Spectrogram 60Hz Preset

[Dan McDonald]

I just recently "discovered" the 60 Hz preset for the spectrogram. When I used it, it seems to take maybe 10 minutes of a recording to resolve into the 60 Hz line, and I basically get the lines on the far right, with a lot of 'smear' on the left up through the middle, taking the full screen (59-61 Hz). I'm just wondering what is different about this preset - I sometimes see about 3-4 lines on the far right, sort of emerging out of a cloud of smeary material. The preset uses the Bessel window type with frequency resolution as far over as possible. In the past, I've used the resolution in the middle. I just was wondering if I'm interpreting it correctly. I didn't see a lot of explanation in the manual. It seems quite handy if I'm seeing what I think I'm seeing - essentially very clean signal around 60 Hz, but only emerging around 12 minutes in for some of the recordings.

[Craig Maier]

Hi Dan,

I recall that preset as producing an extremely high level of frequency resolution centered around 60 Hz. Depending on the length of material highlighted, it will take quite a long time to calculate the fft since it is set for the highest fft count possible in that preset mode. I do recall some smearing on short samples which I think is related to the time doimain impulse respons of the system, but I am not positive. I will ask Rick to see if he remembers more about it.

Craig

[Dan McDonald]

OK - thanks. It seems to work great if you have a long enough sample. I guess you could make the sample longer by copying and pasting the file to append at the end also, if you had a shorter clip.

[Craig Maier]

Hi Again,

I think that you can speed up the process a bit by moving the frequency resolution control (decimation) a notch or two towards the right (instead of x 64 to maybe x 32 or x 16).

Craig

[Dan McDonald]

I'm sorry - I'm not clear about what I meant by "Time" - I was just thinking that I probably phrased it wrong. It only takes about 30 seconds with this computer and a 24-bit file, but I mean the time axis in the spectrogram. That is, it sort of looks like a cloud until you get to about 10-12 minutes in (in terms of the spectrogram) then after that are just a few straight lines.

[Craig Maier]

I understand - - - I think that is the transient response of the system, because it is operating at the total extreme of its frequency resolution by virtue of two parameters:

1. Max fft count

2. Max decimation (max frequency resolution)

Maybe try changing the fft count to 65k and the decmiation to divide by 32 or 16 - - - that may reduce the artifacting. You might also try a different window.

Craig

[Dan McDonald]

OK - It actually works great this way, but I was just concerned about shorter samples. Will try those. Thanks Craig. This is much easier to use than the way I did it before.

[Craig Maier]

Dan,

That is a very popular Forensics feature - - - the combination of high fft count and decimation (x by 64).

Craig

[Dan McDonald]

On a related question, as you move the far right slider down, different lines show up near the 60Hz value. When I check a cassette I made, which I know has only been recorded on one time and using a battery power source, then transferred to digital using the flat transfer method, I get a value at 60.06 Hz, at about -62.63dB. When I move the slider further down, I get 58.59 at -82dB and a -61.45 at -73dB. If I add the 2 'odd values together, I get 120.04, so those two values are almost exactly the same distance from the 60 Hz value, within rounding, the average of the 2 is 60.02. Is that typical to get this kind of 'balance' effect? I am just trying to figure out what those other lines would be. If the 60.06 (strongest signal) is from the transfer to digital, are the two on each side from when the recording was made, when it picked up the mains power through the band's amplifiers? Should those be averaged together to get a better estimate of the value, for checking speed? I know in this case it doesn't really make any practical difference, but it gets me curious about what those values are. For some recordings, I get a number of these additional lines. It almost seems like you can trace a lineage on multi-generation by looking at the signals from weakest to strongest, with the weakest being the original, which would have been copied each time. In those cases, it would be a big help to figure out the original signal. If that is too complicated a question, don't worry about it. I just get curious sometimes.

[Craig Maier]

Hi Dan,

I really am not sure, but it could be possible that the recorder had some wow or flutter and a resultant FM (frequency modulating) effect on the recording. That may produce those other spectral lines. But, I really am not sure about this.

Craig

[rickc]

Back on the subject of the 60hz spectral resolution, There is a minimum amount of audio required to resolve a spectral component. If you don't have enough source file, then you will get some of this noise or artifacting that you see. So for example if you are trying to resolve 1Hz, you need at least 1 second of music, .1 Hz would be 10 seconds and so on.

[Dan McDonald]

Thanks Rick Dan