Night sky time-lapse with v3 camera

[jbeale]

Here's my first try at a night sky timelapse using the v3 camera (normal lens, no IR-block filter). This is using 12 second exposures with analog gain = 8, and taking 2 frames per minute, but played back at 12 fps.
https://www.youtube.com/watch?v=n5haLed_Wk8

[HermannSW]

Thank you, nice!
What is the bright dot entering from top middle?
How did you select analog gain?
Where did you capture night sky?


P.S:
For others, @jbeale did post details on libcamera-still command he used here:
viewtopic.php?t=348009

[jbeale]

Here is a command that sets exposure time to 12 sec and analog gain to 8 (max gain is 16, which of course has more noise)

Code: Select all

libcamera-still -n --immediate --denoise off --sharpness 1.5 -q 92 --lens-position 0.05 --shutter 12000000 --gain 8 --awbgains 1,1 -o pic.jpg

With my particular NoIR camera outside with temp = -4 C, that lens position seemed best. Infinity focus should be at 0.0 but that looked more blurry.
This timelapse was taken just south of Portland OR. Plane trails near morning were coming from the nearby PDX airport.

Looking at an online sky map https://stellarium-web.org/p/observations I am guessing the bright object entering from upper left around 0:35 in the video is Arcturus, which is one of the brightest stars with a visual magnitude of −0.05 and before that around 0:11 in the video, on the right hand side of the frame, you see all 7 stars of the big dipper, with its handle pointing up. You can even see that Mizar, the middle star in the handle of the dipper, is a binary with its nearby dimmer companion Alcor. In fact, in between the dipper and the handle you see the three-in-a-line stars which are 70,71,73 Ursae Majoris. The middle one is mag 5.88 and would normally not be visible by eye in my location, so that shows this little $25 camera works ok!

[HermannSW]

Thanks for the information, this will get me going with first steps of astrophotography.
I have a normal v3 camera, just ordered non-wide NoiR v3 camera.

just south of Portland OR

Just looked up Portland latitude of 45°31′12″N.
That is even south of my home location in South Germany (Eberbach, 49°28′N).
We will be on vacation in Middle Denmark at latitude a little bit north of 55°37′12″N in May.
There the night sky is so much less light polluted compared to Germany, and it is easy to see milky way just with the eyes.
I never have seem miiky way in south Germany …
And the vacation home estate is 5000m² without trees, so free sight to sky is given.


After starting with v3 camera, I might revive my cheap telscope with HQ camera project.
I prefer to use eyepiece projection, because distance A from eypiece to camera sensor goes multiplicative into focal length of telescope.
And I have > 20 CStoC adapter rings to make A big ...

viewtopic.php?t=290331#p1759065



Because low light sensitivity of HQ camera is better than v2 camera (worse than v3):
viewtopic.php?p=2077594&hilit=hermannsw ... y#p2077397
I will remove IR filter from one of my HQ cameras for that.

[jbeale]

Yes, as I am near an urban area, I think the light pollution may be more of a limitation than the camera.
The last two images on this page include a PiCam v3 view from last night including the Pleiades in the lower left corner, and above that Aldebaran (peeking between tree branches), Mars, Capella towards the middle. https://photos.app.goo.gl/Rfifjak45miuzt8PA

The image was taken at Fri 24 Feb 2023 08:23:31 PM PST. I see the EXIF data in the JPEG file shows 8:23:43 which is 12 seconds later, so apparently that marks the time at end of the exposure, while my script recorded the start of the exposure. The raw image had a lot of background from light pollution and a thin overcast haze, so what I uploaded used Gimp to subtract out a blurred version of the same image from it. That way a curves adjustment could reveal some fainter stars.

Pi Camera v3 IR, 12 seconds, processed, downscaled

2023-02-24_823pm_small.jpg (160.32 KiB) Viewed 1638 times

[jbeale]

If anyone is interested in making a timelapse like this, here are four scripts I'm using. The first one runs on the Pi with the camera, and takes a string of consecutive 12-second exposures through the night, until you stop it. Of course adjust exposure, gain and other settings to suit. It saves images in the /dev/shm ramdisk which is maybe marginally faster than SD card (and maybe avoids flash wearout etc.). It also logs the date/time of each exposure to a csv file for later reference. The second script gets run by crontab every minute on the same Pi, to transfer recent images to a remote computer for later processing.

The other two can run on any machine. One is good at removing light pollution (simply get rid of the low-spatial-frequency background, I'm assuming you won't see nebulas with this camera anyway) and adjusting the contrast and resolution to prepare for the last step, just converting all the still images to a mp4 timelapse video.

