Troubleshooting with the Friend Chip SRC AT
David Mellor dabbles in a spot of SMPTE/MIDI troubleshooting with the Friend Chip SRC/AT processor.
I seem to remember that in a recent article, I prophesied that the dedicated SMPTE/MIDI convertor was a dying species, rendered unnecessary by the development of sequencers with inbuilt SMPTE/EBU timecode capability. I did qualify this remark, however, with the phrase 'except for certain specialised purposes'. Let me clarify...
If, in your studio, you make music with the aid of MIDI and timecode, and all the creative effort is outgoing (ie. to other studios or producers), then a sequencer with a simple integrated timecode unit is ideal.
On the other hand, if your work involves incoming elements - tapes, sequences, etc - then you might find yourself in a position where your sequencer cannot cope with the timecode on someone else's tape. Perhaps it can't match the tempo of the music correctly, or perhaps the code is too low in level, or damaged, and your sequencer can't read it. In this case, a more sophisticated timecode unit is called for.
There is a third situation where a highly intelligent SMPTE/MIDI box is essential. This is when you have an existing piece of music which was recorded without the benefit of timecode, and you want to add extra sequenced parts to it. This is not necessarily an easy task, and you'll need all the help you can get from your machinery.
If you find yourself in either of the last two situations, then a SMPTE/MIDI box with a high IQ is called for. If you never need to handle other people's timecode, and never work with un-timecoded tracks, then a sequencer with a simple integrated timecode capability is easier to work with. (Reasons: control over tempo and meter resides in the sequencer as part of the sequence - easier to handle and store on disk).
To put all this in a nutshell, my current thoughts on SMPTE/MIDI units go something like this: either have a simple unit integrated with the sequencer or a full function 'do everything' unit which will perform miracles strapped to any sequencer.
The latest arrival at the top end of SMPTE/MIDI processing is the Friend Chip SRC/AT, which I assume must stand for SMPTE Reading Clock/Advanced Technology, or something like that. It certainly is very advanced and I had a lot of fun thinking up problems for it to solve. But before I get into top gear and onto the unit itself, let's take a look at the problems. Know your enemy, and all that.
Let's imagine an engineer working in a 24-track studio. In comes a punter (sorry, respected client) with a 16-track tape and a Fostex E16 in his briefcase. He wants to copy the 16-track tape onto 24-track and add some extra sparkle to his song - which he recorded in a smaller studio using a Brand X SMPTE-capable sequencer.
It's no problem to transfer all the tracks across, including the timecode track (but see the side panel for notes on rerecording SMPTE/EBU code). The problem comes when the engineer tries to lock his Brand Y sequencer to the track. It takes ages to get the correct start time, and then it's impossible to find the right tempo for the track. Evidently, Brand Y's idea of 120 beats per minute (bpm) is slightly different to Brand X's.
Unfortunately, converting SMPTE/EBU numbers to MIDI song pointers is not a precise science. Some manufacturers do it one way, some do it another. The result is that an indicated 120 bpm can actually be 120.1 bpm. A small difference perhaps, but it means that two and a half minutes through the song you are a full 16th note (semiquaver) out of time. And you would have noticed the difference long before that.
That's problem number one. And what happens if the tempo changes during the course of the song - especially if the sadistic client (or is it masochistic? After all, he's paying for the time taken) has used a sequencer that can programme accelerandi and ritardandi, producing heaps of tempo changes. Worst of all, maybe there isn't a timecode track - just real musicians, and a drummer who goes in and out of time more often than Doctor Who's TARDIS.
Let's home in on the basics. Assuming the tape contains music recorded at a fixed tempo, and timecode, the Friend Chip SRC/AT needs to know just two things to get on with the job of driving a sequencer with the correct MIDI song pointers: the SMPTE value for the start time, ie. hour, minute, second and frame (one frame is 1/25th or 1/30th of a second depending on the type of timecode selected) and the precise tempo in beats per minute.
There are two ways to set the start time. The first is to watch the SMPTE display as the tape rolls, and estimate its value when the song starts. It can be trimmed forwards or backwards in time if necessary. The other, superior, method is to feed an output from any track on the tape which has a good clean start on beat one (a drum machine, say), to the Audio Input socket of the SRC/AT. The unit's Pickup Start Time function will enter the correct SMPTE value automatically when the song starts. It may still need slight adjustment, but it will be very close. Adjustment, by the way, is to a resolution of one timecode bit. A 'bit' is 1/80th of a frame, which works out to around half a millisecond.
That's the easy bit. Setting the tempo can be more difficult. Perhaps the tempo is known already, from the display of the sequencer or sync box used previously. The Friend Chip manual explains what can actually happen when a sequencer calculates tempo against timecode: Suppose the indicated tempo is 123 bpm. This would be equivalent to 975.6097 bits of timecode (running at 25 frames per second). Suppose this were rounded up to 976 bits - the true tempo would now be 122.9508 bpm. If it were rounded down to 975.5 bits, the true tempo would then be 123.0138 bpm. This is enough to make a difference, so merely setting 123 bpm on the SRC/AT is not necessarily good enough.
