5D and ISO 50

[dwdallam]

Are there any problems using ISO 50 on the 5D of which I should be aware?

[John Sheehy]

Are there any problems using ISO 50 on the 5D of which I should be aware?
[a href=\"index.php?act=findpost&pid=70002\"][{POST_SNAPBACK}][/a]

I don't know the specifics of the 5D, but generally speaking, there are three ways that ISO 50 may be obtained:

1) Through lower amplification.  You may get the same highlight headroom as the other ISOs, but you may not, due to sensor saturation.

2) By a mathematical pull in the camera (the most useless method; always loses headroom and posterizes data compared to) ...

3) By a mathematical pull in the RAW converter.  This simply is the same thing as exposing a stop higher at ISO 100, but FEC and EC work as stated for the exposure index.

If you can link to RAW files of a grey step scale or color-checker at +2 EC in both ISO 100 and 50, I can tell you exactly what it is that the 5D does.

[r42ogn]

Only that IMHO it produces worse results than ISO 100, I use ND filters and ISO100 if I have to reduce sensitivity.

Are there any problems using ISO 50 on the 5D of which I should be aware?
[a href=\"index.php?act=findpost&pid=70002\"][{POST_SNAPBACK}][/a]

[John Sheehy]

Only that IMHO it produces worse results than ISO 100, I use ND filters and ISO100 if I have to reduce sensitivity.
[a href=\"index.php?act=findpost&pid=70032\"][{POST_SNAPBACK}][/a]

Worse in what way?  Do the images clip more readily?

[bjanes]

Are there any problems using ISO 50 on the 5D of which I should be aware?
[{POST_SNAPBACK}][/a]

Here is an analysis by Roger Clark for the EOS 1D Mark II. It probably applies to the 5D as well, since they use similar technology. If you want to explore the matter further, Roger Describes his methodology.

[a href=\"http://www.clarkvision.com/imagedetail/evaluation-1d2/index.html]http://www.clarkvision.com/imagedetail/eva...-1d2/index.html[/url]

"The data for the 1D mark II indicate some important aspects for people making images. The lowest ISO, 50, does have higher signal-to-noise ratios than ISO 100, but at the expense of lower dynamic range and danger of saturating highlights. An ISO of about 75 would be a better match to the sensor range and would not sacrifice dynamic range. Unfortunately, ISO 75 is not available."

[BJL]

Here is an analysis by Roger Clark for the EOS 1D Mark II.
"... The lowest ISO, 50, does have higher signal-to-noise ratios than ISO 100, but at the expense of lower dynamic range and danger of saturating highlights.
[{POST_SNAPBACK}][/a]

That fits with the DR tests at DPReview:
[a href=\"http://www.dpreview.com/reviews/canoneos5d/page22.asp]http://www.dpreview.com/reviews/canoneos5d/page22.asp[/url]
ISO 50 loses about a full stop of highlight headroom, according to those tests, suggesting that the "base ISO speed" of that sensor is close to the typical ISO 100.

[Still, it could be useful for scenes with not much highlight range above the mid-tones that dominate the meter reading, and deep shadows, as an alternative to ISO 100 with +1 compensation and then -1 in post-processing.]

[John Sheehy]

[Still, it could be useful for scenes with not much highlight range above the mid-tones that dominate the meter reading, and deep shadows, as an alternative to ISO 100 with +1 compensation and then -1 in post-processing.]
[a href=\"index.php?act=findpost&pid=70089\"][{POST_SNAPBACK}][/a]

If it's done by halving the RAW values of an ISO 100 readout, after metering with an unstated +1 bias, then it is actually inferior, however slightly, to exposing +1 higher at ISO 100  yourself, as everything will be quantized from 12 bits to 11.

Yes, Canon does things like this.  The 1/3-stop ISOs on the 30D, for example, are just mathematical pushes and pulls from the 100-200-400-800-1600 series from the 20D.  Half of the extra ISOs lose 1/3 stop of dynamic range; half lose 1/3 stop of highlights but keep their dynamic range (except for some slight posterization to 3200 as levels).

They double the RAW values at ISO 3200 on many cameras, instead of leaving them as 1600 RAW data, losing a stop of extra highlights.

[BJL]

If it's done by halving the RAW values of an ISO 100 readout, after metering with an unstated +1 bias, then it is actually inferior
[a href=\"index.php?act=findpost&pid=70127\"][{POST_SNAPBACK}][/a]

True: the only hope is that it actually lowers the pre-amplification. There are hints that the 5D does does ISO 50 the good way, in comments that shadow noise levels are somewhat lower than with ISO 100.

