W6LVP wrote:More progress.
If you tick the Auto box above the two graphs in the lower right corner, there is a good depiction of the state of the resampler servo.
As discussed throughout this entire topic. Have your read the topic from the beginning?
You don't have to use the Auto box, you can manually adjust the scale with the value in the box to the right of it. Adjust either/both settings to taste to see usable information in the graphical display.
The amplitude of the variation is a direct indication of the Doppler shift caused by the pointer moving.
This is not "Doppler", this is the "Var Ratio" or, more properly, the ratio between the two, different 48KHz clocks in the two separate clock domains (PC and ANAN hardware). The graph merely shows the fractional part as a difference from perfect, where perfect is a ratio of 1.0 exactly.
For example, if you see a Var Ratio of 1.000010, the graph will show +0.000010, and if one clock is exactly 48KHz the other clock will be at 48.0005KHz (0.5Hz difference).
My problem is patience. Most any change to the VAC settings will disturb the resampler servo.
It's worth noting at this point it's not a "servo", it's a feedback loop with a moving average.
Initial Thetis launch will also restart the process. It seems that ticking the two Old VAR boxes on the Advanced page might accelerate the post launch stabilization process by using the values from the last session.
Yes, any change to VAC settings will completely restart the process.
With my less than optimized configuration, the time to reach acceptable servo stability is several minutes. Seems to me that if you create a reasonable VAC resampler configuration and give it time to stabilize, it works fine. If you launch Thetis and give the servo time to stabilize, all is good.
Again, this is well discussed throughout this entire topic. Have you read from the beginning?
Ultimately, for me personally, I settled on only making two changes in the Advanced tab:
Feedback Gain: 0.0000001 (1E-07)
FF Alpha: 0.001
I found these to be a good compromise between stability of the Var Ratio (as you note, the absolute value of this number is not important as long as it is correct, it is any variation in this number that causes problems) and the settling time of the resampler feedback loop.
The following workflow is recommended:
1. Without changing default settings on the Advanced tab, find the buffer/ringbuffer/portaudio settings that keep under and overflow counters at zero, or at least close to zero (a few counts per hour is acceptable).
2. After achieving stable under/overflow performance only then attempt adjustments on the Advanced tab. Such adjustments will not strongly affect under/overflow performance but will act to stabilize variations in Var Ratio. If you have read this topic in its entirety you will know that the key adjustments are Feedback Gain and FF Alpha.
As discussed in my posts in this topic, I have the capability to lock my ASIO sound interface to an external GPSDO. With both the ANAN hardware and sound interface both locked to their own GPSDO, that is a best case situation. As you can see from my posts, the difference in performance between this best case situation and a properly adjusted resampler, while measurable, is functionally nil from any practical perspective:
viewtopic.php?f=13&t=3901&start=90#p16667viewtopic.php?f=13&t=3901&start=120#p16729Any frequency variations due to a wandering Var Ratio will be <<0.01Hz assuming proper resampler settings. This is far, far less than ionospheric induced Doppler, and therefore should be unnoticeable.