Page 13 of 18

Re: Foundation Camera Board information

Posted: Tue May 14, 2013 8:42 pm
by madrian
Thank, we'll see. I want to build this camera module into a dummy security camera case. :)

Re: Foundation Camera Board information

Posted: Wed May 15, 2013 6:48 am
by simplesi
I want to build this camera module into a dummy security camera case
Its going to be a lot cheaper to put a dummy camera in a dummy case rather than wasting a real one :)
Simon

Re: Foundation Camera Board information

Posted: Wed May 15, 2013 9:01 am
by Grid
I ordered two on launch day, can't wait!!

But while I do wait, is there a technical drawing with the dimensions of the PCB, and the positioning, size, height of the sensor so that I can get started on an enclosure?

Thanks in advance

Re: Foundation Camera Board information

Posted: Wed May 15, 2013 5:32 pm
by zoukimando
Just got mine, and wanted to say what a great job you guys have done. Far beyond my expectations. Well done guys

Re: Foundation Camera Board information

Posted: Wed May 15, 2013 10:26 pm
by wallarug
Grid wrote:I ordered two on launch day, can't wait!!

But while I do wait, is there a technical drawing with the dimensions of the PCB, and the positioning, size, height of the sensor so that I can get started on an enclosure?

Thanks in advance

This looks like a cry for wiki maintenance and updating (by the community).

Re: Foundation Camera Board information

Posted: Fri May 17, 2013 3:45 pm
by madrian
Mine is arriving at Monday, because of the weekend.

Fast :)

UPS

How many GB's are needed for 24 Hour hi-res footage?

Re: Foundation Camera Board information

Posted: Fri May 17, 2013 4:07 pm
by jamesh
madrian wrote:Mine is arriving at Monday, because of the weekend.

Fast :)

UPS

How many GB's are needed for 24 Hour hi-res footage?
Timelapse stills? VIdeo? Resolution? Framerate? Quality? All affect the amount of time you can record per GB.

Re: Foundation Camera Board information

Posted: Fri May 17, 2013 9:18 pm
by fastmapper
I have focal length estimation results.

My Raspberry Pi camera was delivered yesterday. Today I setup my measuring tapes and took photos to estimate the camera focal length. I measured the scene width at 5 different distances from the camera and I got a consistent result. My estimate of the focal length is 6.50+/-0.01mm. I suspect that means the nominal focal length is 6.5mm. This corresponds to a 31.19+/-0.05 degree horizontal field of view at 2592 horizontal pixels and a 23.65+/-0.04 degree vertical field of view at 1944 vertical pixels.

I also noticed that the EXIF tags produced by raspistill indicate a focal length of zero. This is impossible for any camera. I suggest the software be updated to indicate a focal length of 6.5mm until a better focal length estimate is available.

Re: Foundation Camera Board information

Posted: Fri May 17, 2013 9:38 pm
by jbeale
Thank you for your careful work; I appreciate that you indicated the error bounds on your estimate! +/- 0.15% seems quite good.

There is some reason to suspect http://www.raspberrypi.org/phpBB3/viewt ... 43&t=44077 that we have not yet seen the full frame that this sensor produces, but this is just my own suspicion, by no means confirmed. If I am right, we are seeing a 1440x1080 crop of the full 2592x1944 frame, which is then upscaled by 1.8x to get 2592 x 1944. JamesH has indicated he will look into this matter next week, maybe we will know more then.

Meanwhile, since you have a camera, I'd be curious to know if you can get a crisply-focused, nearly-but-not-quite horizontal, high contrast edge or line, and see what you make of the aliasing.
This image http://www.raspberrypi.org/phpBB3/viewt ... 50#p329599 is what is making me think there is an upscaling step between the sensor and the output we see.

Another hint is how the field of view varies when the camera is fixed in place, and switched between 1080p video, and full-res stills. If additional scene detail around the edges is not visible in the 2592x1944 still, then either the sensor is cropping and upscaling in stills mode, or it is downscaling in video mode. If the latter was true, you should see more aliasing / moire effect on fine high-contrast detail in the 1080p video mode, than in still mode.
Now according to http://www.raspberrypi.org/phpBB3/viewt ... 43&t=44077 the stills mode gives you a smaller field of view than 1080p video, which is the opposite of what it should be if both modes are using a 1:1 mapping of image pixels to sensor pixels.

PS. If you need some test patterns, I have a few here: http://www.bealecorner.org/red/test-patterns/index.html

Re: Foundation Camera Board information

Posted: Fri May 17, 2013 11:34 pm
by fastmapper
The focal length estimates I made presume that the image has not be scaled, so if there is a 1.8x scaling, that would imply a focal length closer to 3.6mm.

I have also made some related physical measurements and I think the distance from the lens cover front to the focal plane is about 5.2mm. It would be very unusual for the focal length to be longer than that.

If we can get the raw Bayer form of the image, then there should be no image scaling. I'll see what happens when I try the related option of raspistill.

