Filter "warping" nearing the Nyquist Frequency

Chester01 · Mar 16, 2009, 03:31 AM

Hey, I was wondering if the dsp filters within the kX project account for the non-symmetrical "warping" that occurs as the frequency being filtered approaches the Nyquist limit. I have recently learned of this effect/issue and am considering figuring out a "correction factor" or a way to make a second "compensating" EQ to make things more accurate.

For those who have not heard of the "warping" effect, at the graph on this page: <http://www.sonoris.nl/catalog/equalizer-p-31.html> about 3/5 the way down. There is also some information in this PDF: http://www.musicdsp.org/files/EQ-Coefficients.pdf

I can try and run some tests to figure out if the parametric filters suffer from this or not however I figure someone more versed in DSP coding could better shed some light on the subject. Thanks!

Max M. · Mar 16, 2009, 03:53 AM

well, in short:
The ufx EQs have "symmetrical" response (see http://www.ece.rutgers.edu/~orfanidi/ece521/peq.pdf, http://web.media.mit.edu/~tristan/Pa...MaxFilters.pdf)
The other EQs have not (they use the approximation proposed by RBJ - e.g. "The Equivalence of Various Methods..." #3)

Chester01 · Mar 16, 2009, 04:26 AM

So the UFX EQ's use the "Orfanidis" calculation method? In the Electri-Q manual http://www.aixcoustic.com/uploads/media/Manual.pdf (this is where I first learned of the symmetry issue) they show a comparison between RBJ, Orfanidis, and "S-plane"; the Orfanidis method is not completely symmetrical either... Not to say that this graph applies to the kX dsp, just checking

Max M. · Mar 16, 2009, 04:37 AM

>So the UFX EQ's use the "Orfanidis" calculation method?

Nope, i was not able to get Orfanidis formulas to give the expected response too, so i had to reinvent the wheel on my own.

Chester01 · Mar 16, 2009, 04:53 AM

Ok, excellent

so, basically, the UFX filters are a "perfect" transfer function, or as close to a "simulated"/optimal response as can be achieved?

My next question then; I have not been using the UFX filters (no offence

) because I cannot enter in my coefficients in as precisely (frequency and bandwidth) as with the "peak" filters. Is there a way I can "key" in parameters to get frequencies and bandwidths more precise? I suppose I could round however as someone who "reinvented the wheel" to create a more precise response, you can probably appreciate my desire to enter in exact values

BTW, is the "Max/MSP" filters paper your work? If so, awesome job!

EDIT/NOTE: I already know that I can click on a parameter in the UFX EQs and use the keyboard to be a bit more precise than with the mouse however the BW section seems to make some "large" jumps... Also I could use some larger BW's too... seems crazy doesnt it

I do not mean to demean or make light of your EQ, it is very well done.

Max M. · Mar 16, 2009, 05:05 AM

>so, basically, the UFX filters are a "perfect" transfer function, or as close to a "simulated"/optimal response as can be achieved?

yep. sorta. (see comments below though - that's not so deadly-important)
- that's just what was to my taste.

>BTW, is the "Max/MSP" filters paper your work?

No, it's not (sorry to disappoint

)

>because I cannot enter in my coefficients in as precisely

Well, there're some design compromises behind that. Either way i don't believe in such kind of "precision", i believe in tuning the EQ by ears and not "by eyes and by pleasant looking numbers" - and the ufx EQs are precise enough to my taste (the key is that personally i don't believe in differense between (for example) 100 and 101.2 Hz (it's only 1/5 of semitone delta) - so is there any reason you need a peak exactly at 100Hz and not just 1/60 octave higher?)
Anyway this is not something i did intentionaly - but it's OK to me (all of other EQs i used to use do the same)

>I do not mean to demean or make light of your EQ, it is very well done.