[Ray]

Never had much desire to use an ISO with reduced performance. I used it once in Thailand when I borrowed the tripod of a complete stranger to photograph a waterfall. I wanted to try the silky smooth effect but didn't have an ND filter, so ISO 50 was the next best thing.

Now, if my data base was in order, I could find that image and show it to y'all.  

[bjanes]

I don't know the specifics of the 5D, but generally speaking, there are three ways that ISO 50 may be obtained:

1) Through lower amplification.  You may get the same highlight headroom as the other ISOs, but you may not, due to sensor saturation.

2) By a mathematical pull in the camera (the most useless method; always loses headroom and posterizes data compared to) ...

3) By a mathematical pull in the RAW converter.  This simply is the same thing as exposing a stop higher at ISO 100, but FEC and EC work as stated for the exposure index.

If you can link to RAW files of a grey step scale or color-checker at +2 EC in both ISO 100 and 50, I can tell you exactly what it is that the 5D does.
[a href=\"index.php?act=findpost&pid=70009\"][{POST_SNAPBACK}][/a]

John,

If you look at Clark's data, I think most of what you need is there for the 1Ds2. The sensor saturates at arount 80,000 electrons, and maximum signal for ISO 50 is 79,900 electrons. At ISO 100, the maximum signal is 53,000 electrons. The gain at ISO 50 is  26.03 electrons/data number (ADU number) and it is 13.02 electrons/data number at ISO 100. The amplification is the same. Doing the math indicates that the maximum data numbers are 3069 for ISO 50 and 4071 for ISO 100.

To get a maximum data number of 255 in the converted 8 bit space, a pull in the raw converter or in camera jpeg engine would be necessary. If an ISO of 75 were available, the maximum signal would be 79,900 electrons and the data number would be near 4095 and no pull would be needed.

Of course, the 5D may behave differently, but the pixel size is the same, 82. microns.

Your comments would be appreciated.

[John Sheehy]

Never had much desire to use an ISO with reduced performance. I used it once in Thailand when I borrowed the tripod of a complete stranger to photograph a waterfall. I wanted to try the silky smooth effect but didn't have an ND filter, so ISO 50 was the next best thing.

Now, if my data base was in order, I could find that image and show it to y'all. 
[a href=\"index.php?act=findpost&pid=70143\"][{POST_SNAPBACK}][/a]

I guess the "ISO expansion" should serve as a warning, but the reduced DR should be stated in the literature.

[dwdallam]

I take all of this to mean don't use ISO 50, use an ND filter, which I have. Thanks.

[Ray]

I take all of this to mean don't use ISO 50, use an ND filter, which I have. Thanks.
[a href=\"index.php?act=findpost&pid=70204\"][{POST_SNAPBACK}][/a]

dwdallam,
Why don't you compare ISO 100, with ND filter of one stop, with ISO 50 without ND filter? Show us the results and you will have contributed to the sum total of human knowledge.

[dwdallam]

dwdallam,
Why don't you compare ISO 100, with ND filter of one stop, with ISO 50 without ND filter? Show us the results and you will have contributed to the sum total of human knowledge.
[a href=\"index.php?act=findpost&pid=70217\"][{POST_SNAPBACK}][/a]

Haha, yeah. Well, given the technical information here, I don't want to take the time. First, it is clear that using ISO 50 is not going to give you a denser picture, like film ISO 50 would compared to film ISO 100. Second, the files are noisier than ISO 100. Third, in the face of one and two, the only reason to use ISO 50 would be to reduce light, which you can accomplish just fine with an ND filter (+polarizer if needed.) Those are all good enough reasons for me.

[BJL]

the files are noisier than ISO 100.
[a href=\"index.php?act=findpost&pid=70224\"][{POST_SNAPBACK}][/a]

Where did you get that? The above quotes from www.clarkvision.com say "The lowest ISO, 50, does have higher signal-to-noise ratios than ISO 100 ..." which indicates less visible noise. And the DPReview DR tests also indicate a hair more DR below midtones: even though I am skeptical of the DPReview methodology (based mostly on standard out-of-the-camera JPEGs) this goes against the claim of more noise at ISO 50.

[r42ogn]

I haven't done scientific tests but the the ISO 50 pictures I took seemed noisier and less  snappy than the ISO 100+ ND.  I tried to load a couple up but frankly I couldn't see the difference on the web that I can see on screen.  It's pretty small.

