> A lot of this discussion seems to center on the idea of using a VCO
> generated sawtooth wave, putting it into an A to D converter and using the
> output essentially as a counter. The accuracy of this particular scheme is
> highly dependent on the sawtooth ramp being a perfectly straight line. Any
> curvature in the ramp would cause the counter to count faster at some times
> and slower at others as the slope varies.
Whilst this is true, the linearity of the charge/discharge is actually
pretty linear. The biggest problem is normally maximum peak level, this
is normally deliberately reduced as the VCO frequency gets higher to
compensate for discharge timing. Failure to arrive at the correct peak
level means certain output codes missing.
Any oddities at the peak caused by the discharge circuit can be
problematic for the ADC too. The Oakley VCO would spike to +10V
momentarily and this would may cause some problem to any level detecting
circuit.
One method, used by Grant Richter, is to incorporate the ADC in the VCO
core itself. Thus discharge is controlled by the ADC and not a
comparator. Another method of high frequency compensation would need to
be used of course, but this can be done in the expo and/or CV summer.
> A better way would be to start with the triangle wave, run it through a
> series of rectifiers followed by capacitors to eliminate the DC component.
> This would get you a frequency multiplied triangle wave.
This method is actually harder to do. Whilst the number of pulses
derived per cycle of the input frequency is correct, the spacing between
the pulses is entirely dependant on the accuracy of the full wave
rectifiers. The FWR circuits will become less accurate as the frequency
of operation increases.
Both active methods, that is; FET switches or diodes in feedback loops
do have problems with the speed at which they operate. ie. the time it
takes for them to conduct in each direction is limited by the switch
speed or slew rate of the op-amp. Odd things happen to the output at the
point of phase reversal.
Also, noise is a problem, since each folding of the wave requires a gain
change of 6dB to bring it back up to the same level as the input, or a
resultant increase in sensitivity of the final pulse generator. Noise
from the switching of the diodes or FET switches may then be amplified
sufficiently to produce erroneous glitches in the ADC output.
I looked at both systems when I was making my WalshBank digital VCO and
revisited them when designing the 'two up two down' module. This former
relies on a high frequency clock which is them decoded down by binary
counters. In the end I settled for a very fast triangle based VCO. This
would operate at 32 times the normal frequencies of a musical VCO.
However, I never really got much further than breadboarding this one.
Another thought is a tracking sync slave VCO. Slaved to the master CV,
it would operate at 32 times to a master VCO. A hard sync connection
would then keep it effectively locked in.
Tony
www.oakleysound.com