Pages: [1]   Go Down

Author Topic: ISO metadata and ETTR  (Read 4204 times)

lowep

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 568
    • http://sites.google.com/site/peterlowefoto/
ISO metadata and ETTR
« on: December 05, 2011, 02:13:51 am »

Guillermo writes here:

"ETTR based on the histogram in a camera where ISO is pure metadata, is a guarantee to be underexposing by at least as many stops as the difference between camera's base ISO and the ISO set by the user. In other words, a ruin."

I am no expert but attempts to ETTR with my MFDB leads me to see merit in this conclusion. Do others have similar experiences?

Is it only in a few older MFDB's and other digital cameras that the ISO is pure metadata?
« Last Edit: December 05, 2011, 02:17:27 am by lowep »
Logged

ondebanks

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 858
Re: ISO metadata and ETTR
« Reply #1 on: December 05, 2011, 09:10:35 am »

Pure metadata ISO the only sensible way to implement different ISO settings in a camera which cannot employ analog gain to reduce overall readout noise. As far as I can tell, all high-end cameras/backs with CCD sensors fall into this category.

I have two older CCD-based units (Kodak DCS645M MFDB, and Kodak DCS720x DSLR) where the readout noise I measured matches the "floor" in the sensor datasheets, and so nothing is to be gained in S/N terms from higher ISO settings. I have confirmed that they use ISO as a metadata flag only. But this is not a bad thing. I use higher ISOs on these cameras when I need faster shutter speeds and/or smaller apertures, in the full knowledge that I am simply underexposing, as if using negative exposure compensation; but on the plus side, increasing ISO maintains the brightness of what I see on the LCD (whereas negative exposure compensation does not), and I still know that I am retaining all of the DR that would be available at base ISO, and in fact gaining as many more stops of DR at the highlight end as the number of stops that I am underexposing by and losing at the shadow end.

So yes, I expect that pure metadata ISO is the case with the majority of MFDBs. It really should be the case, because if there are some which are instead digitally multiplying the signal before writing it to RAW, they are just throwing away those extra highlight levels without any payoff improvement in S/N.

Ray
Logged

Anders_HK

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1010
    • andersloof.com
Re: ISO metadata and ETTR
« Reply #2 on: December 05, 2011, 08:38:31 pm »

Pure metadata ISO the only sensible way to implement different ISO settings in a camera which cannot employ analog gain to reduce overall readout noise. As far as I can tell, all high-end cameras/backs with CCD sensors fall into this category.

Is this not a matter of providing algorithms within camera/back versus in the raw converter? Thus that for optimum results we wish to control it per it being applied in the raw converter and not have "Canikon" determine for us the end results by providing algorithms in our camera or back?

Why then change ISO as compared to not ETTR? Because the algorithms can be assumed to be based in part on the ISO we use?

Best regards,
Anders
Logged

Wayne Fox

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4237
    • waynefox.com
Re: ISO metadata and ETTR
« Reply #3 on: December 06, 2011, 01:19:26 am »


Why then change ISO as compared to not ETTR? Because the algorithms can be assumed to be based in part on the ISO we use?

Best regards,
Anders
I saw this idea earlier and didn't really believe it so I took two identical shots on my IQ180, changing only the ISO from 200 down to 35.  While the ISO 35 looked dramatically underexposed, and the ISO 200 looked like a good ETTR exposure, adjustments in Capture One yielded virtually two identical images, including noise and detail.  I'd post the results but not sure where I stored them.

So the main advantage of increasing ISO seems to be a workflow one ... the starting point and work required for the ISO 200 file was far easier than the underexposed ISO 35 file.  Those that stay at ISO 35 no matter what to lower noise really aren't doing so - the only way you can lower noise is let more light into the sensor.  If you are as slow and wide as you want to go, increasing the ISO might make post processing easier, but end results will be pretty much the same.
Logged

lowep

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 568
    • http://sites.google.com/site/peterlowefoto/
Re: ISO metadata and ETTR
« Reply #4 on: December 06, 2011, 02:22:59 am »

So if I understand this right ETTR feeds more light to the sensor but increasing ISO does not?
Logged

Anders_HK

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1010
    • andersloof.com
Re: ISO metadata and ETTR
« Reply #5 on: December 06, 2011, 08:30:19 am »

Or...

The purpose of ETTR is to optimize image information captured, alias at base ISO.

