New CCDCIEL release initial notes - DSLR only

[stash_old]

Patrick Chevalley has just released a new version of CCDCIEL 0.9.60 and as I have been following this and hoping I thought I would give it a go again.

Test

1. a.running Windows version of CCDCIEL on old Vista test laptop connected to Ascom Eqmod sim

    b. Connected to Canon DSLR on old RPI3 (not + or 4) running Indiserver from command line

Results:-

    I was able to connect to both the Canon DSLR (on RPI) and Ascom Mount Eqmod sim no problems - obviously the camera connection has a slight time lag of maybe 1/2 sec before it connects.    

    I did straight forward captures (all dslr images come as FITS mono or colour) of varying exposures 0.001 to 30secs - all worked and images were saved on the Vista box and display on the Vista screen inside CCDCIEL. About 20-30secs from mirror closure o download,convert to Fits and display/save over wired ethernet.

   So moving on and upwards I started some more captures using the very simple to use Plan and sequence editor in CCDCiel.

   This time CDC is loaded to do 2 things :-

       i. Watch the scope(ok sim scope) travel to the correct object.

      ii. Be able to create objects in the Plan/Sequencer just by clicking on"Planetarium" button (this pops up a window showing any object your click on in CDC)

On my test I had 2 objects and 2 capture plans and overrode rise times etc. Both worked without a hitch - very simple and very easy to do.

Pros:-

1. The ability to use a Canon DSLR remotely on a non Windows PC (RPI Linux) using Indi - this cannot by done by any other software simply. APT or anything else I know cannot do this at this moment in time.

2. As well as using the DSLR remotely I was also using Ascom so I could connect to any local Ascom device - mount,CCD,Focuser etc. So best of all worlds for me using DSLR

3. I could have mixed devices even more - e.g. Mount (or other devices) via EQMOD etc  

4. I can run Platesolving - Platesolve2,ASTAP etc on my fast kit taking the load of smaller,older RPI's

5. The Sequencer/Plan editor is very simple and powerful but may have some hidden problems as this was a quick test in daylight.

6. Lots of options - Dome/roof,Weather etc etc

CONS:-

1. You cannot set DSLR ISO from the Capture control screen - this has to be done via the Indi settings. Hopefully this will change soon. Patrick did mentionsome option which showed the ISO on the capture screen but I couldn't find it.

2. The DSLR has to be on Indi/Indigo as no direct connection of a DSLR on Windows is possible (as far as I can see)  CCD's are well supported via both Indi and/or Ascom but no native connections appear.

So for DSLR users waiting for APT maybe this option could be an option giving best of Ascom and Indi/Linux that appears to work (even though its still Beta) AND IMHO far simpler and more flexible than Kstars - so long as it works in the real world 🙂

So more testing needed !

 

 

[stash_old]

OK ISO can be set - need to untick "gain from camera driver" in edit-preferences-camera. Now you can use the "Gain" drop down menu to set ISO 🙂

[stash_old]

FYI - As per above also works using Indigo_server on a RPI3b+ (Buster) and CCDciel on Windows. So no need to compile anything just install Indigo Server on Buster load and go. 🙂

[don4l]

Another thumbs up for CCDCiel from me.

I've been using it for about 5 months now.  I upgraded last night and everything went very smoothly.

Edited August 30, 2019 by don4l

[stash_old]

ok found a bug - when using Indigo you cant set exposure to above 30secs - on DSLR at least. No such problem when using Indi tested ,indoors,360 secs exp no problem.

If it keeps this up in real testing I will be replacing APT as the author seems to have "forgot" about DSLR users on Linux(shame) . As CCDciel has 1 important game changer when compared to APT - The Sequencer which IMO is far better ,interface wise, then Ekos. 

Still needs some "real testing.

[stash_old]

First light for real - Indiserver running from command line with just canon dslr connected and CCDCiel running on Win 10 I5 connected to Ascom for mount and Canon connected via Indiserver.

All well and good to start with.

Slew ok and camera images being taken. However could not get Platesolving to work via CCDCiel using Platesolve2 - odd thing I can run APT at the same time connected to the same EQMOD and load the image taken by CCDCIEL and it solves in under 9secs. Therefore the problem must be the parameters being passed by CCDciel to Platesolve. The only diff I can see BUT cant alter is the Arcminutes/Degree's size and regions. On APT it sets this at 33x22 Arcminutes CCDciel sets it at 162x166 and regions 9999 on APT 999 on CCdciel. Also tried Astromery.net (local) still not solving. Both are extracting stars ok.

