[sdiy] HF VCOs and tracking problems

[Richie Burnett]

> Any favourites you'd care to share?

Analog Devices AD9834 is the one i've used most recently at work.  They're 
about 5GBP each in volume for the 20-lead TSSOP part.

> Although, if you do have any digital, SPI is about as simple as you can 
> get, even if you have to bit-bang it!

Yep, it's easy if you've got a half-decent micro there already.  You can 
also hang a number of them off an SPI bus and use a seperate chip-select 
line to enable each one in turn, so well suited to applications where you 
want to synthesise several different frequencies accurately at the same time 
:-)

> What's the actual frequency accuracy like? From what you say, it sounds 
> like they're very "true to themselves." Or do they use a crystal, so as 
> good as is the crystal?

They basically divide down the frequency of the master oscillator.  So the 
absolute accuracy and drift is down to the spec of the crystal used for the 
master oscillator.  The frequency resolution is down to the number of bits 
that you can set the frequency to.  In the case of the AD9834 part the 
maximum frequency for the master oscillator is 75MHz and the frequency 
registers are 28bits, so that lets you generate any frequency from 0 to 
37.5MHz in steps of 0.2794Hz.  If you don't need to go up as high as 37.5MHz 
you can drop the master oscillator down to something like 1MHz and get an 
output range of 0 to 500kHz in steps of 3.73mHz!  It's a deterministic 
digital system so the output frequency is as repeatable as the master 
oscillator used to clock it.  The output is a sinewave of decent purity but 
it contains a built-in comparator that you can use to square it up to make a 
digital clock.

For my application at work the master oscillator is 10MHz coming from a GPS 
locked atomic clock, so I can generate any frequency I need from this to 
within +/-19mHz.  Great as a reference for characterising the drift of 
analogue synth VCOs, but I digress...

-Richie,