Previous news story    Next news story

Canon releases sample video from high-sensitivity full frame CMOS sensor

By dpreview staff on Sep 13, 2013 at 17:38 GMT

Canon has released a short video produced with a unique full frame CMOS sensor. Announced in March, it's a sensor dedicated to extreme low light video recording with unusually large pixels. The test subject? A group of Yaeyama-hime fireflies on Japan's Ishigaki Island. The sample video was captured without artificial lighting in just 0.01 lux, a light level that challenges the naked eye. Beyond natural and astronomical applications, Canon says the high sensitivity sensor may one day be used for medical and security purposes.

Canon staff used a prototype camera carrying the new CMOS sensor to produce a sample video featuring an extremely dark scene. The 35mm full frame CMOS sensor uses a 1920 x 1080 pixel array with individual pixels measuring a whopping 19μm.

The high sensitivity chip uses large pixels and low readout-noise circuitry, allowing it to capture light around 10 times less bright than current CCDs used for astronomy. Each pixel is 7.5 times larger than those of the Canon EOS 1DX.

Via: Canon Rumors, Source: Canon Global

Comments

Total comments: 228
12
falconeyes
By falconeyes (7 months ago)

This is one possible approach to decent low light capablity in video.

The other approach is to stop the nonsense to subsample sensors in video mode, reading out maybe 1 out of 6 pixels. This is what creates noise and aliasing artefacts in video. Unneccessarily so, as a few cameras (Panasonic, Nokia) show which don't subsample but create a video signal from all pixels. I.e., it is quite feasible.

Therefore, Thumbs Down for Canon to work around a problem they rather should solve.

5 upvotes
Roland Karlsson
By Roland Karlsson (7 months ago)

What? Are you sure? Subsampling for video? Hmmmm ... have to google it.

EDIT:

Do you mean chroma subsampling?

Comment edited 7 minutes after posting
0 upvotes
falconeyes
By falconeyes (7 months ago)

I mean that the typical SLR from Canon or Nikon only reads out a small subset of pixels (aka subsampling) used to create the video signal from.

This is contrasted by supersampling where all pixels are read and the video is a full downsampling from the frame.

P.S.
Some call it line skipping (esp. for Canon) but line skipping really is a special case of subsampling.

Comment edited 1 minute after posting
2 upvotes
Frank_BR
By Frank_BR (7 months ago)

I think you meant spatial subsampling. Subsampling requires relatively few computing resources, so Canon and Nikon use subsampling probably because they do not have video processors powerful enough to process all the pixels from the sensor as Panasonic does.

2 upvotes
DSPographer
By DSPographer (7 months ago)

The 5D mark III does NOT throw away pixels during video read-out. Instead it uses both horizontal and vertical binning to read the sensor in video mode. That is why its low light sensitivity is so much better than the D800:
http://falklumo.blogspot.com/2012/04/lumolabs-nikon-d800-video-function.html

0 upvotes
DSPographer
By DSPographer (7 months ago)

The per-pixel read noise of CMOS sensors has been improving over time with process improvements, and the pixel size has been shrinking over time: but, for a given process it is possible to design a large pixel with about the same per-pixel read noise as a small sensor. Right now that read noise is about 1.5 e- or 1.5 h+ (for pmos sensors). Using large pixels then reduces the read noise per area. With a large pixel and a read noise per pixel of about 1.5 it becomes unnecessary to use photo-multipliers for night vision sensing: that is the purpose of this sensor. Here is another large pixel sensor for this purpose:
http://www.imagesensors.org/Past%20Workshops/2013%20Workshop/2013%20Papers/05-12_029_Tower_Paper.pdf

Comment edited 2 times, last edit 1 minute after posting
0 upvotes
ccs_hello
By ccs_hello (7 months ago)

Now that the demo video is out, it would be helpful that Canon also discloses the optics used (e.g., fl=180mm f/0.9) in the shooting. Also with (estimated) MSRP as a reference for such full-frame image circle lens, it would help some folks starting to size up the required piggy bank.

Comment edited 57 seconds after posting
0 upvotes
RRSKDA
By RRSKDA (7 months ago)

0.01 lux... I believe that's about the same as the black level on a current gen Panasonic plasma screen, maybe slightly more. Pretty amazing sensor.

