Strange things happening with parametric EQ...

wingfeather · Sep 12, 2003, 09:51 PM

As mentioned in previous posts I use the PEQ effect to get a flat bass response as low as possible. I like to cut it off at about 25Hz. Since all 5 bands of one PEQ were used to get my nice EQ curve and steep bass slope, I added another with a 'brickwall' highpass (4 bands set to Highpass and the final one one set to Bell to shape the response) at about 22Hz. However, even though the response line on the PEQ display never goes above the 0dB line, this makes the signal clip when a frequency around the cutoff is passed. When I bypass the effect, there is no clipping. How is this possible? Is there an error in the PEQ code somewhere? I regret that I have no knowledge of programming so I can't investigate.
Any thoughts?

Max M. · Sep 12, 2003, 10:53 PM

check out http://www.hardwareheaven.com/showthre...threadid=21815
the answer is similiar... - "higher peak" does not mean "higher level" (it's like.. hmm... signal just somewhat "explodes")

also i can say that such low frequencies and such high-order filtering (four lowpasses with high slopes) are very critical even for 32/64 bit precision - probably some DC offset is also takes place...

(btw... "tubesound" plugin can be used as "DC offset remover" - when its depth set to 0 "tubesound" does nothing except killing DC)...

wingfeather · Sep 13, 2003, 12:24 AM

hmmmm I wouldn't mind the fact that it's clipping so much if I couldn't hear it, but when it goes past about +3dB I get very loud pops and clicks. Is the only way to get rid of this to reduce the overall gain (as I'm doing now)? It just seems that a reduction in a band should never give an increase in response, peak or otherwise. Is this clipping due to a limitation of the chipset accuracy or am I just plain wrong?

Max M. · Sep 13, 2003, 12:55 AM

>It just seems that a reduction in a band should never give an increase in response, peak or otherwise

It is not correct... try for example a little expirement:
In you favorite waveeditor (WaveLab, Forge, CoolEdit whatever) generate square wave at (for example) 440Hz with "peak level" (for example) -6dB... now take your favorite EQ (DX, VST or built-in one) and tweak it to highpass with cutoff at 100Hz (for example) with high slope/Q...
"Logic" says you that such filter should not affect our square at all since 440Hz square has nothing below that 440Hz... Now just apply filter to that square and check the result (yep, just look at wave you'll get and it's maximum peak value)...

---

---

That happens due to phaseshifting (which every filter (except very special ones) introduces)...
Of course square is a quite synthetic example but in practice the music written on cd's (or mp3s coded from them - whatever) are so hardly compresed, limited, maximized that their "waveforms" have same "huge power in little range" property as square... so the result maximum peak change direction is not predictable for any filtering
(and many other kinds of signal processing)

So the only right way in this situation is to reduce overall gain... no more no less... nothing criminal...

Sep 12, 2003, 09:51 PM	#1
wingfeather DriverHeaven Newbie Join Date: Jul 2003 Location: UK Posts: 18 Rep Power: 0	Strange things happening with parametric EQ... As mentioned in previous posts I use the PEQ effect to get a flat bass response as low as possible. I like to cut it off at about 25Hz. Since all 5 bands of one PEQ were used to get my nice EQ curve and steep bass slope, I added another with a 'brickwall' highpass (4 bands set to Highpass and the final one one set to Bell to shape the response) at about 22Hz. However, even though the response line on the PEQ display never goes above the 0dB line, this makes the signal clip when a frequency around the cutoff is passed. When I bypass the effect, there is no clipping. How is this possible? Is there an error in the PEQ code somewhere? I regret that I have no knowledge of programming so I can't investigate. Any thoughts?

Sep 12, 2003, 10:53 PM	#2
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	check out http://www.hardwareheaven.com/showthre...threadid=21815 the answer is similiar... - "higher peak" does not mean "higher level" (it's like.. hmm... signal just somewhat "explodes") also i can say that such low frequencies and such high-order filtering (four lowpasses with high slopes) are very critical even for 32/64 bit precision - probably some DC offset is also takes place... (btw... "tubesound" plugin can be used as "DC offset remover" - when its depth set to 0 "tubesound" does nothing except killing DC)... __________________ Last edited by Max M.; Sep 12, 2003 at 11:05 PM.

Sep 13, 2003, 12:55 AM	#4
Max M. h/h member-shmember Join Date: Dec 2002 Location: Evil Empire Posts: 2,639 Rep Power: 69	>It just seems that a reduction in a band should never give an increase in response, peak or otherwise It is not correct... try for example a little expirement: In you favorite waveeditor (WaveLab, Forge, CoolEdit whatever) generate square wave at (for example) 440Hz with "peak level" (for example) -6dB... now take your favorite EQ (DX, VST or built-in one) and tweak it to highpass with cutoff at 100Hz (for example) with high slope/Q... "Logic" says you that such filter should not affect our square at all since 440Hz square has nothing below that 440Hz... Now just apply filter to that square and check the result (yep, just look at wave you'll get and it's maximum peak value)... --- --- That happens due to phaseshifting (which every filter (except very special ones) introduces)... Of course square is a quite synthetic example but in practice the music written on cd's (or mp3s coded from them - whatever) are so hardly compresed, limited, maximized that their "waveforms" have same "huge power in little range" property as square... so the result maximum peak change direction is not predictable for any filtering (and many other kinds of signal processing) So the only right way in this situation is to reduce overall gain... no more no less... nothing criminal... __________________ Last edited by Max M.; Sep 13, 2003 at 01:43 AM.

Sep 13, 2003, 12:24 AM	Thread Starter #3
wingfeather DriverHeaven Newbie Join Date: Jul 2003 Location: UK Posts: 18 Rep Power: 0	hmmmm I wouldn't mind the fact that it's clipping so much if I couldn't hear it, but when it goes past about +3dB I get very loud pops and clicks. Is the only way to get rid of this to reduce the overall gain (as I'm doing now)? It just seems that a reduction in a band should never give an increase in response, peak or otherwise. Is this clipping due to a limitation of the chipset accuracy or am I just plain wrong?