At 10:51 PM 10/24/2013, Malcolm Rockwell wrote:
>There may not be a major problem here. What has printed through is the 
>audio from the next layer of tape, correct? With digital manipulation 
>being what it is today it should be simple enough to grab the full volume 
>layer of audio, attenuate it, flip the waveform and apply it "over" the 
>printed through signal. There will probably be artifacts but if you fiddle 
>with various parameters for a while, such as EQ, you will probably be able 
>to find an acceptable solution to your problem. I'd apply this to softer 
>passages and leave louder material well enough alone, though.
>It's worth a try.
>Comments?

There are a number of problems to consider.

First, the printed recording is not the same length as the original.  The 
delay between original and print changes as tape pack diameter 
changes.  Seems to me that for a tape which has been stored tails out, the 
print is longer than the original.  (Delay from original to print increases 
as tape pack diameter on the take up reel increases.)  The original 
recording can be speed-shifted, but you need to figure out how much to 
shift it.

Second, the frequency response of the printing process is not 
flat.  According to
http://www.aes.org/aeshc/docs/3mtape/printthrough.pdf
"The worst print-through occurs at a wavelength equal to 27 * C. C is the 
total tape caliper in mils. For a typical 2
mil mastering tape, the worst wavelength for print-through would be about 
12.6 mils. When running at 15 ips, this
would be a frequency of about 1200 Hz."
Note:  there's an error in their formula, as printed in the on-line 
document.  It should read 2 * Pi * C.  But they're correct that it's worst 
at about 1200 Hz.  This is confirmed by the October 1980 JAES article "The 
Print-Through Phenomenon" by Bertram, Stafford and Mills.  It includes a 
graph of print-through vs. frequency.

BTW, this article also states that "print-through ... can be reduced if 
[the tape] is repeatedly rewound.  The amount of print reduction ...can 
reach as much as 7 dB."  In their tests, this required 6 rewindings.  "The 
rewindings should be consecutive with an optimum storage time between 
rewindings to achieve maximum reduction.  The optimum storage time may 
depend upon the individual tape."

Third, is the printing process linear or non-linear?  The 3M document cited 
above says it's linear.  Camras, in the 1988 edition of "Magnetic Recording 
Handbook", says it's not, and that the ratio of the original to printed 
signal varies with the level of the original signal.  I think Bertram et 
al.  are saying it is linear, but I must admit that I have not yet entirely 
digested this long, complex article.  Hopefully it is linear, because 
modelling a non-linear transfer function will not be much fun.

So....  If the printing process is linear, the other problems seem 
manageable.  But it will no doubt require a fair bit of fiddling to get the 
cancellation signal lined up in time and amplitude with each objectionable 
echo.

-- John Chester