Suppose you have an analog music tape that plays at varying speeds (or
perhaps was recorded at varying speeds) so that the frequencies are shifting
higher/lower than they should be, are there mathematical techniques to
detect and correct the shifts? Even though the frequencies in the music are
changing all the time, I thought it still might be possible to detect
synchronized shifts in all frequencies over time, which is what speed
changes would cause. There is no correct reference recording to compare
with.

David

Unless there is a reference tone recorded along with it, the only way would
be to manually adjust speed as it plays, and this will surely be imperfect.

Mark Z.




"David White" wrote in message
...
Suppose you have an analog music tape that plays at varying speeds (or
perhaps was recorded at varying speeds) so that the frequencies are
shifting
higher/lower than they should be, are there mathematical techniques to
detect and correct the shifts? Even though the frequencies in the music
are
changing all the time, I thought it still might be possible to detect
synchronized shifts in all frequencies over time, which is what speed
changes would cause. There is no correct reference recording to compare
with.

David

"David White" wrote in message


Suppose you have an analog music tape that plays at varying speeds (or
perhaps was recorded at varying speeds) so that the frequencies are
shifting higher/lower than they should be, are there mathematical
techniques to detect and correct the shifts?

Their must either be a fixed reference in the recording, or presumptions
must be made about the content of the recording.

Even though the
frequencies in the music are changing all the time, I thought it
still might be possible to detect synchronized shifts in all
frequencies over time, which is what speed changes would cause.

This would seem to be an interesting problem in signal processing and
artificial intelligence.

There is no correct reference recording to compare with.

But, you have now presumed that a reference of sorts exists. Your
presumption is that any frequency shift that affects all tones equally is
due to failings in the transmission medium, and can therefore be removed
without diminishing sound quality.

To do this right you'd have to process a lot of tones concurrently, detect
the FM in each, and determine which FMs correlated, and which did not. The
simple part would be applying corrective FM to the over-all signal. based on
the FM that correlated.

The relevant process that does work involves recovering the origional tape
recorder's bias signal. The signal is corrected as indicated by FM of this
bias signal. Tape heads with very small gaps were used.

In , on 09/21/04
at 01:55 PM, "David White" said:

Suppose you have an analog music tape that plays at varying speeds (or
perhaps was recorded at varying speeds) so that the frequencies are
shifting higher/lower than they should be, are there mathematical
techniques to detect and correct the shifts? Even though the
frequencies in the music are changing all the time, I thought it still
might be possible to detect synchronized shifts in all frequencies
over time, which is what speed changes would cause. There is no
correct reference recording to compare with.

As others indicated, this is a difficult problem, but you could get
lucky.

If there is a recognizable signal on the tape, perhaps 50 or 60 cycle
hum (or a harmonic), you could lock on an "correct" it.

----

The pitch changes on the original tape are probably not random, but are
likely to be obscure. Below are some possibilities. (note that any or
all of these could be a property of the record and/or playback
machine):

Many cassette deck motors change speed after being off for a while. In
my observation they usually slow down after a few seconds.

There could be some speed variations related to the size of the tape
pack on one or both reels.

Temperature.

Power line frequency (if one or both machines reference the power line
and it is unstable).

Power line voltage (only for low quality machines)

Tape condition. (The adhesive on some older tapes may be leaching to
the surface and change the coefficient of friction on an inch by inch
basis. -- physically cleaning or conditioning the tape may help if it
is only a playback problem) Keeping the playback heads clean is also a
help.

I don't know your level of experience, but it is common for beginners
to confuse high frequency level changes (due to worn playback heads or
wandering tape) with speed changes. Be sure you are barking at the
right dog.

---

Once you've figured out if you have a record and/or playback problem
and can generate some sort of function to track the changes, applying
the function to the playback deck is not so hard.

"Pitch extraction" algorithms are being refined, but I'm not aware of
any that would be foolproof for this application.

If the tape is extremely important and you have the time, you could
construct a computer file that incrementally controlled the playback
pitch (either on a real tape deck or a computer file) and keep fussing
with things till you are satisfied or run out of time.

-----------------------------------------------------------
spam:
wordgame:123(abc):14 9 20 5 2 9 18 4 at 22 15 9 3 5 14 5 20 dot 3 15
13 (Barry Mann)
[sorry about the puzzle, spammers are ruining my mailbox]
-----------------------------------------------------------

"Barry Mann" wrote in message
om...
In , on 09/21/04
at 01:55 PM, "David White" said:

Suppose you have an analog music tape that plays at varying speeds (or
perhaps was recorded at varying speeds) so that the frequencies are
shifting higher/lower than they should be, are there mathematical
techniques to detect and correct the shifts? Even though the
frequencies in the music are changing all the time, I thought it still
might be possible to detect synchronized shifts in all frequencies
over time, which is what speed changes would cause. There is no
correct reference recording to compare with.

