RAW files: 1Ds3 and Phase P30+

[Ray]

Canon cameras do not have particular ISO gains for the 1/3 stop or 1/2 stop settings. If you select a +1/3 ISO, then the next lower full-stop ISO will be applied and the result multiplied by about 1.27. This reduces the dynamic range without absolute any advantage if you are recording raw data.
[a href=\"index.php?act=findpost&pid=202914\"][{POST_SNAPBACK}][/a]

Gabor,
There's total confusion here. Emil wrote on the previous page, the following:

My understanding is different. There is fairly good evidence that Canon employs a standard variable gain amplifier (VGA) off the sensor to implement ISO; two in fact, one for the "main" ISO's 100-200-400-800-1600 etc, and a second one to implement "in-between" ISO's

My understanding is, the VGA is at each pixel site. As we know, the actual light-receiving photo-receptor on CMOS sensors is considerably smaller than the pixel pitch, in order to make room for other processing devices.

Canon generally don't advertise such details about their sensor design, but years ago Michael provided some information about Canon's first DSLR, the 3mp D30, which he got from some technical notes provided by Canon.

The D30, as a result of its small number of pixels has a very wide pixel pitch, around 10 microns, which is wider than the pixel pitch of most MFDBs, yet the actual light-receiving photodiode is apparently only 5.25 microns in diameter.

However, the microlens covers almost the whole of each photosite (with narrow gaps between each microlens). It's the microlenses and their particular design which ensures that as much of the light as possible, which falls on the sensor, is directed to the photodiodes underneath rather than the other processing devices next to the photodiodes.

[Ray]

The MFDBs in question (those which I looked at) do not carry out numerical adjustment of pixel values to imitate higher ISO like the DSLRs do. They always record at their maximum ISO gain. They can afford it with 16bit depth.

The P45+ behaves rather like the DSLRs in ISO, though it does not offer the fake high ISOs.
[a href=\"index.php?act=findpost&pid=202830\"][{POST_SNAPBACK}][/a]

Gabor,
Can you amplify on that explanation, please? I can't envisage how a camera could always record at its maximum ISO gain, unless it has only one ISO gain.

[Panopeeper]

Gabor,
I can't envisage how a camera could always record at its maximum ISO gain, unless it has only one ISO gain.

That's right; I did not express it correctly. What I meant is, that there is only one ISO gain, and that is the maximum possible with reasonable noise.

For example my 40D yields quite good image in the 8th and 9th stop at ISO 1600. The 14bits are too many, keep only 13 of them, and add three bits for the now cut off stops at the high end (at 1600): that would mean 11-12 stops DR. That would suffice me for a while. The cost of 16bits is all right with me: lower frame rate. It would require even less storage than now, because I shoot very often with bracketing.

[Panopeeper]

Re 1/3 stop ISO on Canons

I don't have samples from all cameras (who uses these ISOs?), but I just shot a set with the 40D. This has the particularity, that the pixel value range changes with ISO; that and the "holes" in the pixel values respecively the staggering demonstrates the effect.

I marked the saturation levels with a yellow ellipsis on each capture. The histograms are "zoomed in", each column/bar represents one pixel value.

ISO 200

ISO 250

ISO 320

ISO 400

[ejmartin]

Low noise at high ISO is a result of low noise in the circuitry involved in reading out the sensor to the VGA, it seems to me.

Emil,
If this is the case, then why does the low noise in the circuitry not improve S/N at base ISO?
[a href=\"index.php?act=findpost&pid=202806\"][{POST_SNAPBACK}][/a]

Because the noise floor is controlled by the noisiest element in the signal processing chain for any given ISO.  At low ISO, the sensor noise is not amplified much, and the biggest noise source comes from the VGA and ADC.  At high ISO, the sensor noise is highly amplified and becomes a bigger noise source than these two.  The low sensor noise can't help if it is being masked by bigger contributions from downstream in the signal processing chain.

[ejmartin]

Canon cameras do not have particular ISO gains for the 1/3 stop or 1/2 stop settings. If you select a +1/3 ISO, then the next lower full-stop ISO will be applied and the result multiplied by about 1.27. This reduces the dynamic range without absolute any advantage if you are recording raw data.

If you select a -1/3 ISO, then the next higher full stop ISO will be applied, and the result divided by about 1.27. This at least does not reduce the dynamic range, only the number of levels, but with 16bit that should not be an issue. However, this too is only fooling yourself.

[a href=\"index.php?act=findpost&pid=202914\"][{POST_SNAPBACK}][/a]

This depends on the model.  Canon's "pro" line (the 1 series and 5D) have "true" intermediate ISO gains (such as 1250) that are implemented in hardware via a second stage amplification downstream of the main ISO amplifier.  It is only the "consumer" line (xxxD and xxD series) that implement intermediate ISO's by digital multiplication of data from the nearest main ISO 100/200/400/800/1600.

