Using an observatory and only polar aligning once or twice a year. When I do and recheck it can be difficult to get consistent results.

I am currently using a 1650mm focal length scope, seeing is generally poor to average, images every 3 seconds. Two changes helped the guide results go from wild swings and frequent loss of the guide star to averaging around 0.5 RMS.

The first was Doug's suggestion in the first response. I changed from internal to PHD2 with its default settings and learned it just enough to get it working. The main thing was using its drift align tool which indicated the polar alignment was off. After using that to correct polar alignment it did not loose the guide star. The errors and corrections were still jumping around but it was able to guide well enough not to show streaks in the image.

Changing back to use the internal guiding and the results were similar. Next increasing minimum pulse from 30 to 50 seemed to calm things down a little. But what really brought all the activity down was to use 2x2 binning.

Also have same issue with guiding and AVX.

I polar align using QHY PoleMaster. Can't say enough positive things about this.

I then plate solve so scope knows where it is pointed. It has little issue slewing to a target although to get within 11 arc seconds it has to do a few plate solve / adjust cycles.

Once on target it tracks well enough that I can take 60 second exposure on 610mm scope without guiding.

But when I try to use the guiding software it either fails during calibration saying "failed to reach starting point after 21 iterations.... possible mount backlash problem"
or
it starts guiding and the performance seems pretty good for 5-10 minutes but then it drifts outside of the +/- 2 arc second window and even though correction pulses are being sent
it keeps drifting away. When the guiding seems to be working I can see it issuing correcting pulses and scope location in the drift plot does move back and forth like one would expect.

The star field image from the guide scope / camera seem fine. There doesn't seem to be any issues with that as far as I can tell. Certainly no complaints from software.

I've tried playing around with the guide settings but nothing seems to make it work.

Honestly wondering if the older ST4 cable might work better but have not tried this.

Sending pulses through the RS-232 interface seems suspect to me. Not sure how the software in the hand controller handles that.

Anyway would really like to get guiding working so I could take longer than 60 second exposures.

Some thoughts of mine on this:

1. I remember coming across some threads on CN forums where users were mentioning that autocorrection function on AVX mounts counteract with guiding software correction signals. They reccommended to switch off the auto error correction when using the mount with a guide cam.

2. Polar alignment. Here i have a favourite routine:
First part is what i always do first when i get a new tripod (at home):
A. I set up the tripod to point to the north
B. I balance the tripod using a bubble/spirit level
C. After i attached the mount, i am leveling the RA counterweight bar on both sides, and mark the positions using a waterproof pen (green and red for east and west)
D. Using the RA scale by setting it equally exact to 90/-90 on both east and west, i find the zero point and mark that on the motor housing too, using a black pen.
E. Zeroing and marking Dec the same way.

I might have confused some details, but you can look on youtube for videos, there are some available.

Now wherever i set up the mount, i only have to level the tripod, pointed to the north, and use the black marks to adjust the tripod so that it's already pointing to the CP, i don't even have to have polaris visible. Then i start PA with what best suits the situation, may it be the manufacturers 3-star alignment, PHD2, Stellarmate (EKOS) or even manual via Scheiner's.

I found out if i am stuck with one way to PA, i'll get in troubles more likely. In the beginning, i had to often give up a session because i was only relying on the one method of PA. This wasn't worth all the effort, so i adapted.

Hope that helps