QE and read noise of different DSLRs

[gregbradley]

I found this link I read ages ago and was wanting to find it again.

Its interesting to see the development of DSLRs sensors with increasing QE and lower read noise as later models come out.

One of the best is the Toshiba sensor in the Nikon D5200, that is really high QE and really low read noise. No wonder Nikon chose that sensor over the Sony equivalent.

Of the Canon's 1DX, 6D and 5D3 are the best with the 6D seeming to be the best performance for buck for astro and low light. Not 100% sure how this performance translates over to RAW images as these may all be for jpeg. If you look at dpreview ISO comparison tools jpegs of many cameras are often way cleaner than their RAWs due to in camera noise reduction processing. Usually at a cost of detail and sharpness as most noise reduction involves blurring.

http://www.sensorgen.info/

Greg.

[swannies1983 (Dan)]

Thanks for that Greg. While my two cameras aren't in the list (30D and 400D), it gives me a ballpark figure.

[gregbradley]

Out of interest I looked up the total noise of KAF8300 and KAI11002 which are 16 and 30 electrons respectively.

I am not sure the read noise of the DSLRs are measured the same way (read noise may not be the only source of noise) but the latest DSLRs are around 1.4 to 4 electrons. So unless we are not comparing apples with apples here they are considerably cleaner.

Greg.

[rally]

Greg,

Nice find
Copied the list and calculated out the dynamic Range
I suspect the cameras are all performing in-camera post processing

Simply cannot get this to format - spaces are removed
So this will have to do.

Brand Camera Qe Read noise Well Depth Dynamic range
Canon EOS-1D_X 47% 1.3 90367 96.84
Nikon D4 53% 1.9 117813 95.84
Canon EOS_6D 50% 1.6 76606 93.60
Canon EOS-1D_MkIV 44% 1.5 48702 90.22
Nikon D3s 57% 2.8 83732 89.51
Canon EOS_5DMkIII 49% 2.4 67531 88.98
Nikon D600 53% 2.8 76231 88.69
Pentax K-5 46% 1.9 47159 87.89
Canon EOS_5D_MkII 33% 3.2 64600 86.10
Nikon D5200 65% 1.4 27517 85.86
Nikon D7000 48% 2.5 49058 85.85
Sony A580 47% 2.5 44983 85.10
Nikon D800 56% 2.6 44972 84.75
Canon EOS_60D 40% 1.5 24322 84.19
Canon EOS-1DMkIII 29% 4.7 66480 83.01
Nikon D5000 35% 2.1 28411 82.62
Nikon D3X 35% 3.7 48975 82.43
Canon EOS_5D 25% 4.2 55297 82.38
Canon EOS_20D 26% 3.7 48255 82.30
Canon EOS_1100D 36% 2.5 31787 82.08
Olympus OM-D_E-M5 53% 2.0 25041 81.95
Nikon D3200 43% 2.3 28729 81.93
Sony SLT-A33 24% 2.1 25994 81.85
Canon EOS_550D 40% 2.0 23408 81.36
Sony NEX-7 47% 2.2 25443 81.26

Added later

Some SBIG Astro cameras for comparison

SBIG ST-402ME 83% 17 100000 75.39
SBIG ST-1603ME 83% 17 100000 75.39
SBIG ST-3200ME 86% 10 50000 73.97
SBIG ST-7XME 83% 15 100000 76.47
SBIG ST-8XME 83% 15 100000 76.47
SBIG ST-9XE 67% 15 150000 80.00
SBIG ST-10XME 86% 8.8 77000 78.84
SBIG ST-2000XM 56% 7.9 45000 75.11
SBIG ST-2000XCM 45% 7.9 45000 75.11
SBIG ST-4000XCM 45% 8 40000 73.97
SBIG STL-1001E 72% 15 200000 82.49
SBIG STL-4020M 55% 7.9 40000 74.08
SBIG STL-4020CM 45% 7.9 40000 74.08
SBIG STL-6303E 68% 13.5 100000 77.39
SBIG STL-11000M 50% 13 50000 71.70
SBIG STL-11000CM 50% 13 50000 71.70

I do wonder if the extremely low read noise calculations for the DSLRs are only imputed from the image data that has already been noise reduced by the cameras internal processing system - rather than the actual CCDs real read noise ?

Of course dark noise and ability to cool, blooming specs, image scale, mono Vs colour, ability to filter etc are all equally important considerations - so Dynamic range is just one of many things to compare not the B-All

[DavidNg (David)]

Hi all,
Looking at these figures. What would be the "best" choice for astrophotography? eg the moon, nebulae. Thanks.
David

[Merlin66 (Ken)]

Christian Buil has done a lot of work (primarily for spectroscopy) on DSLR QE noise etc:
http://www.astrosurf.com/buil/50d/test.htm
http://www.astrosurf.com/buil/isis/noise/result.htm
Worth reading.....

[rmuhlack (Richard)]

forgive me if i'm wrong, but doesn't one need to factor in the gain and bit depth as well?

For example, from Roger Clarks page a 40D has a full well of 43400 and a minimum read noise of 4.3. The 40D is a 14-bit camera, which gives a max DN of 16384.

However that figure of 4.3 is achieved at ISO1600, where the gain (electrons per DN) is 0.21. The electrons required to clip the AD converter is (i believe) given by maxDN*gain = 16384*0.21 = 3441.

If my analysis is correct, under these conditions it wouldn't matter if you continued to expose the sensor, once 3441 electrons have been recorded the AD is fully saturated.

Similarly, a 7D has a full well of 24800 (see here), and a read noise at ISO1600 of 3.0, however the gain at ISO1600 is 0.15. from my calcs this means that the electrons required to clip the sensor = 16384*0.15 = 2458.

So sure, the read noise is very low but due to the gain used the electrons required to clip (and therefore the dynamic range) is much lower than the full well depth might suggest.

[rmuhlack (Richard)]

have just seen that the sensorgen link Greg linked to in the OP has similar data when you actually click through to a particular camera (they refer to it as "saturation")

[gregbradley]

Also of interest Sony has developed a back thinned stacked CMOS chip that has 100% pixel getting light as the circuitry that normally surrounds each pixel is stacked underneath the thinned pixel layer.

Its Exmor R and I know of a 13mp variety. I believe they are the hot new sensor in upcoming smartphones.
That sensor would have to be performing on a higher level than the current crop of True Sense Imaging chips.

Every year there are no new developments in CCD sees CMOS more and more likely to be the sensor of choice in future Astro cams.

Greg.