by R37, and is compensating for reset time only.
The exponential convertor causes errors due to ‘bulk emitter resistance’ or Rbe in the NPN pair, the
top half of the THAT300. Now the THAT300 is a very good matched pair with low values of Rbe,
but it still has an effect at very high frequencies of the VCO. Rbe is not a real resistor, but it acts as
if a resistor was connected in series with the emitter of the transistor.
In the early issue VCO boards, this loss of high frequency performance was cancelled out by the
effect of the Franco resistor. But really, this is a bit of a cludge since the Franco is designed to
compensate for a fixed timing error (the FET reset time), whilst Rbe causes an error related to
frequency. Over the limited range of an audio VCO, this is usually perfectly fine and Franco is used
in many commercial synth VCOs as the only high frequency compensation technique. For issue
three I decided to go one better. The newer issue VCOs use a fixed Franco resistor for the reset
time, and also use the Rossum technique to compensate for the exponential generators error.
Dave Rossum, founder of E-Mu systems in the 1970s, was an important player in the development
of the SSM chips. The SSM2030 VCO chip, which was not as good in comparison to the later
Curtis CEM3340, was ground breaking when it came out. Dave’s method of HF compensation for
the SSM2030 involved gently pulling the base of the first transistor of the NPN pair lower when the
collector current through the second NPN gets bigger.
In the issue 3 VCO I slavishly copied Dave’s idea as presented in Electronotes and indeed
implemented in the Prophet V many years ago. However, due to work done by René Schmitz this
simple method can be improved upon with the same number of components. As in the original
Rossum idea a fixed emitter resistor, R17, can be used to effectively measure the emitter current,
since the voltage on its bottom end will fall as collector current rises. Previously we took this
voltage, passed it through a diode to compensate for the base-emitter voltage drop (Vbe) of the
matched pair and then fed a small proportion of the resultant voltage back to the base of the left
hand transistor. This works well enough.
However, René’s idea improves on this on two counts. Firstly, the diode in the pure Rossum
method doesn’t exactly compensate for the Vbe because the current flowing in the diode and the
base-emitter junction are very much different. In the new circuit, both semiconductors see a similar
current as they are fed from identical resistive sources; R17 and the HFT trimmer are both the same
value. Therefore, the voltage drop across the 'diode' and Vbe should track each other reasonably
well, even over a change in ambient temperature. The 'diode' in the issue 5 VCO is actually the
base-emitter junction in a spare NPN transistor from the THAT300 array. One of the other benefits
of René’s method is that we feed the base of the left hand transistor with more or less the same
resistance which was not true of the original Rossum technique. So now the extra resistance due to
the HFT circuitry the left hand transistor base ‘sees’ is pretty much fixed at the value of R20 only.
Thus altering the HFT trimmer should not result in a change in the overall operating frequency of
the VCO as it did in the old method.
Going back to the VCO core: The sawtooth output from U5/Q2 is amplified by the other half of U5
(pins 5, 6, 7), before being sent to the output pad. The final output is r5V to -5V, ie. 10V p-
p.
The triangle shaper is essentially a full wave rectifier, whose operating point is about 2.5V. If the
operating point is not exactly half the peak value of the sawtooth output of U5, then the triangle
wave will have discontinuities in it. This leads to a slightly brighter or harsher sound than the
8
Summary of Contents for VCO 5U
Page 2: ...2 ...