Good DSPs for Home-Cooked Coefficients?

Frank Koenig

Frank Koenig

Sophomore
Mar 7, 2011
187
0
16
Palo Alto, CA USA
www.dunmovin.com
I'd like to get a speaker-processor style DSP box that lets me easily enter my own FIR and IIR filter coefficients. FIR with 240 taps at fs = 48kHz or 480 taps at 96kHz would do nicely.

I'm aware of Xilica. FourAudio looks interesting but I don't see much of a US presence. I have a Lab Gruppen PLM10000q but don't know how to get into the processor. Obviously speaker manufacturers, as Fulcrum, get to do this, but how about mortals? A board-level product would be fine, too. I really appreciate any suggestions of where to look and especially any tales of direct experience.

All this is so I can do some experiments, the likely outcome of which is that I'll buy good speakers that come with settings designed by people who know what they're doing. But, as with pushing plywood through saws, the process builds deeper appreciation for what's involved. And even a blind squirrel finds a nut now and then.

As an aside, I've implemented the Gunness-Hoy spectrogram representation in R (reading the impulse response captured by ARTA). Let me know if you want a copy to play with. It's a rather preliminary (and slower than molasses) work in progress.

--Frank
 
Re: Good DSPs for Home-Cooked Coefficients?

Hi Frank,

If you have a Lab.gruppen PLM amplifier, surely you can use the Lab.gruppen LAKE software to get in there and work with the DSP in the amplifier.

Otherwise, if you want to go the external route, the MiniDSP product seems to be a reasonable price for the power it can give you.
 
Re: Good DSPs for Home-Cooked Coefficients?

Soundweb London allows direct coefficient entry. A BLU-100 is obviously the cheapest but can only be clocked at 48k. The BLU-160 can run a 6700 Tap filter, but takes up 50% of the DSP horsepower at 48k. A 3000 tap filter takes up 52% at 96k.
 
Re: Good DSPs for Home-Cooked Coefficients?

David Sturzenbecher said:
Soundweb London allows direct coefficient entry. A BLU-100 is obviously the cheapest but can only be clocked at 48k. The BLU-160 can run a 6700 Tap filter, but takes up 50% of the DSP horsepower at 48k. A 3000 tap filter takes up 52% at 96k.
Click to expand...

Just because you CAN, does not mean that it is a good idea.

Have you calculated or measured the latency through a filter with over 1000 taps?

Here is a hint, it is not insignificant... in fact, it is quite significant... let alone running over 6000 taps...

I'm thinking that if you really feel you need that many taps to accomplish something, you are probably doing it wrong.

Incidentally, the MediaMatrix NION platform will also allow you to do direct entry for up to 1024 taps and it will support clock rates of 22.05kHz, 24kHz, 32kHz, 44.1kHz, 48kHz, 64kHz, 88.2kHz, and 96kHz. Even more interesting, however, is the ability to run different algorithms at different sample rates within the project. There are sample rate converter algorithms that allow you to upsample and downsample things so your critical audio path can be maintained at one sample rate, while your sidechains and other non-critical paths can be a lower sample rate. Really, you can make it as complicated as you want.
 
Re: Good DSPs for Home-Cooked Coefficients?

Frank Koenig said:
I'd like to get a speaker-processor style DSP box that lets me easily enter my own FIR and IIR filter coefficients. FIR with 240 taps at fs = 48kHz or 480 taps at 96kHz would do nicely.

I'm aware of Xilica. FourAudio looks interesting but I don't see much of a US presence. I have a Lab Gruppen PLM10000q but don't know how to get into the processor. Obviously speaker manufacturers, as Fulcrum, get to do this, but how about mortals? A board-level product would be fine, too. I really appreciate any suggestions of where to look and especially any tales of direct experience.

All this is so I can do some experiments, the likely outcome of which is that I'll buy good speakers that come with settings designed by people who know what they're doing. But, as with pushing plywood through saws, the process builds deeper appreciation for what's involved. And even a blind squirrel finds a nut now and then.

As an aside, I've implemented the Gunness-Hoy spectrogram representation in R (reading the impulse response captured by ARTA). Let me know if you want a copy to play with. It's a rather preliminary (and slower than molasses) work in progress.

--Frank
Click to expand...

I could make you a re-configurable version of our commercial DSP but it would cost you significantly more than other options:)

I would suggest you look into the analog devices EZ-Kit demo boards:
SHARC 21364 EZ-KIT Lite Evaluation Board | Analog Devices

I have found them on eBay for less than 100$US. Any board with a CPU/DSP better than a ADSP-21364 will suit your needs. Then download VisualAudio and a demo of VisualDSP++ from the analog devices website.

