Cornet Chop Suey: a recording in two keys

There is a controversy regarding the correct mastering speed - and therefore the correct key - of Louis Armstrong’s classic 1926 recording “Cornet Chop Suey”. One Columbia/Legacy release has gone so far as to include two masterings of it, one in the key of F (with a length of 2:58) and one in E flat (with a length of 3:18).

The guidelines say that “variations in playback speed” is not a reason to keep separate recordings. However, it seems to me that this case, with a 20 second difference and a whole tone between them, should be an exception to that.

There’s also a question of what to do with the many copies of “Cornet Chop Suey” currently in MB as I am trying to clean up some of the Armstrong catalog. It seems reasonable to me to assume that a recording that is within 2-3 seconds of the 2:58 time could be merged with the “key of F” version, and likewise for a recording within 2-3 seconds of the 3:18 time merged with the Eb version. (This also assumes that the recording can be identified as the 1926 Hot Five performance, and not one of Armstrong’s later recordings of the tune.)


There is one demi-tone between E and F.
I don’t really dig the fact that we should merge recordings when different speed.
But at least we should keep recordings with one demi-tone difference, or more, separate.

PS. BTW, I didn’t know flat was bémol so between E bémol and F, there is even one full tone, as you said.
But I think one demi-tone would also be a good reason to keep separate.

I queued two edits to the F key version as it seems to be presented by the compilation as the main version (no version names):

If you want to vote them or propose better edits.


I absolutely love the education I am getting from the posts by all the wonderful editors here in MB, This post is another example. While I have always noticed the tone variance in turntable speeds, it never dawned on me that recording companies would change the speed to change the key to “create” a new recording to create more sales.

So far this discourse is leading me toward separation. But would this separation simply be in the addition of a disambiguation or is more being proposed?

Edit: @jesus2099 Thanks for showing me this post is not about creating different recordings. I had skipped over the “controversy” link. Duh. Irregardless, MB eds are great.


That’s not exactly the case here:

In doubt (I have read the interesting linked article), a compilation included both, which is good, IMO.


@jesus2099 addressed something that I was thinking about but forgot to add to my pos - what is an appropriate relationship between the recordings. I guess the “edit of” is the closest we have, but it seems a little imprecise.

@Llama_lover this has been an education to me as well; speed calibration of the disc-cutting lathe isn’t something I’d thought about much. :slight_smile:

To paraphrase the Norman Field article that I linked above, the root issue seems to be that due to speed variations, the original 78 would actually play (on a properly calibrated turntable) more or less in the key of E, which is not a key used by early jazz bands. Moving it either up or down a semitone results in a more plausible key.

The fact that there are recordings in MB that split the time difference between the two - right around 3:08 - suggests that there are also releases mastered with no correction in either direction, effectively in the key of E. At some point, though, it becomes impossible to identify by time alone (and nearly impossible to conclusively identify even by listening, given the possibility of infinite variations in mastering speed).


If separate Recodings are being created for the keys either side of the uncorrected Recording then it would seem better to have a Recording for the uncorrected version too.
Which would give us 3 Recordings.
A graph showing the tempo of known Releases would be interesting to see - hopefully 3 discrete groups would be shown.

Edit: Speed calibration, especially in playback, has been used to criticize the fidelity of piano rolls. Others have been prepared to discount the skills of the artists to explain the level of musicality now heard.


In principle I agree with this, but I’m concerned that the groupings will not be a discrete as we might like.

1 Like

I have merged several recordings clocking in at 3:08 (or within 1 second) to represent the “unadjusted” recording. I chose 3:08 as that is the midpoint between the lengths of the known adjusted recordings.

I have then changed the “E flat” and “F” recordings to both be edits of this unadjusted version: