Hi All,
Been educating myself on photons, electrons and camera flow as it relates to exposure and I’ve come across some very interesting information. Many of you likely already know some or all of this so I’m sorry if this question seems naive
When we think of “noise” in a digital image there are two sources, one is upstream which includes the image and sensor, the second is downstream which includes the amplification due to the ISO setting.
In emulsion the ISO was part of the exposure setting. In digital the shutter speed and aperture are all there is to exposure. The ISO is simply an amplification ratio of the signal from the sensor. Now intuitively one may think this means that the lower the ISO, the less noise. But this is NOT the case in most digitals because the S/N ratio actually goes UP as ISO increases until a “break point” is reached. However, those sensors of the new Sony “Exmoor” type are different. They are linear in amplification and S/N does not increase with increasing ISO
So the take home point is that for most cameras you want to push the ISO to reduce noise but keep it below the breakpoint. But in the exmoors it is suggested that you leave the ISO at its native setting and just use the speed/aperture you want. Use post processing in LR or whatever you use to trim exposure since you gain no benefit from upping ISO settings.
So my question is; does anyone know the ISO S/N curve for the S2 (and M9 for that matter)? At what point does the ISO amplification begin to degrade the S/N ratio? Or is it like the Exmoors and it really doesn’t matter what ISO you set it to as long as it isn’t too high. Making it essentially an “ISO-less” camera
So my question is; does anyone know the ISO S/N curve for the S2 (and M9 for that matter)? At what point does the ISO amplification begin to degrade the S/N ratio? Or is it like the Exmoors and it really doesn’t matter what ISO you set it to as long as it isn’t too high. Making it essentially an “ISO-less” camera
The sensor-type (CMOS or CCD) is more relevant for the noise-type than the processor. Leica uses CCD-sensors that have other characteristics than CMOS used in 35 mm DSLR.
And even with CMOS-sensors the noise is increasing with higher ISO, but not as sharp as with a CCD.
Compared, for example, and starting wirh 100 ISO, the S2 behaves almost the same at 200, 400 iso, compared to a Nikon D3x, the spread increases rapidly from 640 ISO upwards. I would say from experience that 800 ISO with the S2 compares to 1200-1600 ISO on the Nikon D3x (and maybe 3200-6400 ISO on a D3S).
Also, resolution and post-processing is an important factor.
btw, ISO-Quality is not a question of S/N-ratio, it is more a stastistical problem of missing and eraticcal data. This is why a CCD-sensor with a lot of pixels can be as good as a CMOS-sonsor with less pixels (up to a certain level, though).
stephan;1219 wrote: The sensor-type (CMOS or CCD) is more relevant for the noise-type than the processor. Leica uses CCD-sensors that have other characteristics than CMOS used in 35 mm DSLR.
And even with CMOS-sensors the noise is increasing with higher ISO, but not as sharp as with a CCD.
Compared, for example, and starting wirh 100 ISO, the S2 behaves almost the same at 200, 400 iso, compared to a Nikon D3x, the spread increases rapidly from 640 ISO upwards. I would say from experience that 800 ISO with the S2 compares to 1200-1600 ISO on the Nikon D3x (and maybe 3200-6400 ISO on a D3S).
Also, resolution and post-processing is an important factor.
btw, ISO-Quality is not a question of S/N-ratio, it is more a stastistical problem of missing and eraticcal data. This is why a CCD-sensor with a lot of pixels can be as good as a CMOS-sonsor with less pixels (up to a certain level, though).
thanks for a reply but perhaps I could understand you better;
First Emil Martin of U of C fame has published a set of curves that plot ISO vs noise here; http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p3.html#ETTR
and here is a discussion concerning particular cameras and how read noise and ISO effect final S/N; http://www.luminous-landscape.com/forum/index.php?topic=56906.msg460646#msg460646
So for my new Sony A-77 ISO setting is not terribly relevant, Just keep it @ 160 and push the histogram. But I’m wondering about considering the curves in that second reference, what does an S2 or M9 curve look like?
Or do CCD’s produce a different family of curves? In which case, what do they look like and do we benefit more from ETTR or just follow the ISO? If so, where do we max out?
