Autofocus with DSLR + H-Alpha filter


Hi, I have been doing autofocus runs with my mod. Canon 550D and Robofocus (on an FSQ) without any problem (I usually get quite good V shaped curves). The settings I use are 2 sec. @ 1600 ISO. I have recently bought an EOS clip-on H-Alpha 12 nm filter. I have tried many settings (from 2 sec. to 10 sec. exposures and ISO 1600 to 6400; I also use darks) but continually get zig-zag curves (the HFR values are very inconsistent). I eventually have to focus “visually” (take one > check frame visually > change focus position…etc.) but even so, often it is difficult to decide what the best focus position is. I am aware of the difficulties H-Alpha filters create with DSLRs (Bayer matrix issues) but is there a recommended method/technique for autofocusing a DSLR with an H-Alpha filter?

Thanks for your help.



I do not have a solution but I am struggling with the same problem. I just use a broadband light pollution filter with my dslr. My first two point in the run is normally great but then I also get the sig zag Patten. Initially I though it might be a mechanical problem. So I tried to rework my focuser and it connection. This weekend I still got the same so I was thinking it might be a mechanical problem related to the position of the scope. But now I am wondering if the filter on the dslr is not affecting the calculation. I ill need to try and experiment this weekend.


It is nearly impossible to use AF with 2 filters in front of the CCD. You are only working with 25% of a stars data and the AF routine cannot handle this.


Thanks Ken. The second filter you refer to is the IR cut one or the Bayer matrix? I have succesfully used AF in SGPro with the mod. Canon 550D (with the IR cut filter installed) plus an IDAS LPS filter (two -or three?- filters in total). I assume the problem for AF is the narrow band of the H-Alpha plus the Bayer matrix (only 25% of all pixels recieve signal). Is this correct?




While Ken is right, you might try increasing exposure to see if you can get a workable image. 10sec sounds very short. When I ran Ha filters on my mono camera to get filter offsets, I’m usually taking 30-35s exposures on a Tak FSQ106.


Yes, this is correct. Just friendly advice (as I can’t say we officially support this type of configuration)… Bump your exposure time to 45-60s and then software bin AF frames at 2x2 to get rid of the bayer artifacts (software binning is found in the Canon settings dialog).


When I used to try to focus with my narrowband filters, I used 45 seconds with my STF8300 and a 3-nm HA filter. I’ve since given up on it and gone to offsets.

It doesn’t work very well.


Just for my own clarity please. Does the filter stack somehow effect the star profile or the calculation or is it simply a matter of less light and a reduced number of star. I use a board band filter and easily get enough star with a 6 t o10 sec exposure but I still struggle wit the zigzag pattern. I initially though that this is mechanical in nature but now with the comment made here it seem there is more to it. This got me slightly worried since I ordered a clip filter with a field flattener for my new configuration.


You’re not exposing the picture long enough to generate stars that SGP can do a HFR calculation on. So, it starts hunting and finds the noise in your picture and assumes it is stars. This is what gives you the ‘zig zag’.

Personally I’d just use a Bhatinov mask. Or, focus manually.


With my mod. Canon 550d and an IDAS LPS filter (at f5), I get nice V curves with 2sec. exposures @ 1600 ISO. No problem there (I just had to figure out the correct step size).

With the clip 12 nm H-Alpha filter, I have tried with 5 and 10 sec. exposures @ 6400 ISO and I can see perfectly plenty of stars in the image, but I still get zigzag curves (?)

I have tried 30 sec. exposures @ 6400 ISO and got a descending line (from right to left), no zigzag or curve.

I finally have decided to focus visually because it doesn’t seem to work with the H-Alpha filter, no matter how much exposure I use ( I don’t really understand why since I can see the stars in the image…)



It’s related to the noise at low exposures. You can increase your nebula rejection and reduce the number of stars required and give that a try.

They’re working on some new focus techniques. Maybe they’ll solve your issues. It’s been this way for a long time. Most of us just use offsets.


This is not relevant to determination of viable HFR metrics. Your eyes are not tuned to the anomalies introduced by using this setup. Centroids shift… data that is supposed to be bright is not (or is missing entirely). The stars you see are circular representations, but they are a mess when it comes to centroid determination and when determining the radius in which half the flux sits (even single pixel anomalies affect this calculation…). The data is just not there.


Ok, I understand. Thanks for the explanation Ken.