Pitch correction methods

[paul62]

maul
Feb 20, 2020 at 9:55pm

Hi Paul, there's a question i always wanted to ask you: how do you calculate the exact amount of pitch correction a track needs? I always adjusted it by ear, but on BZ i noticed you often indicate it with millesimal precision

thanks

[paul62]

mojofilter
Feb 21, 2020 at 3:34am

The way I've done it for years: it requires good relative pitch, and the ability to tell when the pitch playing back is sharp or flat of a reference tone or instrument tuned to A=440 Hz. You play the recording off the drive and mix with it the sound of your digital keyboard, playing along to determine if the source is fast or slow. Then you highlight a selection with at least one chord change - A to D or E to B, and play it along with your keyboard, changing the value each time by increments, until your piano and the record match in pitch. Then you can resample the track to the pitch that matches your piano and have it in concert pitch. It ceases to matter whether the recording on the vinyl record it played from ran fast or slow, or the turntable ran fast or slow, or whether the various tape decks the tape has been played or dubbed on and then played one last time for its digital transfer ran the same speed (chances are good that they did not).

If there is now software that can calculate the degree to which a recording is playing back off-key, and where to adjust it to match concert pitch, I have no knowledge of it. But my way works unfailingly.

[paul62]

blackbird
Feb 20, 2020 at 5:49am

how do you speed matched with another source?

I would like to try to AR 2019 Remix speed match with 2009 Remaster or CP35-3016.

[paul62]

paul62
Feb 21, 2020 at 10:12am

Great questions!! I know that a lot of the mono versions of Beatles tracks have different speeds to the stereo versions, for example, so if I want to speed-match one version of a track with another version of the same track, I will open up the first track in Audacity and then import the other version (which sits underneath the first track I've opened up). I will look for a drum beat near the start of the two tracks and then carefully cut away the excess that comes before the beat: the beat will be the exact start of each track (and will be the same beat for each track!). I will highlight the track I wish to speed up and then it really becomes a matter of trying different rates of speed correction until you come upon the right rate. As I come to the point where the tracks more or less correlate, I will zoom in and finesse the correction until they match (and I will listen to the tracks play as well: sometimes I will have the left channel of one open and the right channel of the other open and get a pleasing new stereo mix!). Once you have found the percentage rate of speed-correction, the track being corrected will be opened again, in full and on it's own, with the rate of correction being applied to that.

If you want to get, for example, a BBC version of a Beatles song to match the pitch of the EMI version, you can open both in Audacity (one under the other, as above) and try different rates of speed-correction on the BBC version until the two versions sound like they are more or less in the same pitch. Having good "relative pitch" helps a lot when doing this!! Because two different performances are being matched to each other, the listening experience during the matching process isn't so pleasant (the drum beats won't be matching, for example, when you are trying this!).

[paul62]

mojofilter
Feb 21, 2020 at 3:34pm

If you're tuning a track on one CD to the length of the same track on another CD and there is a discrepancy, how do you know which one is in the correct pitch? Your formula turns out to have everything to do with mathematics, but nothing to do with music. You are making lengths match, but not pitches. What if the first tape was off? What if the machine they transferred it to digital with in 1987 or 2009 ran a slightly different speed slow or fast from the machine the tape was recorded on in the 1960s?

In Recording Sessions, Mark Lewisohn quotes an EMI engineer who said that on cold winter nights, the speed of the EMI tape recorders would slow down because of the drain on the electrical grid from heaters in peoples' homes consuming so much power. This means that while the instruments were most likely in tune with the EMI studio piano when they were recording, the tape speed would be slow, or gradually slow down, and the next day and thereafter, it would play back at a higher speed when the electricity level was back up to normal, and be sharp of concert pitch. This does not mean that they meant for it to be sharp. It is a technical fault. There was no such thing as quartz-lock for tape drive motors then. There are hardly any two tape decks side by side that will run at exactly the same speed unless they have quartz-lock. I've worked with recording equipment since the 1960s, and I've never encountered any that would match.

Let's imagine Paul's Bluthner piano on "Let It Be." You may or may not know that "Let It Be" is played in C. What if you took this track from two different discs and put them in Audacity and the length you made both tracks was based upon one being sharp or flat, but you weren't sure which? You would then have two tracks matched for time that are between two keys on the piano, to some degree. If you have no concert-tuned musical reference tone, how can you know if it's playing at the right speed?

We can count on one thing: music with concert grand pianos after the 435 Hz standard was abandoned in 1936 is tuned to A=440. All modern instruments like synthesizers and electric pianos and organs and clavinets etc. are made to adhere to A=440 tuning. Any music containing pianos and/or orchestral instruments made after 1947 when German tape recorders were redeveloped and improved for the US recording industry, was recorded at A=440 but may not play back there due to unregulated speed of tape deck motors. In a perfect world, the studio machine would have recorded at exactly 15 or 30 IPS; the deck that transferred the tape to the record lathe or digital recorder was dead on 15 or 30, the lathe was running at exactly 33.3 or 45 RPM, and the turntable you listened to the record on at home ran the same. CD playback speed is governed by the 44.1 KHz sampling rate. If you send a recording to digital that plays fast or slow, that's how it stays (until you resample it) (Hello, 99% of bootleggers!). But it didn't work out that way. That's why I've been fixing the speed of record and tape playback for 30-odd years.

After the advent digital recording, everything is in A=440 unless the instruments are deliberately detuned, and pop music isn't normally detuned. It's played in A=440 and locked there for all time by the digital recording standard. You can't mix sampling rates in the same recording. This is why Frank Zappa could no longer appear to sing in a bass range not achievable by humans, or speed up singers to near-chipmunk range when played back at normal speed. You could speed up or slow down tape recordings and record on one track out of 8, 16 or 24, to give that certain effect at normal speed, but you can't change the pitch of anything on a computer without resampling the whole recording. No mixing of rates is possible.