Also, when you say ISO is not about S/N but missing or erratic data, well yes the missing data is the problem but isn’t the result more noise/less picture data? Or is there more?
Sorry to ask such complex questions but I want to get the most out of the camera’s range
OK, this is what I mean;
http://sensorgen.info/LeicaM9.html
wish I could upload it to make it easy but it seems that this site wont take reasonably sized attachments… or maybe I’m screwing it up. In any case;
you can see that raising the ISO of an M9 compresses DR and does NOT decrease read noise. That it is a CCD is not the issue, read noise is. Further, as ISO increases, saturation decreases (makes sense of course)
the implication here is that the M9 is also an ISO-less camera. Increasing ISO is actually deleterious to image quality from compression. One is better off nailing ISO at 160 and shoot using the stop or speed one needs for a given shot. Or, if it doesn’t matter, just ETTR as much as you dare. The resultant underexposure, if present, can be upped in PP.
I just wonder how the S2 works… no data yet.
allegretto;1235 wrote: OK, this is what I mean;
http://sensorgen.info/LeicaM9.html
wish I could upload it to make it easy but it seems that this site wont take reasonably sized attachments… or maybe I’m screwing it up. In any case;
you can see that raising the ISO of an M9 compresses DR and does NOT decrease read noise. That it is a CCD is not the issue, read noise is. Further, as ISO increases, saturation decreases (makes sense of course)
the implication here is that the M9 is also an ISO-less camera. Increasing ISO is actually deleterious to image quality from compression. One is better off nailing ISO at 160 and shoot using the stop or speed one needs for a given shot. Or, if it doesn’t matter, just ETTR as much as you dare. The resultant underexposure, if present, can be upped in PP.
I just wonder how the S2 works… no data yet.
Not sure if this translates to real world results. An easy test would be to shoot a test target at ISO 160, underexposed 3 stops and at ISO 1250 properly exposed. Boost the ISO 160 shot 3 stops in LR and then compare. Without doing the test (yet) my guess is that the 1250 shot will look better. If I get a few free minutes today, I will do the test with an M9 and an S2.
David Farkas
Red Dot Forum
Leica Store MiamiDavid Farkas;1239 wrote: Not sure if this translates to real world results. An easy test would be to shoot a test target at ISO 160, underexposed 3 stops and at ISO 1250 properly exposed. Boost the ISO 160 shot 3 stops in LR and then compare. Without doing the test (yet) my guess is that the 1250 shot will look better. If I get a few free minutes today, I will do the test with an M9 and an S2.
was going to do that too this weekend, but you have more cred, so please do
also, try 160 and ETTR a bit too…
Oh Boy, theory meets reality
Remember; …in theory, there is no difference between reality and theory…
Alright… here’s the test:
Testing methodology:
M9 with 35 f/2 ASPH on tripod shooting into open shade. I first shot 1250 ISO at 1/125th @ f/8 (proper exposure) then changed ISO to 160 at the same shutter speed and aperture.
Imported into LR 3.5 using my standard M9 preset, synced the WB on both shots to 5400K +12 tint. Boosted exposure slider to +3 on ISO 160 shot. Exported to web sized and two 100% matching crops.
ISO 1250
Full image
100% crop – please click for full size
100% crop – please click for full size
ISO 160 – Pushed 3 Stops
Full image
100% crop – please click for full size
100% crop – please click for full sizeHere’s my take:
ISO 160 pushed three stops looks better than I expected it to. It does pick up a fair amount of contrast vs. the straight ISO 1250 shot, but this can be tweaked in LR to match if so desired. The 1250 shot definitely has an edge with regards to noise, though. So, the theory doesn’t pan out 100% in reality…. but it is pretty close. Make sure to click on the thumbnails for full sized images.
David Farkas
Red Dot Forum
Leica Store Miamithanks David, that’s what I was looking for but I’ll have to check it on my large monitor to know. they look close on my Mac Air
Now what if you just left it @ 160 and ETTR for a stop or two it might be even better…
David Farkas;1246 wrote: Alright… here’s the test:
Testing methodology:
M9 with 35 f/2 ASPH on tripod shooting into open shade. I first shot 1250 ISO at 1/125th @ f/8 (proper exposure) then changed ISO to 160 at the same shutter speed and aperture.
