Raspberry Pi Camera Module 3

[Slayingripper]

I'm thinking about using the PI camera module 3 for allsky but im not 100% sure if it will work as it uses libcam instead of raspi-cam. Looking through the source code, I've seen mentions of libcam, but I'm not sure if that only works with USB cameras(becomes enabled when connected).
The new module looks decent, and if it does work with all sky, it takes up way less space, so it would fit my use case.

[EricClaeys]

@Slayingripper The Module 3 camera will not work with the current release of Allsky but we plan to support in in the next release whis is being tested now.

I assume you meant libcamera and not libcam?
The HQ camera has been working with libcamera sine Bullseye came out 14 months ago, and with raspi-still for a couple years before that.

All the RPi cameras plug into the camera port on the Pi and do not use USB.

I think the biggest question will be whether or not people will use a 120 degree lens (with the "wide" version) for an allsky camera. That's not very wide.

[Slayingripper]

Still looks better than the Pi cam module 2 that I use. But I see your point!

[EricClaeys]

@Dentonknifeworks Those stars look like out of focus stars from a telescope that has a central obstruction like an SCT, where the dark part in the middle is a shadow from the obstruction.

Might I suggest you take a picture during the day when you know it's in focus? Or wait until the moon is out. It's possible the auto focus is "focusing" on hot pixels which won't be an issue with brighter images.

[Dentonknifeworks]

You have a valid point I wonder how well the autofocus is working at night. The stars do look out of focus that is what I am seeing as well. The daytime photos on clouds look good.
As soon as the rain stops I wil try more ideas with the autofocus.

[EricClaeys]

@Dentonknifeworks. Let's keep good notes on the best way to auto focus an allsky camera. I would like to add them to our documentation since other people will have the same issues and I suspect the libcamera team won't address it since long exposure night images are not their target market.

[Dentonknifeworks]

There may be hope for this after all this is with the focus set to Infinity. When the auto is on it seems to be hunting for a focus. This makes some sense to me

[EricClaeys]

@Slayingripper libcamera is only supported on the Bullseye operating system.

[stuartsmith-nasstar]

Hi chaps, was this resolved ? I'm using a camera 3 module, and its focused close. I can't find anything in the docs to change that ?
I've tried from the command line using libcamera-still --autofocus-mode=manual --lens-position=0 too, and that seems to make no difference.

[EricClaeys]

@Dentonknifeworks, Brad, what command(s) did you use to auto-focus your Module 3 camera?

[stuartsmith-nasstar]

I had a bit of experimenting last night and though no joy - maybe some of this is useful.

I installed https://github.com/Gordon999/Pi_LIbCamera_GUI which is a great gui which supports the Pi 3 camera and others.

Auto focus works during the day time and I get sharp clouds, etc.

I can't run this and allsky at the same time, but stopping it and starting allsky gets me the same blurry pic - it seems that autofocus does not work at all in allsky - day or night - and just goes to default.

I tried creating a custom imx708_wide.json but with the autofocus range set to 0 to 0 and changing default to 0.0 (from 1.0) (i.e. infinity to infinity), and pointed allsky to use that in advanced settings, but no difference (attached file for reference)

I tried Pi_LIbCamera_GUI again and put it in manual and worked out where best focus was - seemed about 1020 on their scale if that helps.

But quiting that, and starting allsky results in the same blurry image, clearly defaulting to something much closer.

So it seems that allsky is not picking up new json changes, or something else is stopping it. It would be great if the libcamera lens-position variable was exposed in allsky - the we could simply set it to the value we want.

In the meantime, it certainly looks like @Dentonknifeworks got something working - I just have to hope he sees the notification and replies here - though it was a long time since he commented last.
imx708_wide.json

[stuartsmith-nasstar]

OK chaps - here is the answer - the documentation for the camera 3 is totally wrong. It states infinity is 0.0, but it is not - it is 1023/100 = 10.23. I've pulled that back a bit to try to hit hyper focus, but will fine tweak tonight.

So, to get the camera 3 to focus on stars, add this to 'extra arguments' in advanced section:

--autofocus-mode manual --lens-position 10.19

Experiment with values from 10.23 down to find what works for you.

