p.6 #3 · Call for help - building sensor readout speed database
Would it be possible to include crop factor in the video section or other things like line skipping? I understand these are probably something that needs looked up instead of measured. Also any thoughts on additional photo speeds, for example the a7Rv does 1/10s in lossless compressed raw but 1/20s in compressed raw, basically the difference between 14 bit and 12 bit.
p.6 #4 · Call for help - building sensor readout speed database
tschopp wrote:
Would it be possible to include crop factor in the video section or other things like line skipping? I understand these are probably something that needs looked up instead of measured. Also any thoughts on additional photo speeds, for example the a7Rv does 1/10s in lossless compressed raw but 1/20s in compressed raw, basically the difference between 14 bit and 12 bit.
I include crop and compressed vs uncompressed resuults in the detailed measurements of many of the cameras - you can get to it by clicking the camera name in the table.
Establishing line skipping/binning would require shooting specific targets and analyzing the resulting output, which would be very involved. Fortunately it's mostly obvious which cameras are line skipping/binning based on their video readout speeds vs still speeds, net of the difference of the lower bit depths associated with video.
Here's a demonstration of the R3's high-speed flicker reduction feature, applied to the Arduino board LED that I'm running at 500Hz for my rolling shutter measurements.
p.6 #12 · Call for help - building sensor readout speed database
Interesting, it is clear then that all full frame stacked sensor cameras from Canon/Nikon/Sony except the A9III has slower readout in video.
Another exception is the OM-1, but that is the slowest readout stacked sensor to begin with.
Wonder if that is because running fast readout to DRAM constantly generates to much heat in a stabilized sensor which probably doesn't have the best heat dissipation. Could also be a reason we don't see more cameras with pre capture.
p.6 #13 · Call for help - building sensor readout speed database
randomguy wrote:
Interesting, it is clear then that all full frame stacked sensor cameras from Canon/Nikon/Sony except the A9III has slower readout in video.
Another exception is the OM-1, but that is the slowest readout stacked sensor to begin with.
Wonder if that is because running fast readout to DRAM constantly generates to much heat in a stabilized sensor which probably doesn't have the best heat dissipation. Could also be a reason we don't see more cameras with pre capture.
Thanks for the excellent work!
Heat would be my guess. Someone in my dpreview thread about this found a note in Sony's datasheet for the A9 sensor and it states the on-chip stacked memory can't be used in video mode due to heat.
I'ts not clear the A9 III has a faster readout for video. Since its global shutter latches the pixel values into separate registers as the method to avoid rolling shutter artifacts it can read out those registers at whatever minimum speed is necessary to support the frame rate.
p.6 #14 · Call for help - building sensor readout speed database
snapsy wrote:
Heat would be my guess. Someone in my dpreview thread about this found a note in Sony's datasheet for the A9 sensor and it states the on-chip stacked memory can't be used in video mode due to heat.
I'ts not clear the A9 III has a faster readout for video. Since its global shutter latches the pixel values into separate registers as the method to avoid rolling shutter artifacts it can read out those registers at whatever minimum speed is necessary to support the frame rate.
By register you mean storing the charge on a capacitor until it is read? Just asking since I usually associate the word with a digital storage.
p.6 #15 · Call for help - building sensor readout speed database
randomguy wrote:
By register you mean storing the charge on a capacitor until it is read? Just asking since I usually associate the word with a digital storage.
Yep. The sensor transfers the charge of all pixels concurrently into intermediate registers, where it's no longer integrating the exposure and thus can be read at whatever speed the camera wants to meet its other demands, for example the fps rate of the video.
Should look identical since the R8 is basically the R6II without IBIS and probably a smaller buffer (or is restricted to only use a portion of the full buffer for purposes of model differentiation).
---------------------------------------------
snapsy wrote:
And here's the Canon R3...had to buy this one myself to get the measurements
p.6 #17 · Call for help - building sensor readout speed database
Using completely different mesuring method I came to almost identical result (4.8 ms for ES for photographs) for R3. I also measured readout time for mechanical shutter (EFC) on R3 end came to unexpected result: 1.34 ms or 1/747 sec. Could someone test this value (for EFC) of some other manufacture's camera - would be interesting to see what value has Sony A1.
p.6 #18 · Call for help - building sensor readout speed database
Mikica wrote:
Using completely different mesuring method I came to almost identical result (4.8 ms for ES for photographs) for R3. I also measured readout time for mechanical shutter (EFC) on R3 end came to unexpected result: 1.34 ms or 1/747 sec. Could someone test this value (for EFC) of some other manufacture's camera - would be interesting to see what value has Sony A1.
I plan to measure the R3's mechanical shutter as well - forgot to in my first pass. One thing you have to be careful when measuring mechanical shutters is accounting for their acceleration / non-linear speed over their travel distance. Measuring at the start of the movement will yield significantly different results than measuring the middle and end of the travel. What's interesting is that the EFCS is timed to account and match this, making for some interesting engineering.
p.6 #19 · Call for help - building sensor readout speed database
Mikica wrote:
Using completely different mesuring method I came to almost identical result (4.8 ms for ES for photographs) for R3. I also measured readout time for mechanical shutter (EFC) on R3 end came to unexpected result: 1.34 ms or 1/747 sec. Could someone test this value (for EFC) of some other manufacture's camera - would be interesting to see what value has Sony A1.
I just measured the R3's EFCS at 1/274 and fully mechanical shutter at 1/253. This roughly matches Canon's R3's stated flash x-sync speeds of 1/250 for EFCS and 1/200 for fully mechanical. It's 1/180 for the electronic shutter.
