I’ll hop back on here and bump this thread again as well, as this remains one of the more frustrating parts of using SGP. I just came from a thread where someone was asking for similar improvements and being told he must be doing things wrong. This thread was the first suggested thread after that, so at least the forum software recognizes there’s a problem! 
Beo
It has indeed been months and no further feedback on this topic. Is it safe to assume that the AF functionality we have in SGP is simply what we have and we should not expect any improvements? Even with one of the most expensive focusers on the market, the NiteCrawler, on one of the better scopes on the market, the FSQ106, I have still had bouts of unreliable focus with SGP. Same exact kinds of problems I have had with other systems, so I am pretty sure the issue is not the hardware. The hardware is mind bogglingly precise and repeatable in my own manual testing. This is the one area where I think SGP’s lack of progress could push me onto other options in the not too distant future… :’(
Well lately AF become once again a big subject, a lot of recent topics related with AF are once again opening the discussion.
Jared told us that something is coming… what we are missing is a concrete timeline, this can happen in the next beta or in 5y from now which once again is kinda of a downer.
I personally love SGP. It has changed everything I do with my astrophotography. That said, I don’t think the software is profitable enough for the SGP developers to make a full time living adding new features. I’m amazed at how much they do responding to issues raised by users.
I see this software as more of a “community” venture than a “commercial” venture.
I wish there was a way we could funnel money to the developers to help them respond to requests for new features. I guess I’m thinking of a “GoFundMe” approach. For example, I would donate $25 in a minute to have the developers work on the autofocus issue.
I’m not sure what a reasonable rate for highly experienced and skilled developers would be in the US but $25 per hour seems low to me.
Also money may not be the issue, If you are holding down a full time job then your spare time may be priceless to you. I don’t know if this applis to Ken and Jared, they may be freelance and see this as part of their normal work but even then they would need to see this in commercial terms and get a reasonable return for the time this takes.
I am a retired professional software developer with a lot of experience in this sort of development and my rule of thumb for estimating is that you always underestimate the size of this sort of problem and the less you know about it the more you underestimate it.
My opinion is that a robust fix to focusing would take of the order of months of work to get a fully developed, tested and released solution. It would probably take two to three weeks just to get an idea of the scale of the problem.
$25 an hour for 40 hours a week is $1000 a week, $4000 a month.
What I’m saying is that you are asking a lot from people who are probaby doing this more as a labour of love even if you are paying a nominal amount, and demanding hard time scales is asking a lot.
You misunderstood me I think. I’m talking about crowd sourcing. I’m thinking perhaps 500 people each donating $25.
See here for some SGPro AF stuff: SGPro 3.1 and Early 4.0 News - Sequence Generator - Main Sequence Software
I understood that. Not sure there’s enough people willing to committ.
Let’s see what 3.1 and 4.0 produce.
Not sure there’s enough people willing to commit.
I think it would be worth trying.
Bob
I already payed for the license… I will not pay again to get bugs fixed.
If this was an open source software then crowdfunding and bounty hunting would make sense.
With all due respect Chris to your lot of experience in this sort of development, I think you are overestimating the time and effort of fixing the problems with the focus routine by maybe a factor of 100. Too many folks think it requires a major rewrite of the routine to fix the problems. This is totally false. The reason it is false is that the current routine is:
The current routine being effective and the best focus routine on the market are well attested to by the numerous users on the forum that continue to tell us the current routine works perfectly for them. It works perfectly for me too, when conditions are good.
So fixing the routine to deal effectively with bad conditions does not require a complete rewrite, it requires a few targeted TWEAKS. I have presented in detail in another post exactly what minor adjustments to the routine will solve the most serious complaints folks have.
The #1 recommendation Ken has already committed to implementing in the current release. This enhancement allows the user to specify a minimum Quality factor that the curve must pass, or some corrective action will occur. I would venture that the hardest part to implement is producing this Quality factor. (see #3 below)
My #2 recommendation would take Ken about 30 minutes to implement. Its effectiveness in fixing the globular problem is unknown, but can only improve the accuracy of all the focus results, not just for globulars.
My #3 recommendation was to change the straight line intersection approach for finding the best focus position to a Quadratic fit( ie. a parabola), which offers major improvements to the accuracy of the best focus result. I estimated this would be the most time consuming enhancement of the three. Sure enough, it took me several hours yesterday to implement the Quadratic fit routine in C++. I will shortly send Ken and Jared my routine, which they should be able to easily plug in to their code in place of the straight line approach. Before doing so I will post here numerous examples of the results it gives compared to the current routine. ie. thorough testing.
A wonderful output of the Quadratic fit is the R factor, which is the correlation coefficient or the perfect measure of how well the focus data points fit the best fit parabolic curve. This is exactly the Quality measure that Ken will need to implement #1.
NOTE: I am also a retired professional software developer with a lot of experience in this sort of development. Well, not totally retired.
Fine, prove me wrong. It will only tae until Monday.
Being pragmatic here… but I think the discussion is going a bit pointless now.
Chris, you might want to re-read the quote you posted. $25 a minute not hour.
Ken, thanks for the update! Glad to hear that you guys are looking into some AF improvements. The improvements mentioned are great… I am curious what you guys might be able to do about speed as well. One of the difficult things about SGP AF is it is very slow, particularly with narrow band. A reliable AF cycle with SGP and narrow band filters currently requires about 15-20 second exposures (at around f/4!!) to get strong enough star signals to avoid noisy curves. It also takes about 7-9 steps, which means the AF routine can take several minutes to complete.
This has been the case with both a Canon 600mm f/4 lens, as well as a Tak FSQ106 f/5 scope. The slow speed of the process is eating into a lot of my imaging time, especially earlier in the evenings when I need to focus once every 15-20 minutes (which is usually about every 2-4 5 or 10 minute frames). Colorado tends to have a very large temperature differential, and can easily go from 90 degree days to 60 degree (or even colder) nights, especially in spring and fall. I’ve had it go from over 90 during the day to low 50s at night in early fall at times, which often forces me to focus fairly often for most of the night.
Efficiency is really the big thing…AF should not just be reliable, but it should also be fast. CMOS cameras with their variable gain and video modes might offer some potential options here. Higher gains with video might allow for some very fast focusing…
If it could help spur development along, I’d be all for some crowd sourced funding. I like SGP. It’s a one-stop shop for sequence operation, and it is not all that complicated compared to some of the alternatives…and the price is difficult to beat (and free software always comes with the risk of abandonment as well.) I think if the AF issues are resolved, it’ll remain a very solid imaging platform for some time to come.
It should be possible to determine the offset between the luminance and narrow band filters, then focus using the luminance and offset to the NB filter position. The shorter exposures needed by using 100% of the light rather than 5% or less will help a lot.
As for the efficiency vs speed debate what looses more imaging time, having a slow autofocus that works or a fast one that doesn’t?
I have offsets configured already. They generally work, however at very fast f-ratios they often don’t work well enough. I image at f/4 a lot, and will be imaging at f/3.6 and f/3 soon enough here. With faster f-ratios, I find that the tolerances with offsets and backlash, even with backlash compensation, is too tight, and focusing with each narrow band filter gets me visibly sharper results. At slower f-ratios, f/5 or slower, offsets work better. I find that at the faster f-ratios, I still need relatively long exposures (still 15 seconds, in fact 20 seconds if often not quite long enough at f/4 but I make do), and I do need to focus with each filter for best results.
Agreed, especially since FM used to be free.
Jon - I’m not quite sure I follow. Backlash (and compensation) is usually present with any move of the focuser - caused by an offset move in one direction or the final move to the focus position in an AF run.