Recentering after flip failure


Hi all,
I’m having a few problems centring the image after a meridian flip and hoping someone can suggest something to improve. I’m using a Losmandy G11 with gemini 2. After the flip, the centring routine fails to do much better than 200 pixels after 12 tries. I suspect it’s the relatively poor model I built using the gemini, as I have a fairly heavily obstructed location, and only use enough stars to get the object somewhere on the chip.
When the centring routine sends the commands, does it make a relative adjustment, or does it just synchronise and attempt a goto to the target spot? If the latter, then I assume no amount of tries will improve it, and I need to build a better pointing model.
Andrew Lockwood.


What is your pixel scale? While the model could have something to do with it after 12 tries it seems fairly unlikely. You should be getting a little closer with each try.

I would guess it’s either epoch related or potentially backlash. Knowing that the G11 only takes JNow I would guess epoch. You can set the G11 to expect J2000 which should help with the solving, but it may cause some plate solve failures as the G11 will only return JNow, however you’re likely already getting back all JNow so it’s probably not that big of a deal.



thanks for your reply Jared - the pixel scale is 0.965" per pixel. It doesn’t have any trouble plate solving at the moment, it just fails to improve the error each time it iterates. It slews, but ends up within a few pixels of the previous effort, and sometimes even worse. It worked well the other night near the equator and did a flip to within about 20 pixels, but last nights object was at -63 declination, where I don’t have any stars in the model and it didnt improve after about the second attempt.
I’ll poke about with the epoch - should I just set everything to Jnow?


For some reason my earlier email reply didn’t come through.

I had this exact same problem a couple months ago. I don’t really understand the epoch thing, but at the time I set it to J2000 since that seems to be the more common default setting for other applications. I didn’t see a great improvement with just changing to J2000. I did a cold start and rebuilt a simple pointing model of three stars on each side of the meridian and haven’t had a problem since.

Jared, your post is interesting to me. Do your comments apply to both Gemini 1 and Gemini 2? Does the G2 only return JNow?


My earlier email reply did not come through either :confused: !? Here is what I wrote:

Hi Andrew,

I have a G11 with Gemini 2 too and my meridian flips have been quite reliable. I never ever build any model at all (cold start every night.) I set the number of attempts to a generously high number, and it always centers even if it has to repeat a few times. I have max attempts set to 8, though I have never seen it go beyond 5, and it usually gets it in 1-3 attempts. I use ~ 20-30 arc-seconds worth of pixels for the tolerance value (“until error is less than”).

On the mount side, I have seen that when there is a lot of backlash it can make the centering fail (i.e. take more attempts). I like to keep a fair amount of play&backlash to get best guiding performance out of the mount, but this only seems to cause it to have to retry one or more times on occasion.

You may want to make sure you have the latest Gemini firmware installed from There were bugs related to short slews in earlier versions, and auto-center induces those bugs. (the symptoms of the bugs were failure to respond to guide pulses and, more rarely, wild slews)

Your seeing failures with 200 pixel tolerance & 12 attempts makes me think something must be seriously wonky. Are you sure you are not experiencing a slipping clutch or slipping worm coupler, or stickage somewhere? (I guess you could rule out those problems with inspection or by testing slews after you have built a good model.)

Hope that helps,


I’m not 100% sure about the Gemini 2. But I believe that both will only return JNow, but can be set to accept J2000.

Really for plate solving it doesn’t matter what the value is that we get back from the scope as long as it’s close. Since we plate solve every frame during the auto center the values that the scope returns is moot provided that they’re close enough for a solve to succeed. However it is important that the value we send for the sync gets translated correctly with the right epoch as this is that the scope will be set to.

I haven’t put together a “proper” model in years. As long as your polar alignment is decent then a sync relatively close to your target should be better than a model. Which is why I don’t think it’s your model (especially with 12 iterations). Do you find that you end up at the same 2 locations after a couple of iterations or does it seem random?



Thanks again Jared. It’s definitely not random, as it drops in one spot, then stays within 10-20 pixels of that spot no matter how many tries it makes. SO I’ll try updating the firmware, although mine is only a few months old (I think) and also set the epoch to J2000 and see what happens.


Thanks Andy and Joel for your help as well - I’m going to update the firmware and do a cold start and see what happens!