If at another ISO ETTR is not that important... but an exposure that capture within the high and low values of the scene is. Low ISO remain important for optimized image information. Thus averaged exposure should work fine.
Logged

ejmartin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: ISO metadata and ETTR
« Reply #6 on: December 06, 2011, 10:42:45 am »

So if I understand this right ETTR feeds more light to the sensor but increasing ISO does not?

Correct.

The major reason to adjust ISO (apart from the desire to chimp) is if the amount of noise added to the image by the camera electronics varies with ISO.  For cameras with CCD sensors, which includes any MFDB, this added noise is approximately constant across ISO in photometric terms, so there is little reason photographically to adjust the ISO in-camera provided your raw converter accurately applies exposure compensation. 

Much of the apparent noisiness of high ISO images is not a function of the ISO setting in the camera so much as it is due to a low photometric exposure which necessitates a large amplification of the low light signal; whether that amplification is done in-camera or in the converter does not change the result much in a camera such as a MFDB (apart from the fact that the use of a lower ISO in the camera leaves more headroom for highlight detail).
Logged
emil

lowep

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 568
    • http://sites.google.com/site/peterlowefoto/
Re: ISO metadata and ETTR
« Reply #7 on: December 06, 2011, 11:19:49 am »

Thanks for clearing this up  :)

Regarding amplification of the low light signal, do some cameras such as Leica M9 or 5DMKII do this better than others or is it just that some camera companies do better marketing about low light performance than others
Logged

Graham Welland

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 722
Re: ISO metadata and ETTR
« Reply #8 on: December 06, 2011, 11:20:30 am »

I saw this idea earlier and didn't really believe it so I took two identical shots on my IQ180, changing only the ISO from 200 down to 35.  While the ISO 35 looked dramatically underexposed, and the ISO 200 looked like a good ETTR exposure, adjustments in Capture One yielded virtually two identical images, including noise and detail.

Wayne,

What level of processing normalized the images? Just raising the exposure slider or did you also need to add highlight / shadow recovery or levels / curve adjustments too? (I assume all in C1Pro?). I assume of course that the non-ETTR image had no shadow clip on the in camera histogram?

It's interesting but I guess not too surprising to see that the same principles apply even with MFDBs as we've seen with smaller sensor cameras where for low light exposures it has typically been preferable to seemingly under expose an image at base ISO and correct in post vs bumping the ISO in camera and resulting in a noisier final image due to in-camera processing even of the raw, although I guess in this case the IQ is producing a practically identical RAW image to start with.
« Last Edit: December 06, 2011, 11:23:53 am by Graham Welland »
Logged
Graham

ejmartin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: ISO metadata and ETTR
« Reply #9 on: December 06, 2011, 11:43:55 am »

Thanks for clearing this up  :)

Regarding amplification of the low light signal, do some cameras such as Leica M9 or 5DMKII do this better than others or is it just that some camera companies do better marketing about low light performance than others

There are several aspects to low light performance:

1) Sensor size -- how much light collection area is there. 
2) Quantum efficiency -- what percentage of the collected light does the sensor record in the raw data.
3) Read noise -- how much noise does the camera electronics add in recording the signal.

Cameras such as the 5D2 and D3s have a big advantage in (3) having a read noise of around 2.5-3 electrons per pixel; the D3s also has high quantum efficiency (57%, as opposed to around 25-35% for most other DSLR's and MFDB's).  MFDB's have the sensor size advantage, of course, but their CCD sensors' high ISO read noise of around 15 electrons is a big disadvantage relative to the best CMOS sensor'd DSLRs which are about 8 times better, at around 2 electrons; quantum efficiency is also a factor of two less than the D3s, which is the king of high ISO.
Logged
emil

lowep

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 568
    • http://sites.google.com/site/peterlowefoto/
Re: ISO metadata and ETTR
« Reply #10 on: December 06, 2011, 01:16:51 pm »

 this is much more useful than the usual "are top end DSLRs now overtaking older MFDBs" claptrap  :)
Logged

Wayne Fox

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4237
    • waynefox.com
Re: ISO metadata and ETTR
« Reply #11 on: December 06, 2011, 06:46:16 pm »

So if I understand this right ETTR feeds more light to the sensor but increasing ISO does not?
I would perhaps word this a little differently ...