CCdciel documentation on Focal Length is poor playing with this item changed the Degree's in Platesolve but still no luck. Dont know if APT crops the image or what. Oh Well will have to do some investigaing.

Astrotoaster worked well using the FITS images from CCdciel although the images were not as clear as CR2 versions but that might be me(is me). See attached M39 live stack - not concerned about the quality at this stage as it was just a proving run.

So a few things to luck at before I can move on to stage 2 - Guiding.

 

[han59]

With respect to the solving problems:

For setting astrometry options, there are only two screens (camera and astrometry) to configure in the menu edit, preferences. See screen shots.

Loading the nightly produced FITS file in a viewer will tell you if the sensor pixel size and focal-length are set correctly by the reported FOV. Both PlateSolve2 or ASTAP will do that for you. CCDciel will write this data in the FITS header.

Personally, I prefer to set these settings independently of the driver.  Focal length is expressed in mm and pixel size in micrometer.  If it all fails, you could try ASTAP as solver.   For DSLR camera set downsampling to 2 en search area to something like 30 degrees rather then the default 5 degrees..

 

[don4l]

I've seen the pixel size reported incorrectly in "IndiGui".  This has fed through to CCDCiel and killed the platesolving stone dead.  As Han has suggested, I hard set the pixel size in the Camera Tab, and I haven't had any problems since.

[stash_old]

@HAN59  Thanks - Yep already hard set my settings manually and I have found out that ,as you say, the FITS header is where the RA/DEC details are retrieved. Which is a pain if you have some CR2 converted files that haven't carried the RA/DEC in the conversion as APT allows you to input RA/DEC manually if required. PC say'she may change CCDCiel to ask for RA/DEC if no co-ords exist in the FITS header. Fine so long as you are not using the Plan Editor in auto mode LOL.

As for " For DSLR camera set downsampling to 2 " this depends on the number of extracted stars (in Platesolve2 and Indi   anyway) and setting "down sampling" to higher values to achieve around 3-400 stars improves the speed in solving. In Platesolve2 (not blind software) I get 4-6secs solve time on most images with a reasonable close RA/DEC coords.

Sorry I just cant on with ASTAP's interface although I must confess I haven't tried it directly from CCDCiel - I am too simple 🙂 Fine software that it is Hans.

Still waiting for clear sky and no moon.:-) 

[han59]

The down sampling factor 2 helps to combine the red, green, pixels (bayer matrix) of the OSC/DSLR raw format.  That helps with star shape and therefore the star detection.  Reducing the number of stars you better do with the solver parameters then higher sampling factors.

If you get solving in 4-6 seconds there is not much to improve.

 

 

[stash_old]

15 minutes ago, han59 said:

If you get solving in 4-6 seconds there is not much to improve.

Totally agree just need CCDciel to call PS2 and do it as fast or even slower - just do it - its on the same PC. Will get there in the end !

[stash_old]

ok finally got o test with changes to settings in platesolving - decided to use ANSVR as it will blind solve as well it seems. 

Set up Sequencer to take images of M31/2 and also loaded Astroaoster to stack live images. With platesolving to check position inside the Sequence Plan.

It worked - forget the image quality as this was a test of platesolving in the Sequencer.

Pretty satisfied.

So to summarise - camera connected via Indi, CCDciel running on Windows, Mount and Focuser connected to Linux but they are used in Windows Ascom via Ser2net and Virtual com software. Quite pleased with DSLR downloading time in CCDciel - VERY pleased with CCDciel Sequencer - sorry i loved APT but with the Sequencer CCDciel rules - still early days but looking promising.

[don4l]

Another release came out yesterday.  This release seems to focus on tidying up the user interface.  This will make it easier for new users to learn the package.

The platesolving and autofocus routines are working brilliantly.  Autofocus through broadband filters is taking 44 seconds.  NB is taking 2 mins.

 

[stash_old]

17 hours ago, don4l said:

NB is taking 2 mins.

Ok I am stumped what is NB ?  Nebulosity software ?

[don4l]

2 hours ago, stash_old said:

Ok I am stumped what is NB ? 

Sorry, Narrowband.  Focusing with a 3nm narrowband filter needs 8 second exposures, rather than the 0.5s that I use for the red, green or blue filters.  I used to spend 20mins or more focusing.  When you have to re-focus several times in an evening, it soon adds up.

 

[han59]

For narrow band you often need longer exposure times to get stars above the noise.  You can speedup focusing by using 1) "stay at position" and 2) dynamic focusing. 3) Limit the number of step to 7  or  8 or 9 steps .  Furthermore you can reduce the exposure time if you 4) limit the focuser steps such that your focus if going not far off.  For my setup the HFD value at focus is around 2.5. I tune the focuser steps such that the maximum HFD is around 5  maximum.  Example for H-alpha 7 nm, I'm using 4 or 5 seconds exposure for focusing.