Re: Foundation Camera Board information

Posted: Fri May 17, 2013 11:56 pm
by jbeale
fastmapper wrote:If we can get the raw Bayer form of the image, then there should be no image scaling. I'll see what happens when I try the related option of raspistill.
That is a good idea. I would be very interested to see such a file with the embedded raw data, to try some experiments. I don't know if any raw data from this camera has been published yet. Something like LibRaw might be helpful to decode it.

Re: Foundation Camera Board information

Posted: Sat May 18, 2013 6:04 am
by Paul Webster
fastmapper wrote: I also noticed that the EXIF tags produced by raspistill indicate a focal length of zero. This is impossible for any camera. I suggest the software be updated to indicate a focal length of 6.5mm until a better focal length estimate is available.
others have reported that the EXIF tags do not match what the source code is trying to set - and investigation is in progress - so this might be a related bug.

Re: Foundation Camera Board information

Posted: Sat May 18, 2013 1:16 pm
by peepo
what is the thread size?

[quote="M33P"]If you unscrew the lens from the default position you can get a minimum ~6cm focal length. The FOV isn't spectacular though - about 30 degrees.
Do I understand the lens remains in place, barely held by a thread?

iirc astrophotography doesn't rely on IR, just tries to gather all available light,
so should be good to go, with lens off.

Re: Foundation Camera Board information

Posted: Sat May 18, 2013 1:56 pm
by M33P
I don't know the thread size - hazard a guess at 6mmx0.5mm. The threaded insert does allow for quite a bit of adjustment.

While you can get pretty good planetary pics with an IR filter in place, it's in wide-field objects (nebulae) that IR filters really hurt performance.

656nm is one of the principal emission lines of various types of nebulae - hydrogen-alpha which unfortunately is in the far red end of the visible spectrum. Many types of IR filter will have a cutoff that starts at about 650nm which means that H-α gets unfairly attenuated.

In addition, most of the dust and non-luminous content of a nebula will actually emit quite a lot in the range 700-1000um (near-IR) which typical CMOS detectors will capture.

Re: Foundation Camera Board information

Posted: Sat May 18, 2013 3:50 pm
by peepo
[quote="M33P"]a modest amount of work..
what did you twist the lens with?

gave myself a real shock,
as the whole lens with ribbon detaches fairly easily.
luckily no harm done, cheesy fit.

but epoxy must be present still,
will leave until morning...

~:"

Re: Foundation Camera Board information

Posted: Sat May 18, 2013 9:03 pm
by fastmapper
jbeale wrote:
fastmapper wrote:If we can get the raw Bayer form of the image, then there should be no image scaling. I'll see what happens when I try the related option of raspistill.
That is a good idea. I would be very interested to see such a file with the embedded raw data, to try some experiments. I don't know if any raw data from this camera has been published yet. Something like LibRaw might be helpful to decode it.
I tried capturing raw Bayer imagery using raspistill, but it appears as if there is no effect when using the option to request it. Perhaps this is related to the general problem of setting EXIF tags.

I did capture some stills in BMP format. I presume there has been no compression, decompression, or significant removal of high frequency from images captured in this format. I evaluated 6 photos that I took to estimate the highest preserved frequency in the 2592x1944 photos and predict the number of independent pixels. This is a very rough estimate (and a lower bound), but accounting for the Bayer arrangement, my experiment suggests that the source image is only about 1272 by 776 pixels. This is significantly lower than expected so I think this is additional evidence that camera images are being scaled and not being produced at the maximum resolution provided by the image sensor.

Re: Foundation Camera Board information

Posted: Sun May 19, 2013 3:25 am
by fastmapper
fastmapper wrote:
jbeale wrote:
fastmapper wrote:If we can get the raw Bayer form of the image, then there should be no image scaling. I'll see what happens when I try the related option of raspistill.
That is a good idea. I would be very interested to see such a file with the embedded raw data, to try some experiments. I don't know if any raw data from this camera has been published yet. Something like LibRaw might be helpful to decode it.
I tried capturing raw Bayer imagery using raspistill, but it appears as if there is no effect when using the option to request it. Perhaps this is related to the general problem of setting EXIF tags.

I did capture some stills in BMP format. I presume there has been no compression, decompression, or significant removal of high frequency from images captured in this format. I evaluated 6 photos that I took to estimate the highest preserved frequency in the 2592x1944 photos and predict the number of independent pixels. This is a very rough estimate (and a lower bound), but accounting for the Bayer arrangement, my experiment suggests that the source image is only about 1272 by 776 pixels. This is significantly lower than expected so I think this is additional evidence that camera images are being scaled and not being produced at the maximum resolution provided by the image sensor.
I also evaluated four BMP format images that were collected at a size of 1296x972 and got similar results. Those smaller images have an estimated source resolution of about 1240 by 824 independent pixels. Changing the image size by a factor of 2 in both dimensions results in only a slightly different effective resolution. I think this is a very strong indicator that not only is the entire array of pixels not being used, but also the images are collected at a fixed size and scaled to produce the result.

It looks like I'm getting 1 megapixel photos from my 5 megapixel camera. It sure would be nice to get this resolved soon.

