CanPi
Posts: 51
Joined: Wed May 20, 2020 12:21 pm

HQ Camera - raspistill inactive switches

Wed May 20, 2020 2:10 pm

Love the new HQ. These raspistill switches are useful but I must be doing something wrong because they seem not to have the desired effect:

-st -> Nothing happens. Is this output directed to a file somewhere?
-x -> Nothing happens. For instance -x FNumber=5.6 still shows the default value of 2 in the jpg.
-t -> setting it to zero with -ex off hangs the cam. What's the minimum practical value with no preview and -ex off?

Also, is the maximum frame rate in --mode 3 10 fps or can it get up to 40 fps per the flyer?

Keep up the good work!

ethanol100
Posts: 637
Joined: Wed Oct 02, 2013 12:28 pm

Re: HQ Camera - raspistill inactive switches

Wed May 20, 2020 5:39 pm

CanPi wrote: Love the new HQ. These raspistill switches are useful but I must be doing something wrong because they seem not to have the desired effect:

-st -> Nothing happens. Is this output directed to a file somewhere?
If you just run raspistill without arguments it will show as:

Code: Select all

-st, --stats    : Force recomputation of statistics on stills capture pass
On github the commit of this feature says: "If a stats pass is done on stills captures, then digital gain
and AWB are recomputed based on the actual capture frame statistics.
Otherwise it uses the values from the preview frame immediately before
the capture."
Therefore, only the output image will look different, as the AWB gains and digital gain will be using the captured image.
-x -> Nothing happens. For instance -x FNumber=5.6 still shows the default value of 2 in the jpg.
This is a known issue, currently the image encoder overwrites these settings. There will hopefully be a fix soon. You will still need to define it as EXIF.FNumber=5.6 then.
-t -> setting it to zero with -ex off hangs the cam. What's the minimum practical value with no preview and -ex off?
t = 0 means run until killed. This just disables the timeout. You still need some frames, take something like 200 or 500 there.

Also, is the maximum frame rate in --mode 3 10 fps or can it get up to 40 fps per the flyer?

Keep up the good work!
Sorry, I don't understand to which flyer you are referring. Mode three takes full resolution images, the pi only uses two lanes, therefore, the interface bandwidth limits the framerate to about 10fps(The sensor can produce up to 40 fps). In reality you will not even be able to do anything useful at 10fps for 12MPix images. This means you get about 120MB/s of data, you will not be able to analyse it, you will not be able to encode it at that speed.

CanPi
Posts: 51
Joined: Wed May 20, 2020 12:21 pm

Re: HQ Camera - raspistill inactive switches

Thu May 21, 2020 8:03 am

ethanol100 wrote:
Wed May 20, 2020 5:39 pm
CanPi wrote: -st -> Nothing happens. Is this output directed to a file somewhere?
If you just run raspistill without arguments it will show as:

Code: Select all

-st, --stats    : Force recomputation of statistics on stills capture pass
On github the commit of this feature says: "If a stats pass is done on stills captures, then digital gain
and AWB are recomputed based on the actual capture frame statistics.
Otherwise it uses the values from the preview frame immediately before
the capture."
Therefore, only the output image will look different, as the AWB gains and digital gain will be using the captured image.
Got it. Just be aware that in the documentation page https://www.raspberrypi.org/documentati ... /camera.md it instead says this

Code: Select all

--stats,    -st        Display image statistics
Displays the exposure, analogue and digital gains, and AWB settings used.
CanPi wrote: -x -> Nothing happens. For instance -x FNumber=5.6 still shows the default value of 2 in the jpg.
ethanol100 wrote:
Wed May 20, 2020 5:39 pm

This is a known issue, currently the image encoder overwrites these settings. There will hopefully be a fix soon. You will still need to define it as EXIF.FNumber=5.6 then.
ok
ethanol100 wrote:
Wed May 20, 2020 5:39 pm
t = 0 means run until killed. This just disables the timeout.

Ok, it would be useful to have this information in the documentation page. What's the best way to kill the process in this case?
You still need some frames, take something like 200 or 500 there.
Ok, though it seems to work even with -t 1. If one goes fully manual, what's the purpose of the timeout frames and does one really need them?
ethanol100 wrote:
Wed May 20, 2020 5:39 pm
CanPi wrote: Also, is the maximum frame rate in --mode 3 10 fps or can it get up to 40 fps per the flyer?
Sorry, I don't understand to which flyer you are referring.
The IMX477 flyer (https://www.sony-semicon.co.jp/products ... _Flyer.pdf) says that in the Normal Drive Mode the sensor is cabable of 40 fps at 12-bits and 60 fps at 10 with full resolution images.
ethanol100 wrote:
Wed May 20, 2020 5:39 pm
Mode three takes full resolution images, the pi only uses two lanes, therefore, the interface bandwidth limits the framerate to about 10fps(The sensor can produce up to 40 fps). In reality you will not even be able to do anything useful at 10fps for 12MPix images. This means you get about 120MB/s of data, you will not be able to analyse it, you will not be able to encode it at that speed.
I understand. So if we use two out of four lanes, can we get half what it says in the flyer or are we limited elsewhere? I am ok with just storing raw data and doing the processing later.

Thank you very much.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 26432
Joined: Sat Jul 30, 2011 7:41 pm

Re: HQ Camera - raspistill inactive switches

Thu May 21, 2020 8:38 am

Just submitted a PR to fix the camera docs issues referenced above. https://github.com/raspberrypi/documentation/pull/1555


With regard to the speed question, you are limited by the speed of the ISP as well as the speed with which data can be pulled off the sensor and transferred down the 2 lane CSI-2 bus. Also limited by speed with which you can write to the SD card/storage medium. Lots of things to take in to account.

The fix for the incorrect F number/aperture is now in rpi-update, BUT, I would not recommend grabbing it just yet - lots of churn in there at the moment and you might break other things.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

ethanol100
Posts: 637
Joined: Wed Oct 02, 2013 12:28 pm

Re: HQ Camera - raspistill inactive switches

Thu May 21, 2020 8:51 am

Yes, I did notice the issue in the documentations, too. If you want to have the settings, you need the "-set" switch.

Just a note about the timeout: There is also another feature coming, which uses the "-st" switch to speed up long exposure captures, I think this should also speed up capturing images with fixed settings(will not really noticeable at the higher frame rates....).

CanPi
Posts: 51
Joined: Wed May 20, 2020 12:21 pm

Re: HQ Camera - raspistill inactive switches

Thu May 21, 2020 1:19 pm

Grand, thank you both.

Return to “Camera board”