In the end it boils down to trial and error. Enter a tempo, set the tape rolling (with the correct start time) a few minutes into the song where any discrepancies in tempo will have accumulated, judge the result, and make appropriate corrections. After a while, the music on tape and the sequencer driven by the SRC/AT should be ticking along in perfect step.
Let's assume a slightly more complex situation, where the tempo of the song varies. The procedure outlined above suddenly becomes too hot to handle, and the learning capabilities of the SRC/AT take centre stage.
If among the percussion tracks there is an instrument, say a bass drum, which keeps a regular beat, the SRC/AT can follow its tempo over the course of a song. That isn't too tall an order if the bass drum always falls on beat one of every bar, and on no other beats. But what if the drum programmer got an attack of the 'creatives' and inserted a few paradiddles at choice moments (ever heard a paradiddle on a bass drum?)?
The Friend Chip uses a concept called the 'Window' together with an approximation of the true tempo to allow it to follow complex drum parts correctly. Suppose the bass drum hits every beat of every bar, and also puts in an occasional 16th note offbeat. If the SRC/AT knows when to expect the bass drum, it need only be actively tempo-seeking around that time. Other beats can thus be excluded.
In practice, this is done by setting a Window percentage. A 15% Window means that after a beat is successfully received, the SRC/AT will not accept more beats for 85% of the expected time before the next beat. It opens up slightly before the next one is due to come in. If the next beat doesn't come in on cue - if it is slightly late - then the Window will remain open for another 15% of the expected beat-to-beat time. In this example, the SRC/AT could cope with a +/-15 bpm variation in tempo about a 120 bpm mean value, according to Friend Chip's calculation, and also reject spurious 16th notes which would confuse the tempo setting.
The Window function will also help eliminate mis-triggering due to any decay on the percussion sounds. Additionally, there is a click output generated whenever a trigger is successfully received. This helps when setting the audio level into the unit, and the Window percentage.
That's the theory of tempo learning, but how easy is it when you actually go to work? Well, in this case, falling off the proverbial log is just a little easier. But with patience, all things are possible. When successfully carried out, the learned tempo tends to vary up and down slightly - even when the input tempo was supposedly constant. But you would hardly notice, and it does actually help to give a more 'human' feel to a strict tempo sequenced track.
If all else fails, there is the final option of entering the tempo by tapping along to the track on something that can provide an audio input to the SRC/AT. If you are a drummer, you'll find this easy enough. If you were attracted to MIDI sequencing because of the 'professionalising' capabilities of quantisation, you might find this a little more difficult...
By the way, tempo learning is good both for sequenced and 'real' music, so those 'extra beats' remixes of old tracks are a definite possibility.
Chapter 2 of the Handbook Of Creative Strategies (a mythical volume, as yet) suggests that one route towards success in human endeavour is to make something, then make it better. In MIDI music, this procedure is known as editing - where you fumble an ugly duckling series of notes into a sequencer, then go to the Edit page and turn them into a beautiful swan.
The Friend Chip SRC/AT has powerful editing features. What I particularly like here is the ability to edit learned tempo sequences. What actually happens inside the SRC/AT when it learns the tempo of a track, via MIDI or audio inputs, is that every 16th note MIDI song pointer is linked to a particular timecode value. For instance, a sequence I have in the unit at the moment has a table (as follows) which can be called up on the display:
It used to be the case that the ordinary muso in the street (ie. you and I) used to be happy if he could buy any unit with timecode functions for less than a king's ransom. But more and more, we want to see these units perform. And by that, I mean that they can operate on something less than pristine condition rock solid code. It's not that we abuse our timecode intentionally - but if someone else brings the code into the studio, then you don't know where it has been, do you? Probably through several generations of rerecording. And in this day and age, timecode units need to be able to ride unruffled over dropouts and edits.
Fortunately, the SRC/AT is provided with dropout compensation. If the code disappears momentarily for any reason, the device can soldier on. The length of time over which it is able to do this is adjustable. With the unit I had for review, it did soldier on magnificently - but at 120 bpm instead of at the tempo I had programmed. The manual offered no help on this, so I assume it must be an oversight in this version of the software that is undergoing rapid correction at Friend Chip laboratories. Even so, it would be possible to edit the tempo sequence to get around this, even if the dropout was quite long. Joins in the tape passed through without hiccup.
This unit certainly could be your friend for life if you need to deal with SMPTE/MIDI problems in your stride, or even if you just have a favourite sequencer that needs a timecode shot in the arm. I certainly haven't seen a better unit yet (although I'm still open to persuasion). The system of editing is very easy to understand, and also very comprehensive. I fear that timing problems, because of their nature, will continue to be difficult to solve. But it doesn't get any easier than with the Friend Chip SRC/AT.
£699 Inc VAT.
The Synthesizer Company, (Contact Details).
Feature by David Mellor
Previous article in this issue:
Next article in this issue:
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!