Re: Foundation Camera Board information

Posted: Sun May 19, 2013 3:54 am
by Bikeman
Does this take into account the nature of the Bayer array? A 5 Mega pixel sensor has only 1.25M pixels that detect blue, 1.25M pixels that detect red, and 2.5 M pixels that detect green. So there can be no more than 1.25 M truly independent pixels in the de-bayered final image. Or am I missing something?

HB

Re: Foundation Camera Board information

Posted: Sun May 19, 2013 4:37 am
by fastmapper
Bikeman wrote:Does this take into account the nature of the Bayer array? A 5 Mega pixel sensor has only 1.25M pixels that detect blue, 1.25M pixels that detect red, and 2.5 M pixels that detect green. So there can be no more than 1.25 M truly independent pixels in the de-bayered final image. Or am I missing something?
Yes. My estimates do account for the Bayer array. In fact, I only examined the red and the blue bands to avoid the more complex possibilities related to the green band. I estimated almost the same number of independent pixels at 2592x1944 (5.0 megapixels) as at 1296x972 (1.3 megapixels).

Re: Foundation Camera Board information

Posted: Sun May 19, 2013 6:37 am
by Bikeman
Ah..ok...but this is exactly what you would expect then, right?? With only 1.25 mega pixels that are sensitive to (say) red on the physical sensor itself, you cannot expect to get more than that in independent pixels in the final image, no matter at which resolution, even 5 M (de-bayered = interpolated) pixels. You should see a difference for green color, but as you said, you didn't test that.

Or to put it differently: if your hypothesis was right that only a crop of ca 1.3 Mega (sub)pixels of the Bayer array was really used for the final image, you would have detected ca a forth , so around 0.3M, independed RGB pixels. That actually proves that the hypothesis is wrong, IMHO.

Cheers
HB

Re: Foundation Camera Board information

Posted: Sun May 19, 2013 7:20 am
by fastmapper
Bikeman wrote:Ah..ok...but this is exactly what you would expect then, right?? With only 1.25 mega pixels that are sensitive to (say) red on the physical sensor itself, you cannot expect to get more than that in independent pixels in the final image, no matter at which resolution, even 5 M (de-bayered = interpolated) pixels. You should see a difference for green color, but as you said, you didn't test that.

Or to put it differently: if your hypothesis was right that only a crop of ca 1.3 Mega (sub)pixels of the Bayer array was really used for the final image, you would have detected ca a forth , so around 0.3M, independed RGB pixels. That actually proves that the hypothesis is wrong, IMHO.
I expect 5 million independent pixels for a 5 megapixel Bayer sensor, but I only estimated about 1 million independent pixels. I did account for the Bayer arrangement. It isn't necessary to look at more than one band of pixels since counting the number from one band can be converted into the number for another band or for all pixels in the array.

The 2592x1944 pixel image and the 1296x972 pixel image covered the same area on the focal plane because the photos show the same area imaged (I have the camera setup in a fixed location). That means that they were sensed over the same physical pixels on the image sensor. There are no visible cropping differences between images of the two sizes; they simply have different numbers of output pixels.

I have worked with similar image sensors and I do have some familiarity with the image scaling capabilities they provide. I still think that the most likely explanation for what I am seeing is that a fixed area of roughly 1 million pixels is being scaled to the image produced. My understanding is that jamesh will be checking into this.

Re: Foundation Camera Board information

Posted: Sun May 19, 2013 8:09 am
by simplesi
Bloomin heck - I didn't know there was so much stuff about a camera! :)

Another RPi thread that have to be required reading for engineering students :)

Simon

Re: Foundation Camera Board information

Posted: Sun May 19, 2013 9:49 am
by jamesh
Before everyone goes completely mental, there does appear to be an issue, so I'll be looking the the problem with the sensor apparently not emitting at full resolution on Monday. No need to keep investigating.

There are only two modes defined in the camera driver at the moment, 1080p30 and full resolution. I think it may be using the 1080p mode instead of full.

Re: Foundation Camera Board information

Posted: Sun May 19, 2013 9:58 am
by poing
fastmapper wrote:It looks like I'm getting 1 megapixel photos from my 5 megapixel camera. It sure would be nice to get this resolved soon.
About 1 MP of data was exactly my guess when I saw the first images coming from my camera. This is only based on experience; not more than a guess but photography has been my day job for 30+ years. I thought it was due to the lens, but that was likely and hopefully a wrong assumption. Very glad to hear this, as we may be positively surprised by the image quality when this glitch is removed.

Re: Foundation Camera Board information

Posted: Sun May 19, 2013 10:00 am
by jamesh
poing wrote:
fastmapper wrote:It looks like I'm getting 1 megapixel photos from my 5 megapixel camera. It sure would be nice to get this resolved soon.
About 1 MP of data was exactly my guess when I saw the first images coming from my camera. This is only based on experience; not more than a guess but photography has been my day job for 30+ years. I thought it was due to the lens, but that was likely and hopefully a wrong assumption. Very glad to hear this, as we may be positively surprised by the image quality when this glitch is removed.
Well, the increase will be about 50% vertically, a bit less horizontally- noticeable but not huge.