Photron announced 1M fps fast camera, the FASTCAM IS-1M, based on ISIS-CCD (In-Situ Image Storage-CCD), better known as IS-CCD. The IS-CCD uses an image storage buffer immediately beside the light sensitive portion of the 10-bit pixel to greatly increase the speed achievable by conventional high speed cameras. The 100 frame in-situ image buffer is said to provide an adequate synchronous record duration for the majority of ultra high speed phenomena such as ink jet droplet measurement, micro PIV, combustion, crack propagation or blast dynamics.
The camera resolution is 312 x 260 pixels and it consumes 150W. Its datasheet has nice pictures of various high speed events.
nice! Congrats to Photron and Prof. Etoh!
ReplyDeletewhat is the scene illumination level needed to operate such camera at 1M fps??
ReplyDelete-yang ni
Yang Ni, surely you can do the math! :)
ReplyDeleteEric, let's go:
ReplyDelete1. 10e RMS noise (suppose)
2. 1% gray level discrimination => 1000e/pixel
3. flux = 1e9 e/pixel/s
4. QE=50% photon flux = 2e9/s/pixel
5. Suppose green light 550nm => photon energy = 2.25eV
6. 720pW/pixel
7. Suppose 10x10um pixel => 7.2W/m2 (1w=683Lumen) => 4900 Lux
8. Suppose F1.4 lens used = 50kLux
What is your comment ? So you can take high quality video at 1M fps under Sunshine in the summer on the beach :)
-yang ni
No fighting, Eric and Yang, or I'll have to ask to have you banned.
ReplyDeleteCMD, I'm checking my calculation. Don't ban me but just check the result for me. It seems to me a little bit optimist for the illumination level, don't you ?
ReplyDelete-yang ni
Here is an alternate and perhaps easier way of getting to what you want. Let's say a 1.4 um pixel has good YSNR at 300 lux at 15 fps.
ReplyDeleteFor ISIS use pixel scale factor of 500 (43 um pixel and high BSI fill factor).
Use an f-no scale factor of 4.
For monochrome use a filter scale factor of 3.
So, this has 6000x improvement in basic sensitivity estimate.
For frame rate, derate by a factor of 1e6/15 = 70,000
So illumination requirement is increased by 70000/6000 = 10x or 3000 lux.
This is similar to your number if you had used a 10x larger pixel area.
So, you won't burn up the subject to get enough light to see something and far less than summer sunshine on a (California) beach.
Hope I did not forget something in this calculation myself!
Eric, it's better to go back to physics and forget a small moment the 1.4um pixel and YSNR :)
ReplyDeleteI don't think that under 1M fps, you can see nothing under 3000 Lux illumination.
-yang ni
nothing = anything
ReplyDeletehttp://hs-movies.de/index.php/archives/nochmal-was-zum-thema-fliegende-geschosskugeln
ReplyDelete3000W lampe used, the illumination should be much higher than 3000lux !!
-yang ni
Gents, you are right that for 1M frames per second you need a lot of light, but why should I need 1M frames for scenes on the beach ? Things you would like to capture on the beach move much slower and you can easily work with 50 fr/s, at least at the beaches in our country. Maybe in California they move faster .... ;-)
ReplyDeleteYeh, the temperature on the California beach is higher than that on Netherland beach, so all moves faster :) Fortunately there is more light also !
ReplyDelete-yang ni
I remember Walter Kosonocky designing a one million frame per second CCD image sensor for Princeton Scientific Instrumentx, Inc back...maybe in the around 1988 or 89. I think the camera PSI made was 100 to 1,000,000 frames per second for 16 frames.
ReplyDeleteand the other comment would be that once you have recorded say 10ms of video at 1Mfps, it will take >5 minutes to watch the resulting video at human friendly 30fps.
ReplyDelete@ yang ni
ReplyDeleteYour numbers are okay. However, since the buffer is full after just 100 frames, the continuous optical power can probably be much less. (Just a little flash of inspiration...) :D