ProSoundWeb Community

Sound Reinforcement - Forums for Live Sound Professionals - Your Displayed Name Must Be Your Real Full Name To Post In The Live Sound Forums => SR Forum Archives => The Basement FUD Forum Archive => Topic started by: Bennett Prescott on January 08, 2011, 12:11:49 PM

Title: DSP Tower of Babel
Post by: Bennett Prescott on January 08, 2011, 12:11:49 PM
Friends, Romans, Countrymen...

I am going to write an article about different DSP platforms, in an attempt to have a semi-definitive source to point people to the next time I get asked why those Driverack settings sound so funny in the KT processor.

I would be touched if y'all would do me an enormous favor and help me get measurements from different processing platforms. I have designed an imaginary loudspeaker preset that I think will stress the differences in their filter definitions. If anyone thinks they have better ideas, speak now or forever hold your peace. I am trying to make it somewhat realistic while fitting within the filter capabilities every processor is likely to have. Any reasonably mainstream processor welcome. These exact frequencies and bandwidths may not be available (on purpose), please get as close as possible.

If you would like to help, please program these filters into whatever DSP you have at hand:
High Pass - Bessel - 97Hz 18db/octave
Bell - 280Hz - 2 octaves/0.66 Q -4.5dB
Bell - 176Hz - 0.6 octaves/2.39 Q +5dB
Bell  - 6.35kHz - 1/8 octave/11.54Q -8dB
Shelf - 8kHz - 1.6 octave/0.86Q/6dB +5.5dB

Measure with Smaart, using the standard internal pink noise source as a reference. Use 16 averages, 48k sample rate, and 24bit, no smoothing. Use auto small to set your delay compensation. Please take a measurement with a 16K FFT and one with FPPO. If you could also send me an IR file that would be invaluable.

Please label your measurements and email them to me, don't post them here... I need to know which processor which measurement came from, but I don't want everyone to know. Anyone who wants to go even further and put in bandwidth instead of Q and take another measurement of that would receive another free Internet.

bennettprescott at gmail dot com

Thanks in advance to those of you with a little time to waste and some measurement chops!
Title: Re: DSP Tower of Babel
Post by: John Roberts {JR} on January 08, 2011, 01:44:20 PM
Bennett Prescott wrote on Sat, 08 January 2011 11:11

Friends, Romans, Countrymen...

I am going to write an article about different DSP platforms, in an attempt to have a semi-definitive source to point people to the next time I get asked why those Driverack settings sound so funny in the KT processor.

I would be touched if y'all would do me an enormous favor and help me get measurements from different processing platforms. I have designed an imaginary loudspeaker preset that I think will stress the differences in their filter definitions. If anyone thinks they have better ideas, speak now or forever hold your peace. I am trying to make it somewhat realistic while fitting within the filter capabilities every processor is likely to have. Any reasonably mainstream processor welcome. These exact frequencies and bandwidths may not be available (on purpose), please get as close as possible.

If you would like to help, please program these filters into whatever DSP you have at hand:
High Pass - Bessel - 97Hz 18db/octave
Bell - 280Hz - 2 octaves/0.66 Q -4.5dB
Bell - 176Hz - 0.6 octaves/2.39 Q +5dB
Bell  - 6.35kHz - 1/8 octave/11.54Q -8dB
Shelf - 8kHz - 1.6 octave/0.86Q/6dB +5.5dB

Measure with Smaart, using the standard internal pink noise source as a reference. Use 16 averages, 48k sample rate, and 24bit, no smoothing. Use auto small to set your delay compensation. Please take a measurement with a 16K FFT and one with FPPO. If you could also send me an IR file that would be invaluable.

Please label your measurements and email them to me, don't post them here... I need to know which processor which measurement came from, but I don't want everyone to know. Anyone who wants to go even further and put in bandwidth instead of Q and take another measurement of that would receive another free Internet.

bennettprescott at gmail dot com

Thanks in advance to those of you with a little time to waste and some measurement chops!