it's OK if you do - guess who can know its shortcomings better

btw. i think i need to express some opinion for the "anti-warping feature". Despite the "analog style bla-bla-bla" booz-words this is not so deadly-important stuff as it might be advertised by some EQ manuals sometimes - that's just a sort of "little little" bonus. One should take into acount that in the digital world the signal itself is already srewed-up near the Nyquist-freq. ("Fn") so you don't get the "analog" kind of EQ response no matter what type of the curve the EQ offers (and normally you *don't* equalizing near the Fn - it's wrong if you do, more over you don't really care of the level of the curve at the Fn simply because there's *no* signal there).
So again, *to me*, the "symmetrical" curve is just a little bonus - it does not render the "non-symmetrical" curve to be "worse" - the *sound* difference between these two types of curves is *very* subtle (guess why the graphs are important in papers and manuals

). One is better to consider both types of the curve as a "style" of an EQ (e.g. a sort-of "the color of EQ")

Chester01 · Mar 16, 2009, 06:05 AM

Well... I SUPPOSE I can "put up with" 1/5 of a semi-tone... since you put it that way

In regards to what you mentioned about the response being messed up near the Nyquist frequency, is that being depicted in Figure 3 (on page 7) the EQ-Coefficients pdf I mentioned earlier?

Basically what I take this all to mean is that I definitely should be using your UFX filter for filters above 48000*.2=9600 hz, where the accuracy of RBJ starts to become "more non-linear" and there will be more asymmetry... (I think I have figured this stuff out a bit more!

Max M. · Mar 16, 2009, 06:18 AM

>In regards to what you mentioned about the response being messed up near the Nyquist frequency,

No not really, i was about the signal itself is being messed near the Fn - so that the precise "symmetry" in the curve becomes not-so important. The Fig.3 shows completely "raw" (uncompensated) response one would get if using initial formulas. But they use the eq.(15) bandwidth compensation/approximation - and the Fig.4 shows what you finally get (you can see the assymetry in the right bell - it's not so dramatical)

Chester01 · Mar 16, 2009, 06:26 AM

Oh, that shows the "amplitude" error as you approach the Nyquist frequency? :/

Max M. · Mar 16, 2009, 06:30 AM

mm, sorry (i'm not completely awaken yet

) - "that" - which one?

Chester01 · Mar 16, 2009, 06:33 AM

Ah my bad; "that" was supposed to be Fig 3... duh

Max M. · Mar 16, 2009, 06:40 AM

ah, the Fig3. shows the bandwidth error when the freq. approaches the Fn
(it becomes narrower and narrower). But this is the error of the uncompensated formulas - no EQ uses such. Instead they use a compensated bandwidth calculation (the one in the Eq.15) - which gives quite fair bandwidth (but still non-symmetrical))

Chester01 · Mar 16, 2009, 06:47 AM

oh ok, yeah I was referring to something like this: 44 KHz (CD) not enough !? (Nyquist etc.) - Hydrogenaudio Forums
but with using no EQ filter... I get what you mean now how its showing the uncompensated bandwidth. Anyways; I am using your UFX plugin now at least for stuff over 5khz

the highs sound much more symmetrical

Thanks for all of your help and patience explaining this stuff.

Max M. · Mar 16, 2009, 06:58 AM

>the highs sound much more symmetrical

hehe, cool said

btw. just a thought - i guess you realize that this applies only to "peaking" (aka "bell") type of EQ - the shelving and pass filters do not have such problem (as they are non-symmetrical by definition) - so that i wonder a bit what would be the use of the peaking type of EQ at *higher* frequencies (where the problem becomes noticable) - actually i would expect the high-shelf and/or low-pass filters to be likely used there.

(although, hah, the hocus-pocus is that the "symmetrical" peaking curve does transform into the high-shelving curve when approaching Fn

)

Chester01 · Mar 16, 2009, 07:15 AM

Thats a good point; should the shelving filters use the same bw values as the bells (in UFX) to get the same response below the frequency it is set to?

