Time for an update. I sort of left my promise of test examples unfulfilled, with that in mind I don’t have any DNG’s or samples from my previous test.
This is going to be sort of a conclusion to this subject, and my opportunity to address some points.
DISCLAIMER: I’m not an engineer or have much of any experience or qualifications in this field, however I believe the underlying points of these statements hold true. As I have stated before and for anyone who understands principles of an image sensor, cooling will improve overall image quality. Low-light or Daylight, (less so on a well exposed image) But results vary on different sensors.
For me cooling did the best job at removing Hot pixels. Sort of makes sense…cooling “HOT” pixels.
Why I didn’t see more improvement beyond that comes down to way the 2.5K and the soon to be 4.6k sensor works.
These sensors achieve their excellent dynamic range using special circuity built onto the sensor that samples each column of pixels at both a Low and High gain, both streams of 11 bit data are then merged to create a 16 bit raw image. The sensors are also readout in halves, that’s what leads to the split sensor artifact.
More about that here:
http://www.andor.com/learning-academy/d ... amic-rangeSplit sensor artifact shown below.
This is important to the way the sensors GAIN or ASA is then controlled by the user. As has been extensively been discussed in the past, the 2.5K (& pocket ) sensor both share a base sensitivity of 800 ASA. In camera adjustment of sensitivity is all digital gain, as opposed to traditional analog gain which is standard in most camera systems. (Not exactly true since there is some better mapping for the shadows that occurs when using 1600 ASA, but for sake of argument it is all digital gain.)
For this reason the Cinema Camera and Pocket Camera will never be as robust as something like an a7s. As you are essentially always shooting 800 and simply stretching the information in camera or in post. This to me is most noticeable in a more recent test I did, where I cooled the camera to near freezing. When looking at near black parts of an image no matter how much I raised the exposure, there was nothing to recover. Analog gain will always be superior vs digital gain, at least in regards to resulting in better SNR.
( and yes there is a radiator hanging out my window, because you can bet I'm going to embrace a Canadian winter to do cooling tests
also that isn't clipped highlights under the water cooler, its ice forming on the peltier/cold plate)
While it is not certain what specific GAIN configuration BMD is using on the sensor, the sensor itself the “CIS 2521f” does have various gain modes. It may be possible for BMD to employ a gain configuration that may be better suited to low light shooting when desirable. This does have a reduction in dynamic range and bit depth because it would essentially be derived from a single channel of the 2x 11 bit stream. It may be possible that BMD could explore this, although very unlikely at this point If they haven’t already done it.
So with this in mind, don’t shoot with these cameras like you would an a7s. It’s not going to work as well, no matter how cool the sensor is. That’s what all this has taught me. So don’t go cooling your cameras with LN2 or putting them in the freezer. It won’t help you shoot in moonlight.
Also as another address to camera thermals, I frequently see sensor heat as a potential limit to High Frame Rate recording. I want to throw my thoughts into this one, because although I agree with this point I don’t believe it’s as severe as it’s put out to be. Heat is generated by the sensor, but to a degree far less than something like the FPGA in these cameras. If we look at what available information we have about the sensors, they operate typically at least than a few watts. Taken directly from the 2.5k datasheet
“Power consumption < 2W at 100 fps”
less than 2 watts in the grand scheme of the entire thermal power design of the camera is very insignificant when it comes to its heat contribution. Your cam gets hot far more from the Peltier, the SSD, the battery and mostly the FPGA.
So to say the “The cam can’t handle it because of sensor heat." is not likely to be as great of a factor as one might think.
I’m not saying the cam can therefore handle HFR the way it is, however to me is suggests is has a lot more to do with the processing pipeline or insufficient power.
With that little rant over, some closing words.
Nothing can change shooting with light, use lots of it. No cooling can substitute a lack of lighting. That’s what I have taken from all this.
The more you know!
Having said this...I'm not done...yet...