tripgood The crazy thing about this performance is that only 3 stars were used!
Hi Lee, if the turbulence on the three stars are independent (no reason to think why they do not), then using three star at 0.5 second exposure is equivalent to using 1.5 seconds exposure on 1 star in terms of measurement error [error when estimating the centroid], assuming that all the stars the same signal-to-noise ratio. I.e., treat the problem as ergodic statistics. In practice, they won't have the same SNR, but you are probably doing better than the equivalent of 1 second exposures as far as measurement accuracy is concerned, but without stars elongating with the longer exposure time, and large periodic error.
In your case, the IR-pass filter is probably helping reduce the measurement too (but at the detriment of getting enough signal -- which is why you need larger aperture, and good objective lens, to do near-IR guiding).
This is why multi-star centroid is so useful. For mounts such as the RST-135 and Avalon single forks, we can't use long exposures. I'll bet the ZWO mount will behave similarly too.
The other thing about getting very accurate measurement of the centroids is that the loop gains (the RA and declination "aggr") can be held moderately low. I see that is true for you too. You can turn the declination loop gain even lower if polar alignment is good -- the declination should not wander except from the initial pulling in, and after a dither in declination.
PHD2 conflates two problems; one is the measurement error and the other is the mount error, and it cannot tell them apart. When the measurement error is high, it thinks that it is caused by the mount, and mistakenly applying a correction to the pulse -- which causes the mount to move when it should not be moved. So, the way to fix that is to reduce measurement error to much less than the mount error. And this is what using multiple star centroids help out (when all stars have the same SNR, the variance of the measurement halves each time you double the amount of stars, and thus the Standard Deviation (same as Root Mean Square, RMS, for ergodic systems )of the measurement error drops by 1.4 each time you double the number of stars. (3 dB per doubling.)
As to the three stars, I think I had explained the "goldilocks region" earlier. There are two "forces" working against you -- one is star saturation on the high side and the other is the SNR on the low side. The lack of dynamic range is what causes you to only get 3 stars. You will find that changing the gain of the camera (since exposure time is fixed) will move this dynamic range up and down, and you may need to experiment by changing the gain perhaps 1 dB at a time to search for the optimal gain.
All that said, if the Dawes limit, coupled with atmospheric turbulence is giving you a point spread function ("star size") of larger than 4 arc seconds on your main camera, there is no difference if your guiding is 1" RMS or 0.1" RMS -- in one case, the star would bloat to 4.12 arc seconds, and the second case to 4.001 arc seconds. Focusing error is much greater.
Good to see you are having fun.
Chen
