Determining a source

[Dan McDonald]

Here's one that's a little whacky, but is there any way to determine whether a tape has come from an original cassette source or an open-reel source? It's not a major issue, but the topic has come up with an old concert recording. In this case, the only surviving copies are all very hissy and are probably multiple generations off the original.

Dan

[Craig Maier]

Hi Dan,

I think that one can determine the number of dubs by looking at the slope of the noise above 10 kHz. To view this slope accurately, it is best to transfer the recording at 96 kHz.

If I recall correctly, the noise should roll off at 3 dB/octave per recorder in the pathway. So, if that noise slope is just 3 dB/Oct, then it is an original and if it rolling off at 6 db/Octave, then it is second generation and so forth. You will need to find a quiet passage between performances to determine this slope accurately.

At least, that is my recollection.

Craig

[Dan McDonald]

That is interesting, Craig - and I'll use that in this project also, but I was actually wondering if there was a way to tell if the original was cassette or open-reel.

But back to your other point, is that 3dB per octave, regardless of the original source?

Thanks Dan

[Craig Maier]

Hi,

I think that the 3dB/Octave is a physics based slope having something to do with the intrinsic power spectral density of a tape heads interaction with tape emulsion, but again, I am a little fuzzy on that. I have not thought about that topic for years and I just do not precisly remember the exact physics involved.

As for the original being a R-R vs a Cassette, if you are many generations away from the source, I would think that it would be almost impossible to tell. Clearly, if there was no dubbing going on inbetween, the noise signature of a R-R is distinctive compared to Cassette tape. Cassette tape sort of brick walls at around 17 kHz or so and R-R tape (at 15 ips) brick walls at a much higher frequency.

[Dan McDonald]

Thanks very much, Craig - that's very interesting to me. I have a multi-generation cassette transfer that has almost nothing above 5 kHz, and most of it is around the 500-2kHz range. Since it was a rock band, I am guessing that the instruments (apart from cymbals, etc.) probably peaked at around 17-18kHz in the original tape, and the bulk of the music was around 3-4kHz, based on what I've seen with other tapes.

Using that as a baseline guess, the peaks would appear to have dropped from 18kHz to 5kHz, or the bulk from about 4kHz to 2 kHz.
Could I use that information to estimate (roughly) the number of generations?

Dan

[Craig Maier]

Hi,

I do not know if you can tell simply by the highest peak since you do not know the "fidelity" of the original recording.

I would put the spectrum analyzer into averaging mode and let the file preview for a few minutes and measure the slope of the line created above 10KHz.

[Dan McDonald]

I would.. but basically there's nothing above about 5kHz except random noise. The spectrum analyzer drops below -95db around 5kHz, except for an odd noise spike around 14 kHz.

Well, I'll look to see what I can find if I zoom in as much as possible.

[Craig Maier]

Hi Dan,

Is that the case with a 96 KHz, 24 bit transfer? There should still be something there - - - what you are looking for is the random noise produced by the tape(s). To be sure, do a transfer with the tape player off and another one with the tape player on so that you can compare to the baseline noise of your soundcard.

Craig

[Dan McDonald]

OK - will do. Thanks Craig.

[Dan McDonald]

At 3kHz, it's at -91 dB
at 12kHz, it's at -116 dB

So, if I remember right, doubling the frequency is an octave, right?
so that's two octaves (3 to 6, then 6 to 12)?

and the signal moves -25, or 12.5 per octave, is only about 4 generations, but the signal is so poor, it almost has to be more than that. Above 10kHz though, there is so little information that it looks like it's just a flat line at -116 dB, with very minor fluctuation around the line.

I guess I'm just trying to figure out if it's a lot of generations or a very poor original source. Maybe there's not enough information in the file to be able to figure that out.

Dan

[Craig Maier]

Hi,

Yes, an octave is a doubling in frequency. So, the next octave above 3 kHz is 6 kHz and the next one above 12 kHz is 24 kHz.

I would try to measure the change in amplitude between 10 kHZ and 20 kHz. If that value delta is 12 dB, that would likely mean 4 generations.

Also, if they are cassette tape generations, that would not sound too good. Of course, as you point out, the final fidelity may have been poor to start with; you have no way of knowing what type of microphones were used or what type of tape deck was involved. It is even possible that a microcassette recorder was the master (I sometimes sneaked those microcassette recorders into concerts a long time ago).

Craig

[Dan McDonald]

OK - fun stuff...
I've been trying to expand the sound a bit, using the DC8 software after looking at it in 7.5 forensics to see what is going on with it.

To eliminate hiss, I'm thinking about a low-pass filter set pretty low because there's almost nothing but hiss above the 5kHz.

Dan

[Craig Maier]

Hi Dan,

Maybe after you brick wall it at 5 kHz, try the overtone synth and the vva to see if you can re-establish some top end. Also, consider using the sub-harmonic synth and the punch and crunch effect.

Craig

[Dan McDonald]

I was thinking along the same lines, except I've never gotten very good at Punch 'n Crunch. I'm seem to be better with the dynamic noise in enhancer mode, and the dynamics processor in expander. I don't know why I've never been very successful with P'nC. I always think I'm going to do it well, but I think I over-work it or something.

[Dan McDonald]

Looking at the file with the forensics tools is kind of fun. I ran the spectrum analyzer and see 5k to 20Hz as a straight line around -97 dB. Except there are a number of spikes at multiples of 60 Hz - around 20220 (337 x 60), 11580 (193 x 60), and 16500 (275 x 60). Not sure what they're from. It looks like they are in the original tape because they don't show up at the very beginning of the file or at one spot where another tape was added in.

I had to go to 2.5 kHz-5 kHz to find an octave that had a slope. That octave changes from -77dB to -97 dB or about 20 dB.

The brick wall filter seems to be what I need with this to cut the hiss out, and getting some expansion and harmonics puts sound out to about 8kHz, which is a pretty dramatic improvement.

Dan