2022-10-04 01:52:42 <KipIngram   > boru: Hey, do you know if PixInsight has a (thorough and complete) Linux CLI?
2022-10-04 01:53:07 <KipIngram   > All these little widgets I slug around through the GUI to set up and run - seems like it ought to be possible to run them from a bash script instead.
2022-10-04 01:53:40 <KipIngram   > I had it crash twice on me tonight, while I was following along with a video tutorial.
2022-10-04 01:53:57 <KipIngram   > First time it just stopped, but the second time it rebooted the whole system.
2022-10-04 01:54:17 <KipIngram   > But man, it really produces some great results.
2022-10-04 06:26:50 <boru        > KipIngram; sort of. You can produce a named pipeline and then invoke that directly. Unfortunately, it's not really designed with that sort of workflow in mind. For that sort of stuff, I wrote some programs myself, use software written by observatories (I used to work at one), and rely on other tools.
2022-10-04 06:27:38 <boru        > I sort of started worked on a graph-based workflow program which, for the moment, relies on imagemagick and friends, but once I have the time, I'd like to turn that into something more serious.
2022-10-04 06:28:43 <boru        > For amateur astrophotography, PI is really great though, yeah. Sadly, FreeBSD support has been dropped because of Qt problems, so I'm a bit behind in processing data at the moment, plus I haven't really had time to get out and do acquisition much either.
2022-10-04 06:30:53 <boru        > In the software I was working on, I was gating the data output from nodes by various parameters e.g. SNR, PSF etc, then passing that data onto the next process. I'm not a GUI guy, but it was fun to work on; each process was a standalone program with a bunch of parameters, and the GUI just built the CLI in the background.
2022-10-04 06:33:26 <boru        > Also, we run a small channel here on libera -- ##astroimaging -- low traffic, but we discuss data and processing a lot.
2022-10-04 10:48:50 <KipIngram   > boru: That's a little sad.  I think "that kind of workflow" should be the STARTING POINT for systems like this.  Then the GUI would get mounted on top of that.
2022-10-04 10:49:36 <KipIngram   > I'm somewhat familiar with ImageMagick.
2022-10-04 10:50:20 <thrig       > Apple tried to bolt AppleScript onto its GUI ...
2022-10-04 10:50:29 <KipIngram   > They do have this script for stacking and calibrating raw images, so that got written somehow.  I think I caught something in the forum, though, about scripts needing to be "signed."  Does that mean only scripts they produce can work?
2022-10-04 10:50:39 <KipIngram   > I don't even remotely see the point of such a limitation.
2022-10-04 10:51:24 <thrig       > does this run in the cloud or something
2022-10-04 10:51:32 <KipIngram   > Anyway, it seems like a perfect opportunity for a command line interface; most of what you're doing is setting a bunch of parameters in a dialog and applying that operation to an input image to get an output image.
2022-10-04 10:51:51 <KipIngram   > So "operation -options input_file output_file."
2022-10-04 10:53:23 <KipIngram   > I do think they've done a fairly good job designing the UI, though.
2022-10-04 10:54:21 <KipIngram   > I'm usually pretty hard to please on such things.  Maybe I'm just in a good mood because I'm on vacation, but I feel like they deserve fairly good marks.
2022-10-04 10:54:24 <boru        > PI is more like a photo editing app for pretty pictures, as opposed to processing scientific data.
2022-10-04 10:54:52 <KipIngram   > Yeah, I'm gradually picking that up.  I delved into it primarily for the stacking, but the other stuff is nice to have too.
2022-10-04 10:54:56 <thrig       > there's also uh ImageJ (never used it)
2022-10-04 10:55:18 <KipIngram   > A lot of the tutorials, though, talk about using Deepsky Stacker for stacking and then PI for the final image.
2022-10-04 10:55:23 <boru        > It's nice to hack around with. PixelMath and wavelet transforms are infinitely useful.
2022-10-04 10:55:28 <KipIngram   > I don't really see why they do that, since PI will stack.
2022-10-04 10:55:36 <KipIngram   > And DSS is windows only (yuck).
2022-10-04 10:55:56 <KipIngram   > Yeah, I have a lot to learn about those aspects.
