[sdiy] Oberheim OB-XA autotune procedure questions

[WeAreAs1 at aol.com]

In a message dated 4/29/05 1:30:14 PM, gene at ixiacom.com writes:

<< I have some dim recollection of how the OB-8 does autotune, which may be

similar to the OBXa...


The autotune button is pressed. The microprocessor sets all VCO's to pulse

wave, then sets all pulse widths to beyond 100% (silence). It then  sets the

pulse width on each VCO, one at a time, to an audible setting. The filters

are opened up all the way, and the final mono audio output is monitored by

the microprocessor through a comparator via an I/O pin (or some such thing).

A set of test CV's are set up, the output frequency is monitored, and the uP

builds up a scaling table for each VCO.


If the feedback path from the output back to the uP is opened somehow, all

autotune will fail. >>

Yes, that feedback path is definitely important, as is the ability of each 
VCO to output a near 50% (square) pulse wave (the basic pulse width of each VCO 
must be well-calibrated) when in autotune mode.  It is important to note, 
however, that the OBxa does not build up a scaling table for each VCO during its 
autotune procedure.  It simply sets a single basic tuning offset CV for each 
VCO.  Therefore, it's very important to first calibrate each VCO's scaling and 
hi-track trim as close as you can.  If the VCO scaling or hi-track is off, the 
autotune offset won't really help them all that much.  

In case any readers didn't know, when you calibrate an OBxa, you flip an 
internal "calibration" switch, and that disables autotune and all of its offset 
CV's.  Then you just trim the VCO's manually, as if you were trimming a very, 
very large Minimoog.  When you're done, you re-enable the autotune, and 
hopefully, it will help keep things in line for the user for a while.  As Gene pointed 
out, the OB8's autotune software did calculate a new scaling table in RAM for 
each VCO every time you pressed that button, just like the Rev3.x Prophet V's 
did.  This allowed the autotune to compensate for both scaling and overall 
pitch offset for each VCO, without having to get each VCO's hardware scaling 
trim exactly perfect -- the trimmer just needs to be "in the ballpark", as we say 
in the States.  

Just to be clear, the OB8 "scaling" table doesn't actually adjust VCO 
scaling, strictly speaking.  It too is just a simple pitch offset, but the lookup 
table for each VCO calculates and holds a different offset value for every two or 
three notes up the entire range of the VCO, effectively allowing the autotune 
to build a piece-wise approximation of what you might get by carefully 
trimming that scale trimmer.  

The Prophet V rev3's had this first (Rev 2's had the 
single-tuning-offset-only system), and then Oberheim followed suit with the OB8.  Roland Jupiter 8's 
also had the OBxa-type autotune system (overall pitch offset only), but Roland 
switched over to the "scaling table" system with the Jupiter 6 (and later, the 
Super Jupiter MKS-80).  This may be one reason why JP-8's sound "fatter" than 
Jupiter 6's.  Yes, there are several other reasons, too -- but the JP-8's 
tuning is rarely ever as accurate as the that of the JP-6, so it will always 
sound a little warmer and wetter, at least in terms of the relative tuning of 
VCO's.  A freshly autotuned JP-6 almost sounds digital and clinical in comparison!

Mike B.

P.S. -- the lack of a "perfect" VCO hardware calibration function in the Rev3 
Prophets is why the pitch gets so wacky when you do large pitch bends.  The 
pitch bend CV's are fed directly to the VCO CV summers, and if the hardware 
scaling is off, they will not bend together -- even if the autotune has been 
recently hit, and the VCO's are playing perfectly in tune from the keyboard.  This 
is especially apparent in unison mode, or when you pitch bend large chords.  
The same is true of wide LFO pitch modulations -- unison-tuned VCO pairs and 
notes in chords go further and further out of tune with each other, the farther 
you modulate them from center pitch (try it with a squarewave LFO...).  The 
Prophet designers just didn't bother (and likely couldn't bother, due to 
hardware and software limitations) to run the pitch bend and LFO mod CV's through 
some sort of autotune compensation, like they did with the keyboard and 
transpose CV.  If you think about it for a minute, you'll probably realize why this 
would be really, really hard to do, or at least highly impractical to do in 
1982.  I've never had the occasion to check this on the OB8, but I'm betting that 
it has a similar problem with wide pitch bends, too.