Raspberry Pi HQ Camera 12MP

[WiSi-Testpilot]

Herman, that is a very nice test. Thank you.

Best regards,

Wilhelm

[dabateman]

Here is a interesting detailed analysis of the pi hq sensor:

https://www.strollswithmydog.com/pi-hq-cam-sensor-performance/

 

In the recent MaxMax news ltter,Dan indicated that the V2 camera has been converted successfully to monochrome. I have asked Dan if possible to convert the hq camera.

This maybe the cheapest fully open source monochrome camera available if possible.

I could build a respectable camera and use some fun lenses I built on it.

[WiSi-Testpilot]

Now I‘m playing with the HQ camera too. As Herman has shown, the UV transmission of the 16mm CCTV lens is poor. Because the camera is small and suitable for my drones I will buy a better lens.

The UV picture was made with such a command:

raspistill -o q3.jpg -ISO 800 -ss 200000 -awb off -awbg 0.9,0.3

@dabateman, debayering is a very good idea.

 

Edit: Just tested: the WB works also with raspivid.

raspivid -t 0 -awb off -awbg 0.9,0.5

generates a white balanced full screen video feed.

 

Filter Baader_U; ISO 800; 1/5 s; f/~2

 

 

Setup

[colinbm]

Just love the dongles & doohicky's.....;-)

[Andy Perrin]

The Wollensak Velostigmat 25m lens is a good lens that works with my C-mount devices and has nice UV transmission.

[OlDoinyo]

What is the sensor size? It appears that the sensor is not covering very much of the image circle in some of these.

 

FWIW, I find that white-balancing images taken with a hard black-IR filter such as the Haida 950 is a waste of time, as the images have no chromaticity anyway. If one merely greyscale-converts the unbalanced image, it can act as a sort of internal HDR: the least sensitive channel (usually green) preserves highlight detail which may blow out in the other channels, whereas the most sensitive (usually red) does the same for shadow detail.

[Andy Perrin]

OlDoinyo, I find the same thing with the 850nm and beyond filters. Much better not to white balance and just convert right to grayscale.

[Stefano]

Can you push colors in deep-IR (increase saturation) or your "color-noise" is too much?

[Andy Perrin]

Can you push colors in deep-IR (increase saturation) or your "color-noise" is too much?

No, the problem is that the Bayer dyes themselves are transparent past 850nm. There is no color to be seen by increasing saturation because all channels are equal.

[Stefano]

The green dye still does absorb a bit. Anyway I imagine you get a monochromatic magenta image, so you can't have any colors.

[Andy Perrin]

Probably it's more accurate to say that the transmission is constant for all three channels past 850nm, so since the R:G:B ratio doesn't vary, you can't get anything with colors.

http://www.astrosurf.com/luxorion/photo-ir-uv3.htm

[Stefano]

OlDoinyo, I find the same thing with the 850nm and beyond filters. Much better not to white balance and just convert right to grayscale.

Anyway, I agree too. I would like much more to have a true B&W image instead of a color image that's practically B&W in which you may see little color variations resulting from noise. Converting it to pure B&W gives a cleaner result.

[dabateman]

What is the sensor size? It appears that the sensor is not covering very much of the image circle in some of these.

 

The pi hq camera is a cs-mount with a Sony BSI IMX477 1/2.3 inch sensor. Its capable of 12 bit raw recording at 3040×4056 pixels of 1.55um pitch with maximum of 10 frames per second. Sensor active area is 4.712 x 6.287 mm with a 7.857mm diagonal, implying a 5.51x multiplier.

 

Being on a Raspberry pi, it allows for almost anything. The UV/IR blocking filter is easily removed, but not replaceable. But they are only $50-$60 each, so just buy 2 or 4 for multi spectral photography.

 

There is also a Stereo pi board that can capture stereo images with it.

 

Or get an Arducam multi camera board to switch between 4 cameras or the Arducam stereo board to use two at the same time.

 

If Dan can get it to monochrome, I see this as quite an exciting system to play with.

 

I wouldn't buy any of the pi lenses. They are cheap for a reason. Better to adapt a know UV lens or make one. All my fused silica elements are in c-mount adapters ready for this. But will be very telephoto. I have two planoconvex 75mm, equivalent to 413mm focal length that might be fun. My 0.5x reducers should help though bring that back into the room.

 

 

[WiSi-Testpilot]

Yes, the filter is very easily removed, and I think it can easily be stuck in again with a little bit glue.

 

https://forums.devel...bility/123646/7

RidgeRun is working on a free version of a driver for the Nvidia Jetson famlily.

 

The low cost Jetson Nano and the Jetson Xavier NX have also two ports for the R-Pi HQ camera.

Best regards,

Wilhelm

[colinbm]

All things being equal, & realising that our domestic camera conversions are probably limited to about 320nm or so, what would be the single most useful focal length for UV photography ?

I ask because IR seems most suitable, but not limited to landscape panos & UV suitable, but not limited to macros....?

[dabateman]

All things being equal, & realising that our domestic camera conversions are probably limited to about 320nm or so, what would be the single most useful focal length for UV photography ?

I ask because IR seems most suitable, but not limited to landscape panos & UV suitable, but not limited to macros....?

 

Colin confused by your question.

If you are going to get a Pi HQ camera with 5.51x crop, then depends on what you like. I like the general field of view of about 30mm and 75mm in 135 format. So to get that I need a 5.4mm lens and a 13.6mm respectfully to see those angles. However, a 16mm on a Pi HQ might still work great for me at under 90mm view.