Worse in what way?  Do the images clip more readily?
[a href=\"index.php?act=findpost&pid=70062\"][{POST_SNAPBACK}][/a]

[John Sheehy]

Haha, yeah. Well, given the technical information here, I don't want to take the time. First, it is clear that using ISO 50 is not going to give you a denser picture, like film ISO 50 would compared to film ISO 100. Second, the files are noisier than ISO 100.[a href=\"index.php?act=findpost&pid=70224\"][{POST_SNAPBACK}][/a]

With the same absolute exposure, or the same EC?

Judging from the trends of shadow noise at various ISOs, I would not expect the camera's ISO 50 to be of much ultimate value.  There seems to be a basic flat rate of readout noise at all ISOs in Canon DSLRs, which really keeps the cameras from achieving the sensor's potential DR at low ISOs.  The fact that ISO 50 probably has at least a half stop less headroom than the other ISOs means you can't expose to the right as much, meaning that maximum S/N is going to be reduced at ISO 50, putting it somewhere around ISO 300-400 levels.  If you're going to expose conservatively, where RAW headroom is not an issue (or shoot JPEG), then ISO should work as advertised, but really doesn't have any significant S/N benefits over ISO 100.

[John Sheehy]

Where did you get that? The above quotes from www.clarkvision.com say "The lowest ISO, 50, does have higher signal-to-noise ratios than ISO 100 ..." which indicates less visible noise. And the DPReview DR tests also indicate a hair more DR below midtones: even though I am skeptical of the DPReview methodology (based mostly on standard out-of-the-camera JPEGs) this goes against the claim of more noise at ISO 50.
[a href=\"index.php?act=findpost&pid=70229\"][{POST_SNAPBACK}][/a]

There are two aspects to this.  There is S/N at any given relative exposure, but if there is less RAW headroom at ISO 50, then you can't expose as far to the right, and therefore you maximum S/N will be lower.  Exposure, stop for stop, has far more effect on noise than ISO does.  There is very little difference in deep shadow noise levels at the low end of the ISO spectrum(100-400), generally, and highlight noise is proportional to absolute exposure.

[bjanes]

Where did you get that? The above quotes from www.clarkvision.com say "The lowest ISO, 50, does have higher signal-to-noise ratios than ISO 100 ..." which indicates less visible noise. And the DPReview DR tests also indicate a hair more DR below midtones: even though I am skeptical of the DPReview methodology (based mostly on standard out-of-the-camera JPEGs) this goes against the claim of more noise at ISO 50.
[a href=\"index.php?act=findpost&pid=70229\"][{POST_SNAPBACK}][/a]

The Clark analysis demonstrates some interesting points about dynamic range and noise. Highlight noise is determined primarily by the number of electrons collected and is equal to the square root of that number. Absolute noise is greatest in the highlights, but this noise is not perceived by the eye due to the high signal/noise ratio. Maximal signal to noise ratio is maximal signal / noise or electrons / sqrt(electrons), which simplifies to sqrt (electrons). The shadow noise does not enter into the calculation of signal to noise ratios.

In the shadows, noise is related to the number of electrons collected, as above, but also the read noise. The read noise often predominates. Read noise increases with lower ISOs, and that is why the DR is lower at ISO 50. DR is defined as (electrons)/read noise (electrons). When one goes from ISO 100 to ISO 50 with the D1M2, read noise essentially doubles, but the number of electrons collected is less than doubled, so DR is worse.

On DPreview Phil determines the dynamic range differently by comparing the ratio of highlights to shadow, but the cutoff at the shadow end is determined by what noise level is acceptable.  The total shadow noise is essentially sqrt(electrons in shadows) + read noise.

[dwdallam]

Where did you get that? The above quotes from www.clarkvision.com say "The lowest ISO, 50, does have higher signal-to-noise ratios than ISO 100 ..." which indicates less visible noise. And the DPReview DR tests also indicate a hair more DR below midtones: even though I am skeptical of the DPReview methodology (based mostly on standard out-of-the-camera JPEGs) this goes against the claim of more noise at ISO 50.
[a href=\"index.php?act=findpost&pid=70229\"][{POST_SNAPBACK}][/a]

I got it because I misread that sentence. Can't you read my mind? What I meant to say is that ISO 50 seems to give worse results, even if it gives lesss noise, than ISO 100 in many situations, such as highlights and over saturation. Thanks for pointing that out.