0 upvotes
Caleido
By Caleido (7 months ago)

Interesting stuff. But at this video quality it looks almost the same as a video shot with my smartphone.

Not really the way to go to impress.

0 upvotes
Plastek
By Plastek (7 months ago)

Considering the amount of light available? That's a state of the art technology. That's basically a DSLR sensor with night vision goggles integrated.

4 upvotes
vladimir vanek
By vladimir vanek (7 months ago)

Yeah, but this video at 0.01 lux looks like your smartphone's video at 100 lx... ;)

13 upvotes
Plastek
By Plastek (7 months ago)

Vladimir - see the post below, one by rrccad, before you post something stupid. Thanks.

0 upvotes
rrccad
By rrccad (7 months ago)

this would be a fascinating advancement for astro work. as so many image stackers would take a 30 or 24fps stream and stack the images to improve contrast,etc. usually this is limited to planetary .. but the sensitivity of this is impressive.

that's roughly the same as -8 EV or around f1.4 for 8M exposure at ISO 100. or around 1/30th of a second f1.4 at 819200 ISO for video work.

yikes.

4 upvotes
Joel Pimenta
By Joel Pimenta (7 months ago)

Like your explanation. And the calc.
-8 ev!!!

But there are some guys comparing that sensor to the one in their cellphones... Amazing!

1 upvote
Leandros S
By Leandros S (7 months ago)

You would, however, need longer lenses to get the same amount of detail as with any stacking you may currently do (assuming you already have a higher resolution camera for this purpose).

0 upvotes
deep7
By deep7 (7 months ago)

Having worked in similar conditions, I find this absolutely remarkable. Good on ya, Canon.

1 upvote
jkoch2
By jkoch2 (7 months ago)

A full-frame sensor with 1920x1080 resolution is probably great for HQ video. Nice if Canon would put one in an affordable video camera. But astronomic applications probably mean astronomic prices.

On the other hand, perhaps the 1/2.3" MP sensor in the HF-G30 might provide, in anything but pitch darkness, perhaps 70% of the low light performance for 1/1,000 the cost. In moderate light, the performance differentials probably dwindle.

Balanced color may be almost impossible to coax out of star illumination, no matter what the sensor.

Isn't an owl's daytime vision proverbially not worth two hoots?

0 upvotes
sportyaccordy
By sportyaccordy (7 months ago)

This is a far more exciting development to me than the generally useless MP race. I'd rather have 6MP @ a clean ISO51200 than 100MP at no more than ISO3200

3 upvotes
Samaistuin
By Samaistuin (7 months ago)

If it's up to 3200 ISO, I'm taking the 100MP and matching lenses any day.

5 upvotes
Silvarum
By Silvarum (7 months ago)

Thing is - you will not have sensor with 6MP @ a clean ISO51200. Noise is defined only by a sensor technology (such as quantum efficiency) and a total (!) gathered light. Yes, bigger pixels indeed collect more light, but they also have more readout noise. Several smaller pixels will gather same amount of light and have proportionally smaller readout noise each.
If you downscale 100MP image to 6MP you will get exactly the same amount of total noise in the picture as if you shot it with 6MP in the first place.

4 upvotes
yabokkie
By yabokkie (7 months ago)

actually why people care about pixels is because they are not small enough.

1 upvote
Plastek
By Plastek (7 months ago)

"Several smaller pixels will gather same amount of light " - no, they won't. Not even in BSI sensors.

6 upvotes
Leandros S
By Leandros S (7 months ago)

@Silvarum: Your reasoning seems to ignore one thing - we're talking digital here! Assuming you are recording your 6MP at 14 bit encoding, you'd have to record each of your 100MP pixels at 100/6*14 = 233 bits in order to be able to pick up the same faint light. So rather than a 6k*14 file, you'd end up with a 100k*233 file - an approx. 280 times increase in file size. That's not great news for any further processing you may want to do...

0 upvotes
howardroark
By howardroark (7 months ago)

Yab, that's why you care about pixels. Others know better.