2022-10-04 10:56:07 <boru        > Depends on what you're stacking. PI is bad at solar system data, and depending on your image scale and atmospheric seeing, not very good with deep sky objects either.
2022-10-04 10:56:08 <KipIngram   > The stacking, the histogram stuff, that all makes great sense to me.
2022-10-04 10:56:21 <KipIngram   > But noise reduction and related stuff - I know little about.
2022-10-04 10:56:26 <boru        > Yeah, spatial domain stuff is straighforward.
2022-10-04 10:56:38 <boru        > Wavelets can help with noise as well.
2022-10-04 10:56:57 <KipIngram   > Yeah, those more esoteric things seem to make a huge difference in final quality.
2022-10-04 10:57:32 <boru        > NR is mostly fire and forget. However, the best NR is better SNR, ergo, more data.
2022-10-04 10:57:35 <KipIngram   > It's really pretty remarkable what they can do with these image sensors, even in applications they weren't really designed for.
2022-10-04 10:57:43 <boru        > Indeed.
2022-10-04 10:57:54 <KipIngram   > Consume cameras are meant to be used in situations where you have "plenty of light."
2022-10-04 10:58:02 <KipIngram   > Consumer
2022-10-04 10:58:09 <boru        > Especially newer CMOS sensors e.g. electron noise levels etc.
2022-10-04 10:58:21 <KipIngram   > I guess when you have a lot of light it "covers up" some of the imperfections.
2022-10-04 10:58:51 <boru        > I'm still using an old CCD: KAF-8300. Not many Kodak CCDs still around anymore, sadly. QE is decent around the H-alpha emission line, and others, which is why I bought it; I mostly do narrowband imaging.
2022-10-04 10:59:10 <boru        > More light == more data == better SNR.
2022-10-04 10:59:48 <boru        > You've probably seen amp glow/thermal noise in your data. You can use dark frame calibration to get rid of that, but it's a real problem in older CCDs without a TEC.
2022-10-04 10:59:54 <KipIngram   > A few months ago I bought an inexpensive linear CCD; had it in mind that I might want to try to make a camera of my own that uses Earth's rotation to move the target across the sensor.
2022-10-04 11:00:04 <KipIngram   > Just capture line after line after line, and put them together in software.
2022-10-04 11:00:17 <Zarutian_iPad> Was not there something called Foven 3X sensor?
2022-10-04 11:00:17 <boru        > Ah, yes, survey astronomy is fun.
2022-10-04 11:00:26 <KipIngram   > That would be like having a BIG sensor for the cost of a cheap linear sensor.
2022-10-04 11:00:30 <boru        > I built a radio telescope to do that with a friend.
2022-10-04 11:00:44 <KipIngram   > Ah, yeah - radio astronomy has intrigued me too.
2022-10-04 11:00:50 <boru        > Don't do it
2022-10-04 11:00:51 <KipIngram   > My problem is that too many things intrigue me.
2022-10-04 11:00:55 <boru        > Exactly
2022-10-04 11:00:59 <boru        > This is my problem also.
2022-10-04 11:01:04 <boru        > So many unfinished projects...
2022-10-04 11:01:12 <KipIngram   > Yes.
2022-10-04 11:01:30 <KipIngram   > So, why "don't do it"?
2022-10-04 11:01:36 <KipIngram   > Out of curiosity.
2022-10-04 11:01:38 <boru        > You're welcome in ##astroimaging -- people like to talk shop there a lot.
2022-10-04 11:01:53 <boru        > Because astrophotography is already a major time/money sink.
2022-10-04 11:01:53 <KipIngram   > Cool - I'll definitely check that out.
2022-10-04 11:02:04 <KipIngram   > Ah.
2022-10-04 11:02:29 <KipIngram   > I have a CCD camera I got with my 14" Dobsonian, but I've never been able to get it to work.
2022-10-04 11:02:35 <boru        > As you've probably already seen, the sky is the limit (pardon the pun), wrt the cost of "amateur" equipment.
2022-10-04 11:02:38 <KipIngram   > Linux can "see it," but won't actually engage it.
2022-10-04 11:02:54 <KipIngram   > Yes, that's quite evident.
