Ok, I did not say that right I have seen conversion charts, as settings are different on different dsp platforms...thanks in advance, randy
Looking for the chart that converts XILICA or others to others
- Thread starter Randy Frierson
- Start date
Silas Pradetto
Graduate Student
Re: Looking for the chart that converts XILICA or others to others
Probably easier just to get the transfer function of one DSP and copy to the other. Yes, it requires you have both DSPs and Smaart (or similar), but I bet not all the differences between DSPs are linear and therefore can't be represented properly in a conversion sheet.
Probably easier just to get the transfer function of one DSP and copy to the other. Yes, it requires you have both DSPs and Smaart (or similar), but I bet not all the differences between DSPs are linear and therefore can't be represented properly in a conversion sheet.
Re: Looking for the chart that converts XILICA or others to others
There is no "standard" between different DSP's. I have seen all kinds of "differences", mainly the Q of the filter. But have also seen freq and level differences.
I have also "heard" that on at least one manufacturer-the same settings cannot be "copied" to different models.
Of course just "copying settings" will get you closer than nothing. But may not be "right". That is where the actual measurement comes in.
That is not an easy thing to "make". As silas says-the bet way is to get a transfer function and then "trace" it on whatever platform you want.
There is no "standard" between different DSP's. I have seen all kinds of "differences", mainly the Q of the filter. But have also seen freq and level differences.
I have also "heard" that on at least one manufacturer-the same settings cannot be "copied" to different models.
Of course just "copying settings" will get you closer than nothing. But may not be "right". That is where the actual measurement comes in.
Re: Looking for the chart that converts XILICA or others to others
Here are -6 dB cuts at 1 kHz and 15 kHz as measured in 14 different DSPs (yes, the Xilica is amongst them). Both cuts are 0.48 BW / 3.0 Q. The Q conversion formula is what I found to be the most commonly used. I'd say a simple chart is darn near impossible. Creating one would certainly be WAY more time consuming that simply setting down and making measurements of a reference and target processor.
Here are -6 dB cuts at 1 kHz and 15 kHz as measured in 14 different DSPs (yes, the Xilica is amongst them). Both cuts are 0.48 BW / 3.0 Q. The Q conversion formula is what I found to be the most commonly used. I'd say a simple chart is darn near impossible. Creating one would certainly be WAY more time consuming that simply setting down and making measurements of a reference and target processor.
Re: Looking for the chart that converts XILICA or others to others
Randy,
Here's my article on the problems here: http://soundforums.net/hub/240-tale-two-processors.html
And my slides from a later presentation on the same topic, with much measurement support from Rich: http://www.bennettprescott.com/downloads/DSP_Differences.pdf
Randy,
Here's my article on the problems here: http://soundforums.net/hub/240-tale-two-processors.html
And my slides from a later presentation on the same topic, with much measurement support from Rich: http://www.bennettprescott.com/downloads/DSP_Differences.pdf
Re: Looking for the chart that converts XILICA or others to others
I approached the AES standards body several years ago about this. They agreed there was a discrepancy or lack of a concise single definition for bandwidth/Q for boost/cut type EQ, but took little action toward resolving it. I suspect that the industry needs to do the work to resolve this by championing one or more standards then maybe the AES will help select between one winner or a small group of options. There can't be that many different ways to define bandwidth for that one type of EQ, but apparently there is more than one. ( RANE named their specific BW approach for GEQ years ago).
There are also a handful of other errors associated with DSP platforms (one visible in the difference between the 1kHz and 15kHz cut filter upper skirt). Bennet discusses these and more in his articles.
I suspect many of the errors in practice are probably due to sloppy control interface code, that they could get away with because there is no formal standard to be measured against, so the marketplace does not punish this sloppiness. Center frequency and amount of boost/cut seems pretty clearly understood. This is embarrassing but not new. Analog 1/3 octave GEQ products are equally dissimilar for actual boost/cut bandwidths. Looking at the data Rich provided the CF and depth of cuts looks more accurate now than I recall from even several years ago.
Before I gave up pursuing this I found there are a number of industry people aware of the issue and interested in resolving it. I am not active in this product area so have no leverage or vehicle to promote a standard with. Bennet's presentation does a good job of describing the issues.
Questions as I see it are
A) how many different types of boost/cut EQ are there, and how many are we willing to support?
