all i did was hit it with a STATIC 11,025Hz sine and engaged a HPF @ 20Hz (24db/oct) and it started freaking out...there shouldn't be anything moving around in there...
i tried with 4 other EQs and none did anything like this...
and i think it's only this frequency...(or 5512.5Hz, etc...)
@ 44.1K
__________________
"it's a long way to the top if you wanna rock -n- roll" -Jesus
it looks like aliasing but everything else i tested with this tone just gets folded back on itself and that's it...(for the most part...)
this thing just blew my mind...
this seems to be the only tone that sets it off, too...(edit: i tried a 5512.5Hz and that acts similar...)
i was wondering if other EQ's would do the same thing...
it's not a regular digital EQ...
oh...the reading you see if from Span...i didn't want to show the EQ...
you can attach the file and i'll look if you want...
also...it doesn't look like you engaged the HPF...?
(i just put it in bold if you missed it...)
Here's another interesting pic. I low passed around 250 then cranked L2 after and as you can see from my meters there is still some signal getting thru (even tho HEQ and L2 meters both show nada).
yeh...the file looks fine...
it's normal to get a little signal still with the lpf on there...
it's not a total "cut"...it's a slope...
thanks for checking...
the only plug ins i can get to do something like this with this tone are things like phasers and modulation efx...is this some type of phase modulation...?
edit: i attached the audio file i bounced of it...i checked it in a different audio editor using a different spectrum analyzer and still got basically the same results...
just want to confirm it's not my computer/DAW...??
(it's the Logic channel EQ, by the way...)
so this is what i get when i run a DC source through the EQ with a HPF @32Hz 24db/oct ...
i gained it up 100db for your listening pleasure...
(actual level after EQing was -110db...the DC source file was 0db...)
i couldn't get any other EQ i have to do this...(i tried DMG, Voxengo, and most of Logic's eq's/filters)
also, it seems strange that Logic would have this EQ available on every channel strip when it acts the way it does...it is not transparent and it acts "crazy" to certain type of stimuli...not that it can't sound good, but this is kind of "forcing" a "sound" on the user if they use the EQ on every track...it should be 100% transparent, really...
i don't want my stuff to sound like it came from Logic...
(little joke there...)
well, good news (for me)...some one has actually confirmed this behavior with the Cubase parametric and graphic EQ's...so it's not a bug in my sh!t...(i don't think)...
interesting...maybe...?
edit: i just noticed it makes a big difference if it's mono or stereo...in the video it is mono...stereo will yield milder results, maybe more noticeable with a 12db/oct slope, but not like in mono...
the only thing funnier than the EQ's behavior is the fact that many smart dudes have looked at this thread, but they have nothing to say...
so, am i missing something here...is this common knowledge...or...?
i did get a response from a well respected developer...
this was his speculation...(going on evidence i provided...no testing of his own...so it's just a guess...i guess...)
"This looks like noise-shaped dithering built-in to the EQ. But probably a faulty one due to some reason.
In the older ages when DSP was done in fixed point math, there was a problem of distortion and oscillation, especially with filters tuned to the lower end of the spectrum.
With 64-bit floating point wordlength it's not an issue anymore."
...i also got this reply from someone...
"...This is usually due to "optimized" code to save up on CPU power and/or analog typish modeling and results in a very huge noise floor that even goes up to -40dBFS digital on the analyser..."
...so what do you guys think ?
"analog modeled" EQ or sh!t code in Logic...?
or both...?
(worth discussing, i think, maybe...)
some more findings...
this was a track of silence playing...
looks like harmonics added (location varies on HPF), but when you move the HPF to a certain point, they all disappear...
#1=audio track of silence playing
#2=silence playing with HPF @30Hz 24db/oct
#3= #2 gained up with Free-G
#4= approx. point at which it nulls again...(didn't null at @ 7K, did at 7.5K, didn't bother checking in between...)
i'm actually kind of scared to...
not really...my main EQs seem to work very well...
i just find it interesting...(i am one curious S.O.B. about anything and everything...used to take everything apart when i was little...)
that, and i really don't feel like doing this project right now...
ah, mondays...
