Author 
VCO calibration strategy question 
cygmu 
I have wondered about this for a while  hoping someone smart can help me figure it out.
In this post tojpeters explained:
Quote: 
The problem most people have is turning the trimmer the wrong direction. If you set your oscillator to put out 100Hz at 0v and then go up 4 octaves (4v) and the output is flat (785 Hz) you adjust the trimmer so it goes even flatter. If sharp go sharper. Then go back the 0v,reset your tuning to 100Hz,go up 4 octaves and adjust again,sharper if sharp,flatter if flat.

Ray Wilson also says the same thing in the calibration instructions for his VCO:
http://musicfromouterspace.com/analogsynth_new/VCO20090724REV0/VCO2009 0724REV0.html
So... why should that be the strategy?
I mean, if your VCO responds with more than a 1 octave shift to a 1V CV, so it goes sharp, obviously you want to adjust the trimmer so that it is less responsive. Why should that mean that the frequency you are hearing goes sharper as you adjust?
This should be obvious from thinking about the circuits but to me at least it is not. 

Graham Hinton 
The effect of changing the law to go flatter at 1V makes the tuning at 0V even further flat, hence increasing the interval.
Most exponential stages have the same current in both transistors at the top of their range, although on some it is halfway. The collector currents are equal when both bases are at 0V. The octaves stretch out from that point which does not correspond to 0V external CV, hence if you use 0V, 1V, 2V, etc. the frequencies at each point are all are changing relative to the "fixed point" of Vbe = 0V.
I find it it lot easier to concentrate on the relative tuning of the V/oct adjustment and then use the offset to set the absolute tuning at the end. If you use a tuner just observe the difference between the 1V steps, e.g. if 0V is 20 cents sharp and 1V is 25 cents sharp the difference is 1 octave and 5 cents. I use both a Korg DTR2 (no longer made) and a Universal Counter with reciprocal frequency (it is more accurate and stable to measure the period at low frequencies than count the cycles) in tandem as sometimes it is easier to see two numbers and most tuning meters don't cover the higher octaves. 

cygmu 
Graham Hinton wrote:  The effect of changing the law to go flatter at 1V makes the tuning at 0V even further flat, hence increasing the interval. 
Thanks... I've tried to make that thought work for me but I can't!
In my mental model, the current output of the exponential stage is given by an equation like
current = I_ref * (exp(CV * scale + bias))
where I_ref is the fixed reference current, "bias" is the actual differential voltage on the bases when 0 CV is applied, and "scale" is whatever scaling you're doing, which is presumably what is being trimmed.
The ratio of current at CV value x+1 versus at x is
exp(scale * (x+1) + bias) / exp (scale * x + bias)
which is just exp(scale).
Frequency is linear with this current, so if this factor is greater than 2, then the oscillator goes sharp when we add 1V of CV.
So we reduce the magnitude of the "scale" factor. Now the question is what happened to the overall current, i.e. what do we hear as we reduce the scale. But just looking at the current equation it is obvious that it also reduces. And yet the strategy that appears to work suggests going the other way, so I must have something wrong.


Graham Hinton 
You are not taking into account that the CV adder shifts the voltage too. The CV going in to the exponential stage is not the same as the external CV.
When Vbe = 0 the current is usually maximum and at low frequencies the current is less. The exponent in the equation is signed, but you are considering only the positive part when you are actually operating in the negative.
Clearer or really confused now? Put some real numbers into the equation or plot the graph. 

cygmu 
Thanks for your patience. I think I am there. My equation was somewhere between wrong and not quite general enough. It should have been something along the lines of
I_ref * exp(scale * (CV + bias1) + bias2)
to account for the fact that it is possible (and usual) to set the initial differential voltage "upstream" so that it too is affected by the scaling. Which is of course what you said at the beginning
bias2 is included for generality but I think it is zero normally. It makes no difference at all to the calculations.
"scale" here incorporates all the scaling and differencing that might go on, but it must be overall positive so that increasing CV results in a higher frequency. But the bias1 could be anything, and could be negative, and in fact it must be if zero differential voltage is the highest point on the scale. Which you also said in the beginning.
So if you are working with a design where the highest frequency is at 0 differential voltage, then the strategy is as we have been discussing. If that point is half way up the scale then I think that the strategy will be opposite in the upper half than it is in the lower half.
Excellent, mystery (that only I care about) solved. Thank you! 

TheSlowGrowth 
On a sidenote: You can try my VCOTuner application. Basically you get a graphical representation of the current tuning. It's only a matter of turning the trimmer until the line on your screen is as flat as possible.
It's simple to tune a VCO in just a couple of seconds, really.
https://github.com/TheSlowGrowth/VCOTuner 

cretaceousear 
That looks cool slow growth  but does it require MIDI?
I have a midi keyboard and then a CV modular and CV keyboard  no convertor! 

TheSlowGrowth 

The Real MC 
Never use a guitar tuner to calibrate VCOs, even the rackmount ones. They are not precise enough. 

aabbcc 
TheSlowGrowth wrote:  On a sidenote: You can try my VCOTuner application. Basically you get a graphical representation of the current tuning. It's only a matter of turning the trimmer until the line on your screen is as flat as possible.
It's simple to tune a VCO in just a couple of seconds, really.
https://github.com/TheSlowGrowth/VCOTuner 
I was trying this for tuning an RE303 but never got it to work. Maybe a stupid question but does it require a MIDI to CV interface or is a midi interface enough? I only have the latter. 

TheSlowGrowth 
The software sends out MIDI notes and measures the relative difference in pitch between them to work out if the scaling is right. By measuring only the relative difference, it liberates you from having to keep your oscillator at a fixed pitch during the calibration process.
If your oscillator/Synth has a MIDI input, you're fine with just a MIDI interface. For a voltage controlled oscillator you will need some sort of MIDI to CV interface to turn those MIDI notes into a voltage. That interface should be fairly precise  the weakest link in the chain should always be the oscillator you're trying to tune. 
