Subscriber Discussion

Camera Recommendation For Extreme Low Light


We have a client that requires monitoring of certain photosensitive equipments under extremely low light conditions (around 0.0001 lux). The equipment is extremely sensitive to light and so we cannot use IR illuminators. What would be the best fit here? Is there any surveillance camera similar to night vision goggles?

- Pushkar

If u don't care about resolution use Thermal camera

Thermal is a good option and widely available. We have a number of thermal camera tests.

Less widely available but with potentially higher resolution (and much higher costs) are EMCCD cameras (though Toshiba is the only one I know of in the commercial market). Here's an overview of a FLIR EMCCD offering that was cancelled in 2011.

Perhaps there are some 1/2" or 2/3" CCD cameras that you could try? This sounds like a unique industrial application so you might look into a specialty camera that suits your needs.

I can't vouch for this company, but you might try inquiring to this company.

Or you may try this 1/2" CCD camera (not weatherproof)

Or this 1/2" CCD camera (weatherproof).

I've had good luck with the last camera listedand nothing else I've tried comes close to 1/2" sensor if there are no IR LEDs used.

Hope this helps.

We offer a camera that can be used at very low light levelbut it is not color.Unlike an EMCCD it will not be damaged by accidental exposure to bright lighjt.

This camera uses an sCMOS sensor with very low readout. It should work for most low light applications.

Unfortunately the Toshiba IK1000 is now discontinued.

On a general EMCCD note, I checked with Toshiba's service manager this morning and he indicated that during their run, the IK1000 did not have any substantial number RMAs due to bright light exposure.

If the use case is to see color in extremely low light then EMCCD is the tech to go with.

Thanks, Jon. "EMCCD is the tech to go with" if anyone still makes them? :)

Anyone with other EMCCD recommendations, please share.

This is far from surveillance, but here are some companies who have real sCMOS cameras:


Raptor Photonics


They also offer cooled, intensified and EMCCD cameras. Targeted at scientific use and even more expensive than thermal. They additionally require a CameraLink framegrabber, the cables and a fast lens.

Does the customer need a high frame rate? A high quality CCD or CMOS with a long integration time and some post-processing might also be able to deliver an image.

I've done some tests with the Nox-20 developed by Ampleye. Its performance in a pitch-black room while facing a white wall with a mini laser stage lighting device projecting erratic pulses of light on the wall was really great.

Ps.: John, I wanted to upload an image showing my setup mentioned above, but the "Insert/Edit Image" button at the toolbar of the reply text box has disappeared. Is that a bug or a permanent feature from now on?

Tiago, you are recommending a 20MP camera for extremely low light conditions? I am confused.

As a side note, we talked to them this spring and while the concept looks good, they struggled to give us a straight answer and left us with the impression that they were not ready for production.

As for the insert/edit image button, we just upgraded it to allow inserting from a URL as well as editing dimensions. It is enabled and working now.

Hi John, there is no need to be confused or worried about ;) The Nox-20 is equipped with a 35 mm full-frame sensor. If you are familiar with DSLR photo cameras, those that are equipped with full-frame sensors (e.g., Nikon D800 or Canon 5D Mark III) are well-known by providing excellent image quality (at high pixel counts: 36.3MP and 22MP, respectively) in low light conditions.

Quoting here the Wikipedia article about Full-frame digital SLR:

"In addition to wide-angle photography, another major advantage of full-frame cameras is pixel size. For a given number of pixels, the larger sensor allows for larger pixels or photosites that provide wider dynamic range and lower noise at high ISO levels. As a consequence, full-frame DSLRs may produce better quality images in certain high contrast or low light situations."

Ps.: Thanks for fixing the missing button.

I am still confused. Even with a 35mm sensor, you are still talking about 20 million pixels. An SD camera has ~300k pixels. For the pixel size of a 20MP camera to be larger, the imager would need to be 66x larger than an SD camera. So is Ampleye's imager 66x larger than a 1/3" SD imager? Also, does it have similar or better enhanced AGC algorithms for low light?

Secondly, have you done a low light shootout of this 20MP camera vs a low light SD or 720p camera?

I've never heard anyone make such a strong claim about 35mm low light performance, not even from Avigilon fanbois on their 'PRO' series.

First, one thing here is important to clarify. Sensor size does not necessarily mean proportional pixel counts. For example, Nikon D3X, D800, and D600 all have the same sensor size, 35.9 mm x 24.0 mm, but they also have different pixel counts: 24.5 million, 36.3 million, and 24.3 million, respectively.

Now coming to your assumption, which is partially erroneous by the way. A pixel count of 20 million is indeed approximately 66x larger than a standard definition pixel count (640x480 = 307200). However, a full frame sensor has a square millimeter area of 864 whereas an 1/3" sensor has a square millimeter area of 17.30, approximately 50 times more thus. That shows that image sensor size and pixel counts do not need to walk hand in hand. Answering your question then... no, Ampleye's sensor is not 66x larger than an 1/3" sensor.

I've not concentrated my tests on Ampleye's AGC features, but primarily on its bandwidth consumption. So I cannot unfortunately answer your second question.

