[sdiy] VCO tuning philosophy re-visited

[Richie Burnett]

> ...how to convert an exponential
> current into an accurate clock frequency in the MHz range.  Any ideas?

DDS chips from Analog Devices such as the AD9834 are a good way of 
synthesising accurate frequencies in the DC to tens of MHz range with 
milli-hertz resolution.  They are a self contained Numerically Controlled 
Oscillator (NCO), where the frequency is programmed digitally.  DDS chips 
are similar to a Phase-Locked-Loop (PLL) frequency synthesiser but can 
operate over a wider frequency range, with finer frequency step size, and 
can change frequency instantly.  They are common in high-performance 
communications equipment and medical instrumentation.

The AD9834 device that i've used most recently was clocked at 24MHz and the 
output frequency could be programmed to anything between 0Hz and 8 MHz with 
0.1 Hz resolution!  Absolute accuracy and frequency drift depend entirely on 
the quartz crystal used to provide the 24MHz clock.

The output of the DDS chip itself is a discrete-time analogue sinewave 
sampled at the clock rate.  In my case with a sampling rate of 24MHz.  This 
needs to be low-pass filtered to remove the sampling images (round off the 
steps in the sinewave if you like.)  Once it has been lowpass filtered the 
sinewave is continuous time and can be "squared up" using a fast comparator 
to make a variable-frequency digital clock signal.  The only part of the 
design that requires any thought is the design of this elliptic lowpass 
filter.  When designed correctly it removes all of the sampling artifacts 
from the sinewave output of the DDS, resulting in a jitter-free 
variable-frequency digital clock from the comparator output.  (The result of 
the analogue lowpass filtering is that the comparator transitions are no 
longer forced to coincide with the transitions of the 24MHz sampling rate 
clock.)

Sounds tempting for driving an alias-free variable playback rate wavetable 
oscillator?

-Richie,