2 upvotes
yabokkie
By yabokkie (7 months ago)

> smaller pixels will gather same amount of light

they may gather more. because smaller microlenses can be more efficient than larger ones. also microlenses on BSI sensors are more efficient and that's why BSI is needed.

behind the microlenses, we can allow some cross talk among pixels if they are small enough. this is accidentally done with BSI sensors but we tolerate it for we still get better result. the sensor becomes more efficient if some cross talk is allowed.

Comment edited 5 times, last edit 14 minutes after posting
0 upvotes
yabokkie
By yabokkie (7 months ago)

on pixel bits, more pixels we have, less pixel bits we need.
a perfect sensor will need only 2.

Comment edited 5 times, last edit 6 minutes after posting
0 upvotes
Plastek
By Plastek (7 months ago)

Yabokkie - they will gather less light. Microlenses got nothing to deal with that.
You want a hint? Sensor doesn't gather light with it's whole surface.

0 upvotes
Silvarum
By Silvarum (7 months ago)

@Leandros S: I was not talking about file size or processing power needed. I wanted to say that pixel size does not matter for noise performance.
I agree, whole purpose of this sensor is to make video processing much more effective and be able to use data from all pixels in video recording. For stills it is useless.
@Plastek: Yes they will (more or less). In this case 7.5 pixels will gather rougly same amount of light as one pixel in this sensor. Have you ever seen a big BSI sensor? Area of supporting circuitry is pretty much insignificant in bigger sensors. That is why Sony said there would be little benefit of BSI technology even in APS-C sensors.

0 upvotes
yabokkie
By yabokkie (7 months ago)

> Sensor doesn't gather light with it's whole surface.

that's why we need microlenses right?

now we have seamless microlenses which means whole surface. only they are no perfect. there are efforts to put smaller more efficient microlenses on larger pixels. you can search the web to teach yourself.

miscounting of microlenses was one of technical reasons behind the stupid Oly 4/3" mount.

1 upvote
tabloid
By tabloid (7 months ago)

Usual negative rubbish ...., written by armchair critics, who's comments are not worth a grain of thought.

4 upvotes
Robert Morris
By Robert Morris (7 months ago)

Yep, that's Canon. New 2K video when the world is moving to 4K, yesterday's tech for todays cameras.

2 upvotes
DT200
By DT200 (7 months ago)

My cable co does not offer 4k. my local bestbuy does not offer any 4k. All my local providers don't offer the bandwidth to adequately offer 2K, LOL!
Everyone said the world ws moving to 3D too. :)

8 upvotes
ntsan
By ntsan (7 months ago)

So does DVD and 1080P when they are first available, who need 1080P when TV station only do 480p :p

Comment edited 2 times, last edit 33 seconds after posting
3 upvotes
pgb
By pgb (7 months ago)

4K on the small screen, I wonder how much better it would
be ? For theatrical release it makes more sense but it's not the
norm yet.

One hour of Arri Alexa 3K raw is 800gB, to get that
down to a distribution size requires a lot of compression !

Good for the manufacturers though.

1 upvote
Robert Morris
By Robert Morris (7 months ago)

The screens are here, the recording media is here, the cameras are here. So we are ready for 4K. Now we only need the content to watch. Who still has Cable TV anyway and who watches movies on cable if that do. I don't own a 70" screen to watch cable TV.

0 upvotes
Plastek
By Plastek (7 months ago)

Nope, we are not. Almost noone got computers capable of displaying smooth 4k video, while 4k TVs are more exotic right now than 3D. Than there's lack of content and high costs and fact that on most of smaller TVs you will barely see any difference vs quality 1080p stream.
Long story short: 4k is overhyped.

0 upvotes
Robert Morris
By Robert Morris (7 months ago)

Isn't everything! But it has to start somewhere and I do believe it will take hold faster than 3D will.

0 upvotes
Beckler8
By Beckler8 (7 months ago)

Do you idiots even know what you're looking at? Yeah, light levels that would produce a black screen from your useless iphone camera.

16 upvotes
Plastek
By Plastek (7 months ago)

At 1/30th of a sec even your DSLR would produce a black screen. (Perhaps with some yellow, barely visible dots - if at all)

0 upvotes
tjbates
By tjbates (7 months ago)

Amazing technology. Sadly in Canon's hands the best we can expect is that it'll be put into a camera body with no headphone jack and lousy downscaling to produce horrible moire and aliasing artefacts.