(...me=slacker...)
(also, i'm not trying to put this EQ down...i have used it many times with great success and i'm sure many others have, too...it's just good to know your tools so you can use them appropriately...seems like this is a slightly "colored" EQ as opposed to 100% "transparent" modern digital...)
edit: this could be one of the reasons some people claim Logic is "mushy"...?
(disclaimer: i have never noticed Logic to be "mushy"...but i do not disregard any claims...)
Hmm this gave an idea for an interesting test..the channel eq gives some low end ripple when engaged. Since I´m hpf:ing everything except for bassy stuff, I wonder what this adds to LF noise in a mix when doing it on lets say 20 channels and it builds up..
My idea is to to an arrangement with 20 channels of sine waves that are well above any LF and engage HPF with the channel eq. Then mixdown, normalise and measure the LF noise level.
i did that with a 12K tone (-6db, 24bit) the other day on one track...(Span's "scaling" is off, i know...)
this is what i got...(nothing like the 11,025Hz tone in terms of movement...bass bump, that's it really...)
would be interesting to see many together compared to one...
also, if you set up the "test" from post #14 and change the frequency of the HPF crazy stuff happens...
edit: tried 10 random tracks of audio real quick and the "noise" bump didn't seem to add up that much...
Looks like standard behavior in a minimum phase Butterworth HPF with a 24dB/Oct slope. Switch to 12dB/Oct and compare to the Flux Epure II (download a trial). It should be lower.
Switch to the linear phase EQ in Logic with the same setting and compare, that would give you a clue about the phase shifting.
Currently working on David Guetta feat. Ne-Yo & Akon Albert Neve Remix (EMI) · Vinnie Who (EMI) · Basim (Sony) · Ida Corr feat. Fatman Scoop (Sony) · Alphabeat (Universal) · Infernal (Warner)
Looks like standard behavior in a minimum phase Butterworth HPF with a 24dB/Oct slope. Switch to 12dB/Oct and compare to the Flux Epure II (download a trial). It should be lower.
Switch to the linear phase EQ in Logic with the same setting and compare, that would give you a clue about the phase shifting.
...i tried Flux Epure II and i got nothing like the results from the Logic EQ, then Logic crashed when opening a new track (i get crashes when i try Flux demos...i will try again, though...curious to see...).
thanks for the suggestion, i have tried a few other EQs to duplicate, just maybe not "the right one" yet.
i have Logic Express so i don't think i have the linear phase option, do i ?
i think i understand the reason for the "bass bump"/noise floor, but i don't understand why the 11,025Hz tone sets it off into a frenzy.
to me it seems that it's because it is half of 22,050Hz which is half of 44,100Hz which is the project rate...which is a problem for this EQ because...?
aliasing first came to mind, then some kind of phase/modulation thing going on, but i haven't put it all together just yet...
Looks like standard behavior in a minimum phase Butterworth HPF with a 24dB/Oct slope. Switch to 12dB/Oct and compare to the Flux Epure II (download a trial). It should be lower.
I could be wrong about this, but I suspect that it's just because the filters aren't calculated at as great a bit depth as most other are these days. The noise/craziness could be quantisation distortion building up in the filter feedback loops, which could explain why it changes with filter centre frequency.
Try this with a fully 64-bit EQ like reaper's reaeq (or probably EPure), and there shouldn't be any noise theoretically. At least this is true for reaeq - totally clean. Using enough resolution these issues should be completely avoidable afaik.