2022-10-04 11:02:55 <boru        > Check out EKOS; it uses indilib under the hood.
2022-10-04 11:03:04 <boru        > It's a part of kstars now.
2022-10-04 11:03:13 <boru        > Probably the most mature of the indilib-based acquisition tools.
2022-10-04 11:03:28 <KipIngram   > I have a little gadget I bought that goes on my tripod, and will rotate the camera.  It has a setting for 15 degrees an hour, which isn't quite right but it's awfully close.
2022-10-04 11:03:42 <KipIngram   > I feel like that should let me at least extend my exposure times some.
2022-10-04 11:03:49 <KipIngram   > It's 99.7% right.
2022-10-04 11:03:58 <KipIngram   > Minus whatever inaccuracies it's got.
2022-10-04 11:04:13 <boru        > Field rotation is less of a problem away from the celestial north pole.
2022-10-04 11:04:15 <KipIngram   > I'm going to try to mount a red dot finder on the side of it for polar alignment, and see how that goes.
2022-10-04 11:04:26 <boru        > What mount have you got?
2022-10-04 11:04:33 <boru        > Is it an EM or Alt/Az?
2022-10-04 11:04:37 <KipIngram   > Oh, on my telescope it's just a big Dob.
2022-10-04 11:04:41 <boru        > If it's the latter, get a wedge.
2022-10-04 11:04:48 <KipIngram   > It does have GoTo, though.
2022-10-04 11:04:50 <boru        > Yeah, you can get a wedge for the rocker box on your dob.
2022-10-04 11:04:54 <boru        > That will fix field rotation.
2022-10-04 11:05:00 <KipIngram   > Yes.
2022-10-04 11:05:11 <KipIngram   > There are also those "equatorial platforms" out there.
2022-10-04 11:05:20 <KipIngram   > With construction instructions.
2022-10-04 11:05:30 <KipIngram   > It's a pretty big scope, though.
2022-10-04 11:05:32 <boru        > Yeah, EMs are...challenging to use with a Newt/Dob riding on there.
2022-10-04 11:05:44 <KipIngram   > The little gadget I just mentioned is just for the camera, though.
2022-10-04 11:05:53 <KipIngram   > Sits on top of a standard tripod.
2022-10-04 11:06:06 <boru        > To be honest, for deep sky, you don't need an OTA with such a long FL. And if you do, you're better off with a catadioptric rather than a reflector.
2022-10-04 11:06:17 <boru        > Yeah, I know them./
2022-10-04 11:06:21 <KipIngram   > I'm interested in pattern matching - using a star catalog to identify the stars in a picture.
2022-10-04 11:06:35 <KipIngram   > I know that's out there in various packages, but I'm interested in implementing it myself.
2022-10-04 11:06:43 <boru        > The term you're looking for is "plate solving" in astronomy nomenclature.
2022-10-04 11:06:51 <KipIngram   > Seems like a challenging "data strucres + algorithm" p roblem.
2022-10-04 11:06:53 <KipIngram   > Right.
2022-10-04 11:07:07 <boru        > I developed some algorithms to do precisely that for star tracking on satellite ADCSs.
2022-10-04 11:07:28 <boru        > Coarse, computationally inexpensive stuff you can do on a microcontroller.
2022-10-04 11:07:39 <KipIngram   > The initial thought I had on that was that if I could do it, I could just take pictures of the sky over a period of weeks / months, as a planet moves by, and reduce the data to determine the planet's orbital elements.
2022-10-04 11:07:40 <boru        > Since you have other inputs on most ADCSs.
2022-10-04 11:07:56 <boru        > With some degree of error for the planets, yes.
2022-10-04 11:08:07 <boru        > For planets, we can only compute them accurately for about 2 years.
2022-10-04 11:08:17 <KipIngram   > Right, but I figured the more observations I took the more I could minimize the error statistically.
2022-10-04 11:08:20 <boru        > Ephemeris is annoying.
2022-10-04 11:08:34 <boru        > You would think so, but sadly not well.
2022-10-04 11:08:44 <KipIngram   > That is sad.
2022-10-04 11:08:56 <boru        > PI can plate solve, fwiw.
2022-10-04 11:09:06 <boru        > It uses astrometry.net
2022-10-04 11:09:11 <boru        > Which you can also run locally.