4 upvotes
Iskender
By Iskender (7 months ago)

As others have said, this is really low quality for a camera promotion video.

However, astro buffs may be interested to know that the bright stars at the beginning are Alkaid and Mizar (+Alcor). Even with the bad compression, I spotted some seventh magnitude stars. The actual performance could very well be a magnitude or two better.

I wouldn't mind a camera that could capture stars that dim while it's basically being waved around.

1 upvote
Plastek
By Plastek (7 months ago)

It is a brilliant quality once you realize the amount of light available. Your Oly would give black picture with similar exposure times on highest sensitivity.

3 upvotes
tkbslc
By tkbslc (7 months ago)

Not talking about the video footage, but the quality of the production. Highly compressed for web and with corny music.

1 upvote
Iskender
By Iskender (7 months ago)

"It is a brilliant quality once you realize the amount of light available. Your Oly would give black picture with similar exposure times on highest sensitivity." Yeah, read the last paragraph of my post where I basically say the same thing.

As tkbslc said, the problem is the added video compression. All the text is surrounded by compression artifacts, and the rendered picture of the sensor has artifacting too.

Basically Canon have this amazing new sensor, but they're making it look bad by re-encoding the video wrong.

0 upvotes
yabokkie
By yabokkie (7 months ago)

one of reasons why we need more pixels.
it's more difficult to make large pixels to beat small ones.

0 upvotes
Asa Spades
By Asa Spades (7 months ago)

I really hope that you are joking!

6 upvotes
yabokkie
By yabokkie (7 months ago)

should read Canon's paper on the new sensor.

Comment edited 1 minute after posting
1 upvote
Jogger
By Jogger (7 months ago)

lol, so much vitriol from the comments.. must be spillover from the EM1 discussion. this is just a proof of concept/engineering prototype.. its not even aimed at consumers or the pro cine market.

5 upvotes
Frank_BR
By Frank_BR (7 months ago)

Just a nice attempt to hide the fact that Canon is losing the battle of sensors to Nikon?

1 upvote
MikeFairbanks
By MikeFairbanks (7 months ago)

Yes, Nikon sensor are so good they even have oil spots on them to give us that retro-dusty-film look. :)

7 upvotes
Shaun_Nyc
By Shaun_Nyc (7 months ago)

Pssst. Nikon does not make sensors, none nada zippo

6 upvotes
Spectro
By Spectro (7 months ago)

Nikon use Sony sensor. As do pentax, Sony, some Olympus, hasselbald. Fuji base sensor spec of xtran might be a Sony too. Canon dynamic range and ISO is behind. Canon image looks nice but doesn't look as natural as the Sony sensor. A tad saturated, very black black or less d range. But most people PhotoShop and sat stuff so raw talk is pointless in the end. This new sensor is a good development. Security camera need better sensor. Can't tell who is who must case.

2 upvotes
Caleido
By Caleido (7 months ago)

psst Shaun_Nyc
Yes Nikon does. And Nikon did. Most are Sony, others have been designed by Nikon. Put even Sony sensors get a proper Nikon treatment and outperform the same sensors in other cameras.

But where the sensors come from is irrelevant. The current line-up clearly gives the golden cup to Nikon.

2 upvotes
Plastek
By Plastek (7 months ago)

"Put even Sony sensors get a proper Nikon treatment and outperform the same sensors in other cameras."

Caleido - FYI according to DXO it won't. But put it through Pentax treatment instead and it will be the winner. (Pentax right now is the best in tuning sensors, Nikon being close second)

3 upvotes
HowaboutRAW
By HowaboutRAW (7 months ago)

Spectro:

I’m sure Renesas can make a very very good 4MP full framed sensor. The sensors they made for the Nikon D4 and D3s are extraordinary. (Note, Renesas is not Sony.)

Now I wish that Sony, Canon, Nikon+Renesas would all consider shipping say a full framed 6MP camera body, Leica too.