EDIT/NOTE: to me it *SEEMS* that I need to multiply the BW by 2x to get the same general "shape" on the graph (below the frequency I am tuning at)

Max M. · Mar 16, 2009, 07:22 AM

mmm, i guess the "BW" (if < 1) value will match (less or more) - but they have different definition of the "frequency" parameter value. So it depends on the response needed. Better to see possible variants yourself (the curve shown in the ufx EQs is a "wysiwyg" thing)

edit: aha, well, tried myself - it hardly depends on the frequency (it's less or more similiar for 8..12kHz range but very different at higher freqs)

Chester01 · Mar 16, 2009, 07:32 AM

ok; yeah ill play around a bit with it... is there any way that a shelving filter can be represented by a linkwitz transform? I usually use WinISD Pro to model my EQ's to see whats going on

Max M. · Mar 16, 2009, 07:50 AM

i'm afraid, no, there's not - only visually - (you know... it's a musical EQ not a speaker-works filter

(for instance, in the EQP5, the BW value for the peaking type is defined in octaves instead of that laboratory "Q" you would rather find more consistent))

Chester01 · Mar 16, 2009, 07:59 AM

I got ya, sounds good

Thanks again!

Chester01 · Mar 17, 2009, 06:26 PM

just wanted to say; after playing around and getting my Q's converted to your bandwidths, (and gains) your filters definitely seem superior... I basically reverse engineered the values that your UFX uses and found the "best fit" curve for what I initially designed in WinISD. The UFX filters have a surprisingly low noise floor when applying a large amount of bass... the "peaking EQ" filters seemed still produce signal after playing music and stopping it where yours still do so (atleast it registers on the Peak plugin) but the level is MUCH lower

NOTE: when i say adding a large amount of bass I mean +27 db @24.4 hz; I used to be using 20 hz but I found a pretty good conversion using a few parametrics, I just cannot get the bandwidth as "wide" as i need so I also have 2 "correction factor" eq's at 60 and 130 hz too

doing that with the original eq's (the gain, they could take the needed Q factor) always caused weird bass effects (on the peak meeter, could not hear em) when the music stopped...

Why do I need 27db @24.4? I model the Phons scale at 80 phons (ISO226:2003 chart), works much better than a "loudness" button on old receivers

Thanks again for all your help - great plugin (now that I get used to it)

Max M. · Mar 17, 2009, 07:01 PM

Aha, this is a DC-offset actually (not really a noise) - known problem
(so called "limit cycles" problem) - at very low frequencies filter becomes too "hard" so that it starts to "generate" the DC under certain conditions.

Chester01 · Mar 17, 2009, 07:26 PM

hmm good to know... the "limit cycles" problem common in most eq equipment or is it something that is more a kX issue? (i'd assume it is a problem everywhere however I figure its best to ask)

Max M. · Mar 17, 2009, 08:26 PM

It's a common problem of recursive digital filters (but depending on many factors - bit-depth, frequency etc.)

Chester01 · Mar 18, 2009, 01:31 AM

so the ufx filters do not suffer from the limit cycle thing either then? or just not as badly? I found on this page: AbVolt - DSP - Embedded - Consulting it mentions that dithering the input can stop limit cycling from happening (or reduce it).

Max M. · Mar 18, 2009, 06:23 AM

>not as badly?

yes, not as badly (it has a special code to reduce it)
Yes, sure, there're several techniques to eliminate the DC (and dithering is not the best) - but either one would eat additional DSP resources - so considering a relative unsignificance of the problem...

Chester01 · Mar 18, 2009, 07:27 AM

ah yeah; I am not too worried about it, not a big deal; I did notice on the last site I mentioned they talked about how the ordering of the eq's can effect things too... which got me thinking (go figure) about 1, which "order" are the eq's applied in EQ-P5, and 2, what is the best way to run things... ex, wide or narrow bandwidth first, high or low frequency first, minimal gain or larger gain first... simply put; any tips?

the paper mentions it being helpful for various filter types however I did not know if there were any other considerations. As always, thanks for your advice

Max M. · Mar 18, 2009, 08:11 AM

EQP5 applies bands "left-to-right".
The better order would be:
* Wider bandwidth first
* Higher freqs first
* Higher gain first (but, taking into account possible clipping, it's better to interleave higher and lower gain bands)
* And for filter types: shelving -> passing -> peaking

edit: but, of course, it's all would depend on a particular setup - as there're contradictory/opposite matters/factors as always.