Imported into LR 3.5 using my standard M9 preset, synced the WB on both shots to 5400K +12 tint. Boosted exposure slider to +3 on ISO 160 shot. Exported to web sized and two 100% matching crops.
ISO 1250
Full image
100% crop – please click for full size
100% crop – please click for full size
ISO 160 – Pushed 3 Stops
Full image
100% crop – please click for full size
100% crop – please click for full sizeHere’s my take:
ISO 160 pushed three stops looks better than I expected it to. It does pick up a fair amount of contrast vs. the straight ISO 1250 shot, but this can be tweaked in LR to match if so desired. The 1250 shot definitely has an edge with regards to noise, though. So, the theory doesn’t pan out 100% in reality…. but it is pretty close. Make sure to click on the thumbnails for full sized images.
OK, had a chance to view the images on side-by-side color matched monitors
– yes, more noise in the 160 push as processed, but not too evident in full-frame. The pixel-peep makes it more obvious. In a typical 8 x 10 (I know, violation of the 2:3 Golden Rule) would be curious to see if they could be told apart by even experienced eyes.
– color however is more nuanced in the 160. Both images have that unmistakable pop of the reds and greens as well as the rich depth and dimension that we Leica fans love. No other camera I’ve used can produce the results of the M9 or S2 in this field. But there is no doubt that the subtle reds of the leaves (not the flowers) are better in the 160-push (at least to my eyes)
Very revealing, no? Don’t know about you, but this makes me want to explore these options more. Consider the data on DR in the graphs. I wonder how a scene that puts more of a premium on DR would be rendered differently with the two techniques.
If you have the occasion to repeat this with the S2 (every time I go to LR, the S2 shots just jump off the screen with so much depth and life) why don’t you just post them without telling us which is which? Take a poll and see what folks think. Probably most here would pick it right, but would be interesting.
Also, three stops is a lot of “apparent” underexposure. But if left at 160 and ETTR-ed, it might be formidable. Further, this may be a way to stretch the DR for those tricky shots where you don’t want to clip but want to preserve shadow detail.
Anyway, thanks a lot for following through on this… good stuff!
Remember that “ISO” is different in the digital world than with film. The sensitivity (“ISO”) of a digital sensor is what it is and cannot be changed; changing a digital camera’s ISO setting changes the readout strategy or “the way the lab processes the roll of film”.
In both tests in David’s experiment the sensor had the same state when the shutter closed since the exposure settings were the same but the readout strategy differed since the ISO setting was changed.
David’s result is as I would expect: amplifying the signal in LR means that the readout noise is amplified by the same amount as the signal but changing the camera ISO setting means the camera can amplify the signal with little or no new readout noise added.
jrv;1281 wrote: Remember that “ISO” is different in the digital world than with film. The sensitivity (“ISO”) of a digital sensor is what it is and cannot be changed; changing a digital camera’s ISO setting changes the readout strategy or “the way the lab processes the roll of film”.
In both tests in David’s experiment the sensor had the same state when the shutter closed since the exposure settings were the same but the readout strategy differed since the ISO setting was changed.
David’s result is as I would expect: amplifying the signal in LR means that the readout noise is amplified by the same amount as the signal but changing the camera ISO setting means the camera can amplify the signal with little or no new readout noise added.
well, not exactly
totally agree, the shutter/f-stop were the same, so yes, the exposure is the same. However ISO does not mean the same thing in each camera (digital ones that is). As the references posted, some cameras actually have lower read noise with higher ISO, but some do not (as in the case of the M9 and the new Sony Exmor chip). In the digital world ISO is an amplification factor, and so-called “read-noise” can be expressed differently by different algorithms.
At least that’s what has been posted by folks who are paid to figure these things out.
Many factors feed into final image noise though and ISO is but one. Personally I like the fact that I can pretty much ignore the ISO-boosting on the M9 and just ETTR as much as I can and still get great shots. Uncomplicates my simple thoughts while shooting. If I had a new Nikon D3x or Canon 1D would have to fuss with that too…
You must be logged in to reply to this topic.