Combing Artifacts at intermediate ISO's. Does it effect the 1D Series Cameras

Combing Artifacts at intermediate ISO's. Does it effect the 1D MarkIII?

The 30D and the 5D, at least, seem to have values in between integers of ISO, made by expansion of a neighboring existing calculation rather than am independant path.

Peter Ruevski said:

Hi Rechard,

So I presume when you said "all ISO settings" you included intermedate 1/3 stop ones like ISO 125? If so and you did not see any combing then the 5D would certainly seem to have "proper" i.e. alanog amplifier driven intermediate ISO settings.

...unlike the 30D. Here is what an ISO 125 histogram looks like from the 30D:

The histogram looks "solid" because of the heavy combing. Here is a zoomed in section:

Click to expand...

The result is a comb appearance in the histogram for the missing values. See here !

So does this apply to the 1D flagshiop series and especially to the 1DIII?

Asher

Asher Kelman said:

The 30D and the 5D, at least, seem to have values in between integers of ISO, made by expansion of a neighboring existing calculation rather than am independant path.

The result is a comb appearance in the histogram for the missing values.

So does this apply to the 1D flagshiop series and especially to the 1DIII?

Click to expand...

I can't say that I remember ever looking at any 1D series "intermediate ISO" RAWs, but I have examined 5D ones, and the 5D appears to use a secondary amplifier before the ADC, to push ISO 100 to achieve 125 and 160 (160 is not pulled from 200 as in the 30D). The only gaps in 5D histograms I've seen are one out of every 16 values is not used in some RAW files (the data is multiplied by 16/15). For normal use, I doubt the gaps will be visible, but if you were to stack many low ISO images for critical work, it could result in visible posterization. You might see it too in an extremely slow smooth gradient in the shadows at low ISOs.

Since I am the "culprit" for the original thread I thought I should reply. As John mentioned above, there are two distinct situations when combing occurs:

- The first is with the intermediate ISO settings of the EOS 30D, which are achieved with (mathematical) scaling of the raw data. People owning the 5D have tried its intermediate ISO settings and it seems to have proper amplifiers - i.e. no combing. Like John I have not seen raw files from 1D series cameras shot at 1/3 stop ISOs (send me some ;-)

- The second (interesting) type of combing seems to occur when using wide aperture lenses, set at wide apertures. This one does not always occur, seems to never occur above F3.5 and is much less dense. However I have seen it with significantly varying density (i.e. period, i.e. the "scaling" is different every time). I have personally observed this one on 30D and 5D photos (you can see the original page for some examples http://www.pages.drexel.edu/~par24/rawhistogram/rawhistogram.html). In addition people have reported seeing it on 20D and 350D. This combing seems to be due to some correction that the camera applies to the raw photos from time to time. I can very easily and consistently reproduce it with the EF50mm f/1.8 on the 30D at f/1.8. I have done controlled tests where one photo was combed and just by unlocking the lens and slightly turning it (i.e. contacts are disengaged, camera does not know what lens is on any more) the combing disappears. So I am starting to think this could be some sort of vignetting correction at wide apertures... but his is just a guess.

If someone is willing to shoot some test shots with his 1D series camera at f/2 or wider at different ISOs including 1/3 step ones and send me the raw files I would be glad to run them through the analysis and post the results.

Please, gentlemen. The continuous mega-pixel-peeping just drives me up the wall. How does an image look on your calibrated monitor? How does it look on the Net? Finally, how does it look in a large print? If it looks acceptable then End Of Story.

Whether a forum is a gear-head forum or an artistic forum the endless discussions of minutia go to depths where nothing matters except lines on a chart. It's like a Corvette owner getting another 10 hp out of an 400 hp engine. Sure, the numbers are better (for those who thrive on horsepower numbers) but really, who cares?

Can we get back to talking about something that is remotely relevant to real-world photography?

Jack Joseph Jr said:

Please, gentlemen. The continuous mega-pixel-peeping just drives me up the wall. How does an image look on your calibrated monitor? How does it look on the Net? Finally, how does it look in a large print? If it looks acceptable then End Of Story.

Click to expand...

Jack, how can one argue with what you have written? This part of the discussion is where we would all agree! The basis for the work is that Canon has chosen to make certain approximations, which can end up in an image that has lost tonal values. Yes, it might not be important. That's what we would like to determine, after all!

When someone spends several to tens of thousands of dollars on a location, models, technicians, makeup artists, stylists, grip, wardrobe and more, the quality of the image should be the optimum.

Jack Joseph Jr said:

Whether a forum is a gear-head forum or an artistic forum the endless discussions of minutia go to depths where nothing matters except lines on a chart

Click to expand...

By definition, we don't like such discussions, either!

This topic, however is not about lines on a chart, per se. It’s about finding the unneeded compromises made to give intermediate ISO values when these add nothing, just deteriorate the image. It's such artifacts we wish to identify and avoid!

Canon does certain adjustments to certain ISO settings and with some other settings like wide f-stop. However, they have not disclosed exactly what they do. Some of the proprietary changes are probably related to things like unique colors that DPP gives, since other RAW software has no access to or knowledge of Canon's little tweaks. So, to some extent, Canon's reticence is understood, although we don't like having to pry open their hidden fudges and approximations to know which settings damage our otherwise perfect files!