2022-10-04 11:09:16 <KipIngram   > Yes.  I haven't gotten it to do that yet, but I saw it in the script for stacking.
2022-10-04 11:09:29 <boru        > It's also useful from framing across several sessions.
2022-10-04 11:09:39 <boru        > s/from/for
2022-10-04 11:09:49 <KipIngram   > Ah, that makes sense.
2022-10-04 11:10:12 <boru        > Most acqusition programs will focus, rotate and frame automatically.
2022-10-04 11:10:26 <boru        > And most use a plate solver for that.
2022-10-04 11:10:43 <KipIngram   > Anyway, this little rotary gadget - if everything was perfect (perfectly aimed, it does exactly 15 degrees an hour, etc.) then it felt to me like it should get me from being allowed, say, one second exposure up to about six minutes.
2022-10-04 11:11:06 <KipIngram   > That was just how long the math said it would take for it to lag one second of arc behind the sky.
2022-10-04 11:11:06 <boru        > It will be a function of your OTA's focal length.
2022-10-04 11:11:18 <KipIngram   > Well, yes - that's where the initial one second came from.
2022-10-04 11:11:24 <boru        > 
ACTION nods.

2022-10-04 11:11:25 <KipIngram   > That was just a sample value.
2022-10-04 11:11:57 <KipIngram   > Obviously, I won't get it pointed exactly right, and its rate won't be exactly right.
2022-10-04 11:12:10 <KipIngram   > But if I could even move up to 30 seconds - 1 minute exposures, that would be nice.
2022-10-04 11:12:14 <boru        > It's neat that it works. You could also try solar system objects, which actually benefit from as short exposures as possible, to account for local seeing and instability. Jupiter is at closest approach in the last 60 or so years.
2022-10-04 11:12:21 <KipIngram   > And I wouldn't have to re-aim the camera as often.
2022-10-04 11:12:51 <KipIngram   > Yeah, I saw that last week.
2022-10-04 11:13:16 <boru        > You can also only bracket Jupiter for about 5 minutes before the rotation becomes apparent in your data.
2022-10-04 11:13:38 <KipIngram   > Yeah, I've seen software to correct for that (as best it can).
2022-10-04 11:14:04 <boru        > If you can issue an RoI readout on your sensor and manage about 80-100fps, you'll get some really nice data.
2022-10-04 11:14:22 <KipIngram   > RoI?
2022-10-04 11:14:29 <boru        > Sorry, Region of Interest.
2022-10-04 11:14:44 <boru        > Subframe programming, essentially.
2022-10-04 11:14:54 <KipIngram   > Ah.
2022-10-04 11:15:20 <KipIngram   > Of couse, my skies are kind of sucky.
2022-10-04 11:15:35 <boru        > You'll really see how soupy the atmosphere is when imaging the planets, and most solar system data reducers are quite good at discarding the mushy data.
2022-10-04 11:15:41 <KipIngram   > A guy once told me, though, that in areas like mine some folks get quite good results using narrow band filters.
2022-10-04 11:15:52 <boru        > It depends.
2022-10-04 11:16:12 <boru        > An atmospheric dispersion corrector helps for planets when they're low in the sky
2022-10-04 11:16:13 <KipIngram   > It always seems to.  :-)
2022-10-04 11:16:48 <boru        > But you'll want to get away from human dwellings since local differences in air density from central heating systems and just solar warmed concrete will really ruin your local seeing.
2022-10-04 11:17:25 <boru        > After that, it's just watching the weather. Low instability is good, but bad if you have an inversion layer.
2022-10-04 11:18:00 <boru        > I worked in adaptive optics for some big instruments for a while. The atmosphere is annoying.
2022-10-04 11:24:21 <KipIngram   > Yeah, I've looked at light pollution maps.  I think my skies are Bortle seven.  There's a place about 12 miles from me that looks a good bit better, at least shooting to the west; that would be reasonable for  "quick outings."
2022-10-04 11:24:46 <KipIngram   > And I found a vacation lodge out west of San Antonio that was Bortle 2; I may occasionally drive out there for a weekend.