With respect to your third question, I've not done a throughout low light shootout (as you normally do), but mine was a bit more subjective comparison (remember, my focus was on bandwidth consumption) between the Nox-20 and some Axis cameras we have it here. My overall impression is that video quality was superb in all situations. Some artifacts were observed with highest compression levels, but nothing that compromised the video quality (opposite to Axis M1054 and M5014 cameras - 1/4” sensor sizes). The Axis P1346 (1/3" sensor size) and P1347 (1/2.5" sensor size) performances were close to the Nox-20, but not at all equivalent.

I'm providing a small video I recorded during my tests for you to have an impression about my setup as well as the Nox-20 image quality.

Tiago, thanks, one other thing with the Ampleye camera. When we talked with them a few months ago, our understanding was that their camera was H.264 for an HD stream but required JPEG2000 for the full 20MP one. Is that what you found? I ask because you mention testing bandwidth of the camera and the demands of JPEG2000.

That's correct, John. At least by the time I tested the camera (about 4 months ago) you could only use the full 20MP potential with JPEG2000. By the way, JPEG2000 is used for the camera's multi-view architecture (Ampleye camera's main feature), which is quite interesting: up to 25 simultaneous views, each one of them programmable with respect to frame rate, compression level, and exposure settings. It is interesting to mention that the camera would be rather an overkill for indoors, but large outdoors areas is where it shows its muscles.

Thanks for the replies...I'll research more using the links provided.

BTW - should have mentioned earlier, thermal camera is not an option. The client needs a "real" image, not worried about the quality but needs more like a conventional camera image. I have looked at the EMCCDs but almost none of the big companies offers them anymore. We want to go with some reputed manufacturer because of warranty concerns.

Though discontinued Toshiba IK1000 seems to be a good fit. It is available at B&H photo for $5600. Would toshiba cover the warranty if one buys it from an online retailer?

Pushkar, I asked B&H to respond. Also, I asked FLIR but they did not respond, presumably they have nothing non thermal to address this.

A B&H employee explained to me that a Toshiba camera purchased from B&H comes with a Toshiba warranty. If the customer is worried about things not covered by the warranty happening to the camera, they can buy an add-on protection plan.

The only question that remains to me is what happens if and when Toshiba physically has no more units - i.e., they stopped production, and ran out of inventory for replacements.

Btw, is the equipment stationary? if so, slow shutter

Hamamatsu makes an sCMOS with great low light performance.

The camera you need is the Hitachi KP-DE500 for color, or the KP-E500 for black and white. These are EMCCD cameras in production with many sold to US agencies among others.

Light specs for the KP-DE500:

  • Color in full motion mode 0.0009 lx
  • Color accumulation mode 0.000015 lx
  • Monochrome in full motion mode 0.00005 lx
  • Monochrome accumulation mode 0.0000008 lx

For the KP-E500:

  • Monochrome in full motion mode 0.00003 lx
  • Monochrome accumulation mode 0.0000005 lx

I have evaluated and used thees cameras and the meet or exceed the specifications.

Bill T.

Thanks for the information John. The equipment is stationary and will be used only on a moonless night in near to total darkness. Just wondering, do any of these EMCCD cameras have a "auto shut off" like feature to prevent damage if the light gets too bright to process (like during transition from night to dawn)? I hope they do.

Bill, thanks for the info. I found this youtube video of the KP-DE500 camera. Looks great, I still see some light in the horizon - not exactly sure how it will behave in total dark settings. Also, any idea how much does this camera cost?

- Pushkar

I saw that video but it confused me a little as well.

If the equipment is stationary I bet setting the shutter speed to 1s will provide a great image for any decent IP camera.

Interesting how they are talking about ED (Extra-Low Dispersion) lenses which I'm assuming are on the camera. ED glass is used in a lot of higher end optics so that various wavelengths of light will have nearly the same focal point which will nearly eliminate the IR shift. Expensive, but sometimes necessary depending on application.

To John's point about shutter speed - yes, all cameras look great at night with high shutter speed, until you have a moving object, which then results in a totally blurry image which is useless.

It would help if there were a moving person or moving subject in the video to get a better assessment on the true video quality.

While there isn't any motion in the scene, the camera IS moving and notice there is no streaking.  This is at a full 30 frames per second. I don't know the pricing but i do know it depends on the lens. 


Theese 2 images show the difference between an EMCCD image and a long shutter Image. Note the EMCCD is a lot noiser, that is a function of the EMCCD mutiplying noise as well as the real photons.

I think with a ~30 deg FOV fixed lens the B&W camera(KP-E500) is around $1500 and the color (KPDE500) is around $2500. I do know a fully hardened PTZ color version with ruggedized ballistic hardened enclouser with a Pentax optically stabilized 60 deg FOV to 1.5 deg FOV, F1.2 lens is $25K.

The camera doesn’t need special protection for bright light as a photo multiplier does.  EMCCD isn't damaged by bright light. This camera does require some light; it isn't a thermal imager at all. FLIR announced and demonstrated an EMCCD camera a few years back. I had one for evaluation and I think it was slightly better than the Hitachi model. FLIR never did release the camera and I believe they abandoned the EMCCD business.

I do not work for or represent Hitachi in any way; however I have been involved in several DHS and other 3 letter organizations low light camera evaluations.

Bill T.