Chester01 · Mar 18, 2009, 08:26 AM

ok; should that be read top down also (as in wider bw 'trumps' large gain) or should I just try to balance those things?

Max M. · Mar 18, 2009, 08:35 AM

yeah, probably top-down rule a sort-of applicable - but in the world of balances we never ignore those, do we?

btw, It would be interesting to take a look at some exemplary setup.

Chester01 · Mar 18, 2009, 08:45 AM

do you mean balances as in alternating + and - gain or am i missing something?

Do you have any suggestions to check out (exemplary setups)

EDIT/update: I ended up breaking out excel and made a "ranking system"; I used the following formula:

log(frequency)*Bandwidth*abs(gain)... this seems to create a pretty decent ranking system, I am considering using log(freq)*bandwidth*log(abs(gain)) :/ lol

Mar 16, 2009, 03:31 AM	#1
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Filter "warping" nearing the Nyquist Frequency Hey, I was wondering if the dsp filters within the kX project account for the non-symmetrical "warping" that occurs as the frequency being filtered approaches the Nyquist limit. I have recently learned of this effect/issue and am considering figuring out a "correction factor" or a way to make a second "compensating" EQ to make things more accurate. For those who have not heard of the "warping" effect, at the graph on this page: <http://www.sonoris.nl/catalog/equalizer-p-31.html> about 3/5 the way down. There is also some information in this PDF: http://www.musicdsp.org/files/EQ-Coefficients.pdf I can try and run some tests to figure out if the parametric filters suffer from this or not however I figure someone more versed in DSP coding could better shed some light on the subject. Thanks! __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands...

Mar 16, 2009, 03:53 AM	#2
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency well, in short: The ufx EQs have "symmetrical" response (see http://www.ece.rutgers.edu/~orfanidi/ece521/peq.pdf, http://web.media.mit.edu/~tristan/Pa...MaxFilters.pdf) The other EQs have not (they use the approximation proposed by RBJ - e.g. "The Equivalence of Various Methods..." #3) __________________ Last edited by Max M.; Mar 16, 2009 at 04:20 AM.

Mar 16, 2009, 04:26 AM	Thread Starter #3
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency So the UFX EQ's use the "Orfanidis" calculation method? In the Electri-Q manual http://www.aixcoustic.com/uploads/media/Manual.pdf (this is where I first learned of the symmetry issue) they show a comparison between RBJ, Orfanidis, and "S-plane"; the Orfanidis method is not completely symmetrical either... Not to say that this graph applies to the kX dsp, just checking __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands...

Mar 16, 2009, 04:37 AM	#4
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency >So the UFX EQ's use the "Orfanidis" calculation method? Nope, i was not able to get Orfanidis formulas to give the expected response too, so i had to reinvent the wheel on my own. __________________

Mar 16, 2009, 04:53 AM	Thread Starter #5
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency Ok, excellent so, basically, the UFX filters are a "perfect" transfer function, or as close to a "simulated"/optimal response as can be achieved? My next question then; I have not been using the UFX filters (no offence ) because I cannot enter in my coefficients in as precisely (frequency and bandwidth) as with the "peak" filters. Is there a way I can "key" in parameters to get frequencies and bandwidths more precise? I suppose I could round however as someone who "reinvented the wheel" to create a more precise response, you can probably appreciate my desire to enter in exact values BTW, is the "Max/MSP" filters paper your work? If so, awesome job! EDIT/NOTE: I already know that I can click on a parameter in the UFX EQs and use the keyboard to be a bit more precise than with the mouse however the BW section seems to make some "large" jumps... Also I could use some larger BW's too... seems crazy doesnt it I do not mean to demean or make light of your EQ, it is very well done. __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands... Last edited by Chester01; Mar 16, 2009 at 05:02 AM.