B) What are the other important characteristic DSP platform behaviors that we need to account for in standard curve specifications?
The ultimate goal is to come up with a concise definition of a transfer function that is repeatable. Back in the very old days, they just used a series of pole and zero time-constants (like NAB or RIAA equalization curves), but this only accommodates simple low order curves.
Good luck guys.
JR
I approached the AES standards body several years ago about this. They agreed there was a discrepancy or lack of a concise single definition for bandwidth/Q for boost/cut type EQ, but took little action toward resolving it. I suspect that the industry needs to do the work to resolve this by championing one or more standards then maybe the AES will help select between one winner or a small group of options. There can't be that many different ways to define bandwidth for that one type of EQ, but apparently there is more than one. ( RANE named their specific BW approach for GEQ years ago).
There are also a handful of other errors associated with DSP platforms (one visible in the difference between the 1kHz and 15kHz cut filter upper skirt). Bennet discusses these and more in his articles.
I suspect many of the errors in practice are probably due to sloppy control interface code, that they could get away with because there is no formal standard to be measured against, so the marketplace does not punish this sloppiness. Center frequency and amount of boost/cut seems pretty clearly understood. This is embarrassing but not new. Analog 1/3 octave GEQ products are equally dissimilar for actual boost/cut bandwidths. Looking at the data Rich provided the CF and depth of cuts looks more accurate now than I recall from even several years ago.
Before I gave up pursuing this I found there are a number of industry people aware of the issue and interested in resolving it. I am not active in this product area so have no leverage or vehicle to promote a standard with. Bennet's presentation does a good job of describing the issues.
Questions as I see it are
A) how many different types of boost/cut EQ are there, and how many are we willing to support?
B) What are the other important characteristic DSP platform behaviors that we need to account for in standard curve specifications?
The ultimate goal is to come up with a concise definition of a transfer function that is repeatable. Back in the very old days, they just used a series of pole and zero time-constants (like NAB or RIAA equalization curves), but this only accommodates simple low order curves.
Good luck guys.
JR
Re: Looking for the chart that converts XILICA or others to others
I wonder if there would be resistance to bandwidth definition by some manufacturers because their filter definitions give a particular "sound" to their particular product, errors and all. In theory, if everyone's filter parameters matched, they should sound identical.
I wonder if there would be resistance to bandwidth definition by some manufacturers because their filter definitions give a particular "sound" to their particular product, errors and all. In theory, if everyone's filter parameters matched, they should sound identical.
Re: Looking for the chart that converts XILICA or others to others
Not if your goal is to promote your DSP as the "best one" to achieve the proper results with "XYZ" speaker, or if you only offer "grey box" settings for one DSP platform.
Re: Looking for the chart that converts XILICA or others to others
I took a powerful DSP and matched the responses to within less than 0.5dB. There was a HUGE difference in the overall "sound" of the cabinets-nothing subtle there.
I then removed all the processing and "the sound" difference was still there-although the responses were very different now.
There is a lot more to "the sound" than a simple transfer function.
I just reread your post and realized you were probably talking about the differences in the different DSP's alone.
I disagree. I did a "test" a number of years ago between 2 similar "speced" products-yet very different designs.
I took a powerful DSP and matched the responses to within less than 0.5dB. There was a HUGE difference in the overall "sound" of the cabinets-nothing subtle there.
I then removed all the processing and "the sound" difference was still there-although the responses were very different now.
There is a lot more to "the sound" than a simple transfer function.
I just reread your post and realized you were probably talking about the differences in the different DSP's alone.
Re: Looking for the chart that converts XILICA or others to others
I think you may be giving the manufacturers too much credit. If anything they do enjoy some benefit from the babel since they offer a curve preset that works accurately with "their" brand DSP platform. Using somebody else's DSP platform comes with translation error risk. It is fair to ASSume that accurate speaker equalization should sound better than inaccurate equalization.
More so in the early days than now, but DSP was a little like black magic to mostly analog companies so algorithm design and BW definitions often hinged on some outside consultant or resident in house expert's inclination, or how he learned to do it.
Establishing standards will create instant winners and losers, so ideally this would need to be worked out between those winners and losers to minimize economic pain. Not as binary as the pin 2/pin 3 hot standard reached a wile back, but it will be painful for some, so not trivial. I was working for a company on the losing side of the pin 2 hot standard, so no completely painless way to pick only one.
