PDA

View Full Version : Discovery!



Evan
02-09-2008, 10:19 AM
I have been fiddling with another web cam, a Microsoft Life Cam. I bought it for a project to monitor the night sky for aurora because it has very good low light sensitivity. I have been slightly disappointed that it isn't as good as I would have liked even with maximum fiddling of the camera settings.

EUREKA! While playing with it this early am in our nearly dark living room I discovered a way to increase the sensitivity by an order of magnitude! It's entirely counterintuitive but makes perfect sense. I was pointing the camera at the floor and it showed almost nothing. Then I moved it toward the computer screen and was astounded to see the floor image come into perfect clarity along with a bright diffuse glow from the edge of the field of view caused by light from the screen. The ability to image in the dark is increased many times by simply shining some light on the sensor.

What this does is sort of "prime the pump" of the sensor. It takes the imager out of the non-linear range where there isn't enough light to produce a signal at all into the middle of the sensitivity range. Once the pixels are producing a signal then even slight differences in brightness are visible because it is operating in the middle of the dynamic range. VERY COOL!

I tried it with one of the outside cameras. It doesn't have the resolution or quality of the MS cam but the same effect applies as it should to all digital image sensors.

This is the outside camera that looks out to the driveway from the north side of the house. The driveway is illuminated by a single 50 watt incandescent floodlight only so the low light performance of this camera isn't too bad.

http://vts.bc.ca/pics3/nightcam1.jpg

Then I shined a powerful flashlight into the camera from the side to flood the sensor. Note that the camera is mounted upside down so it looks as though the light is coming from above rather than below. Also note that I did NOT shine any light on the scene. The only change is the amount of total light reaching the sensor which brings it up to the useful part of the dynamic range. See also how well it images the distant tree shadows in the snow, completely invisible in the first picture. The structure in the foreground is a "tent" garage for the car.

http://vts.bc.ca/pics3/nightcam2.jpg

This is going to be very useful. This effect should apply to many types of sensors and in the case of silicon based sensors it should be sufficient to use an infrared LED to flood the sensor. With proper placement it also should be possible to avoid the spot of glare in the image. I shall be experimenting with this.

I should also point out that this effect has been used by astronomers with film in the past. By using a controlled pre fogging of the film the effective film speed can be increased dramatically. I don't know why this didn't occur to me but as I said, the effect is counterintuitive.

A.K. Boomer
02-09-2008, 10:55 AM
Very cool, My sony handycam has a nightvision button that throws off a little infrared and makes a huge Diff...

Somthing kinda related, Sony produced an earlier version of my nightvision camera --- but it was pulled off the market, tests showed that it could actually see through clothing with kind of an X-ray vision, Damn, what a thing to bring to the beach ehh?:D

Evan
02-09-2008, 11:04 AM
This has nothing to do with illuminating the scene as with the infrared illuminators. It's all about increasing the useful dynamic range of the sensor by flooding it with light. It's an effect that should work with almost any sort of solid state sensor and is very easy to implement.

A.K. Boomer
02-09-2008, 11:11 AM
So in effect could you do this internally inside the sensor? and not even have the red lights that I have to use? this would in turn make your unit completely stealth (if im catching what your saying)

Ev, have you heard of that particular X-ray sony im talking about? is it for real, If so how did that happen...

Evan
02-09-2008, 11:20 AM
So in effect could you do this internally inside the sensor?

Exactly. This is about increasing the sensor sensitivity for pennies. I've tried it on two totally different web cams and it works the same on both. I will be trying it on a few more cameras including possibly my Canon Rebel.

Yes, the Sony night shot camera can see "through" some articles of clothing that are transparent in infrared light. Infrared transparency is common for articles that seem opaque in visible light. A piece of exposed and developed color negative film looks completely black to the eye but it is as clear as window glass to infrared light.

Your Old Dog
02-09-2008, 11:42 AM
Very cool, My sony handycam has a nightvision button that throws off a little infrared and makes a huge Diff...

Somthing kinda related, Sony produced an earlier version of my nightvision camera --- but it was pulled off the market, tests showed that it could actually see through clothing with kind of an X-ray vision, Damn, what a thing to bring to the beach ehh?:D

Boomer, I have the camera you are talking about. What it actually sees are the whitners in laundry detergent so a foxy persons vuluptious and pendoulous.........I digress. It shows underwear and not much else. Not worth the $10,000 I paid on ebay for it when the story hit :D Just kidding of course !

Evan, just curious. How often does your wife have to wind you up? :D

rotate
02-09-2008, 12:32 PM
Evan,

I don't think I understand your explanation. Are you saying the photo diodes (the pixel sensors) have a lower quatum efficiency at lower luminosity, and by offsetting into the linear region you can get better sensitivity?

