Topic: Trouble mixing 2 sources (out of sync very quickly)

[Fatah Ruark (aka MIKE B)]

Trying to do a AUD/SBD mix of a Phish show and the 2 sources are wildly out of sync. Back when I used to mix 2 of my own sources (with different decks) I'd get mild sync issues after about 15-20 minutes. These 2 sources are a second off after 5 minutes.

My 2 sources are a 16/48 SBD and a 16/44.1 AUD. I converted the 16/48 SBD source to 16/44.1 for mixing. Both the 16/48 and 16/44.1 SBD sources are the same length after converting so it seems like they converted correctly.

I've included both 16/44.1 files (just track 1 of the show) and also the original 16/48 SBD file.

https://drive.google.com/drive/folders/1iRznQITlSljxHoPlJ2CJc1-2ZmjdtOSG?usp=sharing

Any idea what the problem is and more importantly how to fix? And also how do I tell which source is the correct speed?

[AbbyTaper]

I've done a lot of matrices...and that's by far the worst I've ever seen!  First, I don't think your starting points are exactly aligned (not bad, but not 100% either), so you get off to a bit of a bad start.  But then obviously either your SBD source runs fast or your AUD source runs slow, since as you say they drift apart quite quickly.  You would likely need somebody who has a musician's ear and/or is intimately familiar with Phish's music to tell you which is the case.

[breakonthru]

its not always just clock drift

if one source goes "faster" suddenly and not at a constant pace their could be dropped samples in the stream. what kinds of recorders are you using and have you tested their operation for recording without dropping samples?

[wforwumbo]

The issue isn't clock drift, it's also in the resampler. There is a constant linear phase offset with modern resampling algorithm that is cumulative over time. What you are hearing is something called the "group delay" of a resampling filter.

The only fix for this is to use a time stretching or compressing algorithm. Having done it myself many times, it's frankly usually more trouble than it is worth to do cleanly unless you are getting paid to do so.

Edit: also, both sources are "the correct speed" - as to which is a correct depiction of the speed the show was performed at, I would put my money on the recording operating at its native sample rate.

[Fatah Ruark (aka MIKE B)]

The issue isn't clock drift, it's also in the resampler. There is a constant linear phase offset with modern resampling algorithm that is cumulative over time. What you are hearing is something called the "group delay" of a resampling filter.

The only fix for this is to use a time stretching or compressing algorithm. Having done it myself many times, it's frankly usually more trouble than it is worth to do cleanly unless you are getting paid to do so.

Edit: also, both sources are "the correct speed" - as to which is a correct depiction of the speed the show was performed at, I would put my money on the recording operating at its native sample rate.

Thanks. I suspected it was something strange and difficult to fix. The SBD was recorded to cassette, so I wonder it that may be the culpret (since there is no clock).

I will have to try to find another source of this show (1993-07-25) to figure out if either of these sources are accurate time wise.

This was my first show where a recording exists, so figured it would be fun to make a better sounding source. Both are more than adequate as is, so no big deal if I can't.

[breakonthru]

man i tried but its hard to get right. my only guess is the speed of the cassette is all over the map

[Gutbucket]

I've not listened, but..

The non cassette source is almost certainly the steady one, use that as the constant speed baseline.  If it's that much out of sync before 5 min in, you'd likely need to use a number of sync points just within the first song, shrinking/stretching the cassette source to fit between each. You might check to see if after the first song the drift becomes less egregious and somewhat more manageable. Perhaps reasonable to do if after that point you only need to sync at the start/end of each song. Still a lot of work.  How difficult depends on the tool you use, and your patience.

Like many others here I used to do this regularly when I was using two separate digital recorders, but that usually meant sync points at something like 30 min intervals or more. If both are constant speed, just pick one or the other to use as the baseline.  I used to select the shorter/faster one as baseline just so that I was shrinking the other to fit instead of stretching it, reasoning that slightly raises the hf limit rather than lowering it if the shrink were significant, although given the incrementally small speed/time changes involved it probably didn't matter.  Same process either way.

Musing..
Always wondered about checking pitch of some held synth or piano note to determine which had a better absolute time basis to use as the baseline file.. assuming the band to be in tune. Never did the math to determine what the time basis of a one cent pitch change equates to time-wise.  May not be significant.

The professional way of correcting tape speed variation is the Plangent Process which monitors the bias tone of the tape and adjusts speed on the fly to keep it constant.
https://www.plangentprocesses.com/