FWIW long before that XLR pin standard I argued for clear publication of what XLR polarity standard a given product used. The problem is not so much the different approaches, but not knowing who is using what approach (and how they differ) so we can translate between them.
JR
I think you may be giving the manufacturers too much credit. If anything they do enjoy some benefit from the babel since they offer a curve preset that works accurately with "their" brand DSP platform. Using somebody else's DSP platform comes with translation error risk. It is fair to ASSume that accurate speaker equalization should sound better than inaccurate equalization.
More so in the early days than now, but DSP was a little like black magic to mostly analog companies so algorithm design and BW definitions often hinged on some outside consultant or resident in house expert's inclination, or how he learned to do it.
Establishing standards will create instant winners and losers, so ideally this would need to be worked out between those winners and losers to minimize economic pain. Not as binary as the pin 2/pin 3 hot standard reached a wile back, but it will be painful for some, so not trivial. I was working for a company on the losing side of the pin 2 hot standard, so no completely painless way to pick only one.
FWIW long before that XLR pin standard I argued for clear publication of what XLR polarity standard a given product used. The problem is not so much the different approaches, but not knowing who is using what approach (and how they differ) so we can translate between them.
JR
Re: Looking for the chart that converts XILICA or others to others
Correct, just DSPs. By definition if the filters match from model to model, so should the transfer function and therefore the sound of the filters when used with the same source and speaker system.
Correct, just DSPs. By definition if the filters match from model to model, so should the transfer function and therefore the sound of the filters when used with the same source and speaker system.
Craig Hauber
Sophomore
Re: Looking for the chart that converts XILICA or others to others
The way I see it, that's not in the interest of any manufacturer. The shareholders will frown-upon providing customers with any more reasons to possibly buy the competition's product.
Now they can obfuscate the whole competitive field by championing their particular "sound" as "better/more accurate/warmer/cleaner/more analog-ish" than the competition's product.
It's bad enough that most of the DSP chips are all the same, now if the software was identical as well as the transfer function then basically the difference would be front-panel cosmetics and price. Thus reducing the pro audio DSP market to a commodity-state similar to the desktop PC market.
Now that I think about it, this may not be a bad thing for us users.
~~:smile:
The way I see it, that's not in the interest of any manufacturer. The shareholders will frown-upon providing customers with any more reasons to possibly buy the competition's product.
Now they can obfuscate the whole competitive field by championing their particular "sound" as "better/more accurate/warmer/cleaner/more analog-ish" than the competition's product.
It's bad enough that most of the DSP chips are all the same, now if the software was identical as well as the transfer function then basically the difference would be front-panel cosmetics and price. Thus reducing the pro audio DSP market to a commodity-state similar to the desktop PC market.
Now that I think about it, this may not be a bad thing for us users.
~~:smile:Re: Looking for the chart that converts XILICA or others to others
Randy,
Ask your design team guys behind the toursub, they should be able to match different processors for you with their measurement platform(s) without undue fuss. So should most of your senior system techs.
Randy,
Ask your design team guys behind the toursub, they should be able to match different processors for you with their measurement platform(s) without undue fuss. So should most of your senior system techs.
Mark DeArman
Junior
Re: Looking for the chart that converts XILICA or others to others
Xilica uses a different constant BW method to compute their EQs than is most commonly implemented in commercial off the shelf DSPs. There is no direct conversion chart because there is no direct conversion. You need to make new settings.
Xilica uses a different constant BW method to compute their EQs than is most commonly implemented in commercial off the shelf DSPs. There is no direct conversion chart because there is no direct conversion. You need to make new settings.
Re: Looking for the chart that converts XILICA or others to others
That sounds like a "you can't get there from here". If the old school time-constant pole and zero approach does not cover higher order transfer functions, this begs the question what kind of speaker errors do we need to correct with EQ? My guess would be low order, but I am not the speaker expert here. (I'm just asking about simple boost/cut EQ, where this BW discrepancy exists, not the other sundry DSP tricks).
Perhaps a variant on the old school approach could be used that supports more complex curves. Of course this isn't an easy short term fix, but a more literal description of the transfer function could be more complete and less subject to interpretation. The shorthand descriptions only work when the parameters mean the same thing every time used.