1 upvote
HowaboutRAW
By HowaboutRAW (7 months ago)

Plastek,

I just knew someone was going to bring up the DXO score for sensors used in Nikon DSLRs.

Yet again, DXO scores for sensors are next to meaningless, because they don’t account for the lenses or the onboard processing of the raw data.

The perfect example: The Leica M9, the sensor does not score well, but through ISO 800 the image quality will beat about any other full framed camera. (The yes of course a Zeiss lens will really help the image quality say a Nikon D800.)

So Caleido's point is perfectly valid. Sony makes sensors for some Nikon DSLRs and Sony uses those same sensors in some of their own cameras, but Nikon gets better performance out of those same sensors--even with Sony having some good Zeiss lenses.

Comment edited 2 times, last edit 8 minutes after posting
1 upvote
Dave Luttmann
By Dave Luttmann (7 months ago)

Well congratulations to Canon then for making under-performing sensors all by themselves. What a badge of honour!

1 upvote
Spectro
By Spectro (7 months ago)

howaboutraw

yes, Nikon use many sensor, but sony is the main one generally speaking. Renesas parent company or stockholder is Mitsubishi who is also Nikon. Aptiva is the nikon v-series with phase detection on sensor. Toshiha sensor is on the d5200. Most camera people buy from nikon are sony.

1 upvote
mosc
By mosc (7 months ago)

A sensor that doesn't have enough pixels for me. Haven't seen one of those in like a decade! I mean 2mp is pretty nich but give me a ~9mp sensor sufficient for 4k video and I'll be very happy to use it for stills too. I know it's not popular but I crave lower light capability in a camera, not resolution. 2mp is pretty small though

2 upvotes
AbrasiveReducer
By AbrasiveReducer (7 months ago)

Good for private detectives, I guess. If this was from Nikon it would be one more example of how far they and Sony stay ahead of the competition. But since it's from Canon, it's a fail dude.

1 upvote
JEROME NOLAS
By JEROME NOLAS (7 months ago)

Great, so does it mean Canon will have better sensors? I doubt it...

0 upvotes
guatitamasluz
By guatitamasluz (7 months ago)

Someone over at CR said:
"4K is the leading edge of video, this requires almost 9 MP for 16:9, not 2-MP.
The specifically for video probably mean things like: ways to eliminate moire; possible global shutter, or at least minimized rolling shutter; possible RAW video, pixel binning for lower resolution modes; ultra high readout speeds for slow motion (240 fps or greater) etc."
http://www.canonrumors.com/forum/index.php?topic=16945.0

Comment edited 25 seconds after posting
0 upvotes
Joe Ogiba
By Joe Ogiba (7 months ago)

This sensor is for 1080p so it's about 2MP. Show me were Canon said this is for 4K.

"Each pixel is 7.5 times larger than those of the Canon EOS 1DX."
The 1DX has an 18MP sensor so do the math.

1 upvote
Joe Ogiba
By Joe Ogiba (7 months ago)

I guess it's a 2MP full frame sensor.

0 upvotes
jhinkey
By jhinkey (7 months ago)

Wrong Post!

Comment edited 51 seconds after posting
0 upvotes
ptox
By ptox (7 months ago)

So they're showing off their new, super sensitive CCD featuring extra low noise readout... via a low quality promo video filled with ugly compression artifacts. Great job, Canon PR people!

14 upvotes
tkbslc
By tkbslc (7 months ago)

And with only the finest in 1990 era synthesized background music.

0 upvotes
vladimir vanek
By vladimir vanek (7 months ago)

you guys really don't know what you're talking about. :) this is a video shot in pitch darkness, as you might have read (or not) in the article. you wouldn't be able to see anything with bare eyes in such darkness. that's why it's a really interesting development.

3 upvotes
jhinkey
By jhinkey (7 months ago)

+1 Vladimir! Too many people spouting first before thinking.

0 upvotes
tkbslc
By tkbslc (7 months ago)

So you liked the music, then?

0 upvotes
ptox
By ptox (7 months ago)

@vladimir: well duh. But why not let us see its full capabilities with a high quality video? The compression makes it look like crap.

1 upvote
drumsultan
By drumsultan (7 months ago)

Very nice!

5 upvotes
Total comments: 228
12