Good luck... I hope you get a good response, maybe there are enough people with measurement software out there now.  Some of us have been pissing up this rope for a while.  The subject has been broached before,  but usually with arms thrown up in air, and cries of all is lost, we are damned, time to consider Hari Kari. I hope you are planning to be more constructive than those earlier expositions.

You seem to be hitting most of the known offenders, Q and filter treatment at passband extremes. [insert my well worn rant here about lack of official standards definitions for Q in boost/cut sections, yadda yadda].

Are you attempting to come up with a nomograph and/or equations to help users translate between different DSP Platforms? Or just group them by similar behavior? I think Rane has done the most work in that area to date, but last I looked they only covered a limited number of DSP platforms. http://www.rane.com/note167.html

If trying to come up with reliable translation metrics you might need  more/different data points, but I am not smart enough to tell you exactly what to ask for. I find it remarkable we haven't resolved this yet, but I have only been an uninvolved observer without a strong personal or business motivation to deal with this.

In my judgement we don't really need a firm standard that would create winners and losers (like pin 2 hot), just clear explanations from the sundry DSP sellers, of what "they" mean when they say... XYZ bandwidth or Q...  back when I was writing for RE/P I actually polled all manufacturers asking  which XLR pin they used as hot....and published the results in my column (in the 80s before AES standard).  Since you are writing your article for a magazine, ask your editor to make a similar request of all manufacturers (using his advertising list), like mine did for me back in the day. If the manufacturers refuse to honor this request list them in your article as unresponsive to embarrass them into responding for a follow up (this may require an editor with cohones, some are afraid to tread on advertisers).

I think I even started some work on a questionnaire for this within the last few years, also asking for a contact engineer from inside each company for follow up, but the editor I was working with flaked out on me, so this never happened. Then I got busy with real work.  

I wish you better luck than I had, with this Babel.

JR

Title: Re: DSP Tower of Babel
Post by: Bennett Prescott on January 08, 2011, 02:20:30 PM
JR,

I'm not trying to solve the problem, I'm just trying to document it. I especially do not plan on naming names. I just want people to understand that a difference exists and to get a rough idea of its magnitude. If people will at least convert between Q and bandwidth before entering settings into the wrong processor I will consider it a victory. Hell, there are settings out there that when I plugged them into the correct processor looked like they were for a different loudspeaker, so there's more problems there than I want to address.
Title: Re: DSP Tower of Babel
Post by: Ivan Beaver on January 08, 2011, 03:42:36 PM
What sort of time frame are you looking at for needing this?

I am really swamped right now, but will try to find some time to get a few graphs of some DSP's that I have access to.

As with most journeys-things like this will often raise more questions than provide answers.

But it is all good and helps to "raise the bar" for those who take the time to read and understand Very Happy .



Title: Re: DSP Tower of Babel
Post by: Bennett Prescott on January 08, 2011, 03:55:02 PM
Ivan,

This month?
Title: Re: DSP Tower of Babel
Post by: John Roberts {JR} on January 08, 2011, 04:06:57 PM
Bennett Prescott wrote on Sat, 08 January 2011 13:20

JR,

I'm not trying to solve the problem, I'm just trying to document it. I especially do not plan on naming names. I just want people to understand that a difference exists and to get a rough idea of its magnitude. If people will at least convert between Q and bandwidth before entering settings into the wrong processor I will consider it a victory. Hell, there are settings out there that when I plugged them into the correct processor looked like they were for a different loudspeaker, so there's more problems there than I want to address.


OK, but it is not exactly a new problem...  Another round of arm waving will not move the needle. Waiting for a standards body, is a bit like waiting for Godot. Nobody is enthusiastic about declaring winners and losers.

Box A, B, and D does this,
Box C, E, and F do that.
and Box G does something entirely different.

If you want to document it, actual data and names, will help people, know where they stand with respect to their DSP platform. Even without figuring out translation tables yourself. Showing that there are only a modest number of different DSP "types" should influence speaker manufacturers to publish presets to accommodate the small handful of variants out there. Naming names could help reduce this to a relatively simple  matter of determining which flavor preset to use with which DSP box.