Code: Select all

#!/bin/bash
# runs on RPi; time-lapse recording of sky, repeat many long exposures
FLOG="/home/pi/camera/log6.csv"
ctr=1000  # starting number for first image
OUTDIR="/dev/shm"

while true; do
  DATE=$(date)
  HMS=$(date +"%H%M%S")
  libcamera-still -n --immediate --denoise off --sharpness 1.1 -q 92 --lens-position 0.05 --shutter 12000000 --gain 8 --awbgains 1,1 -o $OUTDIR/$ctr.jpg
  echo $DATE", "$HMS", "$ctr >> $FLOG
  ctr=$((ctr+1))
done
# =======================================================================

#!/bin/bash
# transfer all .jpg and .dng files to remote storage
# then rsync (transfer) from storage dir to a different machine, and delete local copy

RECDIR="/run/shm/"
STOREDIR="/run/shm/OLD"
RMACHINE="john@john-Z83-4.local"  # my remote storage machine
RTODIR="/media/john/Seagate4GB/MINIX-John/Picam2"

echo $RECDIR $STOREDIR $RMACHINE:$RTODIR
find $RECDIR -maxdepth 1 \( -name "*.jpg" -o -name "*.dng" \) -type f -mmin +0.25 -exec mv -t $STOREDIR {} +
rsync --remove-source-files -a $STOREDIR/ $RMACHINE:$RTODIR/
# =======================================================================

#!/bin/bash
# Post-process timelapse images using imagemagick:
# blur each image, to model smooth light pollution background
# subtract background from image, leaving small features (eg. stars)
# brighten image (a whole lot) with an S-curve
# adjust contrast so JPEG black=0 visually matches MP4 video black level
# rescale 12 MP image down to 1080p full-HD

OUT="./proc2" # folder to hold processed still images
for IMG in *.jpg; do
  echo -n "$IMG.. "
  convert \( $IMG -scale 576x324 -depth 16 -brightness-contrast -3x0 \
       -blur 0x7 -resize 4608x2592 -write mpr:blur1 +delete \) \
    \( mpr:blur1 $IMG -compose minus-dst -composite \) \
    \( -sigmoidal-contrast 6,5% \) \
    \( -brightness-contrast 0,-6 \) \
    \( -resize 1920x1080 \) \
    $OUT/$IMG
done
# =======================================================================
  
  
#!/bin/bash
START="1000"  # first filename is "1000.jpg"
# convert still .jpg files into .mp4 video, at whatever frame rate you prefer
# -crf 17 allegedly visually lossless (23 is default)
ffmpeg -framerate 24 -start_number $START -i %04d.jpg -crf 19 -vcodec libx264 lapse.mp4
# =======================================================================

[jbeale]

Here is a sky video I took last night, compressing 10 hours into ~ 2 minutes, using the above scripts to take the images and process them. https://youtu.be/HFTi3XzmBnc

(I only just looked at the video on YT- the compression is awful, looks like VideoCD quality. I would guess they recompressed to maybe 1/20 the bitrate of the 1GB mp4 file I uploaded.)

[HermannSW]

Thanks for the information, this will get me going with first steps of astrophotography.
I have a normal v3 camera, just ordered non-wide NoiR v3 camera.

v3 NoIR arrived today, and I wanted to use it with Pi0W2.

Steps needed to build mjpg-streamer on Pi0W2 (only 512MB RAM) are in this posting (for setting the scene with headless Pi0W2 v3 camera):
viewtopic.php?p=2085900#p2085900

v4l2 command to change v3 camera focus from here:
viewtopic.php?p=2075130#p2075130

I used mjpg-streamer only for very first test of capturing half-moon.
And through a fly screen.
With this command:

Code: Select all

pi@raspberrypi:~ $ for((f=320; f<=960; f+=64)); do v4l2-ctl -d /dev/v4l-subdev1 -c focus_absolute=$f; sleep 0.25; done
pi@raspberrypi:~ $ 

This is the result of "peek" screen recording, either half-moon is too bright, or fly screen is the problem, but this does not look like half moon:
(just looked, the lower half is real)



Next will be real capturing outside (and not on window sill) with @jbeale's commands.
And not capturing moon, it is too bright.

[HermannSW]

If anyone is interested in making a timelapse like this, here are four scripts I'm using. The first one runs on the Pi with the camera, and takes a string of consecutive 12-second exposures through the night, until you stop it. Of course adjust exposure, gain and other settings to suit. It saves images in the /dev/shm ramdisk which is maybe marginally faster than SD card (and maybe avoids flash wearout etc.). It also logs the date/time of each exposure to a csv file for later reference. The second script gets run by crontab every minute on the same Pi, to transfer recent images to a remote computer for later processing.
...

Code: Select all

#!/bin/bash
# runs on RPi; time-lapse recording of sky, repeat many long exposures
FLOG="/home/pi/camera/log6.csv"
ctr=1000  # starting number for first image
OUTDIR="/dev/shm"

while true; do
  DATE=$(date)
  HMS=$(date +"%H%M%S")
  libcamera-still -n --immediate --denoise off --sharpness 1.1 -q 92 --lens-position 0.05 --shutter 12000000 --gain 8 --awbgains 1,1 -o $OUTDIR/$ctr.jpg
  echo $DATE", "$HMS", "$ctr >> $FLOG
  ctr=$((ctr+1))
done
...

You did not specify "-t". I thought it would capture 5 seconds then, but it did capture less.

