Restoring at double speed...

[cb831]

Hi

I have some old tapes recorded at speed 4.75cm/s and my tape deck is only capable of 9,5 and 19cm/s.

I consider digitizing at a 88.2ksamples per second and replay at 44.1ksamples per second to compensate for the double speed.

In other words I do not want to convert samplerates I just want the player to believe that the 88.2 stuff should be played @ 44.1.

How do I do this in DC6 ?

Claus

[Craig Maier]

It sounds like you want to correct the speed of the tape transfer which can be accomplished with the change speed algorithm. Using that routine, no matter what sampling rate that you recorded at, the resultant file can be as high as doubled or low as halved in terms of speed. In your case, I think that you want to record the tape at 9.5 cm/s and then half it using that routine.

Does that help?

[cb831]

But don't Change speed recalculate new samples from interpolation ?
What I want is that the 88.2 file is played at 44.1 with no change to the actual samples. Maybe it is just a chage of header for the wav file ?

Claus

[Craig Maier]

Is that not the goal? Essentially, that is what the speed change filter does plus more. If you just change the file header, you will be introducing digital artifacts and harmonic distortion into the result. The change speed system accounts for that by using the appropriate brick wall anti-aliasing filter and interpolation into the signal chain in order to do the job correctly.

At any rate, that is how it is done with Diamond Cut Software. If you want higher performance, just transfer at a higher bit depth (use 24 bit rather than 16 bit). Or, use 176.4 KHz Sampling Rates (or 192 KHz) and allow the software to do its job on the conversion. Or, do a combination of each.

In either case, you will never really hear much of difference no matter what sampling rate you use, since your source material is a R-R tape recorded at such a slow speed. The quality of your source recording is going to be marginal at best because of its format (3 3/4 ips, reel to reel). That is the limiting factor, not the sampling rate or the method of down-conversion.

[Dan McDonald]

Yes - I've gotten very good results doing this (when you consider the source). I have several tapes recorded at 3.75ips, and my current r-r recorder only plays 7.5 & 15 ips, so I have to transfer at a higher sampling rate and then cut the speed in half.

There is a thread on the forum somewhere where Craig talks about the problems with this kind of transfer. It goes something like, if the original recording went up to about 15kHz at 3.75 and you transfer using an open reel that can reproduce at, say, 20kHz, then whatever was at 15khz is not going to get reproduced in the transfer. You're essentially limited to around half the original high-frequency material, if that material is above what your playback unit is capable of reproducing.

Did I get that right? Anyway, there's a thread on here that talks about this issue. The bass is not affected because you're moving it into the area that is reproducable - e.g., 50 hz becomes 100 Hz.

[Craig Maier]

Dan,

That sounds about right; it depends on the performance of the playback deck. I believe that my Tascam deck is capable of a little more than 22 KHz @ 7.5 ips with a flat response, so the situation would not be quite as bad as you describe given high grade equipment. On the other hand, most 3.75 ips recordings do not have much more than around 10 KHz of top end bandwidth* to start with before radically dropping off, so the problem is more limited by the recording than the playback equipment, unless the playback equipment is marginal in quality or not in good condition (worn playback heads or in need of a good alignment). Also, I agree that the bass-end would not be adversely effected.

*Note: One could ask why Cassette Tapes which only use 1 7/8 ips have bandwidth up to around 14 KHz when R-R's using double that speed are limited to around 10 KHz. The answer is that the tape head gap width for Cassette decks are smaller than that of R-R machines. Somewhere on this forum, I wrote something up about the relationship between playback tape head gaps and frequency response.

I just found the link - - - here it is:

http://www.diamondcut.com/vforum/sho...uency+response

[cb831]

Thanks for the feedback.

I did read that thread before asking and yes my frequence range on the tape is already limited so thats no problem.

I asked because I thought that using the raw 88.2 sample at half the speed would introduce less errors which apparently is not true (I have to find my old study books on DSP ).

Thanks
Claus

[Craig Maier]

Using 88.2 KHz vs 96 KHz just makes the calculation of the least common denemonator a bit easier; since a computer is doing it, the penalty is just a bit more processing time.

Personally, I would make that kind of transfer at 44.1 KHz and run the change speeed routine at the baseband rate. I am confident that no appreciable signal loss would occur on the subject material vs. operating at higher sampling rates. But, if you have any doubt about it, just do the transfer and speed conversion at 88.2 or 96 KHz sampling rate.

[Dan McDonald]

Cb:

This is the thread I was talking about. It was a post by Craig about minimizing aliasing artifacts. You won't hear the difference, but there is a measureable difference:

"Although the DC6 Sample Rate converter produces exceptional performance, the best results will occur when you work at integer multiples of the final products sampling rate. In other words, if you plan to make a CD of the material in Red Book Audio format, that will ultimately be a 44.1 KHz sampling rate. If you want to transfer and operate the software at a higher sampling rate, 88.2 KHz or 176.4 KHz are optimal rather than 96 KHz or 192 KHz. I will mention that these differences are only measureable but not readily observed in listening tests, nonetheless, there is a slight difference here.

...
Again, following this guideline is really nitpicking and certainly nothing to loose any sleep over if you do not follow the recommendation. Your ears will not detect the difference; only a sensitive THD meter will. But, if you want every last ounce of performance from the system, the outlined process will help."

The whole thread is called Minimizing Aliasing Artifacts when Downconverting, from March, 2006, but this is the main information in it.

[Craig Maier]

I think that you can run the experiment yourself if you would like. The Diamond Cut % THD meter is about as good as they get, so you can try it both ways on a "Make Waves" sine wave and compare the various rates. You will measure a small difference in % THD when using non-integer multiples of the final sample rate. As I recall, the reason for this delta are the rounding errors that occur in the math.

[cb831]

Hmmm,

My process is actually an exact integer conversion 2->1 so I'm not certain that I understand why my procedure:

1. Sample tape @ 88.2
2. Play it @ 44.1

introduces more artifacts than:

1. Sample @ 88.2 ( n*44.1)
2. Change speed 2->1 (-50%)
3. Convert 88.2 -> 44.1
4. Play it @ 44.1

I mean sampling my tape @ 2x speed @ 88.2 rate should produce the same samples as 1x speed @ 44.1 ?!?

Claus

[cb831]

If I go in and divide Sample rate and Average bytes per second in the WAV file header directly I get a very good result

[Dan McDonald]

Sorry -

I just re-read your post. This one:


"I asked because I thought that using the raw 88.2 sample at half the speed would introduce less errors which apparently is not true."


I thought you were comparing 88.2 to 96kHz sampling rates, not 88.2 to 44.1.

We were talking about different things. I'll bow out of this one because I simply don't know.

[Craig Maier]

Since you think that you have the answer to your own question, then why not just use your own process? Why are you bothering us asking questions about it? Why are you wasting our time; you seem to have all of the answers anyway.

Note: When you are transferring something (a low quality reel to reel tape in this case) that contains 5 percent harmonic distortion to start with and then you are debating the merits of processes that add 0.005 % vs 0.007% harmonic distortion, then I am completely miffed.

What is your point?

ps - if you record at 88.2 and run the Diamond Cut speed change routine to half the playback rate, it performs the routine in a similar manner as yours, except that it also provides the proper brick wall filter needed for a proper down conversion - - - a step that you are missing.

I am closing this thread.