If you decide to go this route fell free to message me any time as I have lots of experience with this platform.

Mark
 
Last edited:
Re: Good DSPs for Home-Cooked Coefficients?

Thank you all for your suggestions. I now have lots to look into.

Josh: On the Lab.Gruppen PLM, as far as I know the Lake software (which I find pretty ponderous, but that's another story) does not let me enter raw coefficients. I know the internals can do a 5 ms FIR, since I'm using that as a crossover filter. I'll try contacting LG.

The MiniDSP looks really cool. Nothing I found on their Web site reveals how long an FIR it supports or what, exactly is involved in loading coefficients. They are out of Hong Kong. We'll see what there is in the way of support.

David: I definitely will check out Soundweb London. I did not know that they do raw coefficients. I'll barely scratch the surface of a 3000 tap filter for what I want to do. I wonder who uses such long filters and what for? An advantage is that this is a real piece of sound equipment that can be dropped into the rig when I'm done fooling around.

Silas: It's good to know the Itech HD will do that. For this I'm leaning toward a free standing processor, unless I can get the Lab.Gruppen (the bird in hand) to play.

Mark: This, too, I shall investigate. It is no doubt the most flexible solution but likely also the most effort to get going. If I'm serious about this stuff it may be what I should do. And thank you for the offer of some support. I shall not abuse it :)~:)~:smile:

--Frank
 
Last edited:
Re: Good DSPs for Home-Cooked Coefficients?

Frank Koenig said:
Silas: It's good to know the Itech HD will do that. For this I'm leaning toward a free standing processor, unless I can get the Lab.Gruppen (the bird in hand) to play.
Click to expand...

I don't believe the Lab can do it, at least not efficiently. Check out the scenario with EAW Greyboxes, it locks out ALL the inputs and does not allow any routing. I guess if you don't mind having only one input in use on a 4 channel amp, you can do it.
 
Re: Good DSPs for Home-Cooked Coefficients?

Silas Pradetto said:
I don't believe the Lab can do it, at least not efficiently. Check out the scenario with EAW Greyboxes, it locks out ALL the inputs and does not allow any routing. I guess if you don't mind having only one input in use on a 4 channel amp, you can do it.
Click to expand...
BTW, EAW's Single Mode went a long way to fixing this issue on the UX8800. Now greyboxes are defined on the outputs, not the "input" processing blocks.
 
Re: Good DSPs for Home-Cooked Coefficients?

Hi Frank,

Just wondering if you were able to find prices for the FourAudio gear?

Frank Koenig said:
I'd like to get a speaker-processor style DSP box that lets me easily enter my own FIR and IIR filter coefficients. FIR with 240 taps at fs = 48kHz or 480 taps at 96kHz would do nicely.

I'm aware of Xilica. FourAudio looks interesting but I don't see much of a US presence. I have a Lab Gruppen PLM10000q but don't know how to get into the processor. Obviously speaker manufacturers, as Fulcrum, get to do this, but how about mortals? A board-level product would be fine, too. I really appreciate any suggestions of where to look and especially any tales of direct experience.

All this is so I can do some experiments, the likely outcome of which is that I'll buy good speakers that come with settings designed by people who know what they're doing. But, as with pushing plywood through saws, the process builds deeper appreciation for what's involved. And even a blind squirrel finds a nut now and then.

As an aside, I've implemented the Gunness-Hoy spectrogram representation in R (reading the impulse response captured by ARTA). Let me know if you want a copy to play with. It's a rather preliminary (and slower than molasses) work in progress.

--Frank
Click to expand...
 
Re: Good DSPs for Home-Cooked Coefficients?

Frank,
Some that I am aware of, many of which have already come up:

Biamp Tesira
Crown iTech HD
Mediamatrix Nion
MiniDSP OpenDRC
Powersoft K
Soundweb London
Symetrix Jupiter and Symnet Edge

AFAIK, the Xilica box implements FIR differently, I don't recall the specifics, but folks like Fulcrum apparently can't use them.
Same deal with Lab Gruppen/Lake - FIR is implemented differently. However, I recall seeing some information that they were working on adding support for arbitrary coefficients, not sure where that is at though. Last I had heard they were aiming for late 2011, so not sure if that has stalled, been abandoned, or what.

As you are likely already aware, in the interest of latency, you want to minimize the number of taps as much as possible, and be aware of the time vs frequency relationship.

What is your intended application?
 
Re: Good DSPs for Home-Cooked Coefficients?

Jeff,

Thanks for that list and the heads up on the Xilica and Lab/Lake. I still have to make up my mind between a free-standing processor and processed amps. There are multiple requirements I'm trying to satisfy. I downloaded Powersoft's Armonia control application. It looks like what I want is in there and, while a bit bloated, it's much more to my liking than the Lab/Lake app, which makes me crazy.

