The Linear Canvas
This journal is about the wrongs and rights of the world, as I see them.

The Linear Canvas

Digital Re-Synchronization of Analog Recordings

April 29th, 2010 . by Alexander Fisher

After I finished working on my digitally reconstructed analog recording, Sorry It Came To This, I wanted to write down some of the steps I took to get to the finished project, or at least my thinking on a few of the variables. Unfortunately I was a little brain dead from all of the ‘cypherin’, so I put it off. After I processed my next analog restoration, Mean and Cruel, in a similar manner, I finally decided to write a few thoughts about the whole process, before I forget it all, again.

——–

The process of reconstructing these tracks is heavily software based and not that much to do with any one of my technical abilities. It isn’t like the old days when you needed better hand-eye coordination for cutting the tape and splicing it back together. Though the process is not completely dissimilar in it’s application.

I think many people could acquire the skills to reprocess analog recordings, with the automatic processing built into my Digital Audio Workstation (DAW), Cakewalk Sonar v8.5, under most circumstances. When I reprocessed Sorry It Came to This, I did use some of the automatic synchronizing capabilities of the software. When I began working on the song, using a MIDI drum controller, I recorded several drum tracks based on the rhythm of each progressive mix down during the original analog process.

——–

After determining the original tempo of the recording, I linked the internal drum synchronization to these original tracks via MIDI processing. I was able to reestablish a common beat between all of the tracks using automated audio stretching through quantization, beat grooves, and also I manually stretched or compressed portions of the audio signal when the software could not properly recognize whether it was a component of the beat or not. Doing this allowed me to replace the original MIDI created analog drum track with a MIDI drum set included with Sonar. Drum samples have improved greatly since these original recordings were made and are nearly identical to a real drum kit. The improvement in beat timing was evident as well.

The amount of synchronization that I had to do on these tracks is directly related to the recording method that I used in Yamaha MT-100 II the beginning. I would record the initial four tracks on my Yamaha MT-100-II multi-track cassette recorder. It was designed to record on chromium dioxide cassettes (CrO2) and I always recorded at the higher speed of 3 3/4 inches per second that it also featured. The reasoning behind using the higher speed is that it enhances the audio quality and reduces the tape noise in the recordings. In addition the Yamaha recorded with dbx noise reduction.

The aural quality allowed using the dbx process was similar in many respects to the quality achieved in a digital compact disc (CD) recording. An advantage would have been the constant stream of audio as opposed to the digital sampling inherent in the CD recording process. Some have commented to me over the years that dbx, because of its use of compression in its methodology, called linear decibel companding, had an effect on the aural envelope, that you could hear. Compression/decompression technologies can be audible if it is used improperly, but I believe that like the opponents of the compact disc originally who thought they could hear sounds going on and off 44,100 times per second, they were a bit delusional as well. It’s probably the same people that complained about the inferiority of the minidisc that are now completely satisfied with the sound of the much,much inferior MP3 player.

——–

The next step in my process was to mix the original four tracks down, to either one or two tracks. I would hook the Yamaha output up to a dbx encoder/decoder input. Then I connected a standard audio cassette recorder to the dbx unit’s tape loop. I would then play the mix of the first four tracks into the two track recording capability of the cassette recorder. As soon as the playback ended, I would stop the Yamaha and add a few inches of extra tape to separate the next recordings from the first. Then I would rewind the standard cassette and hook the output of the dbx unit to inputs on one or two of the Yamaha’s input channels, depending on whether the mix was stereo or monaural. I would then record the mix back to the Yamaha.

At that point I would either have two or three empty tracks to add more instruments or voices to the song. Generally I used a 90 minute cassette in the Yamaha and could get about three or four of these sessions per tape. That was usually enough space for the entire project. If I ran out of tape on the Yamaha, I used a second cassette tape.