Going to a dedicated bright star for focusing takes more time then "stay at position"

If this is abracadabra for you, ask me for more information.

Han

 

 

 

[don4l]

3 minutes ago, han59 said:

For narrow band you often need longer exposure times to get stars above the noise.  You can speedup focusing by using 1) "stay at position" and 2) dynamic focusing. 3) Limit the number of step to 7  or  8 or 9 steps .  Furthermore you can reduce the exposure time if you 4) limit the focuser steps such that your focus if going not far off.  For my setup the HFD value at focus is around 2.5. I tune the focuser steps such that the maximum HFD is around 5  maximum.  Example for H-alpha 7 nm, I'm using 4 or 5 seconds exposure for focusing.

Going to a dedicated bright star for focusing takes more time then "stay at position"

If this is abracadabra for you, ask me for more information.

Han

 

 

 

That isn't "abracadabra" at all Han!  You've mentioned all the settings that  I've been afraid to play with  in case I break it.  Every time that I change something, I seem to break it, and I have to do the initial calibration again.  This takes valuable time, so I haven't dared change anything once it is working.

I get the impression that the "Start focus HFD" is telling CCDCiel roughly how far to move for it's first sample.  Is this correct?  One problem that I had was that it seemed to move too far and the star became too faint.  I have had the same problem with the initial "File -> Calibrate focuser" routine.  The default was 20, and I had to reduce it to 15 before it would work at all.

I suppose that I do not understand if CCDCiel needs a large range to correctly calculate where it is on the V curve...  so I have been leaving this figure as high as I can.

 

I am also a bit confused about the "Near focus HFD".  The figure in my settings is 8.4.  The focus star needs to be about 1.2, or less at correct focus, so I don't know what the term "Near focus" actually means.

 

 

[han59]

Your using the V-curve option to focus. This is the oldest method. I have used V-curve option in the past till dynamic was introduced.  The Start focus HFD (2)  default setting is 20 but for me 15 or even 10 was required to get good snr values. The Near focus HFD position (3) setting I had set at 6 or 7.   It uses point 2) and 3) to find the focus at 4) So V-curve  focusing only works if 3) still lies in the linear area. I think you can select for 3) a value twice the mininum so the HFD at focus. 

The problem with this method is that you have to calibrate the V-curve in advance and in most case you have to slew to a bright calibration star for a good SNR value. Too much saturation could also be a problem at focus.  So the star should bright but not too bright.

 

Later the hyperbolic curve fitting (=dynamic) was developed and since this is implemented, I haven't used the V-curve focusing anymore. I would suggest to try this dynamic focusing + stay in place. In the dynamic method there is less need to de-focus and therefore it is better suited to "stay in place" for focusing: I have tried to describe the performances with plus and minus signs  in the table below:

 

To change focusing, you have to select "dynamic" and "stay in place". (For sequences "stay in place" is also an option you should set). The minimum snr in this screen shot is probably too low. I would set it at 6 :

 

It tries find the best hyperbola fit on the measured points. Below a quick sketch I made to illustrate the principle. Hyperbola fitting can work in the non-linear area of the curve. Therefore less defocusing is required. The inputs are focuser positions and the corresponding HFD values, the seven stars in the sketch. The curve is the best fit it can find and the routine knows the bottom position so the best focus position.

The minimum HFD is about 2.5. For focusing you don't have to defocus to a HFD value more then two or three times the minimum value. So in this setup to HFD 5 or 6.

 

This is how it looks in reality:

 

-- Han

 

 

Edited December 6, 2019 by han59

[don4l]

On 06/12/2019 at 11:14, han59 said:

Your using the V-curve option to focus. This is the oldest method. I have used V-curve option in the past till dynamic was introduced.  The Start focus HFD (2)  default setting is 20 but for me 15 or even 10 was required to get good snr values. The Near focus HFD position (3) setting I had set at 6 or 7.   It uses point 2) and 3) to find the focus at 4) So V-curve  focusing only works if 3) still lies in the linear area. I think you can select for 3) a value twice the mininum so the HFD at focus. 

-- Han

 

 

 

Thank you, Han

You've more than answered my questions.  I was reluctant to change the results that came from the Calibration routine because I didn't really understand what they meant. 

 

I'll have a proper play with Dynamic focusing next time that I am out.  I had assumed that V curve was the recommended option.