My application is to try my hand at coming up with a set of tools to clean up some of the linear time and spatially invariant aberrations of speakers in the HF range.
I have a background in signal processing so have some idea of what may be involved. I've been simulating some things in R, which appears to be a pretty nice environment for this sort of work even though it's really for statisticians.

--Frank
 
Re: Good DSPs for Home-Cooked Coefficients?

Frank Koenig said:
Jeff,

Thanks for that list and the heads up on the Xilica and Lab/Lake. I still have to make up my mind between a free-standing processor and processed amps. There are multiple requirements I'm trying to satisfy. I downloaded Powersoft's Armonia control application. It looks like what I want is in there and, while a bit bloated, it's much more to my liking than the Lab/Lake app, which makes me crazy.

My application is to try my hand at coming up with a set of tools to clean up some of the linear time and spatially invariant aberrations of speakers in the HF range.
I have a background in signal processing so have some idea of what may be involved. I've been simulating some things in R, which appears to be a pretty nice environment for this sort of work even though it's really for statisticians.

--Frank
Click to expand...

Long ago when I went to visit Camco in Germany I saw the now FourAudio guys demo their DSP, way before it was FourAudio. They for sure know what they are doing but their GUI still needs a lot of work. I still can't believe they are pushing their DOS based measurement system:) It should be easy now a days to port it to windows. I would say that is your best commercial off the shelf option. But I have heard through the grapes that it is 3700$ dealer. If you are still having issues I of course might recommend you contact me. I do have a processing platform which I could adapt to allow you to insert your own FIR coefficients if required for your application.

Mark
 
Re: Good DSPs for Home-Cooked Coefficients?

Frank..all

Slightly related. I received an E-mail today from AFMG regarding release of a program called FIRmaker. There are little details in the e-mail, but something to potentially look at in the future.
 
Re: Good DSPs for Home-Cooked Coefficients?

David Sturzenbecher said:
Frank..all

Slightly related. I received an E-mail today from AFMG regarding release of a program called FIRmaker. There are little details in the e-mail, but something to potentially look at in the future.
Click to expand...

David,

FIRMaker is targeted at directional lobe shaping of arrays like column speakers, based on everything I've seen about it.
 
Re: Good DSPs for Home-Cooked Coefficients?

Jeff Babcock said:
AFAIK, the Xilica box implements FIR differently, I don't recall the specifics, but folks like Fulcrum apparently can't use them.
Same deal with Lab Gruppen/Lake - FIR is implemented differently. However, I recall seeing some information that they were working on adding support for arbitrary coefficients, not sure where that is at though. Last I had heard they were aiming for late 2011, so not sure if that has stalled, been abandoned, or what.
Click to expand...
The issue with Xilica and Lake processing, at least in the past, has been that the FIR coefficients are not "randomly assignable," so I've been told.

The new Firmaker software is intriguing, but I think Phil is correct -- it seems like it's intended for column and line array aiming. It would be really nice is AFMG would describe their products in a slightly more down to earth fashion. If they did, then we would all know exactly what it is and isn't for. Although they do say it's "for any loudspeaker array."
Software - AFMG FIRmaker
 
Re: Good DSPs for Home-Cooked Coefficients?

Just getting started on this concept of using FIR coefficient files, can someone elaborate on the process of creating these files and what the best use case would be?

Grant Conklin said:
The issue with Xilica and Lake processing, at least in the past, has been that the FIR coefficients are not "randomly assignable," so I've been told.

The new Firmaker software is intriguing, but I think Phil is correct -- it seems like it's intended for column and line array aiming. It would be really nice is AFMG would describe their products in a slightly more down to earth fashion. If they did, then we would all know exactly what it is and isn't for. Although they do say it's "for any loudspeaker array."
Software - AFMG FIRmaker
Click to expand...
 
Re: Good DSPs for Home-Cooked Coefficients?

Frank Pajak said:
Just getting started on this concept of using FIR coefficient files, can someone elaborate on the process of creating these files and what the best use case would be?
Click to expand...

Designing custom filters for a loudspeaker setup is just that, the design of custom filters. No one is going to have suggestions on what you should be creating. Good measurements are the key to making a good set of custom filters for your setup. What ever filter types you end up choosing, you should attempt to achieve equal delay between all the bands. The use of FIR linear phase filters might make that easier or might not. If you are trying to create correction FIR filters, which try to inverse the measured response of a certain driver, then there are a number of papers in AES which will provide suggestions on best practices.
 
Top Bottom