Amplitude Increase ?

[Marc Hildebrant]

I am using a "passive" multifilter to remove noise from a cylinder recording. The source file has amplitudes less than 26 K. yet the destination has signals greater than 26 K ?

I have attached a picture to show the result...why are signals higher after filters that have no gain ?





Note: This doesn't happen on all of the songs. Just sometimes.

Marc

[Dan McDonald]

Hi Marc
It looks almost like it's clipped. If that happened, then I think the impulse filters would try to rework the signal to where it's supposed to be, so that might be part of it. If some of those big spikes are clicks or pops, that that might be part of it.

Also, I think the cnf can do some things in certain modes. I would click off the impulse filters, and see if it still does it. If it does, I'd put them back on and click off the cnfs. Anyway, bypassing them like that will tell you which filter is responsible.
On the other hand, you can just wait til Craig sees the post because he probably knows the answer already.
Dan

[Marc Hildebrant]

Dan and All,

If I remove the Hi Pass and bandpass filters the amplitude is about the same before and after. Some high points have been removed though, due to the filters.

I have no idea why this is true....

Marc

[Craig Maier]

Hi,

I think that is just the math. LP and HP filters of higher order than first order produce a peak (as I recall from my college days). So, that is likely what you are seeing. Chebyshev would produce an even higher elevation than Butterworth. I think that the amplitude of the peak varies with the IIR's filter order and filter type.

Craig

[Marc Hildebrant]

Some more info...

In the multifilter shown (it's for cylinder records), I had a High Pass filter and a bandpass filter. The high pass is not needed, so I removed it and tried running the filter again. Without the HP filter, the signal is not increased.

The high pass filter I used was set to 99 Hz, a butterworth type and a slope of 18 dB.

Craig, you are correct that filters can have ripple in the pass band, but the butterworth type is maximally flat so there should not be any amplitude increase with this HP filter.

The curious part is that cylinder records do not have much low end energy and the amplitude increase doesn't always happen ?

Marc

P.S. It's been a long winter here in Minnesota and I have too much time on my hands!

[Craig Maier]

Hi Marc,

18 dB/oct is pretty steep. Although HPF Butterworth higher order filters are maximally flat, I do not think that precludes them from having a peak of something like 3 dB (square root of 2 - - - ~1.414) when cacatanated with another filter such as a LPF of Butterworth design, but I would have to go back to my college textbooks to prove that.

Try putting a swept sine wave into it and find out where the peak occurs in frequency.

Craig

[Marc Hildebrant]

Craig,

I tried using the "make waves" for a swept sine wave...everything works as it should.

So, the mystery continues.

Marc

[Craig Maier]

I am going to guess that the HPF is interacting with the BPF in some manner. What does the system do when excited with random noise (Make waves generator has that function)?

Craog

[Craig Maier]

I wonder what Impulse Response of the multifilter is. To make an impulse, make waves for 10 seconds (any signal - does not matter). Then, highlight it entirely and then mute it with the editor (edit menu \ mute).

Zoom in until you see the samples. Using the pencil tool, create a full scale impulse lasting for 2 samples and then apply that to the filter lineup you created.

Look at the resulting file produced and see if there is ringing and at what frequency it is ringing (assumiing that it is).

Craig

[Marc Hildebrant]

The beat goes on.....

I'm now getting some increased amplitude without the HF filter in the multifilter shown in the picture. I'll try some of the experiments, but I'm wondering if the problem comes from the original song having some "flat tops" from saturation ?

Marc