I, for one appreciate that we have folk who have the patience and aptitude to address these issues.

Jack Joseph Jr said:

Can we get back to talking about something that is remotely relevant to real-world photography?

Click to expand...

I moved this thread to the purely technical forum so this need not hit the eyes of the uninterested.

Asher

Still, your objections are well intended and reminders are, of course, in order when and if we get lost!

While I agree endless pixel peeping gets tedious, in this case I think the information is valuable enough to warrant discussion... Kind of like teaching each other what all the CF menu's do

Jack Joseph Jr said:

Please, gentlemen. The continuous mega-pixel-peeping just drives me up the wall. How does an image look on your calibrated monitor? How does it look on the Net? Finally, how does it look in a large print? If it looks acceptable then End Of Story.

Click to expand...

But of course it is "mega-pixel-peeping" (I like this new term that is why when I first posted about it on DPReview I started with:

Click this quote ;-) said:

So I have finally turned into an equipment measurbator (http://www.kenrockwell.com/tech/7.htm) at least for a couple of days

Click to expand...

I am an engineer; photography is just a hobby, so when I found something interesting (for me) I thought other people may be interested and shared it. I enjoyed the discussion and when it was exhausted I just shut up (that was nine months ago).

But look what happened with that thread - which has had no posts for nine months - it shill has the most views and the most replies in the "Other Eos DSLRs" forum...

Oh and by the way, I was perfectly satisfied with my "severely combed" camera and had taken thousands of photos measurbation free when Asher tricked be back into it ;-)

Jack Joseph Jr said:

Can we get back to talking about something that is remotely relevant to real-world photography?

Click to expand...

I don't know how you got trapped inside a thread that doesn't interest you, but maybe some can figure a way for you to escape and find threads that interest you.

Asher Kelman said:

Canon does certain adjustments to certain ISO settings and with some other settings like wide f-stop. However, they have not disclosed exactly what they do. Some of the proprietary changes are probably related to things like unique colors that DPP gives, since other RAW software has no access to or knowledge of Canon's little tweaks. So, to some extent, Canon's reticence is understood, although we don't like having to pry open their hidden fudges and approximations to know which settings damage our otherwise perfect files!

Click to expand...

The tweaks actually make the data more consistent for all converters, as they put the whitepoint or middle grey at the same level in all RAW files. The problem is that whenever data is scaled so that there are gaps in the histogram, highlight headroom is lost to clipping, when a better solution would be just to note the change in whitepoint or greypoint in the RAW metadata, so a converter can compensate exposure accordingly, if it wants to.

There are some positive side effects to the 30D method of achieving the extra ISOs. DPP, for example, ignores some of the highest highlights in the normal ISOs, and by using 160/320/640/1250, the data is pushed down into the range where DPP operates. I tend to use these ISOs more than others (I never use the 125/50/500/1000 group at all; I wish I could make them disappear from the selection). If I were to do something where I was going to stack many images, I would only use the normal ISOs, as the bunching in the histgram would become more visible in the deep shadows once you got the noise down. The main issue in normal practice, however, is the change of highlight headroom between the three groups of ISOs, and the erratic patterns to read noise, with ISO 125 having the same read noise as ISO 640!

Peter Ruevski said:

People owning the 5D have tried its intermediate ISO settings and it seems to have proper amplifiers - i.e. no combing.

Click to expand...

It doesn't seem to be at the same stage as the amplification of the other ISOs, at the sensor, though. It is just a simple low-gain amplification before digitization. The read noise pretty much scales with ISO in groups of three. The same photosite amplification is used for ISO 100, 125 and 160. I have the exact values here from one 5D, but I've misplaced it.

John Sheehy said:

It doesn't seem to be at the same stage as the amplification of the other ISOs, at the sensor, though. It is just a simple low-gain amplification before digitization. The read noise pretty much scales with ISO in groups of three. The same photosite amplification is used for ISO 100, 125 and 160...

Click to expand...

This is interesting. So you are assuming that the camera has programmable "per pixel" (or at any rate "on sensor") amplifiers that can only do full stop ISOs (e.g. 100). And that in the 5D an extra low-gain amplifier was added before the ADCs to achieve the 1/3 stops (e.g. 125 and 160).

If this is indeed the case it could explain the noise being a step function jumping every three steps.

I am curious; is your idea of how the amplifiers are organized based solely on observations of the noise or is it documented somewhere?

Peter Ruevski said:

This is interesting. So you are assuming that the camera has programmable "per pixel" (or at any rate "on sensor") amplifiers that can only do full stop ISOs (e.g. 100). And that in the 5D an extra low-gain amplifier was added before the ADCs to achieve the 1/3 stops (e.g. 125 and 160).

If this is indeed the case it could explain the noise being a step function jumping every three steps.

I am curious; is your idea of how the amplifiers are organized based solely on observations of the noise or is it documented somewhere?

Click to expand...

I'm just observing the fossil record, and applying logic to it.

Canon doesn't officially comment on these things; Chuck Westfall said it was not something that he would discuss.