Mar 16, 2009, 05:05 AM	#6
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency >so, basically, the UFX filters are a "perfect" transfer function, or as close to a "simulated"/optimal response as can be achieved? yep. sorta. (see comments below though - that's not so deadly-important) - that's just what was to my taste. >BTW, is the "Max/MSP" filters paper your work? No, it's not (sorry to disappoint ) >because I cannot enter in my coefficients in as precisely Well, there're some design compromises behind that. Either way i don't believe in such kind of "precision", i believe in tuning the EQ by ears and not "by eyes and by pleasant looking numbers" - and the ufx EQs are precise enough to my taste (the key is that personally i don't believe in differense between (for example) 100 and 101.2 Hz (it's only 1/5 of semitone delta) - so is there any reason you need a peak exactly at 100Hz and not just 1/60 octave higher?) Anyway this is not something i did intentionaly - but it's OK to me (all of other EQs i used to use do the same) >I do not mean to demean or make light of your EQ, it is very well done. it's OK if you do - guess who can know its shortcomings better btw. i think i need to express some opinion for the "anti-warping feature". Despite the "analog style bla-bla-bla" booz-words this is not so deadly-important stuff as it might be advertised by some EQ manuals sometimes - that's just a sort of "little little" bonus. One should take into acount that in the digital world the signal itself is already srewed-up near the Nyquist-freq. ("Fn") so you don't get the "analog" kind of EQ response no matter what type of the curve the EQ offers (and normally you don't equalizing near the Fn - it's wrong if you do, more over you don't really care of the level of the curve at the Fn simply because there's no signal there). So again, to me, the "symmetrical" curve is just a little bonus - it does not render the "non-symmetrical" curve to be "worse" - the sound difference between these two types of curves is very subtle (guess why the graphs are important in papers and manuals ). One is better to consider both types of the curve as a "style" of an EQ (e.g. a sort-of "the color of EQ") __________________ Last edited by Max M.; Mar 16, 2009 at 06:38 PM. Reason: yak, sorry, too much typos

Mar 16, 2009, 06:05 AM	Thread Starter #7
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency Well... I SUPPOSE I can "put up with" 1/5 of a semi-tone... since you put it that way In regards to what you mentioned about the response being messed up near the Nyquist frequency, is that being depicted in Figure 3 (on page 7) the EQ-Coefficients pdf I mentioned earlier? Basically what I take this all to mean is that I definitely should be using your UFX filter for filters above 48000*.2=9600 hz, where the accuracy of RBJ starts to become "more non-linear" and there will be more asymmetry... (I think I have figured this stuff out a bit more! __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands...

Mar 16, 2009, 06:18 AM	#8
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency >In regards to what you mentioned about the response being messed up near the Nyquist frequency, No not really, i was about the signal itself is being messed near the Fn - so that the precise "symmetry" in the curve becomes not-so important. The Fig.3 shows completely "raw" (uncompensated) response one would get if using initial formulas. But they use the eq.(15) bandwidth compensation/approximation - and the Fig.4 shows what you finally get (you can see the assymetry in the right bell - it's not so dramatical) __________________ Last edited by Max M.; Mar 16, 2009 at 06:28 AM.

Mar 16, 2009, 06:26 AM	Thread Starter #9
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency Oh, that shows the "amplitude" error as you approach the Nyquist frequency? :/ __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands...

Mar 16, 2009, 06:30 AM	#10
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency mm, sorry (i'm not completely awaken yet ) - "that" - which one? __________________

Mar 16, 2009, 06:33 AM	Thread Starter #11
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency Ah my bad; "that" was supposed to be Fig 3... duh __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands...