First picture direction north of Eberbach/Germany 4.5 seconds, roughly 45% slope:

Code: Select all

$ time libcamera-still -n --immediate --denoise off --sharpness 1.1 -q 92 --lens-position 0.05 --shutter 12000000 --gain 8 --awbgains 1,1 -o tst.jpg
[0:04:10.382561816] [872]  INFO Camera camera_manager.cpp:299 libcamera v0.0.3+40-9b860a66
[0:04:10.428932489] [873]  WARN CameraSensorProperties camera_sensor_properties.cpp:205 No static properties available for 'imx708_noir'
[0:04:10.429033542] [873]  WARN CameraSensorProperties camera_sensor_properties.cpp:207 Please consider updating the camera sensor properties database
[0:04:10.544693970] [873]  INFO RPI raspberrypi.cpp:1425 Registered camera /base/soc/i2c0mux/i2c@1/imx708@1a to Unicam device /dev/media3 and ISP device /dev/media0
[0:04:10.546801034] [872]  INFO Camera camera.cpp:1028 configuring streams: (0) 4608x2592-YUV420 (1) 4608x2592-SBGGR10_CSI2P
[0:04:10.547378860] [873]  INFO RPI raspberrypi.cpp:805 Sensor: /base/soc/i2c0mux/i2c@1/imx708@1a - Selected sensor format: 4608x2592-SBGGR10_1X10 - Selected unicam format: 4608x2592-pBAA
Still capture image received

real	0m4.533s
user	0m0.838s
sys	0m0.445s
$

12MP v3 NoIR camera photo:




Second command completed after 3.9 seconds, directed slightly more than horizontal.
You can see a hill right bottom, and light pollution from city of Eberbach in center:

Code: Select all

$ time libcamera-still -n --immediate --denoise off --sharpness 1.1 -q 92 --lens-position 0.05 --shutter 12000000 --gain 8 --awbgains 1,1 -o tst2.jpg
[0:10:12.349070108] [921]  INFO Camera camera_manager.cpp:299 libcamera v0.0.3+40-9b860a66
[0:10:12.394866532] [923]  WARN CameraSensorProperties camera_sensor_properties.cpp:205 No static properties available for 'imx708_noir'
[0:10:12.394968200] [923]  WARN CameraSensorProperties camera_sensor_properties.cpp:207 Please consider updating the camera sensor properties database
[0:10:12.511605963] [923]  INFO RPI raspberrypi.cpp:1425 Registered camera /base/soc/i2c0mux/i2c@1/imx708@1a to Unicam device /dev/media3 and ISP device /dev/media0
[0:10:12.513783891] [921]  INFO Camera camera.cpp:1028 configuring streams: (0) 4608x2592-YUV420 (1) 4608x2592-SBGGR10_CSI2P
[0:10:12.514363529] [923]  INFO RPI raspberrypi.cpp:805 Sensor: /base/soc/i2c0mux/i2c@1/imx708@1a - Selected sensor format: 4608x2592-SBGGR10_1X10 - Selected unicam format: 4608x2592-pBAA
Still capture image received

real	0m3.872s
user	0m0.852s
sys	0m0.371s
$

12MP v3 NoIR camera photo:



I am impressed from the photos, but adding -t 12000 should make capturing really 12 seconds, right?
(just tried, adding "-t 12000" does stop capture with less than 5 seconds as well)
https://stamm-wilbrandt.de/en/forum/nightsky/1/tst3.jpg


Btw, I am not able to match both photo to current night sky from a German website (the photos were taken direction roughly north).
Any idea?
https://www.timeanddate.de/astronomie/n ... l/@6558051

nightsky.eberbach.jpg

nightsky.eberbach.jpg (46.15 KiB) Viewed 1302 times

[HermannSW]

I forgot to show the setup.
I use Pi0W2 powered from powerbank, v3 NoIR camera and "Raspberry camera stepper_pan manual_tilt camera holder":
https://www.printables.com/model/378001 ... -camera-ho
I did not connect the stepper motor, it was used just as a heavy base for the camera holder.
Bath room heading nearly north is the only upstairs room without fly screen.

20230302_090530.16%.jpg

20230302_090530.16%.jpg (46.11 KiB) Viewed 1232 times

This is right bottom part at 100% size of 3rd 12MP photo captured with v3 NoIR camera last night:



P.S:
This seems to be the right bottom edge of nightsky photo in daylight photo at 100% size:

rb.100%.png

rb.100%.png (185.08 KiB) Viewed 1177 times

[HermannSW]

I remembered a bug regarding v3 camera shutter time, and found @6by9's advice to rpi-upgrade to a specific version with the fix:
viewtopic.php?p=2086460#p2081953

I did that on my Pi4B with v3 camera, and @jbeale's commands now takes slightly more than 12 seconds to complete.
I used same setup as on low light sensitivity thread, with less than 1lux light:
viewtopic.php?t=345411#p2076558



This time v3 camera is lying on desk and looking just upwards.

