Salvaging off-pitch recordings? Options

[stephan_g]

I have the following problem:

At one time I digitized a handful of records on my slightly fast-running (about 1.5%) vinyl setup. I recorded these in 24/88, downsampled to 44.1 kHz, cut and dithered to 16/44. The resulting FLACs are all I have now.

Does anyone have a smart idea re: how one might fix that without too much work or loss of quality? I suppose I'll have to combine the individual tracks again (no big deal) and re-cut afterwards (more annoying) to avoid artifacts between tracks? And then use some kind of resampler and mess with sample rate in the WAV header?

What I have in terms of tools: Foobar2000 1.1.11 w/ SoX resampler, SSRC 1.30, Audacity 2.0.1-alpha. Anything else that might be of use?

I will have to re-record the better-sounding stuff on a better setup anyway, but I figured that the present quality would do fine for CCR and such.

This post has been edited by stephan_g: Jun 23 2012, 11:34

[dhromed]

I was going to recommend Playback Rate Shift here:
http://www.foobar2000.org/components/view/foo_dsp_effect

But it only does whole integer %. Nonetheless, it's by the simplest solution and maybe it's good enough for you.

[Apesbrain]

Using foobar2000 you may not even need to combine the tracks. Pick one album, drop the tracks on fb2k's playlist, highlight them and run the component* dhromed suggested. Afterwards, you may want to use Traders Little Helper to check for any SBEs.

*Convert the files to FLAC with that particular DSP in the chain.

This post has been edited by Apesbrain: Jun 23 2012, 13:36

[bandpass]

CODE

sox track1.flac fixed/track1.flac speed 0.985

The speed change is performed internally by SoX using its well-known high-quality resampler. No restriction on the resolution of the speed change.

Had to do this once myself for the same reason. I had one of the tracks on a "Best of" CD, so I used that to precisely gauge the correct speed adjustment.

If the tracks are segued, then rejoin, change speed, recut; otherwise, working with individual tracks should be fine.

[lvqcl]

Another way:
sox -r 43440 input.flac output.flac + resample back to 44100 (or to 48000, or anything else) at playback time.

[stephan_g]

Thanks for all the input.

What I did now was:

Use Foobar's Convert function to resample to 44760 Hz via SoX resampler (making sure that "Reset DSP between tracks" is unchecked, output in 24 bits, WAV format this time but FLAC would also work). This yields individual tracks with no glitches.

Import the whole bunch into Audacity.
Select sample rate of 44100 at all tracks (they gain a bit of length). Change project sample rate to 44100.
Export multiple tracks as individual files. 16 bit w/ ATH dither.

Drag result into Foobar, copy back correct metadata, RG scan.

Ta-Dah!

[Apesbrain]

Just out of curiosity, can you tell me how you set SoX Resampler to that value? My version of the component only has preset values. Thanks.

Later: Got it! I never realized you could just type a random number into the box.

This post has been edited by Apesbrain: Jun 23 2012, 16:25

[Arnold B. Kruege...]

I was going to recommend Playback Rate Shift here:
http://www.foobar2000.org/components/view/foo_dsp_effect

But it only does whole integer %. Nonetheless, it's by the simplest solution and maybe it's good enough for you.

If you are off 0.5% then you are within the limits of human perception.

That justifies the use of products that work to the nearest percent, since you can always tune them to get residual error <=\ 0.5%

[stephan_g]

Still, 0.5% is hardly "perfectionist-proof" when the just noticeable pitch difference has been established to be about 0.3%. When I tweaked my record player's speed, I obtained about -0.1% and decided that I'd be happy with that.

Going through my recordings, estimated speed offsets indeed were about +1.5 to +1.6% (even back when I did the first ones in '99) - except for one. That one started out at +1.0%, then by the end of side 1 (GDR pressing...) it was at 1.2-ish %, and near the end of side 2 it was almost 1.4% fast when compared to a CD issue. Now speed on my trusty Technics, entry level as it may be, has been quite stable in my experience, certainly within +/- 0.1% from average. Since we are talking about an album with a slightly altered track list from the mid/late '70s that was edited on the Eastern side of the iron curtain, I am guessing that some tape machine had an issue that made it run slow, slowly coming up to proper speed as it was used.

[Porcus]

except for one. That one started out at +1.0%, then by the end of side 1 (GDR pressing...) it was at 1.2-ish %, and near the end of side 2 it was almost 1.4% fast when compared to a CD issue. Now speed on my trusty Technics, entry level as it may be, has been quite stable in my experience, certainly within +/- 0.1% from average. Since we are talking about an album with a slightly altered track list from the mid/late '70s that was edited on the Eastern side of the iron curtain, I am guessing that some tape machine had an issue that made it run slow, slowly coming up to proper speed as it was used.

Could it be friction in the cutter? Brake force is applied over a circumference proportional to the radius. Work = force x distance = constant x mass x radius.

[stephan_g]

Then the speed profile would have been different. Here the end of side A was about as fast as the beginning of side B.