As others indicated, this is a difficult problem, but you could get
lucky.

If there is a recognizable signal on the tape, perhaps 50 or 60 cycle
hum (or a harmonic), you could lock on an "correct" it.

----

The pitch changes on the original tape are probably not random, but are
likely to be obscure. Below are some possibilities. (note that any or
all of these could be a property of the record and/or playback
machine):

Many cassette deck motors change speed after being off for a while. In
my observation they usually slow down after a few seconds.

There could be some speed variations related to the size of the tape
pack on one or both reels.

Temperature.

Power line frequency (if one or both machines reference the power line
and it is unstable).

Power line voltage (only for low quality machines)

Tape condition. (The adhesive on some older tapes may be leaching to
the surface and change the coefficient of friction on an inch by inch
basis. -- physically cleaning or conditioning the tape may help if it
is only a playback problem) Keeping the playback heads clean is also a
help.

I don't know your level of experience, but it is common for beginners
to confuse high frequency level changes (due to worn playback heads or
wandering tape) with speed changes. Be sure you are barking at the
right dog.

---

Once you've figured out if you have a record and/or playback problem
and can generate some sort of function to track the changes, applying
the function to the playback deck is not so hard.

"Pitch extraction" algorithms are being refined, but I'm not aware of
any that would be foolproof for this application.

If the tape is extremely important and you have the time, you could
construct a computer file that incrementally controlled the playback
pitch (either on a real tape deck or a computer file) and keep fussing
with things till you are satisfied or run out of time.

-----------------------------------------------------------
spam:
wordgame:123(abc):14 9 20 5 2 9 18 4 at 22 15 9 3 5 14 5 20 dot 3 15
13 (Barry Mann)
[sorry about the puzzle, spammers are ruining my mailbox]
-----------------------------------------------------------


If the tape was made on a high enough quality machine, and at a high enough
speed, you may be able to recover the bias signal by slowing the tape down
enough - 100khz - 50kHz - 25kHz - 12.5k not likely to actually work very
well... perhaps finding a 60Hz buzz would be a better idea

Thanks for the responses. There are some promising ideas there. Locking onto
a buzz or the bias would do the trick very well. Other than that, I can see
that assumptions would have to be made about the content (after all, you
could intentionally produce such effects for the master recording, so to
"correct" them would be wrong). The recording I have in mind is a live
recording of a rock song. You can probably make some assumptions about the
kinds of patterns you would not expect to see in a flawless recording of
such a sound, but I guess that would require some pretty sophisticated
processing.

David

"Carl Valle" wrote in message
news

If the tape was made on a high enough quality machine, and at a high
enough speed, you may be able to recover the bias signal by slowing
the tape down enough - 100khz - 50kHz - 25kHz - 12.5k not likely to
actually work very well... perhaps finding a 60Hz buzz would be a
better idea

Bias on legacy machines was often in the 35-70 KHz range. If the tape was
originally recorded at 7.5 or 15 ips, at 1 7/8 the bias would be at 1/4 or
1/8 the origional frequency. Other than modest losses in SNR there's no
unforgivable sin associated with using a head built for too-narrow tracks
like a cassette machine playback head, to get response @1 7/8 going up to
maybe even 15 KHz. 15 KHz @ 1 7/8 corresponds to 60 KHz at 7.5 ips and 120
KHz at 15 ips.

So find an open reel machine that played 1 7/8, shoehorn in some cassette
heads and transcribe away!

"Barry Mann" wrote in message
om...
In , on 09/21/04

As others indicated, this is a difficult problem, but you could get
lucky.

If there is a recognizable signal on the tape, perhaps 50 or 60 cycle
hum (or a harmonic), you could lock on an "correct" it.

----

The pitch changes on the original tape are probably not random, but are
likely to be obscure. Below are some possibilities. (note that any or
all of these could be a property of the record and/or playback
machine):

Many cassette deck motors change speed after being off for a while. In
my observation they usually slow down after a few seconds.

There could be some speed variations related to the size of the tape
pack on one or both reels.

Temperature.

Power line frequency (if one or both machines reference the power line
and it is unstable).

Power line voltage (only for low quality machines)

Tape condition. (The adhesive on some older tapes may be leaching to
the surface and change the coefficient of friction on an inch by inch
basis. -- physically cleaning or conditioning the tape may help if it
is only a playback problem) Keeping the playback heads clean is also a
help.

I don't know your level of experience,

I am an amateur, but I am a computer programmer and I've written some audio
processing software, which I'm sure no editor on the market would provide,
to improve certain recordings of mine, sometimes with wonderful results.


but it is common for beginners
to confuse high frequency level changes (due to worn playback heads or
wandering tape) with speed changes. Be sure you are barking at the
right dog.