Code: Select all

time libcamera-still -o tst5.jpg

resulted in this photo (sclaed to 16% size):

tst5.16%.jpg

tst5.16%.jpg (8.08 KiB) Viewed 1137 times

Now @jbeale's command took 12 seconds after rpi-update and reboot:

Code: Select all

pi@raspberrypi4B2:~ $ time libcamera-still -n --immediate --denoise off --sharpness 1.1 -q 92 --lens-position 0.05 --shutter 12000000 --gain 8 --awbgains 1,1 -o tst4.jpg
[0:10:44.246233086] [1950]  INFO Camera camera_manager.cpp:299 libcamera v0.0.3+40-9b860a66
[0:10:44.366003978] [1951]  INFO RPI raspberrypi.cpp:1425 Registered camera /base/soc/i2c0mux/i2c@1/imx708@1a to Unicam device /dev/media4 and ISP device /dev/media0
[0:10:44.367151086] [1950]  INFO Camera camera.cpp:1028 configuring streams: (0) 4608x2592-YUV420 (1) 4608x2592-SBGGR10_CSI2P
[0:10:44.367479339] [1951]  INFO RPI raspberrypi.cpp:805 Sensor: /base/soc/i2c0mux/i2c@1/imx708@1a - Selected sensor format: 4608x2592-SBGGR10_1X10 - Selected unicam format: 4608x2592-pBAA
Still capture image received

real	0m12.921s
user	0m0.416s
sys	0m0.139s
pi@raspberrypi4B2:~ $

Scaled to 16% size a little too bright

tst4.16%.jpg

tst4.16%.jpg (24.87 KiB) Viewed 1137 times

[HermannSW]

I can only say wow for the very first 12 second shutter time photo with @jbeale's command I got with v3 NoIR camera on bathroom window sill:

Code: Select all

pi@raspberrypi:~ $ time libcamera-still -n --immediate --denoise off --sharpness 1.1 -q 92 --lens-position 0.05 --shutter 12000000 --gain 8 --awbgains 1,1 -o tst6.jpg
[0:01:26.316368030] [828]  INFO Camera camera_manager.cpp:299 libcamera v0.0.3+40-9b860a66
[0:01:26.393097437] [829]  WARN CameraSensorProperties camera_sensor_properties.cpp:205 No static properties available for 'imx708_noir'
[0:01:26.393239692] [829]  WARN CameraSensorProperties camera_sensor_properties.cpp:207 Please consider updating the camera sensor properties database
[0:01:26.531526862] [829]  INFO RPI raspberrypi.cpp:1425 Registered camera /base/soc/i2c0mux/i2c@1/imx708@1a to Unicam device /dev/media3 and ISP device /dev/media0
[0:01:26.533665070] [828]  INFO Camera camera.cpp:1028 configuring streams: (0) 4608x2592-YUV420 (1) 4608x2592-SBGGR10_CSI2P
[0:01:26.534249510] [829]  INFO RPI raspberrypi.cpp:805 Sensor: /base/soc/i2c0mux/i2c@1/imx708@1a - Selected sensor format: 4608x2592-SBGGR10_1X10 - Selected unicam format: 4608x2592-pBAA
Still capture image received

real	0m14.529s
user	0m0.893s
sys	0m0.354s
pi@raspberrypi:~ $ 

12MP photo (2.2MB) captured with v3 NoIR camera:




I copied out bottom right corner of the photo with gimp, this is 100% size part:

tst6.rb.100%.jpg

tst6.rb.100%.jpg (97.97 KiB) Viewed 1106 times

[jbeale]

I think that is looking very good! At least as good as any of the images I got. Now when you run those through the post-process script which subtracts out a blurred background to reduce the light pollution/ overcast haze/ sensor offset, then you should see some nice dark-background sky images. The line that uses ImageMagick 'convert' to do the work is this one (which is hard-coded to the 4608x2592 input resolution, and then reconverting assuming the output will be frames of a 1920x1080 video)

Code: Select all

  convert \( $IMG -scale 576x324 -depth 16 -brightness-contrast -3x0 \
       -blur 0x7 -resize 4608x2592 -write mpr:blur1 +delete \) \
    \( mpr:blur1 $IMG -compose minus-dst -composite \) \
    \( -sigmoidal-contrast 6,5% \) \
    \( -brightness-contrast 0,-6 \) \
    \( -resize 1920x1080 \) \
    $OUT/$IMG

I had been hoping to take some more star images myself, but I live in an area (Pacific Northwest) where I may not see another clear sky for a month or more.

[HermannSW]

I think that is looking very good! At least as good as any of the images I got. Now when you run those through the post-process script which subtracts out a blurred background to reduce the light pollution/ overcast haze/ sensor offset, then you should see some nice dark-background sky images.

Yes, I used your scripts last night.
But first I realized that after 23 minutes no more images got stored because Pi0W2 /dev/shm filesystem was full.

Anyway, with 14 seconds between frames I got 100 frames, which result in slightly more than 4 seconds of 24fps video with your last script.
A frame every 14 seconds and 24fps video result in 14*24=336× speed increase.