...Since I´m hpf:ing everything except for bassy stuff, I wonder what this adds to LF noise in a mix when doing it on lets say 20 channels and it builds up...
hey, maybe you know already, but for those who don't, the Apple AUHipass filter seems to be very clean and effective...i didn't do extensive testing but it seems normal in a quick comparison to some other digital EQ HPFs...that would be the way to go it seems if "transparency" were of concern...
seems a little counter-productive...everybody seems to use HPF everywhere, so it should be a "clean" one on the channel strip...you'd think
I could be wrong about this, but I suspect that it's just because the filters aren't calculated at as great a bit depth as most other are these days. The noise/craziness could be quantisation distortion building up in the filter feedback loops, which could explain why it changes with filter centre frequency.
Try this with a fully 64-bit EQ like reaper's reaeq (or probably EPure), and there shouldn't be any noise theoretically. At least this is true for reaeq - totally clean. Using enough resolution these issues should be completely avoidable afaik.
I'd agree with this explanation. The quantization can occur from filter coefficient calculation, and/or performing the filter operation itself. For work once I had to build a high order filter (8th order), and mistakenly used single precision (32 bit) floating point resolution for both the coefficients and intermediate calculations. The damn thing blew up (unstable) every time I tried to run it. All was fine when I switched to doubles (64 bit) everywhere.
...so it's more because it's a little outdated and not really "faulty"...?
and i "broke it" with that specific tone...
it seems to be ok with most other tones...some minor distortions whatever but that tone is pretty funny...
and the fact that the EQ will enage with a track of silence playing i find odd, too...none of my others do that...
that is some weird, wild stuff...
...so it's more because it's a little outdated and not really "faulty"...?
Seems like a fine line to me - the way to know which side it's on is to listen to it. In all likelihood the distortion is not audible under most conditions - but I don't know for sure
I had a little bit more time to think about this....
What level was the tone at? Does the crazy behavior go away if you reduce the volume of the tone (maybe use a trim plug ahead of it)?
A tone at the frequency 11,025 Hz, and a sampling rate of 44.1kHz will alias down to DC in the event of overload. High pass filters tend to result in some boost in the resultant signal, which may be causing it to saturate, and alias to DC. The high pass filter then attempts to remove the DC, which changes the nature of the aliasing since it's a recursive filter (it's current output depends on it's previous output).
Out of curiosity, try engaging the HPF on a track with pure DC. Or how about a track with DC and sine tone.
I had a little bit more time to think about this....
What level was the tone at? Does the crazy behavior go away if you reduce the volume of the tone (maybe use a trim plug ahead of it)?
A tone at the frequency 11,025 Hz, and a sampling rate of 44.1kHz will alias down to DC in the event of overload. High pass filters tend to result in some boost in the resultant signal, which may be causing it to saturate, and alias to DC. The high pass filter then attempts to remove the DC, which changes the nature of the aliasing since it's a recursive filter (it's current output depends on it's previous output).
Out of curiosity, try engaging the HPF on a track with pure DC. Or how about a track with DC and sine tone.
Cheers
Kris
level of the tone in the video was -6db, and it doesn't do it as much at -12db...hard to get consistency, though...
post #11 is audio from a DC test...i trashed that .wag (ha) file but i attached another...(be careful that sh!t's hot...)
i couldn't bounce it out without getting a weird pulse at the beginning...doesn't look like anything clipped, though...
hey, maybe you know already, but for those who don't, the Apple AUHipass filter seems to be very clean and effective...i didn't do extensive testing but it seems normal in a quick comparison to some other digital EQ HPFs...that would be the way to go it seems if "transparency" were of concern...
seems a little counter-productive...everybody seems to use HPF everywhere, so it should be a "clean" one on the channel strip...you'd think
...i tried Flux Epure II and i got nothing like the results from the Logic EQ
But remember that the Flux Epure II uses a 12 dB/Oct slope. Did you remember to change the Logic EQ to a 12 dB/Oct slope?
Do that and test again.
Quote:
i have Logic Express so i don't think i have the linear phase option, do i ?
As far as I can tell it's a very different story with the linear phase eq. And no, you don't have that in Express. I could send you a bounce but that would mean dithering or truncation as I render the file. I assume you're measuring on the 32 bit floating point signal.
New Reply 
Facebook
Twitter
LinkedIn
