Well I have ran MSO again with a correction range of 10Hz to 200Hz & there is still a hump below 20Hz of around 5dB & no matter what I try I can get it flat.

I can upload my MSO project for you to take a look if you like.
 
@AdamAttewell a couple of observations

* the reason it isn't tackling the <20Hz area is because the filters are only allowed a centre frequency of 20-150Hz

upload_2019-3-16_20-12-1.png


* you have given it a lot of filters to work with, this can lead to v long optimisation times so you may want to reduce this to 4-5 filters per channel and see what happens next
* you've only allowed it to use PEQ, other filter types are supported and can be useful (LS, HS)
* your optimisation was set to a v low, relative to the measurements, fixed reference level of 80dB which is why you would have ended up with that big hump (it is trying to squash it all down to 80dB but only down to 20Hz)
* you have no target curve in MSO, do you use one? if so, I would add it so that you're not filtering to flat only to then add more EQ later to reshape it again
* you have allowed it a v large delay range (+/- 40ms), I would reduce this significantly (40ms is an entire cycle at 25Hz so is a really really large shift)

I would do the following

* set the optimisation lower limit to 15Hz (and upper to probably 100Hz)
* set the optimisation level target to somewhere around 90dB (+/-2 dB, trial and error is required for best results)
* remove all the PEQ filters, add one that allows a centre frequency of 14-150Hz and then copy it so you have 5 on each channel
* add a target curve if you use one
* do a quick optimisation (10mins) to see what you get

is measurement group 4 the main seat btw? and are all seats equally important? you have 2 subs but 4 positions so you may find it worth experimenting with giving priority to certain seats
 
Last edited:
@AdamAttewell a couple of observations

* the reason it isn't tackling the <20Hz area is because the filters are only allowed a centre frequency of 20-150Hz

View attachment 1129967

* you have given it a lot of filters to work with, this can lead to v long optimisation times so you may want to reduce this to 4-5 filters per channel and see what happens next
* you've only allowed it to use PEQ, other filter types are supported and can be useful (LS, HS)
* your optimisation was set to a v low, relative to the measurements, fixed reference level of 80dB which is why you would have ended up with that big hump (it is trying to squash it all down to 80dB but only down to 20Hz)
* you have no target curve in MSO, do you use one? if so, I would add it so that you're not filtering to flat only to then add more EQ later to reshape it again
* you have allowed it a v large delay range (+/- 40ms), I would reduce this significantly (40ms is an entire cycle at 25Hz so is a really really large shift)

I would do the following

* set the optimisation lower limit to 15Hz
* set the optimisation level target to somewhere around 90dB (+/-2 dB, trial and error is required for best results)
* remove all the PEQ filters, add one that allows a centre frequency of 14-150Hz and then copy it so you have 5 on each channel
* add a target curve if you use one
* do a quick optimisation (10mins) to see what you get

is measurement group 4 the main seat btw? and are all seats equally important? you have 2 subs but 4 positions so you may find it worth experimenting with giving priority to certain seats

Many thanks for having a look, I will implement your suggestions & post my results.

I dont use a target curve in MSO, I let Dirac do that.

Any recommendation on min & max delay?
 
Last edited:
I dont use a target curve in MSO, I let Dirac do that.
the reason to add the same curve to mso is to reduce the probability that it filters down in one direction in MSO then goes the other way in dirac. The final curve will still be produced by dirac in any case, it can just work better if they work in harmony (to the same target). It's another option to try anyway, trial and error is the way forward with MSO for sure.
 

The latest video from AVForums

Is 4K Blu-ray Worth It?
Subscribe to our YouTube channel
Back
Top Bottom