I uploaded to youtube, and it looked really bad compared to "mplayer lapse.mp4" locally.
Then I realized that youtube auto quality selected only 360p.
This video is best viewed with right clicking and selecting "Loop", and then click on settings->Quality and select 1080p (because the "convert" script creates 1080p).

Here is youtube video, discussion below:
https://www.youtube.com/watch?v=aynzcj4gGIA

eberbach.nightsky.timelapse.336x.jpg

eberbach.nightsky.timelapse.336x.jpg (73.38 KiB) Viewed 1027 times

Your "convert" command does a pretty good job in getting a dark background.

I was not aware that Frankfurt airport, 4th largest airport of Europe, was only 71km straight line distance away:
https://www.gmap-pedometer.com/?r=7702114

This night sky time lapse shows lot of airplane activity.
In the youtube screenshot above you can see 4 airplanes, three horizontal lines, and a vertical line.
I did capture bottom left corner with youtube fullscreen on full HD display, so this is 100% size:

eberbach.nightsky.timelapse.336x.airplanes.jpg

eberbach.nightsky.timelapse.336x.airplanes.jpg (28.89 KiB) Viewed 1027 times

The view has hill in bottom right corner, but no distracting houses with light.
It shows the same bright "glow" around the hill as around your trees.
Maybe this is caused by "convert" not being able to subtract to black in that area.

eberbach.nightsky.timelapse.336x.hill_glow.200%.jpg

eberbach.nightsky.timelapse.336x.hill_glow.200%.jpg (12.14 KiB) Viewed 1027 times

I just used magnetic compass and saw that I did record direction 10° nearly North.
The center of rotation is a bit left above top of video, stars rotate around Polaris.
Next time I will aim higher to get rid of the low in the frame airplane activity.
Then only airplanes flying south will cross the view.
And I will use magnetic compass to direct North (0°) in order to get rotation center in horizontal middle of view.
(just measured again, Frankfurt airport is 28km west of Eberbach, 71km distant)


P.S:
Just dialed back in night sky webtool
https://www.timeanddate.de/astronomie/n ... l/@6558051
to time during reccording last night.
Polaris is just top North, in constellation Ursa Minor or Little Bear ("Kleiner Bär" in German):

[jbeale]

Video looks good! Yes, I had created a /dev/shm/OLD directory as a staging area, and I ran the 2nd script from a cron job starting every minute to semi-continuously transfer all the recently saved files to another machine and then delete them from /dev/shm to avoid it filling up.

You can play with the initial smoothing parameters to the Imagemagick convert script to get less ghosting around hills, trees and obstructions, if you want, that was a quick experiment and I didn't spend much time tuning it. Anyway it's fun to see that even this little camera can see many stars some of which are not so easy to see just with your eye directly.

[HermannSW]

Thanks again for your script and videos, and paving the way in the other thread to get 12s shutter time for v3 camera after rpi-update.

I went further, wanted to install on first floor bathroom outside window sill.
That has a slight 6° slope, so I did dremel three 2.8mm diameter holes through platsic and window sill aluminum base.
Then I used three 3.0x16mm screws to fixate:

20230303_180531.part.20%.jpg

20230303_180531.part.20%.jpg (39.58 KiB) Viewed 931 times

After closing the box I raised the view so that hopefully most airplane traffic seen last night keeps below field of view.
And I used magnetic compass to direct camera to North (I did not capture directed to 10° last night, but to 25° compass showed):

20230303_181649.part.20%.jpg

20230303_181649.part.20%.jpg (29.59 KiB) Viewed 931 times

Next I tested your script and redirected the stills captured onto SD card:

Code: Select all

pi@raspberrypi:~ $ cat 12s 
#!/bin/bash
# runs on RPi; time-lapse recording of sky, repeat many long exposures
FLOG="/home/pi/camera/log6.csv"
ctr=1000  # starting number for first image
#OUTDIR="/dev/shm"
OUTDIR="/home/pi/camera"
...

It turned out that the 14 seconds between stills did not change, 14 seconds between writes to SD card seem to be good enough for writing stills.
Although the stills I currently capture are only 189KB because image is all white because it is not dark enough.
I will stop and restart when this command shows more than 2MB for biggest still captured:

Code: Select all

pi@raspberrypi:~ $ ls -lS camera/|head -3
total 19936
-rw-r--r-- 1 pi pi 189413 Mar  3 18:20 1000.jpg
-rw-r--r-- 1 pi pi 189413 Mar  3 18:21 1001.jpg
pi@raspberrypi:~ $ 

P.S:
I calculated that 2MB stills should give me more than 15 hours of capturing with free space on the SD carrd.

Regarding "convert" script conversion, I did convert 100 frames on my Pi400 in less than 8 minutes for last night video.