Why bother with collecting accurate measurements if all you want to say is that they aren't the same. We already know that. There were posts and plots on the LAB here years ago showing how different they can be. I'm sure you could get permission to reprint those if all you want to do is demonstrate that a difference can exist.

Sorry I'm just weary of all this after several years of running this up the flagpole and being ineffective myself. You are young and might be able to get something accomplished in what's left of your lifetime.  Laughing  

More consumer uncertainty IMO will only drive them to cognitive paralysis or packaged solutions and away from generic DSP, and not help the market work more effectively.  

You have an opportunity to do more. I don't think there are that many variants to nail down. Just enumerating the sundry Q approaches could win you a LABEE at the awards ceremony. I'd vote for you.  

JR

Title: Re: DSP Tower of Babel
Post by: Ivan Beaver on January 08, 2011, 05:40:25 PM
Bennett Prescott wrote on Sat, 08 January 2011 15:55

Ivan,

This month?

I'll try to find some time.
Title: Re: DSP Tower of Babel
Post by: Ivan Beaver on January 08, 2011, 07:06:59 PM
Which version of Smaart do you want the files in.  I have 5 and 7.  I never bought into V6
Title: Re: DSP Tower of Babel
Post by: Bennett Prescott on January 08, 2011, 09:02:45 PM
7 Please, whenever possible.
Title: Re: DSP Tower of Babel
Post by: Chad Johnson on January 08, 2011, 11:56:53 PM
Bennett
Do you think there's any benefit to a similar approach to the EQ sections on digital consoles? I know it would be another project to undertake, but I always thought it would be interesting to know how different units would respond to different settings.

I will be able to contribute V7 shots of some DSP units. Are you coordinating different models in offline conversations? No sense taking the time if someone else is doing the same model.

-Chad
Title: Re: DSP Tower of Babel
Post by: Kemper Watson on January 09, 2011, 04:15:38 AM
Ivan Beaver wrote on Sat, 08 January 2011 17:40

Bennett Prescott wrote on Sat, 08 January 2011 15:55

Ivan,

This month?

I'll try to find some time.

Do you have a Driverack 260? I could let you use mine if it would help.
Title: Re: DSP Tower of Babel
Post by: Ivan Beaver on January 09, 2011, 01:57:10 PM
Kemper Watson wrote on Sun, 09 January 2011 04:15

Ivan Beaver wrote on Sat, 08 January 2011 17:40

Bennett Prescott wrote on Sat, 08 January 2011 15:55

Ivan,

This month?

I'll try to find some time.

Do you have a Driverack 260? I could let you use mine if it would help.

The actual measurement Bennett wants does not take any time.

It is all in the setup.

I don't have any DBX products.
Title: Re: DSP Tower of Babel
Post by: Ivan Beaver on January 09, 2011, 02:00:44 PM
Chad Johnson wrote on Sat, 08 January 2011 23:56

Bennett
Do you think there's any benefit to a similar approach to the EQ sections on digital consoles? I know it would be another project to undertake, but I always thought it would be interesting to know how different units would respond to different settings.

I will be able to contribute V7 shots of some DSP units. Are you coordinating different models in offline conversations? No sense taking the time if someone else is doing the same model.

-Chad

I feel consoles are a completely different situation.  Yes they may be different-but so what?  They should be adjusted by ear/taste-not trying to dupilcate settings-like you would do with a DSP.

I feel that the  console falls into the "artistic" side of things, while DSP's fall into the "scientific" side of things.
Title: Re: DSP Tower of Babel
Post by: Bennett Prescott on January 09, 2011, 03:16:08 PM
Chad,

I'm with Ivan on this one, I don't know anyone who takes settings from one console and manually programs them one channel at a time into another console... maybe I just haven't met them.

I have done no offline coordination as of yet, but I also have only one measurement so far!
Title: Re: DSP Tower of Babel
Post by: John Roberts {JR} on January 09, 2011, 03:53:59 PM
Chad Johnson wrote on Sat, 08 January 2011 22:56

Bennett
Do you think there's any benefit to a similar approach to the EQ sections on digital consoles? I know it would be another project to undertake, but I always thought it would be interesting to know how different units would respond to different settings.

