The Declination backlash on my Takahashi EM-400 mount has been a bit of a headache.
Dec backlash measures as large as 1.5 seconds at 100% guiding rate, or equivalently as large as 22 arcseconds. The amount of backlash varies. I do not understand why it varies as much as it does.
Dec backlash can result in poor autoguiding and compromised image quality, as guiding errors remain uncorrected during the periods of time needed to unwind backlash.
Here are the things I do to mitigate Dec backlash:
First, I misalign my mount's polar axis by ~2 arcminutes in azimuth. This misalignment results in a slow declination drift in one direction during exposures within ~2.5 hours of the meridian. Polar misalignment in altitude has a small influence on declination drift near the meridian, the azimuth misalignment dominates the declination drift direction. The accuracy of my mount's polar alignment telescope easily sufficies for this purpose. For my setup's field of view and off-axis guiding angle, field rotation has not been an issue with this relatively small polar misalignment.
Second, when dithering between exposures, I dither only along the Dec axis during one night, and only in the direction of declination drift due to the azimuth polar misalignment. Hence, dither offsets appear to the autoguider as larger declination drifts, and so backlash never needs to be unwound during dither repositioning nor during the next exposure. I dither subsequence nights along the RA axis once on the first exposure, the remaining exposures during each night use this base position. Each dither offset, both in RA and Dec, corresponds to a linear distance of 1 +/- 0.5 image pixel using a random number generator. I specify dither offsets manually using MaximDL's guide star position fields.
Third, after the target reacquisition slew prior to the first exposure on each night, I offset the guide star's current position manually by ~3 image pixels in the direction of declination drift due to azimuth polar misalignment, and then run the autoguider until the guide error is near zero. This procedure unwinds any residual Dec backlash due to the slew prior to the start of the first exposure of the night.
Typically, after the target reacquisition slew, the guide star's current position is not consistent with my desired dither position relative to the prior night. So I manually nudge the guide star into an appropriate position prior to the backlash unwinding step described above.
Fourth, image shift due to refocusing during the exposure can introduce guiding errors counter to the direction of declination drift due to the azimuth polar misalignment. I solved this problem by making relatively small refocusing operations more frequently, as described in a prior post. With this change, focuser image shifts are typically buried by seeing variations and don't result in backlash problems.
Finally, I disable Dec autoguiding corrections in the direction counter to Dec drift. Such anti drift direction errors are due to seeing and hence needn't be corrected.
With these procedures, the RMS guiding error in Dec is about one half that of RA typically. The additional guiding error due to the polar misalignment is small compared to that due to RA periodic error.
Usually the overall RMS guiding error is between 0.5 and 1.0 arcseconds on my 40 minute exposures, using a 6 second autoguiding exposure. RMS guiding errors are strongly correlated with seeing conditions, with 40 minute exposure RMS guiding errors near 0.5 arcsecond occurring during conditions of excellent seeing.