[HermannSW]

I can only say wow again.
I copied over the first 90min or 1372 frames for getting an understanding what I might be harvesting at end of night.
So cool, camera view high enough for not seeing Frankfurt airport left/right airplane traffic.
But I captured between 7:15pm and 8:46pm, still many north/south airplanes crossing the scene.
And few thin clouds.
And high ennough for capturing Polaris, the center of rotation (40%/20% from left/top).
Early in the video Polaris did jump spontaneously some pixels right.
The reason for that was that I closed bathroom window shutter, and stepper motor base of camera is not fully fixated.

Given the length of the airplane, it did fly low above Eberbach on approach for landing at Frankfurt/Germany airport:

eberbach.nightsky.timelapse.336x.airplane.jpg

eberbach.nightsky.timelapse.336x.airplane.jpg (95.16 KiB) Viewed 870 times

16 second video best viewed when choosing "Loop" and selected 1080p quality:
https://www.youtube.com/watch?v=X7vsHgRwR8Y


This video was created from first 372 stills captured in 90 minutes.
I just looked, currently I already have 872 stills in total, for 3.5h of recording until now ...


P.S:
Polaris is part of "Little bear" constellation as discussed in a previous posting.
I just spotted "Big bear" ("Großer Bär" in German) constellation on right side of new video

big_bear.jpg

big_bear.jpg (49.96 KiB) Viewed 837 times

[Gavinmc42]

Astrobiscuit has got some great shots in the middle of London by using filters to get rid of the skyglow.
He uses astro camera sensors which are a bit better than the Pi camera.

Starvis cameras are getting available now for Pi's but how much further can the V3 NOIR and HQ be pushed?
Filter out the skyglow on the V3 NOIR, will the blue filter it comes with help?

[jbeale]

New video looks very nice, neat to see the North Star included as well. It's true that an IR longpass filter will tend to reduce at least some of the skyglow. I didn't try the blue filter that the NoIR version of the Pi camera comes with, but maybe it has that effect. I did try a 780 nm longpass filter (that visually looks completely black) and it definitely reduced the effect of light clouds; I didn't use it in photos I showed because I assumed I would want the best possible sensitivity and didn't want to throw away any photons. Also it's neat to see some of the real color, eg. Mars did look a bit reddish.

There's no shortage of better cameras for astrophotography out there, I was just pleasantly surprised that you can get a start on that hobby with a $25 camera, and see if you want to go up from there. I might be tempted, if I had more than about 1/2 of one clear sky night per month on average in my location!

By the way, I did not know this, but light pollution in the sky globally is now increasing at a rate of 10% each year Global population growth rate is 1.1% per year, but outdoor LED lights are now cheap and plentiful.
https://www.science.org/doi/10.1126/science.adf4952

[HermannSW]

Astrobiscuit has got some great shots in the middle of London by using filters to get rid of the skyglow.
He uses astro camera sensors which are a bit better than the Pi camera.

After I did move camera field of view up to catch Polaris, the light from city of Eberbach did not play a big role anymore.
Also John's ImageMagick script does magic in subtracting out unwanted stuff, resulting in nice stars on very dark background.
That is good enough for me for now, as I am newbie of astrophotography.
John's good deed is to create the scripts and start this thread.
Then astrophotography newbies like myself can just start with a Pi and v3 NoIR camera — no more equipment needed !

There's no shortage of better cameras for astrophotography out there, I was just pleasantly surprised that you can get a start on that hobby with a $25 camera, and see if you want to go up from there. I might be tempted, if I had more than about 1/2 of one clear sky night per month on average in my location!

Exactly, now any astrophotography newbie seeing this thread can just start with v3 NoIR camera — no excuses not to do so


In the morning I took a photo to capture the slope of camera before unmounting the plastic box.
It was later that I installed "Simple Inclinometer" on my smartphone, that would have allowed to exactly measure v3 NoIR lens slope:
https://play.google.com/store/apps/deta ... clinometer

20230304_080407.part.33%.jpg

20230304_080407.part.33%.jpg (27.05 KiB) Viewed 658 times

From outside, 1st floor window sill plastic box and camera are so small — and had no problems with below 0°C temperatures during the night:

20230304_081152.part.jpg

20230304_081152.part.jpg (51.98 KiB) Viewed 658 times

I started capturing 7:15pm, and ended in morning.
I took 2442 stills until clouds covered the stars 4:35am after 9:20h!
That is slightly less than 14 seconds per still.
Here is new (this time 1:40min) video (Polaris is bright star at right border, 20% below top, and center of rotation).
Video best viewed with forcing Quality to 1080p:
https://www.youtube.com/watch?v=qaTP1tb-k2U

eberbach.nightsky.timelapse.336x.100s.jpg

eberbach.nightsky.timelapse.336x.100s.jpg (34.72 KiB) Viewed 658 times

P.S:
For viewing images I normally use "eog".
I installed "feh", because "feh -F 1000.jpg" alllows to see the processed 1920x1080 frames fulllscreen on HDMI display (completely covering X11).
"feh -F *.jpg" allows to move through the stills with right/left arrow keys.

For viewing created lapse.mp4 I use "mplayer -F lapse.mp4", because that does fullscreen play the (nearly 1GB) video.