Evan
02-09-2008, 12:38 PM
Yes. It is more likely the amplifiers though, not the actual quantum efficiency of the detectors. CMOS detectors, which is nearly all cameras now (all webcams), have an amplifier per pixel. The amplifiers almost certainly have a built in low level cutoff to reduce noise in an ordinary low light situation. Pumping in some photons will take the sensor+amplifier into the linear range.

aostling
02-09-2008, 04:39 PM
This is going to be very useful. This effect should apply to many types of sensors and in the case of silicon based sensors it should be sufficient to use an infrared LED to flood the sensor.

I'll be very interested in the results of your experiments with this. Can the proposed infrared fogging can be effective if held to a level just below the threshold of visibility by the sensor? Your experiments may quantify how little (or how much) is needed to just get into the linear-response range. I think visible fogging must degrade image contrast, but perhaps that could be tweaked. After all, there are contrast adjustments on higher-end cameras.

halac
02-09-2008, 08:37 PM
.....Somthing kinda related, Sony produced an earlier version of my nightvision camera --- but it was pulled off the market, tests showed that it could actually see through clothing with kind of an X-ray vision, Damn, what a thing to bring to the beach ehh?:D

I just recently bought a Sony DCR-TVR80 off of eBay. It has the night vision feature. Would this be the model you are talking about?

Evan
02-09-2008, 10:27 PM
First experiment is a success. I place a 9 watt ultraviolet CFL bulb a couple of feet in front of my back door web cam. It doesn't blind or fog the image at all but shows up like an ordinary not very bright bulb. However, the area around the back porch for at least 30 feet is now clearly visible with the only illumination being a 25 watt incandescent bulb. Without the UV shining into the camera only the back porch can be seen. The difference is like , ahem, night and day. :D

JRouche
02-09-2008, 11:14 PM
I so dont get what you are talking about. Over my head. But I bought some cheap cameras from HF and they are supposed to be low light types. I put them in place and nuthin, no image, it was dark.

While playing around with them I found if I shined a bright light toward the lens (from a distance) it would all of a sudden activate the sensor and it would "turn on".. I would get a totally washed out picture till I moved the light away then it would "stay on" and I could see all the area in great detail, all the dark areas.

Dunno if its the same deal as you are dealing with.

For mine it seemed like it needed a "trigger" to bring the sensor into a new state, a low light state..

Sounds like yers needs to have the high light level constant? To keep it "On".. JRouche

Evan
02-09-2008, 11:36 PM
Not exactly. It's really common for the AGC (automatic gain control) to not work unless there is enough light for it to kick in. That's what you are seeing. This isn't the same at all. I have tried it on 3 different web cams and they all respond the same way. I also have a couple of video cams and will be doing some more experiments to see how they respond. The acid test will be to see if it will make it possible for the MS Life cam to pick up starlight without a telescope. Unfortunately the weather forecast is snow and heavy clouds for the next 10 years.

rkepler
02-09-2008, 11:37 PM
It might be interesting to try an IR diode to illuminate the sensor array.

nheng
02-09-2008, 11:59 PM
Evan, You have several things at work here. I've seen several imagers on cameras I've used in my designs behave the same way. The AGC will definitely move to a different operating point with some "bias" on the signal. Depending on how the AGC is designed, this might impact the observed sensitivity of the camera. The other thing is that the light is also raising the photocurrent produced by the array elements. At very low light levels, the dark current produced by the array may be a significant component of the image and will produce a noisy, washed out image. With some light applied, you raise the photocurrent produced and so the dark current is a less significant component. The result would be less noise, a shift in contrast, but also non-linear response to brightness changes which may or may not be visible.

Probably some combination of the above two effects.

Ditto here on the 10 years of snow. We've run out of places to put it and its almost time to get some heavier equipment to relocate it. The snow total as of Christmas day was fine with us :)

JRouche
02-09-2008, 11:59 PM
Not exactly. It's really common for the AGC (automatic gain control) to not work unless there is enough light for it to kick in.

Yeah Evan, I think yer right with whats going on with mine. The AGC is just not firing up until it gets a huge signal then it operates.. What would be nice is to be able to bypass the AGC and control it with a pot or something similar..JRouche

darryl
02-10-2008, 12:02 AM
That is interesting as I'm currently looking for a point and shoot that is capable of taking pictures in the dark, without too long of an exposure time. From what I've read previously, you can remove the ir filter from a lens and it then is capable of actually acquiring an image if there's some ir light on the scene. Most of what I've found is ir cameras that output a video signal- we want to actually take a picture with it for download later. I can come up with an ir 'ring light' without problem, but I'm not sure how an ordinary digital camera with the ir filter removed will work in terms of using the viewfinder, the zoom, etc. Think in terms of reading a licence plate number in the dark from a distance, and identifying the vehicle.