2022-10-04 11:25:23 <KipIngram   > I am interested, though, in exploring just what I can accomplish from the back yard.
2022-10-04 11:25:28 <KipIngram   > Since it's "what I've got."
2022-10-04 11:26:14 <KipIngram   > Shooting east from that 12-mile away place will probably be non-feasible.
2022-10-04 11:26:22 <KipIngram   > Because... Houston.
2022-10-04 11:26:48 <boru        > We all start somewhere, man. Bortle 1/2 is nice, but you can still get good results depending on where you are e.g. low FWHM narrowband filters will almost completely filter out sodium light pollution, yielding pretty decent results even in mid-Bortle skies.
2022-10-04 11:27:03 <KipIngram   > My first goal is modest - I just want to make a picture that shows the Milky Way.  Even a little will please me.
2022-10-04 11:27:04 <boru        > But at Bortle 7, you can still image solar system objects.
2022-10-04 11:27:26 <boru        > Widefield has its own challenges, but yeah, that's doable.
2022-10-04 11:27:39 <boru        > Check out the DBE tool in PI -- it's great for removing sky gradient.
2022-10-04 11:28:17 <KipIngram   > Yes, I was using DBE last night.  It's pretty impressive.
2022-10-04 11:28:31 <KipIngram   > And will be like step 1 for almost anything I do from here at home.
2022-10-04 11:29:17 <boru        > The only thing that can bite you with the DBE tool is if you have any vignetting in your data. Then I'd recommend using flat frame calibration first.
2022-10-04 11:29:45 <KipIngram   > ah - the tutorials tried to use DBE division for that.
2022-10-04 11:30:13 <boru        > Always calibrate data first before everything, even registration/stacking.
2022-10-04 11:31:04 <boru        > When you get to longer exposures, you'll easily see the need for dark frame calibration.
2022-10-04 11:31:04 <KipIngram   > Yes; before trying the reduction script I made a pass over some raw images by hand, just to get a feel for the flow.
2022-10-04 11:31:23 <KipIngram   > Yeah, all that makes total sense to me.
2022-10-04 11:31:27 <boru        > 
ACTION nods.

2022-10-04 11:31:39 <KipIngram   > I ordered a little tracing pad to use for flats.
2022-10-04 11:31:48 <KipIngram   > Backlit panel.
2022-10-04 11:32:07 <boru        > Keeping a library of calibration frames is good. I have master dark and master bias frames from -40C to -10C in step increments of 5C.
2022-10-04 11:32:08 <KipIngram   > Probably could have handled that with my phone, but it was only like $11, so I figured why not.
2022-10-04 11:32:18 <KipIngram   > Oh, nice.
2022-10-04 11:32:35 <boru        > Flat frames are deceptively difficult, actually.
2022-10-04 11:32:47 <boru        > Getting an evenly illuminated surface is really not easy at all.
2022-10-04 11:32:56 <boru        > I would recommend so-called "sky flats"
2022-10-04 11:33:02 <KipIngram   > The instructions I saw said to rubber band a t-shirt over the lens, and lay the white panel on top of that.
2022-10-04 11:33:12 <boru        > At sun set, point your scope at the zenith, and take your flats that way.
2022-10-04 11:33:22 <KipIngram   > Ah, before any stars show.
2022-10-04 11:33:35 <boru        > Any whilst the zenith still has even polarisation.
2022-10-04 11:33:43 <KipIngram   > I watched a bunch of videos from a guy named Nico Carver.
2022-10-04 11:33:52 <KipIngram   > He's... prolific on the video front.
2022-10-04 11:34:04 <boru        > Hmm, don't know the name.
2022-10-04 11:34:34 <KipIngram   > https://www.youtube.com/c/NebulaPhotos
2022-10-04 11:35:01 <boru        > This guy taught me an awful lot about narrowband imaging: Richard Crisp: https://www.narrowbandimaging.com/home_page.htm
2022-10-04 11:35:17 <KipIngram   > I bought a book a few months ago that goes pretty deep into the engineering and physics aspects of sensors.
2022-10-04 11:35:17 <boru        > He's very responsive, albeit a bit back handed, by email.
2022-10-04 11:35:30 <boru        > He can back it up though. He's a pioneer.