Code: Select all

pi@pi400-64:~/ebb2b $ du -sm lapse.mp4 
969	lapse.mp4
pi@pi400-64:~/ebb2b $ 

It took 24:20min on Pi400 to create lapse.mp4.


Converting the 2442 stills with John's script took 4 hours on a Pi400 with stills stored on SSD.


P.P.S:
I will upload all the captured 4608x2592 stills, as well as all 1920x1080 processed stills (8GB in total) with the created log files.
That will allow viewing and/or processing of the stills without having to capture night sky yourself.

Location of Eberbach/Germany window sill: 49°27'N 08°59'E

[Gavinmc42]

Gee, some of those UAPs are leaving smoke trails.
Is that sky glow on plane exhaust/condensation?

That answers my question if the V3 Noir can take night sky shots.
Now I will have to figure out how to make a system to go on the roof, anywhere else is blocked by the trees or has street lights on it.
Waterproof camera mounting?
Clear hemisphere dome.

I did find some Astro stacking software that ran on the Pi's before the V3 came out.
Never had enough images to try it on.

[HermannSW]

Gee, some of those UAPs are leaving smoke trails.
Is that sky glow on plane exhaust/condensation?

I think so.

That answers my question if the V3 Noir can take night sky shots.
Now I will have to figure out how to make a system to go on the roof, anywhere else is blocked by the trees or has street lights on it.

You can use similar plastic box that I did, just fixate with few screws on roof top.

Waterproof camera mounting?

I took the risk although there is only very little possible entry for water where the camera holder and flat ribbon cable leave the plastic box.
Not much risk here because several 100 kilometers away from sea, and rain does not contain salt, and is harmless even for CPU, see here:
viewtopic.php?t=190407#p1196292

I did find some Astro stacking software that ran on the Pi's before the V3 came out.
Never had enough images to try it on.

First you should use @jbeale's scripts, they are simple and just work.
Tomorrow I will upload the 2422 original 4608x2592 and 2422 processed 1920x1080 stills onto my personal website.
I have enough free space there, have already stored a bit for all possible positions for English/French/3-3-2-2 peg.solitaire (1GB+16GB+64GB):
https://stamm-wilbrandt.de/en/#peg-solitaire


I wanted to see accumulated stills, and wrote a short recursive bash script using netpbm tools:

Code: Select all

pi@pi400-64:~/ebb2b $ cat max
#!/bin/bash
if [[ $1 -lt $2 ]]
then
   let n=$1+$3
   pnmarith -max <(jpegtopnm $1.jpg) <($0 $n $2 $3)
else
   jpegtopnm $1.jpg
   echo
fi
pi@pi400-64:~/ebb2b $ 

This command accumulates all processed stills numbered 1600-1900:

Code: Select all

pi@pi400-64:~/ebb2b $ ./max 1600 1900 1 | pnmtojpeg > step.1.jpg
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
...
jpegtopnm: WRITING PPM FILE
pi@pi400-64:~/ebb2b $

Here is the result, scaled to 50% size.
It captures the circular "move" of the stars around Polaris (in reality Earth rotates).
And it shows some nice airplane crossings of camera field of view (I like the non straight line bright curve from south to west most):

step.1.50%.jpg

step.1.50%.jpg (133.7 KiB) Viewed 568 times

Combining 401 stills is time intensive, so I combined every 100th still of all:

Code: Select all

pi@pi400-64:~/ebb2b $ ./max 1000 3000 100 | pnmtojpeg > step.100.jpg
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
jpegtopnm: WRITING PPM FILE
pi@pi400-64:~/ebb2b $

This is proof that Polaris is NOT the center of rotation (it rotates on bright circle) — it is only very close to it:

step.100.50%.jpg

step.100.50%.jpg (99.1 KiB) Viewed 568 times

P.S:
I searched for the airplane passing Eberbach very low, from south (top) to north (bottom).
Since Franfurt airport is only 28km west of Eberbach (71km straight line distance), I assume this is approach for landing.
This is accumulation of stills 951..955, each frame shows two straight lines, lights left and right that are always on.
In the middle you can see 10 single dots per frame, seems to be light that is periodic on/off and most times off during 12 seconds shutter time:

step.1.b.50%.jpg

step.1.b.50%.jpg (44.29 KiB) Viewed 548 times

[HermannSW]

I did upload all stills data, logs and scripts to my personal website.
If you download the data, you can do the "max" script from previous posting experiments yourself without capturing stills yourself.