It never fails to amaze me the amount of effort some folk put into hardware, but not on the week effort or so to write proper accurate docs.. arg..

Anyway, hopefully this will help others. I attach 2 pics - the blurry one is how allsky works with no extra arguments. the sharp one is with the above.

[EricClaeys]

Thanks for the info. I will add it to the "how to focus" documentation page.

Because very few people have auto focus cameras, and those that do normally focus once and leave it, Allsky doesn't natively support focus. You have to use "extra parameters" like you did as well as the optional "tuning" file.

What happens if you focus once then don't put anything in the extra parameters field?

[stuartsmith-nasstar]

doesn't work. the default for libcamera-still is to autofocus. The optional tuning file isn't required - just the parameters.

[EricClaeys]

Try adding the following to the "Extra Arguments" field in the WebUI:

--metadata /home/pi/allsky/config/overlay/extra/libcamera.txt --metadata-format txt 

the libcamera.txt file will be updated every image and willl include data on the image just taken. Here's an example:

ExposureTime=69432
AfState=0
AnalogueGain=1.122807
AfPauseState=0
FrameDuration=70694
SensorTemperature=31.000000
LensPosition=10.190000
AeLocked=false
DigitalGain=1.010074
Lux=212.761215
ColourGains=[ 3.046685, 1.498411 ]
ColourTemperature=7514
SensorBlackLevels=[ 4096, 4096, 4096, 4096 ]
FocusFoM=3768
ColourCorrectionMatrix=[ 1.678631, -0.392871, -0.285760, -0.371249, 1.730274, -0.359036, -0.096311, -0.417801, 1.514122 ]
ScalerCrop=(0, 0)/4608x2592
SensorTimestamp=3710984790273000

I also have this in the Extra Arguments field:

--autofocus-mode manual --lens-position 10.19

hence the AfState=0 (off/manual) and LensPosition=10.190000.

[Vejbymats]

I started playing with a module 3 tonight and on my example (normal, not wide) infinity is around 0!
For me around -0.4
10 is very close, 1cm or so.

[stevedee]

Setting --lens-position 0.0 also works for me on my new camera module 3 with Pi running Bookworm.

With --lens-position 10.0 my camera is focused around 100mm

I now understand that ${FOCUS} is a focus metric, so the higher this value the better!

[EricClaeys]

I believe at lens position X, focus is 1/X meters, so 1/10 would be 100 mm.

The AllSky ${FOCUS} number is basically how much contrast there is in a picture - the higher the number, the more in focus you are. Don't compare it to anything else - every camera is different.

[aaronwmorris]

I was not super impressed with the Camera Module 3 (IMX708) for the purposes of an all sky camera. I finally kind of identified the metric which would define which cameras are "good" and which might not work well for all sky... the gain.

The IMX708 is only capable of a maximum analog gain of 16. The IMX477 HQ camera max analog gain is 22 which is much more appropriate (for an allsky camera). The IMX290, IMX462, and the IMX327 cameras would likely exceed the performance of the IMX477 because I expect the maximum analog gains are 30-32 for these cameras.

[stevedee]

I don't understand your point.
Why do you consider the analogue gain range to be so important?

Here is a photo from one of last nights 'fireball' meteors...

[aaronwmorris]

It just comes down to sensitivity. A lower available gain just means there is less sensitivity or too much noise to support a larger gain.

I think the challenge with the IMX708, and the IMX477 to a degree, is the size of the pixels. Both cameras have incredibly small pixels which can limit the performance. The IMX708 and 477 with 1.55um sized pixels has a 2.19 sq/um surface area per pixel. The IMX462 has 2.9um pixels which is 4.1 sq/um, roughly double that of of the 477.

A larger surface area for a pixel means more area for photons (i.e. signal) to register on the same pixel.

The IMX708 is an Sony Exmor R sensor, but most astronomy cameras usually use Sony Starvis sensors (which includes the IMX462/290). The Starvis sensors have better low light performance.

What I am trying to do is bring some objectivity to why you would use one camera over another camera. Currently, there is no guidance other than subjective recommendations based on other people's experience.