2022-10-04 11:35:49 <KipIngram   > Haven't really digested it yet, but it seems close enough to my expertise that I ought to be able to follow it all.
2022-10-04 11:35:56 <KipIngram   > And I like that kind of technology deep dive.
2022-10-04 11:36:05 <boru        > It's a deep field, again, pardon the pun.
2022-10-04 11:36:13 <KipIngram   > Heh heh.
2022-10-04 11:36:37 <boru        > Even just photon depletion at different wavelengths in different materials is something you could spend a life time studying.
2022-10-04 11:36:48 <KipIngram   > So I start thinking about stuff like this, and then start thinking about building my own drive and so on, and of course programming it in Forth.
2022-10-04 11:37:01 <boru        > I can relate.
2022-10-04 11:37:09 <boru        > I built a GEM with a mechie friend of mind.
2022-10-04 11:37:20 <KipIngram   > And Forth actually has that pedigree.
2022-10-04 11:37:25 <boru        > Indeed.
2022-10-04 11:38:46 <boru        > Alas, it's the lack of time again.
2022-10-04 11:39:15 <KipIngram   > This was the tutorial that caused PI to crash (twice):
2022-10-04 11:39:16 <KipIngram   > https://www.youtube.com/watch?v=EGw9zFqlmKk
2022-10-04 11:39:30 <KipIngram   > He has the starting image on Google Drive; I just walked along with him.
2022-10-04 11:39:43 <KipIngram   > Oh, I'll tell you one thing I found difficult.
2022-10-04 11:39:47 <boru        > OOM killed?
2022-10-04 11:40:05 <boru        > Depending on your thread stack parameters, PI can run out of memory pretty quickly.
2022-10-04 11:41:02 <KipIngram   > He used ATrouswaveletTransform in one step, and it involved making a mask that showed the stars and the nebula but blocked everything else.
2022-10-04 11:41:18 <boru        > That tool is deprecated now, even though it's so damn useful.
2022-10-04 11:41:19 <KipIngram   > And he got his result by diddling with the arrows in the histogram transform tool.
2022-10-04 11:41:39 <KipIngram   > But I couldn't get there that way - I had to squint at the screen and copy out his numerical parameters.
2022-10-04 11:41:44 <KipIngram   > Then mine looked just like his.
2022-10-04 11:41:57 <boru        > Star masks are pretty straight forward.
2022-10-04 11:42:01 <KipIngram   > But two of those arrows were way over at the left, and involved VERY precise settings.
2022-10-04 11:42:12 <boru        > You can do it visually, also.
2022-10-04 11:42:25 <KipIngram   > Yes, but this was a star and nebula mask, and it was getting the nebula part to match his that was delicate.
2022-10-04 11:42:43 <boru        > Ah, right, so he selected the large structures with the wavelet transform.
2022-10-04 11:42:50 <KipIngram   > Anyway, I just felt like I'd wind up having to use the numeric fields instead of the controls.
2022-10-04 11:42:53 <boru        > That's essentially what the multiscale wavelet transform does.
2022-10-04 11:43:28 <KipIngram   > It's going to take m forever to get familiar with everything PI bundles.  That's a crap load of tools.
2022-10-04 11:43:51 <boru        > It is. At least there's documentation now. I've used it since the early days, and it used to have zero documentation.
2022-10-04 11:44:06 <boru        > You'd need a deep understanding of DSP and image processing to use it back then
2022-10-04 11:44:09 <KipIngram   > But following along the tutorials is fun - each matching step is gratifying.
2022-10-04 11:44:21 <boru        > Understandable.
2022-10-04 11:45:15 <KipIngram   > Also, the way that last guy I linked used previews as input to these various tools was interesting; not something I think I would have figured out on my own.
2022-10-04 11:46:09 <KipIngram   > The Blink tools is fun - I loaded one guy's lights into it and ran it - neat seeing the rotation shift right there on the screen.
2022-10-04 11:46:24 <boru        > That brings me back to what I said about PI earlier; it's a tool which requires visual feedback, ergo, its "pipeline" interface is quite limited.
2022-10-04 11:46:28 <KipIngram   > Also seems like a good way to spot bad frames if you have a large number of lights.
