[time-nuts] A new take on the all-hardware GPSDO concept
nsayer at kfu.com
Tue Sep 27 00:18:25 EDT 2016
At the risk of inviting everyone to say “I told you so,” I’ll report here my experimental results from trying this concept out.
Since there was a great deal of doubt about the outcome, I hedged my bet a bit and designed for the DOT050V rather than the OH300. If it worked out for the TCXO, then I could try with the more expensive one.
The results aren’t very good.
With a short TC loop filter, the PLL does lock up, but obviously the jitter of the Venus’ 10 MHz output comes through.
With a longer TC, the PLL never locks - or at least if it does lock, it’s locking significantly off frequency.
That’s with a 10 µF cap and varied resistors between 10k and 1M. The best I got was at 200k - a TC of 2s. That resulted in this video. Unlike other videos I’ve made comparing two GPSDOs, this one is not a time-lapse. The reference is an OH300 based GPSDO.
A time constant of 10s (1M resistor) just doesn’t work at all.
But the real nail in the coffin here is that the price of the PLL chip is still more expensive than the microcontroller and all of the components it replaces.
In the end, I’m glad I tried, but I don’t think I’m going to invest any more time in the design. I could try configuring the venus for a 10 kHz output and see if it’s better able to phase lock with a divided TCXO output, but I don’t think I have any reason to believe that would be more likely to succeed.
More information about the time-nuts