I will be able to contribute V7 shots of some DSP units. Are you coordinating different models in offline conversations? No sense taking the time if someone else is doing the same model.

-Chad


Yup, I discussed this also in my past rants about lack of a Q/Bandwidth standard.

There is eventual merit in characterizing the actual flavor of console equalizer Q/bandwidth, if and when we start using some quasi-standard interface between either generic control surfaces or generic back ends, or both, and want our EQ to work and sound the same as our favorite old iron.

A much less pressing problem than the babelicious crossover presets that are tangled up here and now.

JR
Title: Re: DSP Tower of Babel
Post by: Ivan Beaver on January 09, 2011, 05:14:16 PM
Bennett Prescott wrote on Sun, 09 January 2011 15:16

Chad,

I'm with Ivan on this one, I don't know anyone who takes settings from one console and manually programs them one channel at a time into another console... maybe I just haven't met them.

I have done no offline coordination as of yet, but I also have only one measurement so far!

And if you were so picky aobut that, then that would also be assuming that the sound systems were exactly alike-which they are not-from room to room.

Maybe similar with like cabinets-but still not the same.
Title: Re: DSP Tower of Babel
Post by: Bennett Prescott on January 09, 2011, 06:35:06 PM
I've got measurements from two different processors already, and thought everyone might find this interesting. This processor natively uses Q to define filter bandwidth. The green trace is what these settings "should" look like. The blue trace is what happens when you plug bandwidth numbers into a Q device, as the two definitions are inverse of each other all the filters become much wider, although apparently this processor doesn't let you have 11+ octave filters.

index.php/fa/34676/0/
Title: Re: DSP Tower of Babel
Post by: Bennett Prescott on January 11, 2011, 03:20:58 AM
One quick update, and I'm sorry for the confusion.

In Smaart 7 the non-16K FFT option is "MTW".
In Smaart 7 the delay finder is not called "auto small", just use the auto delay finder.
Title: Re: DSP Tower of Babel
Post by: Chad Johnson on January 12, 2011, 01:17:32 AM
Ivan Beaver wrote on Sun, 09 January 2011 14:14

Bennett Prescott wrote on Sun, 09 January 2011 15:16

Chad,

I'm with Ivan on this one, I don't know anyone who takes settings from one console and manually programs them one channel at a time into another console... maybe I just haven't met them.

I have done no offline coordination as of yet, but I also have only one measurement so far!

And if you were so picky aobut that, then that would also be assuming that the sound systems were exactly alike-which they are not-from room to room.

Maybe similar with like cabinets-but still not the same.


Not picky at all. Of course console EQ is for artistic elements and not for a reference room tune or speaker processing. I understand the reasoning in DSP standardization and have talked about it here myself.  

I was simply mentioning I myself have been curious in the past how different digital consoles come up with their EQ algorithms. I'm most experienced with Yamaha digital console EQ and compared to a Profile, for instance, I find there is a difference in how I use them. It might have more to do with looking at the displays and mixing with eyes more than ears.

Title: Re: DSP Tower of Babel
Post by: John Roberts {JR} on January 12, 2011, 12:02:35 PM
Chad Johnson wrote on Wed, 12 January 2011 00:17



Not picky at all. Of course console EQ is for artistic elements and not for a reference room tune or speaker processing. I understand the reasoning in DSP standardization and have talked about it here myself.  

I was simply mentioning I myself have been curious in the past how different digital consoles come up with their EQ algorithms. I'm most experienced with Yamaha digital console EQ and compared to a Profile, for instance, I find there is a difference in how I use them. It might have more to do with looking at the displays and mixing with eyes more than ears.



While I don't have first hand experience (witrh digital consoles), I expect the EQ in better digital consoles will be voiced to mimic some successful previous console. In serious consoles this voicing goes beyond just the Q/bandwidth, but also involves control law (speed and range).

The company I was working for once brought in a well known record producer for me to work with when voicing a recording console EQ.

JR