That was not a process that I learned from anyone or from the Yamaha manual. I’m not really sure what my motivations were at the time. Prior to buying the Yamaha, I had been using standard audio cassette recorders to accomplish a crude multi-track system, bouncing tracks between them in a similar fashion. In addition I guess I always felt that somehow I would have that individual performance to re-do at a later time if needed, like after a year or so. I never thought that that later time would be 20 years later.

The process outlined in the manual was basically bouncing tracks around and mostly destroying the original recordings. I’m glad now that I did record in this unusual manner, because it allowed me to reconstruct those sessions as though I had somewhat the same recording capability that I have now, back then.

——–

On my earlier projects reconstructing old analog recordings, I originally theorized that solely stretching or shrinking (in time) each generation of mix down and their accompanying tracks, would result in synchronization of each of those generations. My reasoning for this is; I know that tape has a tendency to not move at the same speed at all times. There are many reasons for this, but one specifically, tension.

The total amount of tension the tape was under included things like which hub had the most amount of tape on it and how clean the tape transport was. Either variable could result in the tape moving at any speed. There was also a natural constant fluctuation of tape speed that was just normal. There was usually enough control of the speed by the player to allow the listener to not detect any great variations, usually. But small changes in speed over a long period can accumulate to large differences in playback time.

Another timing variable is that the mechanics of individual tape machines could vary even within a specific model of the machine. Tape machines could never be possibly set exactly at the same speed at the factory. Maintenance issues like dirty or worn out transport parts are also issues. In addition, many machines were used in many different ways, so while no two tape units were exactly the same even just after manufacture, they certainly are less so after many years of use.

DAW’s have many processing features. My particular DAW (Sonar) features automatic timing corrections. It also allows manual processes called slip editing and slip stretching. Slip stretching allows me to re-time an entire segment by stretching or compressing it so that it occurs over a different span of time. That was the process I originally identified that might have the most effect on the audio that I was trying to recover.

What I would think would be the final answer, was in fact only part of the solution. I found that if you tried to stretch the audio over too long a period, the beat would get lost and even though it would end at the right time, there would be too much variation in the original analog recording to allow a consistent speed decrease or increase. Using slip stretching on a shorter clip can work very well, but over the entire length of the recording, it is only part of the answer.

——–

Slip editing allows you to position a track in a spot where it is supposed to start. Because of the timing issues of the recording process originally, it’s hard to say where it will stop and most certainly will not stop where it is supposed to. That is where slip stretching comes into play. But the automatic process called Audiosnap in Sonar can bring a larger audio selection closer to synchronization.

I just wish that Audiosnap was able to get it perfectly, but unfortunately it does not, normally. As soon as Audiosnap is activated, the best thing to do at first is to increase the threshold setting for the transient markers. Audiosnap always wants to mark every single peak of any audio that it finds. Some of these peaks mean something. Others peaks are unusable beats and accidentals. If you find a transient you need to turn on or off, you can. But having too many that are on, can cause the audio to start and stop and produce a tempo unrecognizable to the original project.

Once all of the transient markers are on-beat and in enough quantity for Sonar to recognize where each note is to be played, a synchronized track should result. Sometimes I detect where transient markers really should be, by just counting 1-2-3-4 with the beat. That can be very time consuming on a large audio clip, but sometimes that is the only way to know which transient markers are needed and which are not. You just need to do this manually sometimes. The Audiosnap feature is great, but is only good at assisting the project’s completion and is not capable of doing all the work by itself.

——–

There are so many variables produced by the mechanics of the analog audio equipment, that no one setting or methodology can recover these recordings perfectly. Only labor and time can resurrect audio recorded in this manner. Fortunately it has become very much easier for me, each project that I have completed. Mean and Cruel took only a day where Sorry It Came To This took more than a week to complete. Sorry It Came To This was a longer song and had many more problems that I had to fix to align the tracks properly. But that experience allowed me to process Mean and Cruel much more efficiently.

I am not sure which old analog recording I will be tackling next, but aided by the experience of the previous two recording projects, it should go pretty rapidly, hopefully.

Leave a Reply

You must be logged in to post a comment.