Mar 16, 2009, 06:40 AM	#12
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency ah, the Fig3. shows the bandwidth error when the freq. approaches the Fn (it becomes narrower and narrower). But this is the error of the uncompensated formulas - no EQ uses such. Instead they use a compensated bandwidth calculation (the one in the Eq.15) - which gives quite fair bandwidth (but still non-symmetrical)) __________________

Mar 16, 2009, 06:47 AM	Thread Starter #13
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency oh ok, yeah I was referring to something like this: 44 KHz (CD) not enough !? (Nyquist etc.) - Hydrogenaudio Forums but with using no EQ filter... I get what you mean now how its showing the uncompensated bandwidth. Anyways; I am using your UFX plugin now at least for stuff over 5khz the highs sound much more symmetrical Thanks for all of your help and patience explaining this stuff. __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands...

Mar 16, 2009, 06:58 AM	#14
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency >the highs sound much more symmetrical hehe, cool said btw. just a thought - i guess you realize that this applies only to "peaking" (aka "bell") type of EQ - the shelving and pass filters do not have such problem (as they are non-symmetrical by definition) - so that i wonder a bit what would be the use of the peaking type of EQ at higher frequencies (where the problem becomes noticable) - actually i would expect the high-shelf and/or low-pass filters to be likely used there. (although, hah, the hocus-pocus is that the "symmetrical" peaking curve does transform into the high-shelving curve when approaching Fn ) __________________ Last edited by Max M.; Mar 16, 2009 at 07:17 AM.

Mar 16, 2009, 07:15 AM	Thread Starter #15
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency Thats a good point; should the shelving filters use the same bw values as the bells (in UFX) to get the same response below the frequency it is set to? EDIT/NOTE: to me it SEEMS that I need to multiply the BW by 2x to get the same general "shape" on the graph (below the frequency I am tuning at) __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands... Last edited by Chester01; Mar 16, 2009 at 07:21 AM.

Mar 16, 2009, 07:22 AM	#16
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency mmm, i guess the "BW" (if < 1) value will match (less or more) - but they have different definition of the "frequency" parameter value. So it depends on the response needed. Better to see possible variants yourself (the curve shown in the ufx EQs is a "wysiwyg" thing) edit: aha, well, tried myself - it hardly depends on the frequency (it's less or more similiar for 8..12kHz range but very different at higher freqs) __________________ Last edited by Max M.; Mar 16, 2009 at 07:32 AM.

Mar 16, 2009, 07:32 AM	Thread Starter #17
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency ok; yeah ill play around a bit with it... is there any way that a shelving filter can be represented by a linkwitz transform? I usually use WinISD Pro to model my EQ's to see whats going on __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands...

Mar 16, 2009, 07:50 AM	#18
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency i'm afraid, no, there's not - only visually - (you know... it's a musical EQ not a speaker-works filter (for instance, in the EQP5, the BW value for the peaking type is defined in octaves instead of that laboratory "Q" you would rather find more consistent)) __________________

Mar 16, 2009, 07:59 AM	Thread Starter #19
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency I got ya, sounds good Thanks again! __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands...

Mar 17, 2009, 06:26 PM	Thread Starter #20
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency just wanted to say; after playing around and getting my Q's converted to your bandwidths, (and gains) your filters definitely seem superior... I basically reverse engineered the values that your UFX uses and found the "best fit" curve for what I initially designed in WinISD. The UFX filters have a surprisingly low noise floor when applying a large amount of bass... the "peaking EQ" filters seemed still produce signal after playing music and stopping it where yours still do so (atleast it registers on the Peak plugin) but the level is MUCH lower NOTE: when i say adding a large amount of bass I mean +27 db @24.4 hz; I used to be using 20 hz but I found a pretty good conversion using a few parametrics, I just cannot get the bandwidth as "wide" as i need so I also have 2 "correction factor" eq's at 60 and 130 hz too doing that with the original eq's (the gain, they could take the needed Q factor) always caused weird bass effects (on the peak meeter, could not hear em) when the music stopped... Why do I need 27db @24.4? I model the Phons scale at 80 phons (ISO226:2003 chart), works much better than a "loudness" button on old receivers Thanks again for all your help - great plugin (now that I get used to it) __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands... Last edited by Chester01; Mar 17, 2009 at 06:31 PM.

