Multi Star Guiding

[MarkS]

My next project is to create a version of PHD2 that supports multi-star guiding.

I've recently been examining the theoretical accuracy of the star centroid calculations in the original PHD.  I came to the conclusion that if the star is unsaturated, the FWHM is at least 1.5 (preferably 2.0) and the signal to noise ratio is 30 or so then the star centroid calculations are usually accurate to at least 1/15 pixel.  If the FWHM is below 1.5 then the accuracy deteriorates very rapidly.  So the effect known as "chasing the seeing" that is often referrred to, may actually be a case of "chasing the centroid calculation errors".  This is why it is often recommended to guide on a very slightly defocused star.  However, my tests suggest that if you have a star with FHWM of say 1.3 and defocus it to FWHM 2.0 then the guiding hardly improves because the SNR has has deteriorated.

In my opinion, best advice is to defocus to an FWHM of 2.0 and then choose a brighter (non-saturated) star to guide on.

Everyone already knows you should never guide on a saturated star but maybe reduce the camera gain instead.

But if you don't have a brighter star then what do you do?  This is where multi star guiding comes in.  If you can simultaneously guide on 4 equally bright stars then the centroid calculation error is reduced by a factor of 2.

If the stars are not equally bright then the results can be weighted by star brightness.

Anyway, this should keep me out of mischief for a while

[Ivor]

An interesting challenge (I think I recall you mention this before) it's been raised a number of times on the google group for PHD2.

Is there a particular reason for choosing 4 stars or this the starting point? I wonder what the criteria is for choosing the optimum number of stars for a given scenero.

Writing an ASCOM dome driver has rekindled my programming interest so once I've finished I'd be interested in helping.

[MarkS]

Is there a particular reason for choosing 4 stars or this the starting point? I wonder what the criteria is for choosing the optimum number of stars for a given scenero.

No, I'm not choosing 4 stars.  It was just to illustrate that, unfortunately, we are bound by the square root law: 4 stars improves the accuracy by a factor of 2; 9 stars improves it by 3; 16 improves it by 4 etc.

From a user point of view, the rough idea would be to calibrate and start guiding on a single star as usual.  Once guiding is under way then click on additional stars.  Each star would have its own rectangle.  The contribution of each star would be weighted by its SNR or similar.

Mark

[MarkS]

Well I downloaded the code plus all the bits and pieces for building the software only to discover that their build environment will not work on my aged WinXP machine.

I already know that PixInsight runs out of memory and crashes on WinXP.

So it looks like time to buy a new machine ...

[The Thing]

FYI I haven't been able to get an FWHM of better than 5 on my guiding setup of LX90"8" UHTC +Baader AGII Telecompressor + Brightstar OAG + SLR lens focuser + QHY5L-II Mono (binx2, 640x480, 95% gain).

The SLR lens focuser means I can easily focus to the level my optical train allows. My SNR is usually around 4 to 7 on a good night probably because of optical edge defects affecting the star image. Also with the OAG I rarely get more than two stars that PHD2 will latch on to.

Ergo multi star guiding may be of little practical use when using and OAG.

[MarkS]

FYI I haven't been able to get an FWHM of better than 5 ...

My SNR is usually around 4 to 7 on a good night

The FWHM of 5 at your focal length is limited by seeing.  From my calculations and simulations I've done, this would give the same guiding accuracy as an FWHM of 2.5 and an SNR of 8 to 14 which will easily give sub pixel guiding.  You have no need to do multi-star guiding.

[MarkS]

I've replaced my XP desktop with a shiny new W7 desktop.  I've successfully built the project and run the code.  Now I can think about how to put together a multi-star prototype.

Mark