I am not smart enough to know if this different BW definition is a useful tool other DSP should support, or just different.
JR
PS: I really like to see engineered presets from the speaker professionals used whenever possible. Speakers should be a black box to end users, even if using multiple boxes to accomplish that.
That sounds like a "you can't get there from here". If the old school time-constant pole and zero approach does not cover higher order transfer functions, this begs the question what kind of speaker errors do we need to correct with EQ? My guess would be low order, but I am not the speaker expert here. (I'm just asking about simple boost/cut EQ, where this BW discrepancy exists, not the other sundry DSP tricks).
Perhaps a variant on the old school approach could be used that supports more complex curves. Of course this isn't an easy short term fix, but a more literal description of the transfer function could be more complete and less subject to interpretation. The shorthand descriptions only work when the parameters mean the same thing every time used.
I am not smart enough to know if this different BW definition is a useful tool other DSP should support, or just different.
JR
PS: I really like to see engineered presets from the speaker professionals used whenever possible. Speakers should be a black box to end users, even if using multiple boxes to accomplish that.
Mark DeArman
Junior
Re: Looking for the chart that converts XILICA or others to others
It's about the shape in BW vs. Cut that is different. There is no analog between the two shapes. The structure, poles-zeros, is the same, but the algorithm used to compute their location is different. There is no help, but to re-EQ the system with measurements. In fact, many older DSPs, and stand alone EQ units, have major differences in their shape. Many digital systems systems copied their code from:
http://www.musicdsp.org/files/Audio-EQ-Cookbook.txt
If you want to take a look.
I'm really not sure what you mean by this. IE. Old approach, complex curves??? Yes I agree with you. For our DSP, (Loudspeaker Processor,) I provide an addendum which has definitions of the filters, available to customers. But like I said above, the parametric inputs to these filter calculations vary, and unless copied from some standard source, don't mean the same thing, unless it is a predefined shape like an LR-24 or something.
Xilica would argue, and I would agree, that their PEQ shape (constant BW) is better than the standard you would find on say an older Ashly or such. I do agree, that I wish the parameters would mean the same thing, but that just won't happen as long as innovation continues in the DSP industry. I don't expect my settings from my Ashly to drop onto my LM26 ?
Anyway, regarding your PS, YES, manufactures who do not provide settings for their products (and I mean complete systems) on their recommended or own DSP solutions are crazy or stupid. Most likely end users purchased the loudspeaker system based on hearing those settings, and should use them as a baseline for their own.
Mark.
It's about the shape in BW vs. Cut that is different. There is no analog between the two shapes. The structure, poles-zeros, is the same, but the algorithm used to compute their location is different. There is no help, but to re-EQ the system with measurements. In fact, many older DSPs, and stand alone EQ units, have major differences in their shape. Many digital systems systems copied their code from:
http://www.musicdsp.org/files/Audio-EQ-Cookbook.txt
If you want to take a look.
I'm really not sure what you mean by this. IE. Old approach, complex curves??? Yes I agree with you. For our DSP, (Loudspeaker Processor,) I provide an addendum which has definitions of the filters, available to customers. But like I said above, the parametric inputs to these filter calculations vary, and unless copied from some standard source, don't mean the same thing, unless it is a predefined shape like an LR-24 or something.
Xilica would argue, and I would agree, that their PEQ shape (constant BW) is better than the standard you would find on say an older Ashly or such. I do agree, that I wish the parameters would mean the same thing, but that just won't happen as long as innovation continues in the DSP industry. I don't expect my settings from my Ashly to drop onto my LM26 ?
Anyway, regarding your PS, YES, manufactures who do not provide settings for their products (and I mean complete systems) on their recommended or own DSP solutions are crazy or stupid. Most likely end users purchased the loudspeaker system based on hearing those settings, and should use them as a baseline for their own.
Mark.
Last edited:
Re: Looking for the chart that converts XILICA or others to others
This is even consistent with what we see.. the CF and boost/cut is correct, just the shape of the EQ varies.
To repeat my question, if EQ in the context of speaker management is trying to correct for a single driver's frequency response, this should (?) be simple. If trying to correct for the summed response of two drivers, no longer so simple.
