Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Sample rates and digital speed correction (Read 3229 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Sample rates and digital speed correction

Will recording at an ultra high sample rate provide any significant benefit to the quality of moderate speed corrections in the digital domain?

I've been doing all my analog transfers at 48K, but I was wondering if sampling at 96K (or perhaps 64 or 88K) would provide any real advantage here?

I know people say that high sample rates are good to use with extreme slow-down effects and such, but the amount of pitch shifting I'm looking at will usually never be more than a couple semitones (+/- 10% in either direction).

My thought was that having more samples available would make the pitch correction more accurate. Is this true?

If yes, keep in mind that I'm not doing anything too extreme here, so where does one draw the line as to how much sampling is necessary for the maximum benefit in this situation?

Sample rates and digital speed correction

Reply #1 –
As no-one else has replied so far, I will stick my neck out.


Analogue transfers: Perhaps from vinyl? Perhaps from open reel tape?

I presume the original material may have been recorded a little fast or a littler slow but the instruments were tuned to standard pitch, so by adjusting the speed to hear standard pitch, you hopefully will get the originally performed tempo, pitch, and timbre. 

If the analogue material had a very limited treble response (e.g. old cassette tapes) I'd imagine 48KHz would probably suffice for the reprocessing.  If the treble range was more extended, I'd be inclined to transfer at a higher rate such as 96KHz.  Performing a resampling to effect a speed and pitch change will slightly compromise quality but the higher the sample rate at which the resampling is performed, the less will be the potential impact.

You could still render the resampled high bit-rate digital file at 48KHz for the final product, if that was preferred for consistency with other transfers, or to save on file size.

So for example, you could do the analogue to digital transfer capture at 96KHz, drop the pitch a semi-tone (would involve a notional custom sample rate of about 101,708Hz in whatever software you are using for the combined speed and pitch change).  The new 96KHz version could then be converted to 48KHz to match other transfers.

Would use of a higher (intermediate) sample rate be a significant benefit?  I'd say it would be a marginal benefit only, and it would depend a great deal on the source material being transferred and how carefully one was intending to listen to the final digital product.

ABXing would be a little difficult but if you wanted to test, it could be done indirectly.  For example, you could resample the original 48KHz file to 44.1KHz and then play it at 44.1Khz and compare with the 48KHz version at 48KHz.  If the resampling was having no discernable impact, the two files might sound exactly the same (depending a little on the exact characteristics of the sound card at different sampling rates).

I note it can be very difficult to hear any difference from conversion to 48KHz of a recording made at 96KHz, as illustrated in this recent post in the thread  Results for 24bit/96KHz test, vs. 16bit/44.1KHz.

Of course, all processing should be at a high bit depth, even if the final output is dithered down to 16 bits.

Sample rates and digital speed correction

Reply #2 –
What MLXXX says is correct.

For example, if your source is 10% too fast, and you properly sample at 48kHz, you will get the right sound by upsampling the resulting samples from 43.2kHz to 48kHz. This means that frequencies below 21.6kHz present in the original source will be lost. If the source is 10% too slow, and you properly sample at 48kHz, you will get the right speed by downsampling the resulting samples from 52.8kHz to 48kHz. Here, frequencies above 24kHz will be lost.

Allowing for the resampler passband of 85%, you will lose frequencies above 18.36kHz during a 10% slowdown. Most resamplers can do better - so as long as you are using a high quality resampler/speed converter 48kHz is plenty for a 10% change. Of course, using a higher rate while processing probably isn't a bad thing to do, but not critical in this case.
 

Sample rates and digital speed correction

Reply #3 –
Quote from: MLXXX on
Analogue transfers: Perhaps from vinyl? Perhaps from open reel tape?

I presume the original material may have been recorded a little fast or a littler slow but the instruments were tuned to standard pitch, so by adjusting the speed to hear standard pitch, you hopefully will get the originally performed tempo, pitch, and timbre.


These are cassette transfers. And yes, adjusting the recordings to standard pitch was the idea I had in mind. I know it's technically better to adjust the speed during playback, but not all high end decks have this feature. My current one does, but as I wrote in another post on here, I've become dissatisfied with it due to a strange technical issue.

I really doubt any of my tapes are more than a full step out of pitch (maybe a few extreme cases). And most of the tapes I recorded myself are close to perfect, so I really wouldn't expect to do much extreme shifting there. I was basically hoping that by recording at 48K I could at least retain "44K quality" after pitch shifting. So it seems that this is a reasonable hope?

Recording cassettes at 96K seems a little extreme to me, but the potential for higher quality speed correction has made me consider it. My sound card can also record at 64K, and it seems this might be an ideal solution in my case. But something just feels "weird" about using such an obscure sample rate. Anyway, that's probably just me...

Thanks to everyone who's replied so far.