I haven't actually heard this recording yet. I've been told that it sounds
"warbly", and that altering the speed of some sections improved it a little.
But I also have a recording of my own containing what sounds very like a
speed change in one spot, so I got to thinking about the problem, and I'll
probably apply it to that spot at least if I can figure out a way to do it.


---

Once you've figured out if you have a record and/or playback problem
and can generate some sort of function to track the changes, applying
the function to the playback deck is not so hard.

"Pitch extraction" algorithms are being refined, but I'm not aware of
any that would be foolproof for this application.

If the tape is extremely important and you have the time, you could
construct a computer file that incrementally controlled the playback
pitch (either on a real tape deck or a computer file) and keep fussing
with things till you are satisfied or run out of time.

Yes, I'd be able to write my own software to do that easily enough. However,
speed changes of motors - if that's the problem - are going to be gradual,
so I doubt that simply suddenly changing the speed by a constant amount at
certain sections would work. I'd have to use curves that closely match the
true function to get good results. It sounds like a nightmare to get right
by trial and error. That's why I hoped it could be done automatically and
corrected almost to perfection by analysing the sound; locking onto a buzz
or the bias, as some responders have suggested, is a possible way of doing
just that.

David

In , on 09/22/04
at 09:32 AM, "David White" said:

[ ... ]

I haven't actually heard this recording yet. I've been told that it
sounds "warbly", and that altering the speed of some sections improved
it a little. But I also have a recording of my own containing what
sounds very like a speed change in one spot, so I got to thinking
about the problem, and I'll probably apply it to that spot at least if
I can figure out a way to do it.

As you work with the tape, try to tie the speed changes with a tape
recorder characteristic. For example, if a reel bearing is tight or the
tape is rubbing a reel, then speed changes will be tied to reel
rotation (remember that the tape on the reel is essentially a spiral of
Archimedes and the spacing between the speed changes is tied to the
current pack radius.) If the supply reel bearing is tight, then the
tape may slow down as the pack shrinks.

"Warbly" suggests that the speed changes are fairly rapid. If this is
true, then your speed control system will need to be able to respond
accordingly.

-----------------------------------------------------------
spam:
wordgame:123(abc):14 9 20 5 2 9 18 4 at 22 15 9 3 5 14 5 20 dot 3 15
13 (Barry Mann)
[sorry about the puzzle, spammers are ruining my mailbox]
-----------------------------------------------------------

"Barry Mann" wrote in message
om...
In , on 09/22/04
at 09:32 AM, "David White" said:

[ ... ]

I haven't actually heard this recording yet. I've been told that it
sounds "warbly", and that altering the speed of some sections improved
it a little. But I also have a recording of my own containing what
sounds very like a speed change in one spot, so I got to thinking
about the problem, and I'll probably apply it to that spot at least if
I can figure out a way to do it.

As you work with the tape, try to tie the speed changes with a tape
recorder characteristic. For example, if a reel bearing is tight or the
tape is rubbing a reel, then speed changes will be tied to reel
rotation (remember that the tape on the reel is essentially a spiral of
Archimedes and the spacing between the speed changes is tied to the
current pack radius.) If the supply reel bearing is tight, then the
tape may slow down as the pack shrinks.

Actually, someone else has the tape, and he'll probably send me a wave file
of the sound as played on his equipment. He's already played the tape on a
number of machines and found that certain problems, such as low volume and
the warblyness, vary between them. He's even found that it sounds better if
he doesn't put the tape all the way in on one machine. So there's hope yet
that he can play it on something that it sounds okay on. That would be
preferable to trying to fix it later.

"Warbly" suggests that the speed changes are fairly rapid. If this is
true, then your speed control system will need to be able to respond
accordingly.

Yes. A buzz or bias lock method should be able to handle any rate of change,
within reason, but I'm not looking forward to any trial and error method.

David

"Arny Krueger" wrote in message
...
"Carl Valle" wrote in message
news

If the tape was made on a high enough quality machine, and at a high
enough speed, you may be able to recover the bias signal by slowing
the tape down enough - 100khz - 50kHz - 25kHz - 12.5k not likely to
actually work very well... perhaps finding a 60Hz buzz would be a
better idea

Bias on legacy machines was often in the 35-70 KHz range. If the tape was
originally recorded at 7.5 or 15 ips, at 1 7/8 the bias would be at 1/4
or
1/8 the origional frequency. Other than modest losses in SNR there's no
unforgivable sin associated with using a head built for too-narrow tracks
like a cassette machine playback head, to get response @1 7/8 going up to
maybe even 15 KHz. 15 KHz @ 1 7/8 corresponds to 60 KHz at 7.5 ips and
120
KHz at 15 ips.