Getting really old school, I've seen RIAA or NAB EQ specifications provided literally as a full table of X dB at Y frequency, with accompanying pole and zero time constants.
How many possible variants are there for boost/cut filters using the same CF and boost/cut? My premise (hope) is there are only a handful but perhaps not. If not, there needs to be some more literal description of transfer function short of the X/Y plot to accurately communicate this transfer function.
This is not the first time the industry has quibbled over standards that will create immediate winners and losers. I see benefits from defining these EQ beyond just speaker management, and it is a little embarrassing to think of this as still undefined concisely. It is the repeatability and portability "promised" by digital audio technology, that highlights this glaring discrepancy.
There has to be a better way.
JR
I'm not sure if we are on the same page or not. I am more familiar with analog design than DSP. In analog all the variant boost/cut EQ BWs are ultimately derived from a recognizable BP filter, but that bandpass gets added or subtracted from the dry signal differently resulting in the different shapes. So the poles could be strictly defined and still not completely define the resulting transfer function.
This is even consistent with what we see.. the CF and boost/cut is correct, just the shape of the EQ varies.
To repeat my question, if EQ in the context of speaker management is trying to correct for a single driver's frequency response, this should (?) be simple. If trying to correct for the summed response of two drivers, no longer so simple.
math is not my native language, but I had some general discussions with at least one DSP designer years ago when I first looked into this.
Yes, LR-24 and entire universe of common HP/BP/LP filter alignments are well defined wrt what BW/Q means. Only the basic boost/cut EQ section suffers from lack of clear concise definition. Note: boost/cut EQ are used in other places than speaker management, so this is a larger problem mostly ignored.
Getting really old school, I've seen RIAA or NAB EQ specifications provided literally as a full table of X dB at Y frequency, with accompanying pole and zero time constants.
That is the crux of the matter. I have designed several different parametric EQ over the decades, all different using subtly different topologies. One even had an intentional interaction between Q and boost/cut designed in, so it got narrower for large amounts of boost/cut, and broader for small adjustments (as a first order correction for apparent loudness change in a hifi application). I didn't even try to put BW numbers on the Q control with that one.
How many possible variants are there for boost/cut filters using the same CF and boost/cut? My premise (hope) is there are only a handful but perhaps not. If not, there needs to be some more literal description of transfer function short of the X/Y plot to accurately communicate this transfer function.
This is not the first time the industry has quibbled over standards that will create immediate winners and losers. I see benefits from defining these EQ beyond just speaker management, and it is a little embarrassing to think of this as still undefined concisely. It is the repeatability and portability "promised" by digital audio technology, that highlights this glaring discrepancy.
There has to be a better way.
JR
Re: Looking for the chart that converts XILICA or others to others
After all, the transfer function is the only way to know...
I would think along those lines as well. Same with desk EQ, with the same parameters all desk EQs should sound the same.
After all, the transfer function is the only way to know...
Re: Looking for the chart that converts XILICA or others to others
In fact console EQ do not sound the same and that is part of their real (not imaginary) sound differences. Different EQ topologies exhibit different BW vs boost/cut making for dramatically different sounding EQ. Same for analog consoles.
Loudspeaker corrective EQ "should" be clearly defined and easy to replicate across multiple platforms.
Console EQ should not sound alike, but at some point we need to get our definitions fleshed out enough so universal or generic control interfaces can give repeatable results when they tell some console engine to "do exactly this" or "do exactly that'.
It's 2013 and we still lack adequate vocabulary to communicate EQ concisely. arghhhhhhhh
JR
In fact console EQ do not sound the same and that is part of their real (not imaginary) sound differences. Different EQ topologies exhibit different BW vs boost/cut making for dramatically different sounding EQ. Same for analog consoles.
Loudspeaker corrective EQ "should" be clearly defined and easy to replicate across multiple platforms.
Console EQ should not sound alike, but at some point we need to get our definitions fleshed out enough so universal or generic control interfaces can give repeatable results when they tell some console engine to "do exactly this" or "do exactly that'.
It's 2013 and we still lack adequate vocabulary to communicate EQ concisely. arghhhhhhhh
JR
Re: Looking for the chart that converts XILICA or others to others
http://soundforums.net/varsity/4972-q-lake-bandwidth-converter.html
http://soundforums.net/varsity/4972-q-lake-bandwidth-converter.html