Issues with star selection during guiding calibration

retman

i have two mini's and one (new version 2024) plus. occasionally when i start guiding calibration the routine just loops. this happens on the plus and on the mini's. no stars are selected at all. other times the air immediately finds 12 stars and starts the calibration routine. i have found that i have to slew back to home, turn off power and then restart. then when i slew to the same target for calibration all is good, 12 stars immediately. i'm using a WO 30mm guide scope and asi120mm mini. there are lots of stars in the field and the focus is very good and, when i finally get calibration to begin none of the stars are clipped (so gain is good)
any suggestions?

w7ay

retman none of the stars are clipped (so gain is good)

No, that is the wrong (worst) gain to use.

Because of stellar magnitude distribution, and ZWO using SNR to weigh the centroids of the multiple stars, you do want to clip away the brighter stars so that they are ignored.

If you allow the bright stars to be unclipped, they will be used, and because they have much better SNR than the dimmer stars, the centroid of the brightest star wil hog the centroid averaging. I.e., you effectively end up with using the centroids from 1 to 2 stars.

Increase the gain until you start to see 12 stars, then add another 10dB of gain ("100" units of gain for the scale that ZWO uses 0.1 decibels per gain unit) to purposely clip away the brightest stars.

With that, the remaining unclipped stars will be more even SNR, and you finally get the result that is equivalent to using 6 to 9 stars worth of centroid averages. Unless all 12 stars have the same SNR, you will not get 12 stars worth of centroid averaging. You can download the the HYG catalog from NASA to see check magnitude distribution [and thus SNR distribution] of stars).

From ergodicity of the star statistics, using multiple stars is the same as using longer exposures, when the aim is to get rid of atmospheric turbulence. I.e., averaging the centroid of 2 different stars taken at 1 second is equivalent to using 1 star at 2 second exposure. Similarly, averaging 4 stars at 0.5 second is also equivalent to using 1 star with 2 second exposures.

But the above is only possible if the centroid is equally weighted. ASIAIR mistakenly uses SNR to weight the centroids, and that is where you need to apply lots of gain so that it will reject the brighter stars, otherwise the centroid of the remaining 11 stars don't help -- i.e., you are back to one star guiding, but at 0.5 second!

If your mount is smooth enough that can take 2 to 3 second type guide updates, stick to using one star if you don't have a good guide camera. Some mounts, especially the cheaper strain wave geared mounts, need very rapid guide updates, and those people have no choice but to use multiple star guiding -- but only if the guide camera field is flat over the entire FOV. I.e., pay less for mount, and you pay more for the autoguiding system.

Chen

retman

Chen, you are great! thank you. i have one other question for you. in another topic (asi585 first light...a very long thread) you referenced a supplier of Bahtinov masks that you liked. can you please advise (so i don't have to read thru hundreds of postings........)

thanks again for the comments re gain and selecting guiding stars

retman

w7ay

retman you referenced a supplier of Bahtinov masks that you liked

Most (especially the cut out plexigalss ones from William Optics, and the 3D printed ones) are pure junk. They are so imprecise that either the diffraction spikes have dispersion, or the central star is a complete messy blob, making it hard to determine when the diffraction spikes actually passes through the central point.

I have been happy with the laser etched plexiglass and the ones that are created on negative films. This is what you get from one of them:

Notice the tightness of a central star.

Instead, the cheap junk produces diffraction that typically look like this (big fat bright blob where a small central star should be):

https://en.wikipedia.org/wiki/File:Bahtinov_mask_example.jpg

Since the lines are etched on plexiglass (or printed on negative film), there are no extra superfluous lines in the mask to act as mechanical support -- those (result of the Fourier Transform) produces confusing diffration patterns, instead of a small tight, central star, which makes it hard to find the correct focus.

The etched masks have very fine structure, so they generate a humongous sized diffraction pattern. If you have not seen one, you will be in for a culture shock. So, you can really see when the diffraction spikes pass through the central star.

Most of these masks are made to be mounted on 77mm or 82mm camera filter threads (I have a 82mm one mounted inside Thousand Oaks solar filter cell that I have cut the solar film away -- makes a nice fit to my FSQ-85). They also come in the 100mm x 100mm variety for 100mm filter holders. So, you need to do a little work to make them fit an astrograph instead of a camera lens.

There are quite a few purveyors of the 100mm x 100mm and 77mm/82mm etched plexiglass ones, and they appear to all come from the same OEM, or else they are shamelessly copied. Among these are ones from Kase, Lonely Speck, SkyLabs, K&F, Kenko, etc. Don't search for "Bahtinov" on Google, but "Focusing Mask."

A Hungarian company ("Focus On Stars," a one man company, I believe) makes a very unique Bahtinov mask that is "printed" onto photographic negatives (those who have been doing astrophotography from the Tri-X days will know what I mean) and the structures are even more precise and smaller than the etched glass ones. The photo negative is then sandwiched between two thin 100mm x 100mm plexiglas plates. The size and brightness of the diffraction pattern from them are nothing short of shocking.