Those focal lengths are very expensive in fused silica from UKA optics, in M12 mount. But might just cover the sensor.

 

Single Bk7 elements have great over 320nm transmission. I now have a fun bk5 lens ball. The number is amount of lead in glass, bk7 has 7% lead and bk5 is under 5%, that seems to relate slightly to uv blocking. Its a 40mm ball, I am not sure about its focal length, but might work on the Pi. Will test it on my Lodestar which is similar 5.4x crop with 1/2 sensor.

 

Update, UKA optics has been busy. There are now 6mm, 9mm, 12mm, 25mm and 35mm focal lengths in C-mount. The 16mm focal length looks to just be a reducer on the front of the 25mm lense.

https://www.universeoptics.com/uvlens_assemblies/

 

 

[colinbm]

Thanks Dave, I had forgotten to factor in the imagers size, or crop factor.

I was half thinking of Pi but also APS as well.....OK call that a senior moment....:-(

But I do fail to grasp the expense of BK or fused silica for use on converted domestic cameras limited to about 320nm.

I do appreciate that you can off-set that expense with your better UV cameras too.

[dabateman]

Sorry wasn't clear. Bk7 or bk5 is very cheap. Don't pay alot for it they are ripping you off. Bk7 is good for 320nm and above, so your UVA lens.

JGS1 is fused silica, much more expensive and 99.9999% pure. JGS2 is equivalent to quartz about 99.99% pure and cheaper.

 

The UKA lenses aren't too bad, under $1000 and quartz with aperture rings and filter threads. But yes would be more ideal with a monochrome sensor conversion taken directly to the silicon. Hopefully the pi HQ when convert this way is sensitive to some UVC. Not sure, Gpixel seems to have the best UV sensitivity of sensors now. Also released a new 60mmx60mm 16Mpixels sensor. Although I would rather have their GSENSE 2020BSI sensor.

[OlDoinyo]

That sensor might be small enough that something in D-mount such as this Kern 5.5mm lens

 

https://www.ebay.com...H0AAOSw9u9e5pAz

 

might cover it. The smaller lenses have thinner glass and better odds of UV transmission, though I cannot make secure predictions about individual lenses that have not yet been tested. I think that lens would be mid-wide-angle on the sensor in question.

[dabateman]

Yes I was wondering about D-mount. The specifications indicate that a window to the sensor is 8.5mm. Thus many D-mount lenses designed for 8mm, might work, just as many C-mount lenses designed for 16mm work for m43rds. Even though the diagonal is 22mm.

But I don't know of any tested successfully in UV. When I experimented with S-mount (M12 mount) board lenses some were ok. But they needed to be extremely close to the sensor. I scratched my EM5mk2 dust shaker with one. So its a bit risky. I rarely use greater than f16 on m43rds so haven't noticed the scratch much. But you do need to be careful with these lenses.

 

I decided to get a Pi to play with. So will see when it arrives. Will be the first camera I have built up. But again my problem will be software. Will have to see whats out there and may need to learn how to code what I want.

What got me on the Pi was that I could also add Astrophotography cameras through USB. So that opens up a portable system with ASI294 and ASI1600mm sensors if I want to go down that route. I have wanted a ASI1600 for a while, but its just an expensive sensor. No software, no battery, no screen, no computer. So building that up to have fully open switchable screens, batteries, computers in a built camera will be fun.

 

Just need to figure out how to get live view with a set punch in zoom in a specfic region of interest. Thats my most used feature for manual focusing. Then snapping photos shouldn't be too hard and Dcraw has been modified for the pi.

[dabateman]

Well d-mount might be out. Unfortunately the pentax q camera came out driving the price of most D-mount lenses really high.

 

But Fortunately the pentax people have tested most D-mount lenses and there mostly all garbage. Some 38mm exceptions. But not interested in a telephoto lens.

 

I still have to dig through the tests to see any good one. But most look like junk.

 

The Pentax Q is perfect model system for the pi HQ. Same 1/2.3 sensor and same resolution. The IMX477 is newer design then the Pentax Q and Q10 sensor. So hopefully better image quality.

[colinbm]

On Flickr....

[Herman1705]

On Flickr....

https://www.flickr.c...83/in/contacts/

 

That's me

[WiSi-Testpilot]

The UKA lenses aren't too bad, under $1000 and quartz with aperture rings and filter threads. But yes would be more ideal with a monochrome sensor conversion taken directly to the silicon. Hopefully the pi HQ when convert this way is sensitive to some UVC. Not sure, Gpixel seems to have the best UV sensitivity of sensors now. Also released a new 60mmx60mm 16Mpixels sensor. Although I would rather have their GSENSE 2020BSI sensor.

 

I made a first test with my new quarz lens.

 

raspistill -o quso9.jpg -ISO 200 -ss 100000 -awb off -awbg 0.9,0.7

Filter Baader U; ISO 200; 1/10 s; f/2.8, a bit cloudy

 

Filter Baader U; ISO 200; 1/5 s; f/2.8, a bit cloudy

 

Weed or flower, a question of the perspective.

 

 

Best regards,

Wilhelm

[dabateman]

I took a scan of the Raspberry Pi HQ UV/IR blocking filters on the camera. I have two and both are similar. Looks to actually be CM500 as claimed.

 

 

I see about 1 stop drop at 335nm.

 

 

If you stack a 1.5mm ZWB2 filter in front I still see some leakage.

 

Under a percent:

 

This ZWB2 filter came with a Convoy flashlight so it fits perfectly inside. I think A U340 or ZWB1 filter might leak too much into above 600nm.