To achieve effective ETTR, you need to feed more light to the sensor by shutter/aperture adjustments, but increasing ISO does to achieve ETTR does not help.
Logged

Wayne Fox

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4237
    • waynefox.com
Re: ISO metadata and ETTR
« Reply #12 on: December 06, 2011, 06:56:36 pm »

What level of processing normalized the images? Just raising the exposure slider or did you also need to add highlight / shadow recovery or levels / curve adjustments too? (I assume all in C1Pro?). I assume of course that the non-ETTR image had no shadow clip on the in camera histogram?

The ISO 200 image had no shadow clip visible on the camera histogram, the ISO 35 had quite a bit.  However, the data really wasn't clipped, both final images had the same level of detail and noise in the shadow.

Just raising the exposure slider wasn't enough.  It took some tweaking with various sliders to match, don't recall exactly what I did.  It took a while. However, this might be a very repeatable process, so a 1 stop bump may be something that can be done with a preset, etc.

Since this test, my current IQ180 capture technique is to go ahead and bump the ISO to get a normalized looking histogram (but not ETTR) if I don't want to let more light hit the sensor.  It doesn't degrade the image quality any worse than the underexposure would and it makes the workflow much faster.  Obviously my ideal is an ETTR histogram at base ISO of 35, but as we all know sometimes depth of field and subject motion just won't let you get there.
Logged

Schewe

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6229
    • http:www.schewephoto.com
Re: ISO metadata and ETTR
« Reply #13 on: December 06, 2011, 07:33:47 pm »

Since this test, my current IQ180 capture technique is to go ahead and bump the ISO to get a normalized looking histogram (but not ETTR) if I don't want to let more light hit the sensor. 

Wayne,

Just to better understand, did you measure dynamic range of the scene? In the most recent PODUS, we discussed the impact of DR regarding ISO setting. Originally, Phase One didn't test the 35 ISO until it was determined that there was more DR potential by using 35 as the base ISO.
Logged

dchew

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1020
    • Dave Chew Photography
Re: ISO metadata and ETTR
« Reply #14 on: December 06, 2011, 08:03:31 pm »

Jeff,
I thought the same question, because isn't that getting at the point Ray made initially in the statement below?
So yes, I expect that pure metadata ISO is the case with the majority of MFDBs. It really should be the case, because if there are some which are instead digitally multiplying the signal before writing it to RAW, they are just throwing away those extra highlight levels without any payoff improvement in S/N.

Ray
I spin myself in circles with this stuff.  But the context of this discussion is essentially about opening up underexposure, so if you started with a high DR scene then shooting at base ISO would presumably be better...?

Dave
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: ISO metadata and ETTR
« Reply #15 on: December 06, 2011, 09:34:58 pm »

The ISO 200 image had no shadow clip visible on the camera histogram, the ISO 35 had quite a bit.  However, the data really wasn't clipped, both final images had the same level of detail and noise in the shadow.

Just raising the exposure slider wasn't enough.  It took some tweaking with various sliders to match, don't recall exactly what I did.  It took a while. However, this might be a very repeatable process, so a 1 stop bump may be something that can be done with a preset, etc.

Since this test, my current IQ180 capture technique is to go ahead and bump the ISO to get a normalized looking histogram (but not ETTR) if I don't want to let more light hit the sensor.  It doesn't degrade the image quality any worse than the underexposure would and it makes the workflow much faster.  Obviously my ideal is an ETTR histogram at base ISO of 35, but as we all know sometimes depth of field and subject motion just won't let you get there.

Wayne,

Thanks for posting your results--it is nice to know that practice conforms to theory. Few us have the luxury to be shooting with an IQ180, but the same principle applies to the newer dSRLs such as the Pentax K5. Users of less efficient cameras can use a similar technique by raising the ISO only to the point that read noise is minimized (shown by Sensorgen for many cameras) and then bump the exposure in the raw converter.