Mar 17, 2009, 07:01 PM	#21
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency Aha, this is a DC-offset actually (not really a noise) - known problem (so called "limit cycles" problem) - at very low frequencies filter becomes too "hard" so that it starts to "generate" the DC under certain conditions. __________________

Mar 17, 2009, 07:26 PM	Thread Starter #22
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency hmm good to know... the "limit cycles" problem common in most eq equipment or is it something that is more a kX issue? (i'd assume it is a problem everywhere however I figure its best to ask) __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands...

Mar 17, 2009, 08:26 PM	#23
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency It's a common problem of recursive digital filters (but depending on many factors - bit-depth, frequency etc.) __________________

Mar 18, 2009, 01:31 AM	Thread Starter #24
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency so the ufx filters do not suffer from the limit cycle thing either then? or just not as badly? I found on this page: AbVolt - DSP - Embedded - Consulting it mentions that dithering the input can stop limit cycling from happening (or reduce it). __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands...

Mar 18, 2009, 06:23 AM	#25
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency >not as badly? yes, not as badly (it has a special code to reduce it) Yes, sure, there're several techniques to eliminate the DC (and dithering is not the best) - but either one would eat additional DSP resources - so considering a relative unsignificance of the problem... __________________

Mar 18, 2009, 07:27 AM	Thread Starter #26
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency ah yeah; I am not too worried about it, not a big deal; I did notice on the last site I mentioned they talked about how the ordering of the eq's can effect things too... which got me thinking (go figure) about 1, which "order" are the eq's applied in EQ-P5, and 2, what is the best way to run things... ex, wide or narrow bandwidth first, high or low frequency first, minimal gain or larger gain first... simply put; any tips? the paper mentions it being helpful for various filter types however I did not know if there were any other considerations. As always, thanks for your advice __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands...

Mar 18, 2009, 08:11 AM	#27
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency EQP5 applies bands "left-to-right". The better order would be: * Wider bandwidth first * Higher freqs first * Higher gain first (but, taking into account possible clipping, it's better to interleave higher and lower gain bands) * And for filter types: shelving -> passing -> peaking edit: but, of course, it's all would depend on a particular setup - as there're contradictory/opposite matters/factors as always. __________________ Last edited by Max M.; Mar 18, 2009 at 08:30 AM.

Mar 18, 2009, 08:26 AM	Thread Starter #28
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency ok; should that be read top down also (as in wider bw 'trumps' large gain) or should I just try to balance those things? __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands...

Mar 18, 2009, 08:35 AM	#29
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	Îòâåò: Filter "warping" nearing the Nyquist Frequency yeah, probably top-down rule a sort-of applicable - but in the world of balances we never ignore those, do we? btw, It would be interesting to take a look at some exemplary setup. __________________

Mar 18, 2009, 08:45 AM	Thread Starter #30
Chester01 HardwareHeaven Senior Member Join Date: Jan 2004 Location: St. Cloud, MN Posts: 492 Rep Power: 0	Re: Filter "warping" nearing the Nyquist Frequency do you mean balances as in alternating + and - gain or am i missing something? Do you have any suggestions to check out (exemplary setups) EDIT/update: I ended up breaking out excel and made a "ranking system"; I used the following formula: log(frequency)Bandwidthabs(gain)... this seems to create a pretty decent ranking system, I am considering using log(freq)bandwidthlog(abs(gain)) :/ lol __________________ COMP: Dual Intel PIII 733 Mhz; GeForce 4 Ti 4600; 1.128 Gb RAM; SB0350 (Audigy 2 ZS Platnium) STEREO(I UPGRADED):Crown Audio K1 and K2 amplifiers (4000 watts at .1% THD ), JL 13w7 Subwoofer (6.5 CF) (2) 18" PR's, Klipsch SB-1's, some cement blocks for speaker stands... Last edited by Chester01; Mar 18, 2009 at 11:05 AM.