Hi Ray,
The only measurements I have seen were on DxO-mark. When the 645D arrived I got the impression that it handled read noise really well. My understanding is that DR that DxO-mark measures is dominated by read noise. It is also my understanding that with CCDs analog readout is not integrated on the chip but is on a separate PCB and I got the impression that the Pentax solution is really good. Since than DxO mark republished data on the P40+ and it seems be even better than the Pentax 645D.
I have played around with raw files from Pentax 645D (from Imaging Resource and Miles Hecker) and Leica S2 (with kind permission of Lloyd Chambers) I was not impressed by shadow noise on the S2.
The Leica S2 stuff is here: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/38-observations-on-leica-s2-raw-images
The Pentax article is not about DR, but it's here: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/51-a-closer-look-at-pentax-645d-image-quality
Your idea on MFD read out noise data base is very interesting. My impression is that MFDBs are surrounded with a lot of mythos, but I guess that only benefit of MFDBs over DSLR sensors is larger size. A larger sensor collects more photons, holds more electron charges and so on. A larger sensor may also have advantage in resolution and allow higher MTF on a given structure size. So no doubt that there are many benefits to MF digital.
Do you have any ideas about sample variation?
Best regards
Erik
Hi Eric,
I'm glad you can see the merit of my MFD read out noise database idea. The first results are coming in, and I will share them in due course. Interesting things are also emerging about how some manufacturers are treating their raw data!
I strongly encourage other MFD shooters to contribute as well. I omitted one instruction in my original post up this thread: without this I can get the noise in ADU, but that's useless unless I can also get the conversion factor to electrons. So the full instructions are:
- Take
two frames at each ISO setting, with the lenscap ON, at the shortest shutter speed on your camera
- Take
two frames at each ISO setting, with the lenscap OFF: camera on tripod, aimed at a white wall/ceiling, telephoto (or failing that, normal) lens set to f/16 and INFINITY focus, exposure set to give a bright but unsaturated/unclipped image (metered exposure + 1 stop compensation should be about right; check the LCD histogram)
- Put the RAW files on somewhere like Dropbox, where I can download them.
This assumes that the black point is not being clipped by the Manufacturer's firmware. If my initial work reveals that it is being clipped in some cases, I'll have to ask those contributors for some more data (basically the white wall shots again, but this time at a large range of exposure times). I am not asking for that at this stage, as I don't want to put busy people (including myself!) through extra work unless it's really needed.
You raise a good point about sample variation. Here's where strong community involvement can help. If more than one person sends me samples from the same model of camera, we can start to assess variations like this.
Look at it this way - every contributor will get a sort of custom analysis report on their sensor. For free. I'm sure that manufacturers and repair houses would charge a lot for that sort of thing.
Ray