Seems to me this method of applying some light to the sensor to enhance its sensitivity would be good for my application as well.

dp
02-10-2008, 02:41 AM
Fascinating discovery, Evan - it's surprising how many devices that work in a square law region of their range are improved by introducing offsetting input to get them out of the curve. This works on the old radar detector crystals, too - with a bit of induced current the signal to noise ration and sensitivity improve remarkably. I'd not have guessed that modern digital sensors would improve in this way. Makes me wonder what supercooling as done with telescope sensors it would do, too. I see you've tried UV and I wondered about IR so as to not flood the device with visible light.

Evan
02-10-2008, 06:38 AM
Here is a test of the Microsoft Lifecam.

These pics are in my living room taken of the TV and entertainment equipment cabinet. There is one small night light providing just enough light to navigate and no more.

I have adjusted the video setting of the camera to provide maximum sensitivity. This includes a number of tricks I have found by using web cams for night photography. Setting for low light is just the start. By setting the flicker control to 50 hz it gives a slightly longer exposure between frames, I have of course disabled all the automatic controls and maxed out the exposure setting, the contrast and the gamma. Also, by reducing the allowed bandwith to the camera the frame rate slows and this allows a longer exposure as well per frame.

This is the best it can do in this circumstance.

http://vts.bc.ca/pics3/nightshot1.jpg

Just to be clear, this is what shining a single white LED flashlight on the scene reveals.

http://vts.bc.ca/pics3/nightshot2.jpg

But, turning off that flashlight and turning on a single infrared LED that is just barely shining into the camera lens gives this:

http://vts.bc.ca/pics3/nightshot3.jpg

To emphasize, the last pic has no additional illumination in the room. The infrared LED is very weak drawing only a few milliamps. The camera sensor is just as sensitive to IR as it is to visible light and so see the IR as a white light.

aostling
02-10-2008, 09:49 AM
The last photo shows the IR LED illuminating the pixels on the edge of the frame. Obviously, this has increased the low light sensitivity of all the pixels. Is it spill-over fogging of these other pixels which does the job?

I'm guessing that you are planning on installing a fogging LED inside a camera which will not be in the imaging area. If so, it will be very interesting to learn what you have bumped the ISO rating up to.

Evan
02-10-2008, 10:02 AM
I purposely put the IR LED in view so that it was clear what it looked like to the camera. It is only necessary and in fact required that it evenly illuminate the sensor for best results. I have done a couple of experiments with my digital cameras as well. My Nikon has a true CCD detector and it doesn't exhibit the effect. That may be because it has a very effective IR filter which is something I have verified in other previous testing. My GE A830 (actually a Fuji Finepix) has a CMOS sensor and it does exhibit the effect.

As I said earlier, a CMOS sensor has one amplifier per pixel. That amplifier is in the form of a transistor that surrounds the active element in the sensor grid. Complementary Metal Oxide Semiconductor (CMOS) sensors are inherently noisy and have much lower quantum efficiency than Charge Coupled Devices (CCD). They accumulate thermal electrons and also suffer from variations in the amplfier gain from pixel to pixel. To reduce the noise the amplifiers have a threshold value below which they produce no output. Introducing some light generally to the sensor lifts all the pixel elements above that threshold.

What I find most surprising is that I cannot find any mention of this effect on the net. I also wonder why the manufacturers haven't included such a capability in their products. It would be both trivial to engineer and very cheap to incorporate.

rotate
02-10-2008, 10:45 AM
I've tried this last night on my Logitech STX webcam. It did improve the sensitivity but it's no where near the level that Evan has seen. I'm wondering whether the sensor in my webcam is CCD and that makes the difference.

dp
02-10-2008, 05:55 PM
To emphasize, the last pic has no additional illumination in the room. The infrared LED is very weak drawing only a few milliamps. The camera sensor is just as sensitive to IR as it is to visible light and so see the IR as a white light.

That is a remarkable increase in sensitivity. Interesting that the IR shows as white - I'd have thought it would show as some other color and therefore filterable in a photo tweaking program. Is the complete lack of color anywhere in the frame a function of the light level generally? The next trick I guess is to illuminate the sensor from the rear. If IR is all it takes then just warming it should work, no?

aostling
02-10-2008, 06:24 PM
If IR is all it takes then just warming it should work, no?

No. CMOS and CCD sensors are sensitive to the Near Infrared region (700-1400 nm). Thermal radiation from heated surfaces is typically in the Long Wavelength Infrared (3000 nm to 1 mm). http://en.wikipedia.org/wiki/Infrared describes these different regions.