p.5 #3 · Call for help - building sensor readout speed database
Had some downtime in between submissions so added logic to the results site to allow models to be highlighted by specifying a matching regex expression for the model in the URL, as well as support for specifying a column to sort and the order. Also added support for specifying the display format to milliseconds vs the default of secs (1/x) For example, here's a link that will highlight all Canon bodies and sort the results by ascending photo readout time and display as milliseconds:
p.5 #6 · Call for help - building sensor readout speed database
This database is really excellent. Great work and congratulations to everyone building it. I would love to help but you've got my cameras on the board already!
Made a surprising discovery on its 60fps/120fps continuous shooting modes - the sensor readout rate drops significantly. Started a thread on Dpreview about it.
Made a surprising discovery on its 60fps/120fps continuous shooting modes - the sensor readout rate drops significantly. Started a thread on Dpreview about it.
That's interesting. As a Canon R6II user, which can shoot at 40fps in e-shutter, there has been anecdotal discussion that its sensor readout rate also drops at this frame rate as a result.
p.5 #9 · Call for help - building sensor readout speed database
rscheffler wrote:
That's interesting. As a Canon R6II user, which can shoot at 40fps in e-shutter, there has been anecdotal discussion that its sensor readout rate also drops at this frame rate as a result.
I haven't had a volunteer for the R6 II but I'm guessing it drops to a 12-but readout, which should actually be faster than the normal 14-bit. The Z8/Z9 dropping for these C60/C120 modes is rather unusual and likely the result of using the video pipeline to achieve.
p.5 #11 · Call for help - building sensor readout speed database
My understanding is the Canon R5/6/6II and probably also the R7 and R8 drop to 12 bits in e-shutter. According to Photons to Photos there is a noticeable DR penalty when using e-shutter with the R5/6/6II.
The Z8's use of the video pipeline would make sense in part due to these modes only allowing Jpeg capture.
p.5 #13 · Call for help - building sensor readout speed database
snapsy wrote:
I haven't had a volunteer for the R6 II but I'm guessing it drops to a 12-but readout, which should actually be faster than the normal 14-bit. The Z8/Z9 dropping for these C60/C120 modes is rather unusual and likely the result of using the video pipeline to achieve.
@rscheffler, I actually did test the R6 II - I've tested so many I'm starting to forget which ones I didn't see any change in readout speed for the Continuous H+ mode (40fps). I know it's documented to drop to 12-bits for this mode. I checked the shadow noise difference vs single-shot but since the ISO for my test images is 12,800 the difference in shadow noise between 12-bit and 14-bit is not discernible.
p.5 #17 · Call for help - building sensor readout speed database
snapsy wrote:
To demonstrate this effect of shutter speed more clearly I've set up my Arduino with two LEDs, a red and a green, and alternate between the two every millisecond (1/1000).
To expand this even further, I've created a timing chart showing the intersection of light cycles with sensor row readouts at different shutter speeds, to depict why faster shutter speeds produce distinct bands with equal-sized light/dark areas and slower shutter speeds produce softer bands with smaller dark areas.
I did extensive video measurements on this camera (including ProRes RAW), since it's such a popular model for video and has the fastest, full-quality 4K sensor read time of all hybrids currently on the market. You can see the detailed tests by clicking on the model name in the live table, which will get you to here:
It's interesting to do the math of how the Sony A7S III's stills readout of 1/48 (20.91ms) increases to 1/113 (8.83ms) for full-quality 4K (no binning/line-skipping):
* Video uses a 12-bit readout vs 14-bit for stills. Based on my A7 III 14 vs 12 bit measurements, this reduces readout time by 56.7%
* 4K is 16:9, so that aspect ratio crop to achieve that from the 3:2 sensor reduces the # rows from 2832 (stills) to 2160 (4K UHD)
20.91ms (stills readout) * 0.567 (12-bit readout) * [2160/2832 rows] (16:9 crop) = 9.04ms, close to my 8.83ms 4K measurement.
What's interesting is my 4.2k ProRes RAW measurements match the 4K UHD, even though 4.2k raw has 2408 rows vs 2160 for 4K UHD.
- I've tested so many I'm starting to forget which ones 
.