2022-10-04 11:46:59 <boru        > Yeah, selecting good rejection parameters for data will save you a lot of time, especially if you have a lot of data to reduce.
2022-10-04 11:46:59 <KipIngram   > Yeah, I figured the scripting would mostly be after I knew exactly what my workflow for a particular type of picture would be.
2022-10-04 11:47:16 <KipIngram   > But of course, if a step relies on previews, those need to be chosen at the time.
2022-10-04 11:47:25 <KipIngram   > Like one of them was a small area with no stars.
2022-10-04 11:47:26 <boru        > I think that will depend heavily on the target, also.
2022-10-04 11:47:33 <boru        > Right, amongst other things.
2022-10-04 11:47:48 <boru        > It's fairly straightforward when the data is still linear, but after that, good luck.
2022-10-04 11:48:13 <boru        > You also have a lot of variables that larger more precise instruments don't have.
2022-10-04 11:48:41 <boru        > Anyway, I have to clock out for a bit. I'll be back later.
2022-10-04 11:48:54 <KipIngram   > Hey - nice chat; thanks for sharing your expertise.
2022-10-04 11:48:59 <KipIngram   > Stay safe!
2022-10-04 12:00:36 <KipIngram   > Oh, you know, I misspoke earlier.  I said something about "DBE division."  But that was actually the ABE tools - I played with both last night, and also so one guy on YouTube say "never use ABE" and another guy say "ABE is great."  :-)
2022-10-04 12:01:04 <KipIngram   > Probably another of those things that depends on what kind of stuff one is shooting.
2022-10-04 12:08:55 <KipIngram   > Anyway, regarding the writing of a plate solver, if I try that I want a database *on my computer*.  I don't want to have to go off to the net to get at data.  But on the other hand, I'm not really looking to do this in a way that will support high magnification shots through the telescope.  I'm thinking mostly of stuff I'd shoot from a tripod with a normal camera lens (maybe a small telephoto, but not huge -
2022-10-04 12:08:57 <KipIngram   > say 200mm-500mm).  So I'd only need to handle the brightest stars.
2022-10-04 13:35:44 <Croran      > is there any potential for Forth on quantum computers?
2022-10-04 13:35:58 <Croran      > i noticed IBM is giving free access to their quantum computers on the cloud
2022-10-04 13:39:12 <Zarutian_iPad> not sure how that would work with Forth
2022-10-04 13:39:23 <eris[m]1    > croran: dont thinka nything about tjem works woth forth
2022-10-04 13:57:46 <Croran      > eris[m]1: ah.
2022-10-04 17:23:04 <KipIngram   > Quantum computers are very, very unlike classical computers.
2022-10-04 17:23:29 <KipIngram   > You can implement any classical digital circuit in quantum form, so you could build a "quantum classical architecture computer."
2022-10-04 17:23:37 <KipIngram   > But that's really not how "quantum computers" work.
2022-10-04 17:23:48 <KipIngram   > They're actually more like... "co-processors" or something.
2022-10-04 17:24:29 <thrig       > it's floating point Jim but not as we know it
2022-10-04 17:24:52 <KipIngram   > What you do is create an initial state, and then in most algorithms you apply a "hadamard" gate to that to make it a uniform superposition.  Then you apply some unitary transformation you'e designe, and then you  measure the outputs.
2022-10-04 17:25:18 <KipIngram   > The idea is to find a unitary that will make the probability of the measured result being the desired solution a lot higher than the probability of other results.
2022-10-04 17:52:19 <KipIngram   > So I guess you could think of it as more akin to an analog computer than a digital one.  With analog computers you establish the initial conditions of differential/integral equations, and then just let the circuit evolve.  In that case you continuously record the outputs; in a quantum computer you only do the final single measurement at the end, and you want it completely isolated from "the world" during the
2022-10-04 17:52:21 <KipIngram   > interim.
2022-10-04 17:52:32 <KipIngram   > That isolation is our main problem with building them; it's hard.
2022-10-04 17:52:58 <KipIngram   > If you "peek" during the interim period you destroy the quantum state.
2022-10-04 17:59:53 <thrig       > many cats may or may not have been slaughtered in the construction of this quantum computer