Subject:Spectrum analysis and EQ

Posted by: Simonm

Date:1/26/2008 1:54:40 PM

I'm not sure if this is a scripting question or not - apologies if I've posted in the wrong place:

I'm trying to rescue a live classical recording with some nasty interference on it. I still don't know exactly what the fault was, but it seems to be two sets of clock frequencies leaking from the digital kit into the analog domain and beating with each other. Anyway, the result is a very nasty 'buzz' spread wide over the audio spectrum, but with very identifiable and specific spikes of frequency, oddly, exactly on 1kHz boundaries, and with very sharp peaks. I've looked at this with various spectrum analysis tools, and it doesn't seem to be an artefact of the analysis, but real.

So I can see what I need to remove, and I have an idea how to do it: I need to construct a very high Q 'comb' filter. I can see what frequencies I need to attack and by how much from the GUI of the spectrum analyser, and I can _nearly_ do it in the GUI with the parametric EQ, but I can only get 0.1 octave filtering, which isn't nearly sharp enough.

So to my questions:

Firstly, can I expose the data set used to generate the spectrum analysis/sonogram plot? I need that as a starting point - read into Excel (or something like that) so that I can operate on the data.

Secondly, is there any way of programatically driving any of the EQ filters, so as to produce something with much higher Q than the GUI presently allows?

I am sure this recording is fixable and that Sound Forge could do it (in theory), but having pored over the documentation of the API, it looks like these things are not exposed - is this right, if so it's a great shame. Alternatively, is there some sort of 'forensic' API that's different and would permit this level of control?

Any thoughts appreciated...

Simonm.

Subject:RE: Spectrum analysis and EQ

Reply by: Kennymusicman

Date:1/26/2008 2:36:37 PM

Iff the artifiacts are continuos - you're probably wasting a lot of time trying your methadology. Simply ope nup file in SForge, and use the noise reduction. A little playing and you should find that you could probably eliminate most/all of it in a simple step, without any playing about with filters/eq's and more. If you're not sure, post a little sample link in the SForge forum and I'll have a look.

Subject:RE: Spectrum analysis and EQ

Reply by: MarkWWW

Date:1/27/2008 4:09:12 AM

I don't think there is anything in the scripting api that can help you with this.
But if the noise isn't treatable with the usual Noise Reduction methods then you might like to look at this thread where a variety of special techniques are discussed. In particular, if the noise you have is synchronous (which by your description it sounds like it may well be) then the time domain subtraction method I describe in the thread may be applicable.

Mark

Subject:RE: Spectrum analysis and EQ

Reply by: Simonm

Date:1/29/2008 8:06:55 AM

Many thanks both.

Mark, I've looked at the thread and it's very interesting. I'm not sure how repetitive the interference actually is. Interestingly, I'm suspicious that my own HP laptop was partly at fault (one of the two recorders!).

I have an idea about using synthesis to artificially create a noiseprint. Basically, I need to knock out the louder harmonics, but not all of them, and I can see what to remove (and its amplitude) from the spectrum analysis. By using pure sine waves of the right
frequency and amplitude, I might create a suitable 'mask' to allow the NR plug-in to do its stuff.

I'll also have a look at the time-domain approach that's outlined.

I've had a spot of bother with my web pages, but I'll try to put up some short clips of the interference for anyone interested to play with.

Many thanks once again.

S.

Incidentally, to answer my own original question, it looks like the API doesn't expose enough of the FX parameters for scripting to help much.

Subject:RE: Spectrum analysis and EQ

Reply by: Kennymusicman

Date:1/29/2008 8:25:56 AM

Look in NR at the noise print - you can specify harmonics to keep/remove without having to create a proxy mask.

Subject:RE: Spectrum analysis and EQ

Reply by: MarkWWW

Date:1/29/2008 11:31:42 AM

> I'll try to put up some short clips of the interference for anyone interested to play with.

Jolly good - I'll see if there's anything I can do with them. You should only need to post a few seconds, which should preferably contain a period of silence (plus noise) and also a period of the wanted signal (plus noise).

Mark