[time-nuts] Software Sawtooth correction prerequisites?

[Jim Miller]

Hi Tom

I thought the 1ns resolution phase detector combined with a significant 
phase error in a PLL could generate a number large enough to need a 4 byte 
representation which would then need signed arithmetic done on it. 
Accumulating many of those over the integration period means even greater 
precision in the PI software. Lots of work for a little PIC. Lesser 
resolution means easier computation although 1 second is a long time.

I finally found where the sawtooth info comes from: the RAIM message...I 
mentioned I was a newbie. Since it is just an signed 8bit number in 
nanoseconds and is in the <50ns range it doesn't seem too useful if a low 
resolution phase detector has already quantized the phase sample. I suppose 
it could be accumulated anyway in the off chance the mean of the errors was 
significantly non-zero.

I'm strictly an amateur looking at building a GPSDO for a workbench 
frequency standard and perhaps for occasional use on a ham rig. My 
oscillator will most likely be a HP10811 or similar.

Your simulator sounds interesting. Would it need to be built on a commercial 
package?

tnx
jim miller
ab3cv