Plugin noise

Discussion in 'Effects and the DSP' started by WxMan1, Jul 16, 2010.

  1. WxMan1

    WxMan1 Active Member

    Joined:
    Dec 26, 2007
    Messages:
    307
    Likes Received:
    11
    Trophy Points:
    28
    I'm noticing a continuous signal apparent from the EQ notch plugin after the DSP processes audio input, i.e, either speaker test or playing MP3 file. Reloading the DSP clears this condition and signal strength is shown to be -inf on both channels.

    After doing a speaker test, the output is -110 dB (asymetrically +/-0.5dB). However, after playing an MP3 file, a vestige signal appears that stabilizes between -68 and -72 dB (asymetrically). This vestige signal undergoes a pronounced fluctation after terminating audio input with stabilzation occuring at that level. Many times a clear real-time fluctation can be seen occuring of a magnitude of about 0.2 dB on one or the other channel, i.e, either at the -110 dB or the -68 to -72 dB levels.

    I've also noticed that if I implement the oscilloscope on either channel, that channel will 'deplete' its vestige signal to -110 dB (while the non-oscilloscope channel will stabilize between -68 and -72 dB). No matter what signal strength is apparent in the peak meter when no audio program is being put through the DSP, no waveform is apparent on the oscilloscope. Moreover, this isn't anything relevent to the peak meter itself - akin to the affect of the oscilloscope - in that the condition will be evident after an audio signal is played through the DSP w/out the peak meter present at the time and the peak meter is wired up after that.

    I'm using v3545b on a Win2003 system w/ CT4780 sound-card.

    Any explanation for that? :confused:
     
    Last edited: Jul 16, 2010
  2. Russ

    Russ Well-Known Member

    Joined:
    Jan 17, 2005
    Messages:
    5,722
    Likes Received:
    13
    Trophy Points:
    48
    Did you check the signal with a peak plugin immediately before the EQ plugin?

    Such noise is typically caused by the Reverb Lite plugin, try disabling or unloading it.

    Aside from the above, what is your DSP config? or how is the EQ plugin connected?
     
  3. WxMan1

    WxMan1 Active Member

    Joined:
    Dec 26, 2007
    Messages:
    307
    Likes Received:
    11
    Trophy Points:
    28
    http://img105.imagevenue.com/img.php?image=11087_DSP_122_148lo.jpg

    Isolate, i.e., all by itself, any, e.g. 4th order x-over, sort of EQ plugin and it does the same stinkin' thing.

    FWIW, disabling the Reverb Lite plug-in has no affect in this matter.

    No noise is evident post any Surrounder+ plug-in connectors.

    Futhermore, I can't reconcile the output of WaveGenerator through any sort of EQ plug-in with peak plug-in; these things are acting like 1st order filters (at best), and at worst sometimes they're showing +6dB gain at the freq cut-off.

    My avatar would properly be black; that's what they do when they're unhappy.
     
    Last edited: Jul 20, 2010
  4. Russ

    Russ Well-Known Member

    Joined:
    Jan 17, 2005
    Messages:
    5,722
    Likes Received:
    13
    Trophy Points:
    48
    Hmm, I do not have kX installed at the moment so I cannot look at it myself (and I do not know what settings you are using, etc), but they are RBJ Biquad filters. They are IIR filters, thus they use previous input/output values in their computation, so I suppose that maybe a rounding error or scaling issue is causing (under certain settings?, etc) some final output values to remain constant, giving you some DC in the absence of a signal? I do not know, just a guess.

    You could try feeding it noise from the Wave Generator plugin, and look at the results using some spectrum analyzer (e.g. a VST spectrum analyzer plugin). If you compare the results with and without EQ, it should give an idea of what the EQ is doing (though it can be hard to tell in the lower frequency range).
     
  5. WxMan1

    WxMan1 Active Member

    Joined:
    Dec 26, 2007
    Messages:
    307
    Likes Received:
    11
    Trophy Points:
    28
    I fingered out the issue concerning the peaking. :lol:

    Its the resonance associated with the Q of the bandstop transition.

    Your explanation for the other issue sounds as good as anything.

    FWIW: your RBJ Biqud link contains info I've seen on other sites concerning EQ filtering. There's a lot of stuff to get one's brain around for sure. I'll review it in greater detail to see if there's any nuggets to glean from it. Thx for the link.
     
    Last edited: Jul 23, 2010
  6. Russ

    Russ Well-Known Member

    Joined:
    Jan 17, 2005
    Messages:
    5,722
    Likes Received:
    13
    Trophy Points:
    48
    Hehe, you do not really need to understand all the Bilinear Transform stuff, etc, the point of posting it was really just so that you would have the info about the specific type of filters (Direct Form I Biquad filers, using the RBJ method to calculate the coefficients, etc) used in these plugins, which should help you to find info on how those particular types of EQ perform, etc.
     
  7. WxMan1

    WxMan1 Active Member

    Joined:
    Dec 26, 2007
    Messages:
    307
    Likes Received:
    11
    Trophy Points:
    28
    Understood. :cool:

    I'm just saying I've seen that technology described in my travels through the interwebs looking into this matter.

    For example:

    A beginner's guide to filter topologies - Maxim

    While pass-band filter Q is pice of caeke, Q gets pretty funky when you're dealing with high / low pass filters; compentancy in the calculus I is insufficient for full comprehension; funky polynomials having real and imaginary roots rear their ugly little demonic heads. :duh:

    That notwithstanding, the information that you provided may prove usefull for toying around with makin' a EQ filter plug-in of utility; given that I'm spoda be competant - proficiency being a matter entirely different - in C++.

    One illuminating factor is the IIR aspect to the filters. Thnx for the feedback though. :)
     
  8. Russ

    Russ Well-Known Member

    Joined:
    Jan 17, 2005
    Messages:
    5,722
    Likes Received:
    13
    Trophy Points:
    48
    BTW: Regarding your observations with the peak plugin and these LP/HP filters, (IIRC) Q basically works out to be gain near the cutoff frequency ( dBgain = 20*log10(Q) ).
     

Share This Page

visited