One thing (and this really only apply to camera lenses, since most OTA are slower than f/4) that you will notice of a typical three-section Bahtinov mask is that at the corners of a full frame sensor with fast (f/2.8 or faster) lenses, instead of the typical 6 spikes, you will only see 4 or two of them, because the light ray bundle does not hit all three sections of the mask. Well, "Focus On Stars" has a solution for that too. Because he can print very small structures, he has a version of his mask (for fast lenses), he created instead of one mask, a NxN mosaic of smaller masks. So the ligh rays from all stars will pass through at least one of the smaller NxN masks. Now, these only work with very fast lenses and do not work on slower OTAs. I have both (fast and slow) versions of his masks, and can confirm that.

So, shop around. I am not a YouTube shill, and refuse to recommend any single manufacturer, so do your own homework. The etched plexiglass ones are pretty much equivalent (at least the 4 that I tried even looked the same). The one from Hungary is more unique, but you don't have to use his if you don't crave even larger and brighter spikes from the etched plexiglass ones, or if you don't need his fast lens version.

Now, all that being said. I have discovered why ASIAIR "autofocus" pretty much consistently fails for certain OTA. Knowing its shortfall allows me to now use the autofocus function very reliably, just absurdly slow. But that is a different subject, not what you asked. I will gladly address it if someone else is interested on how to make ASIAIR autofocus work properly. When I have patience that evening, or too lazy to go outdoors to mount masks, and if I am using ASIAIR, I simply use its autofocus feature now.

Chen

retman

hello chen, ok, the "homework" begins. the "Focus On Stars" mask sounds very interesting.

i also own two EAF's that i use with ASIAIR so any education you wish to pass along regarding improving the autofocusing routines would be apprciated (and likely by others as well)

thanks again

w7ay

retman i also own two EAF's that i use with ASIAIR so any education you wish to pass along regarding improving the autofocusing routines would be apprciated (and likely by others as well)

No sweat.

First, we need to understand where ZWO screwed up. Then the second step is to find how to work around ZWO's algorithm. (This is the falacy of the "Easy as 1,2,3." A neophyte will never be able to figure out the numerous workaround, like the centroid weighting problem in autiuguiding, and why autofocus fails.

OK -- problem: ZWO uses a single exposure time to form the U curve.

As an aside, the U curve is also a big bug. From high school Physics, you will remember that geoemtric optics is basically a set of similar triangles, and as you move in and out of focus, the diameter of the misfocused spot changes linearly. Like like Newton's gravitational law, geometric optics is also a rough approximation of nature. As you get closer to focus, the spot does not become the infinitely small spot that geometric optics predicts, but becomes a point spread funtion in the form of an Airy disc. The smallest spot form a real world OTA is even larger than that predicted by diffraction optics. However, as you move away from focus, the spot does follow a linear change, and thus creates a V-curve. Mathematically minded folks will know of it as a hyperbola from a conic section. ZWO just uses a random second degree curve, and ends up with a parabola. They did not constarin the curve fit so that the hyperbola solution results. As such, they cannot predict the focus location as precisely as the hyperbola.

But, lets assume that the parabola is "good enogh."

The way they generate the parabola (U curve) is to measure the HFD of stars at different positions on the EAF.

But, notice that ZWO chooses to get all the HFD from a constant exposure value (for example 3 seconds). Since you cannot compute the HFD of saturated stars, they need to pick a star that has good enough SNR and yet not saturate. As you defocus a star, all the phons land on more pixels, and become dimmer. As a result, for a very defocused position, ZWO had to choose very bright stars, and for more in-focus locations, they have to choose dimmer stars since better focus wil saturate the brighter stars.

Notice what is happening... they are estimating the U curve by using different stars, at different locations in the frame of the sensor.

Now, imagine that your OTA or camera has a tilted image plane, or the backfocus is not perfect, the true focus of stars at different places in the FOV are different!

So, what to do?

First, make sure that your OTA/camera has no tilt. You can use a tool like Siril to check tilt and adjust. For a quick and dirty look at the tilt, take a 30 second exposure of a star rich region of the sky. Milky Way is good. The 30 seconds is to average out atmospheric turbulence. Now use the Star Detect tool in ASIAIR, and it will give you the HFD of different stars in the frame. If there is tilt, the HFD of the stars will be different. A 5%-10% varaince in HFD is probably "good enough" to do autofocus properly (not you need better to get good stars in ypour photos).

Once you have adjusted tilt, backfocus is now easy to adjust. Change the backfocus around 0.1mm at a time. A fast camera lens (or even the WO Pleiades 68 that I have) will need even finer backfocus adjustments, of the order of 0.05mm. (By the way, my WO comes in at 271mm focal length, and thus f/4, and not the f/3.8 that William Optics claim.) Backfocus is important because an incorrect backfocus will cause a defocused star to look like an oval instead of round, and therfore give bad point for the U curve.

Once tilt and backfocus is adjusted, you will find that the HFD of stars are now close to one another, and ZWO bug of choosing different stars for autofocus is no longer a problem since all the stars have the same HFD at each focus position.

So, get tilt and backfocus adjusted so that a 30 second exposure yields a HFD distribution that is off by no more than 5% to 10%, and ASIAIR will get the focus correct. At least within the accuracy of a U-curve.

Chen