This is from ssh session onto my website:

Code: Select all

-chrootsh-4.2$ cd httpdocs/en/forum/nightsky
-chrootsh-4.2$ ls -l 49°27N_08°59E.March_3-4
total 76
-rw-r--r-- 1 web23.web138 psacln   545 Mar  5 14:30 1st
drwxr-xr-x 4 web23.web138 psacln 12288 Mar  5 11:37 ebb2
drwxr-xr-x 4 web23.web138 psacln 53248 Mar  5 12:58 ebb2a
-rw-r--r-- 1 web23.web138 psacln  1471 Mar  5 14:52 index.html
-chrootsh-4.2$ du -sm 49°27N_08°59E.March_3-4
7935	49°27N_08°59E.March_3-4
-chrootsh-4.2$ du -sm 49°27N_08°59E.March_3-4.mp4 
969	49°27N_08°59E.March_3-4.mp4
-chrootsh-4.2$ 

You can download all data (7935MB) with the bash script shown here (for some reason my personal website allows no recursive downloads):
https://stamm-wilbrandt.de/en/forum/nig ... March_3-4/
The stills not used for creating the 9:20h video were moved into "x" directories and are not downloaded by the script, but are present if you need them.

1st, 3rd and 4th are the adapted @jbeale scripts.
So why are there two directories?
I started capturing at 7:15pm on my headless Pi02W.
After some time I downloaded few frames and inspected.
Then I wanted to convert a single frame with "3rd" script.
That did draw too much power and rebooted my Pi400, which was powered from laptop as KVM switch:
viewtopic.php?t=321840

So I used original power supply and rebooted.
Unfortunately the Pi0W2 capturing command was interrupted.
So this time I executed "nohup ./1st" in Pi0W2 ssh session on Pi400, so that an interrupt will not stop recording.

Less than 7 minutes are missing:

Code: Select all

-chrootsh-4.2$ tail -2 ebb2/log6.csv 
Fri 03 Mar 2023 08:45:54 PM CET, 204554, 1370
Fri 03 Mar 2023 08:46:08 PM CET, 204608, 1371
-chrootsh-4.2$ head -2 ebb2a/log6.csv 
Fri 03 Mar 2023 08:52:19 PM CET, 205219, 1000
Fri 03 Mar 2023 08:52:38 PM CET, 205238, 1001
-chrootsh-4.2$ 

With 14 seconds between stills capturing that is 27 frames, so from final 24fps youtube video only 1 second is missing.

All 363 frames converted with v3 for ebb2 4608x2592 stills are in ebb2/proc2 (1920x1080).
All 2059 frames converted with v3 for ebb2a are in ebb2a/proc2.

You can downloaad the 969MB .mp4 (I created from all 363+2059=2422 stills with modified v3) I uploaded to youtube here if you want:
https://stamm-wilbrandt.de/en/forum/nig ... ch_3-4.mp4
Sometimes my website is not that fast, so download and then view with "mplayer -F ..." locally gives better viewing experience.
Or view on youtube, best with quality forced to 1080p:
https://www.youtube.com/watch?v=qaTP1tb-k2U


I found two constalllations in a frame of ebb2a/2574.jpg, "Kassiopeia" (German) and "Kepheus":
I determined the time that still was created:

-chrootsh-4.2$ grep ", 2574" ebb2a/log6.csv
Sat 04 Mar 2023 02:46:04 AM CET, 024604, 2574
-chrootsh-4.2$

Then I went back to that point in time here:
https://www.timeanddate.de/astronomie/n ... l/@6558051

I colored the constallations black and blue, and copied in the relevant part from the link (12MP still scaled down to 20% size):

2574.20%.jpg

2574.20%.jpg (41.24 KiB) Viewed 453 times

P.S:
I colored the two constallations with gimp in same processed frame ebb2a/proc2/2574.jpg (also scaled to 50% size):

2574p.50%.jpg

2574p.50%.jpg (78.46 KiB) Viewed 425 times

[HermannSW]

I tried to max all 2422 processed with @jbeale's 4th script stills into one 1920x1080 image.
First I got errors because user open file limit is 1024 files.
So I used "ulimit -n 3000" and tried again.
Then recursive bash script stopped around 1700 stills with recursion depth error.
So I did rewrite max script so that it does use exactly two files sequentially:

Code: Select all

pi@pi400-64:~ $ cat maxf 
#!/bin/bash
jpegtopnm $1 > /tmp/1.pnm
shift
for f in $*
do
   mv /tmp/1.pnm /tmp/0.pnm
   pnmarith -max <(jpegtopnm $f) /tmp/0.pnm > /tmp/1.pnm
done
cat /tmp/1.pnm
pi@pi400-64:~ $ 

After slightly more than 4 minutes on Pi400 with SSD the result was created:

Code: Select all

pi@pi400-64:~ $ time (./maxf 49°27N_08°59E.March_3-4/*/proc2/*.jpg 2>err >all.pnm)
real	4m15.309s
user	3m34.675s
sys	2m22.960s
pi@pi400-64:~ $

We have seen composition of 401 frames some postings ago.

This is 830KB 1920x1080 max for all processed 2422 frames between last Friday 7:15pm and Saturday 4:35am night:
https://stamm-wilbrandt.de/en/forum/nig ... l.2422.jpg

Here it is scaled to 50% size to fit forum attachment size — nice.
Smallest radius bright circular arc is Polaris, very close to center of rotation of northern hemisphere:

all.2422.50%.jpg

all.2422.50%.jpg (183.34 KiB) Viewed 313 times