To obtain a better view on the LCD and to facilitate PP, it makes sense to increase the ISO on the camera somewhat, but still leave ample headroom for highlights in order to avoid clipping. I presume that the view was quite dark with your "underexposed" shot.
Regards,

Bill
Logged

Wayne Fox

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4237
    • waynefox.com
Re: ISO metadata and ETTR
« Reply #16 on: December 08, 2011, 02:06:36 am »

Wayne,

Just to better understand, did you measure dynamic range of the scene? In the most recent PODUS, we discussed the impact of DR regarding ISO setting. Originally, Phase One didn't test the 35 ISO until it was determined that there was more DR potential by using 35 as the base ISO.
I didn't measure the dynamic range but in this case I'm guessing it was only 6 or 7 stops at the max.  Late soft light, shutter speed was already pretty slow.  I set it 200 so I can get a little more speed ... breeze blowing the leaves a little bit, but in this case because of the previous discussion about this, I put the ISO back to 35 and took a couple of shots -I was just curious. 

I'm gathering from all of this is the concept that the sensors used in backs like the IQ180 can't amplify the analog signal at the chip in the same way that some CMOS sensors do, so the ISO setting on the back is basically just metadata used by the raw processor - so the "amplification" if you will is done in post processing and not electronically.  I could believe that going to a lower number removed some noise which basically would increase the dynamic range so the logic there makes sense to me  I assume you want to get to some "native" ISO which is optimum.  As to whether increasing ISO lessens the dynamic range, I'm not clear.  I do know my "clipped" shadows were completely fine and identical to the image taken at ISO 200, but I didn't have anything approaching the highlight clip point in either capture.

Seems if the dynamic range is strong enough you would fight highlight clipping you might not have a scene that would benefit from a higher ISO anyway? (not sure if that makes sense, just kinda thinking "out loud" if you will ...)

I'd love for Dr. Claus at Phase to offer some thoughts on this ...

Logged

madmanchan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2115
    • Web
Re: ISO metadata and ETTR
« Reply #17 on: December 08, 2011, 03:19:26 am »

Yes, DR is really about how noisy your shadows get, or how much highlights you clip.  With good sensors with low noise floor, you can "afford" to leave a lot more highlight headroom.  (This is one reason why DSLRs meter a lot more conservatively now than they used to.)
Logged
Eric Chan

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2005
    • http://www.guillermoluijk.com
Re: ISO metadata and ETTR
« Reply #18 on: December 08, 2011, 07:02:57 am »

For some the following may sound obvious, but I am pretty sure many users consider DR as an advanced parameter only useful for certain pplications. But the usefulness of a high dynamic range sensor, i.e. a sensor that allows a wide range of RAW exposure values still keeping a good SNR (visible noise), goes beyond facing high dynamic range scenes. Other contexts are possible:

1. To take pictures of high dynamic range scenes: this is the obvious. Where other cameras could need to shoot more than once, or sacrifice highlight/shadow detail, a high DR sensor can capture an entire scene in a single shot.
DR -> enhanced captured DR, easier photographic procedure.

2. Make exposure mistakes and still have a good quality image: in a high DR sensor the user can make strong underexposure mistakes and still be able to obtain a good image by lifting the shadows, while other cameras would ruin the image because of noise. This is not only of interest to novice users, action photography is prone to produce exposure errors.
DR -> compensate errors made by the photographer.

3. Prevent motion blur/lack of DOF: if we are in the limit to achieve enough shutter speed and/or DOF, a high DR sensor allows to expose less by reducing exposure time and/or aperture, with a higher guarantee to avoid trepidation or insufficient DOF thanks to being able to underexpose the RAW file and still recover a high quality image.
DR -> ensuring sharp images and extended DOF

4. Finally if we can rely on our sensor's ability to lift the shadows in conventional scenes, we could even afford the contrary approach to ETTR, i.e. to deliberately underexpose by default all our shots so that we can forget about optimum exposure. The high DR sensor will guarantee a good image while we can focus on other photographic aspects such as framing, composition, focus,...
DR -> focus in creative and technical aspects such as composition, perfect focus,...
« Last Edit: December 08, 2011, 07:10:52 am by Guillermo Luijk »
Logged

lowep

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 568
    • http://sites.google.com/site/peterlowefoto/
Re: ISO metadata and ETTR
« Reply #19 on: December 08, 2011, 10:36:21 am »

There are several aspects to low light performance:
1) Sensor size -- how much light collection area is there.
2) Quantum efficiency -- what percentage of the collected light does the sensor record in the raw data.
3) Read noise -- how much noise does the camera electronics add in recording the signal.


Apart from the physical size of a sensor and the number of MPs packed on it, are some MFDBs better than others at handling low light in terms of quantum efficiency, read noise or other factors?

Logged
Pages: [1]   Go Up