If you can't play your (especially digital) piano along with a piece of recorded music without them clashing in dissonance, that music is playing back at the wrong speed. I'm curious about anyone's thoughts on that.

[paul62]

blackbird
Feb 21, 2020 at 3:48pm

uh, I thought there is magical software to correction pitch.
Maybe I did it in same way.

[paul62]

mojofilter
Feb 21, 2020 at 3:54pm

If there was, I would be using it. According to the Wikpedia article on pitch correcting software and plugins, these are pretty much exclusively used on vocal recordings, to make a singer who can't sing on pitch, sing on pitch. I can find nothing about something that will correct a piece of music that plays back sharp or flat and return it to concert pitch.

[paul62]

paul62
Feb 22, 2020 at 9:30am

The Capstan software program enables music to be speed-adjusted to A=440Hz: this software is an off-shoot of the developer's Melodyne software program. The program can also determine an average pitch for the piece of music being restored, so there is a chance to make a choice.

[paul62]

paul62
Feb 23, 2020 at 12:26pm

Here's something I posted at a few places just over a year ago, ideal if you want to create yourself slightly different stereo versions of the '64 tracks from the 2009 remasters: the huge span and variety of speed differences was quite astonishing to discover and could be accounted for by performances being captured by two different tape recorders (a mono "delta' tape recorder and a stereo tape recorder).

I was checking the 2009 remastered mono versions against the 2009 remastered stereo versions of the "A Hard Day's Night" LP tracks, the "Beatles For Sale" LP tracks, the "Long Tall Sally" EP tracks and the "I Feel Fine"/"She's A Woman" single tracks: in all but one case, the stereo version runs substantially faster than the mono version.

When the remastering team at Abbey Road put together the remastered sets in 2009, they used tape equipment that was recently serviced and took great care (one song at a time was transferred to digital with the tape heads being cleaned after each song was transferred). I understand that the original stereo master tapes and the original mono master tapes were used for the '64 remastered editions, so any speed variations must have been "baked into" the mixes back in '64 (except in the case of "She's A Woman": the mono and stereo mixes more or less correlate).

If you are one of those people that believe that the mono version of a Beatles track is the "official" or "correct" or "right" version then you can use Audacity (which can be downloaded for free) and make the speed corrections needed to make the stereo version "right" or "correct". I'll list each track and give the speed correction percentage needed to bring each stereo track "down to speed" to correlate with the mono version:

CAN'T BUY ME LOVE (-0.730%)
YOU CAN'T DO THAT (-0.925%)
I SHOULD HAVE KNOWN BETTER (-0.735%)
AND I LOVE HER (-0.511%)
TELL ME WHY (-0.835%)
IF I FELL (-0.655%)
I'M HAPPY JUST TO DANCE WITH YOU (-0.664%)
LONG TALL SALLY (-0.164%)
I CALL YOUR NAME (-0.628%)
A HARD DAY'S NIGHT (-0.712%)
MATCHBOX (-0.075%)
I'LL CRY INSTEAD (-0.691%)
SLOW DOWN (-0.256%)
I'LL BE BACK (-0.561%)
THINGS WE SAID TODAY (-0.454%)
WHEN I GET HOME (-0.529%)
ANY TIME AT ALL (-0.604%)
BABY'S IN BLACK (-1.137%)
I'M A LOSER (-0.824%)
I DON'T WANT TO SPOIL THE PARTY (-0.783%)
EVERY LITTLE THING (-0.383%)
NO REPLY (-0.422%)
EIGHT DAYS A WEEK (-0.450%)
SHE'S A WOMAN (-0.012%: minimal difference)
KANSAS CITY/HEY HEY HEY HEY (-0.626%)
MR MOONLIGHT (-0.647%)
I FEEL FINE (-0.591%)
I'LL FOLLOW THE SUN (-0.788%)
EVERYBODY'S TRYING TO BE MY BABY (-0.731%)
ROCK AND ROLL MUSIC (-0.681%)
WORDS OF LOVE (-0.630%)
HONEY DON'T (-0.434%)
WHAT YOU'RE DOING (-0.564%)

[paul62]

LesBaxter
Jul 7, 2020 at 3:55am

Pitch correction is the wrong way to look at it. You need speed correction. Pitch and speed are locked together in the analog domain.

In the digital domain if you manipulate the digital sample length of a recording to match another so that the beginning and end are in sync then pitch will automatically follow.

Altering pitch and tempo as separate items by ear if attempting speed correction is not the accurate method that preserves the speed/pitch ratio IN MY OPINION.

As Paul points out there's not really a correct speed anyway with Beatles recordings at least with what was released and ended up on record. I suppose you COULD digitally bring everything into concert pitch but then you are changing history and what people have heard for 50 years.

If they recorded a song one day and the British line current was running at AC 49.8 cycles and then during mixing two weeks later the British line current was running at AC 50.2 cycles then you are going to have speed differences.

Personally I think a lot of the AHDN album is sped up but I can't prove it without software that finds where the notes are actually sitting. Someone should work on that in a scientific fashion. No way in heck can "Any Time At All" be the actual preformed speed. The voices are soo chirpy. Slow it down and John & Paul sound natural. Now at the time they did not have the varicycle method (as far as we know) but I wonder if they used the slightly off size capstan method that was done since the 1950s. Yes, those great rock and roll songs we know and love were pepped up a little.