Camera Light Frequency Sensitivity


10 posts
by poiqwe » Tue Jul 02, 2013 7:03 am
Without the IR filter on; what is the range of frequencies of light that the camera module can receive? The sensor is an OmniVision OV5647.
Posts: 1
Joined: Tue Jul 02, 2013 6:58 am
by jbeale » Tue Jul 02, 2013 6:53 pm
I have no idea what the exact OV5647 spectral sensitivity curve is, but most cameras are similar in this regard. Silicon-based detectors are typically sensitive to light from about 350 nm (ultraviolet) out to about 1000 nm (near infrared). Those are the limits; the sensitivity is greatest somewhere around 700 to 800 nm (which is red to near infrared).

Image

If you need to detect light wavelengths outside those limits, you need to use other materials besides, or in combination with silicon.
User avatar
Posts: 2688
Joined: Tue Nov 22, 2011 11:51 pm
by fuffkin » Wed Jul 10, 2013 5:58 pm
Does anyone know what the curve / characteristic of the IR filter is?

Btw I asked Omnivision for a spectral response curve for the sensor, they were all happy to help until I told them I wasn't building a new camera, then they told me they were too busy to help me and to get lost. I did glean that they won't give out a spectral curve without an NDA. In the spirit of Open Hardware, their attitude and that of Broadcom saddens me :(
Sleepy Pi - Extend RPi battery life with a plugin Power Management & Arduino I/O board
http://www.spellfoundry.com
Posts: 88
Joined: Thu Mar 07, 2013 10:50 pm
Location: United Kingdom
by leol » Thu Jul 11, 2013 12:31 pm
As far as I know, no one has ever said the Pi or its Camera board were open hardware.


fuffkin wrote:Does anyone know what the curve / characteristic of the IR filter is?

Btw I asked Omnivision for a spectral response curve for the sensor, they were all happy to help until I told them I wasn't building a new camera, then they told me they were too busy to help me and to get lost. I did glean that they won't give out a spectral curve without an NDA. In the spirit of Open Hardware, their attitude and that of Broadcom saddens me :(
User avatar
Posts: 130
Joined: Fri Jan 13, 2012 4:27 pm
Location: Reading, England
by jamesh » Thu Jul 11, 2013 12:47 pm
fuffkin wrote:Does anyone know what the curve / characteristic of the IR filter is?

Btw I asked Omnivision for a spectral response curve for the sensor, they were all happy to help until I told them I wasn't building a new camera, then they told me they were too busy to help me and to get lost. I did glean that they won't give out a spectral curve without an NDA. In the spirit of Open Hardware, their attitude and that of Broadcom saddens me :(


However, in the spirit of staying in business, it seems an attractive approach.

Not sure who makes omnivision modules, but if you find the curve for ones made in the same place, it's likely to have the same response. The response is dictated by the pigments used in the microlenses, and I *think* unless you specify them you probably get a fairly standard set.

You could probably work out a partial response from the CCM used in the ISP - take a picture and extract the makernote data and you can get the CCM.
Volunteer at the Raspberry Pi Foundation, helper at Picademy September, October, November 2014.
Forum Moderator
Forum Moderator
Posts: 14843
Joined: Sat Jul 30, 2011 7:41 pm
by fuffkin » Thu Jul 11, 2013 1:04 pm
You could probably work out a partial response from the CCM used in the ISP - take a picture and extract the makernote data and you can get the CCM.


Jamesh,

Lol, you've completely lost me there! CCM? ISP? Makernote data?

Over my head I'm afraid.

Jon
Sleepy Pi - Extend RPi battery life with a plugin Power Management & Arduino I/O board
http://www.spellfoundry.com
Posts: 88
Joined: Thu Mar 07, 2013 10:50 pm
Location: United Kingdom
by jamesh » Thu Jul 11, 2013 1:38 pm
fuffkin wrote:
You could probably work out a partial response from the CCM used in the ISP - take a picture and extract the makernote data and you can get the CCM.


Jamesh,

Lol, you've completely lost me there! CCM? ISP? Makernote data?

Over my head I'm afraid.

Jon


CCM = colour correction matrix. The matrix used to compensate for the spectral response of the filters in the sensor- all RGB values from the sensor (after a lot of debayering etc since the sensor isn't RGB) are multiplied out by this 3x3 matrix
ISP = Image system pipeline - the path the image data takes from the sensor to finally being a image that looks like an image. Lots of stages here.
makernote data = data stored in the JPEG file which contains supplier specific information (ie not in the standard EXIF data). In our case, this is a string added by the ISP to record certain setting used when taking the picture.

If that still doesn't make sense, then you probably don't need the spectral response curve anyway.
Volunteer at the Raspberry Pi Foundation, helper at Picademy September, October, November 2014.
Forum Moderator
Forum Moderator
Posts: 14843
Joined: Sat Jul 30, 2011 7:41 pm
by fuffkin » Thu Jul 11, 2013 2:00 pm
Jamesh,

Thanks for that, I have now attained a state of partial enlightenment! :D

Jon
Sleepy Pi - Extend RPi battery life with a plugin Power Management & Arduino I/O board
http://www.spellfoundry.com
Posts: 88
Joined: Thu Mar 07, 2013 10:50 pm
Location: United Kingdom
by gtoal » Fri Nov 13, 2015 8:12 pm
Obviously the camera works with typical 880nm IR LED illumination. However I'm interested in the farther IR range such as say 1300 or so. We have a beehive project that we want to record in IR and our first experiments seem to show that they reacted badly to the IR illumination overnight. (Although we still have experiments to do to eliminate whether it was the IR frequency that was the problem or just the raw heat from the IR lights) Anyway, we're looking at running a 1300nm IR LED instead. Before buying one (they're about $20 and hard to find, so not a no-brainer...) I thought I should check with anyone who knows, as to whether there's a chance that the RPi NoIR camera will even see those LEDs???

thanks

Graham
Posts: 34
Joined: Sun Nov 18, 2012 12:02 am
by jbeale » Fri Nov 13, 2015 11:14 pm
There is no chance that 1300 nm illumination will be visible to the R-Pi camera, because the silicon bandgap controls the response to light and it is almost nothing beyond around 1100 nm for any silicon-type detector. The RPi camera is a silicon CMOS sensor. For that wavelength you need a different detector material like InGaAs (indium-gallium-arsenide), or germanium, etc.

Image

If you want to be precise, there is a "long tail" in the IR cutoff but the silicon absorption at 1300 nm is five orders of magnitude(!) lower than at 1100 nm, which is already very small relative to the normal, visible light. In practice you'll never see it.
Image
User avatar
Posts: 2688
Joined: Tue Nov 22, 2011 11:51 pm