So find an open reel machine that played 1 7/8, shoehorn in some cassette
heads and transcribe away!



Yeah that might actually work - I dont know how old the machine was, my
akais have narrow gap glass heads but the bias is 120kHz and it won't run at
1.825. Actually I can make my 630 run at that speed, but it isn't equalized
for it and it's unstable. The problem with most reel machines is you double
the wow and flutter when you halve the speed, unless they are direct drive.
You could copy up to higher speed tape to slow it down. Say take a 3 to 15
then drop back to 3. Then I guess you would have to have some way of
continuously varying the sample rate based on the pitch change. What happens
when the flutter and the wow are different frequencies, which they should
be. I guess you could use multiple passes correcting every periodic you can
find. I think i would just use a pitch control and try to fix gross pitch
problems by ear and call it a day.
Carl

On Wed, 22 Sep 2004 09:07:43 +1000, "David White"
wrote:

Thanks for the responses. There are some promising ideas there. Locking onto
a buzz or the bias would do the trick very well.

I was going to suggest using the bias signal, but had assumed that it
was too high in frequency to get recorded. Is it in fact recorded?
Would it need a special playback head to retrieve it? If possible,
this would seem the ideal system.

On Tue, 21 Sep 2004 22:37:20 -0400, (Barry Mann) wrote:

As you work with the tape, try to tie the speed changes with a tape
recorder characteristic. For example, if a reel bearing is tight or the
tape is rubbing a reel, then speed changes will be tied to reel
rotation (remember that the tape on the reel is essentially a spiral of
Archimedes and the spacing between the speed changes is tied to the
current pack radius.) If the supply reel bearing is tight, then the
tape may slow down as the pack shrinks.

"Warbly" suggests that the speed changes are fairly rapid. If this is
true, then your speed control system will need to be able to respond
accordingly.

I doubt anyone was contemplating a MECHANICAL speed-correction system
:-) But it's an intrieging idea!

In , on 09/22/04
at 02:36 PM, "David White" said:

[ ... ]

Actually, someone else has the tape, and he'll probably send me a wave
file of the sound as played on his equipment. He's already played the
tape on a number of machines and found that certain problems, such as
low volume and the warblyness, vary between them. He's even found that
it sounds better if he doesn't put the tape all the way in on one
machine. So there's hope yet that he can play it on something that it
sounds okay on. That would be preferable to trying to fix it later.

[ ... ]

I wonder if his playback deck is complicating the issue.

-----------------------------------------------------------
spam:
wordgame:123(abc):14 9 20 5 2 9 18 4 at 22 15 9 3 5 14 5 20 dot 3 15
13 (Barry Mann)
[sorry about the puzzle, spammers are ruining my mailbox]
-----------------------------------------------------------

On Wed, 22 Sep 2004 09:07:43 +1000, "David White"
wrote:

Thanks for the responses. There are some promising ideas there. Locking
onto
a buzz or the bias would do the trick very well.

I was going to suggest using the bias signal, but had assumed that it
was too high in frequency to get recorded. Is it in fact recorded?
Would it need a special playback head to retrieve it? If possible,
this would seem the ideal system.


The bias signal is on the tape and i have been able to hear it on machines
played back at very slow speesd such as while scrubbing on a reel-reel.

Recovering the bias is another thing altogether.

Richard H. Kuschel
"I canna change the law of physics."-----Scotty

"Richard Kuschel" wrote in message
...
Recovering the bias is another thing altogether.

It may be possible. If the record head can get the bias frequency onto the
tape, it should be able to recover some of it. You could feed the signal to
a separate amp with band filtering to the desired frequency range.

TonyP.

"Laurence Payne" wrote in
message
On Wed, 22 Sep 2004 09:07:43 +1000, "David White"
wrote:

Thanks for the responses. There are some promising ideas there.
Locking onto a buzz or the bias would do the trick very well.

I was going to suggest using the bias signal, but had assumed that it
was too high in frequency to get recorded. Is it in fact recorded?

Yes, and there are accounts on the web of commercial audio applications of
bias recovery for the purpose of speed correction.

Would it need a special playback head to retrieve it?

I believe that the specific example I am thinking of used a special playback
head. However, playing the tape at a low speed could work as well.

If possible, this would seem the ideal system.

In the cases I'm thinking of, the demos showed dramatic improvements.

After some searching around, I found the web page I was thinking of:

http://www.plangentprocesses.com/


Here's the RAP thread that brought this to my attention:

http://groups.google.com/groups?selm...nix2.panix.com

Credit to the encyclopedic audio knowledge of Scott Dorsey (again!). ;-)