[Panopeeper]

In fact, I have an ISO 1000 and 1250 image from the 5D, and these do not show the typical signs of fake ISOs.

[Ray]

In fact, I have an ISO 1000 and 1250 image from the 5D, and these do not show the typical signs of fake ISOs.
[a href=\"index.php?act=findpost&pid=202940\"][{POST_SNAPBACK}][/a]

 
This is a good example of the sort of problems we face on the internet. The notion that Canon DSLRs do not have 'real' intermediate ISO values has been around for quite a while. John Sheehy mentioned this characteristic of Canon DSLRs quite often.

Whilst I have a tendency to disbelieve almost everything I read, I simply don't have the time to check everything, and comparing ISO 160 on my 5D with ISO 100 (at the same exposure) was something I couldn't be bothered doing. I just avoided using those intermediate ISOs.

It now seems there was no good reason to avoid them.

[Ray]

Because the noise floor is controlled by the noisiest element in the signal processing chain for any given ISO.  At low ISO, the sensor noise is not amplified much, and the biggest noise source comes from the VGA and ADC.  At high ISO, the sensor noise is highly amplified and becomes a bigger noise source than these two.  The low sensor noise can't help if it is being masked by bigger contributions from downstream in the signal processing chain.
[a href=\"index.php?act=findpost&pid=202935\"][{POST_SNAPBACK}][/a]

This is still confusing for the layman, Emil. You wrote that the the lower noise at high ISO was due to the lower circuitry noise before analog pre-amplification.

If this is the case, whatever the analog gain later applied, the lower circuitry noise, prior to gain and with or without gain, should also have an effect on lowering noise whatever the ISO setting, which is merely an instruction to apply gain.

However, as Canon improves its technology, I can see that lower circuitry noise prior to pre-amplification, coupled with lower noise pre-amplifiers, could result in the dramatic improvement we now see in high ISO noise compared with Canon's earlier models.

[ejmartin]

This is still confusing for the layman, Emil. You wrote that the the lower noise at high ISO was due to the lower circuitry noise before analog pre-amplification.

If this is the case, whatever the analog gain later applied, the lower circuitry noise, prior to gain and with or without gain, should also have an effect on lowering noise whatever the ISO setting, which is merely an instruction to apply gain.

[{POST_SNAPBACK}][/a]

Sorry, I didn't want to resort to using equations.  The point is that the gain is not applied to all noise sources, only to those upstream of the amplifier; downstream noise sources (including the noise of the amplifier itself) are not amplified.  

As I understand it, the sensor readout has an associated noise N; this feeds into the amplifier that does ISO, which applies a gain G and adds its own noise N'.  The output has two independent contributions to the noise, G*N from the amplified sensor readout noise, and N' from the amplifier itself.   For instance, take the 1D3; the noise G*N is about .8*ISO/100 and the noise N' is about 4.2 raw levels; at ISO 100, N' is the bigger contributor, while at ISO 1600, G*N~13 is bigger.  

Because the low ISO and high ISO noise are determined by different circuit elements, reducing one of them does not necessarily result in improved performance for all ISO.  And note that it is the noise from circuit elements *before* the ISO amplification that determine the high ISO noise performance; this I think is why CMOS provides better high ISO performance, by having lower noise in the sensor element readout.

BTW, I probably shouldn't have said the ISO amplifier is off-chip.  It can be and often is on the sensor chip itself, but I don't think it's part of the pixel circuitry itself, and that is the only distinction that was important for what I was saying...

[a href=\"http://en.wikipedia.org/wiki/Active_pixel_sensor]http://en.wikipedia.org/wiki/Active_pixel_sensor[/url]

[ejmartin]

This is a good example of the sort of problems we face on the internet. The notion that Canon DSLRs do not have 'real' intermediate ISO values has been around for quite a while. John Sheehy mentioned this characteristic of Canon DSLRs quite often.

Yes, it was John who discovered this property (in the context of the 30D, IIRC).  But in his defense, he is always quite specific in qualifying what models have this property (he was also involved in verifying that the 5D has true intermediate ISO's); the problem is the usual one of human communication, that a piece of information relayed multiple times loses a fraction of its accuracy with each transmission and eventually becomes garbled.

Whilst I have a tendency to disbelieve almost everything I read, I simply don't have the time to check everything, and comparing ISO 160 on my 5D with ISO 100 (at the same exposure) was something I couldn't be bothered doing. I just avoided using those intermediate ISOs.

It now seems there was no good reason to avoid them.
[a href=\"index.php?act=findpost&pid=202951\"][{POST_SNAPBACK}][/a]

Well, there is still a good reason to avoid intermediate ISO's on the pro Canons, they are still quite a bit noisier in absolute terms than their "main" ISO neighbors.  If I have a choice in terms of Av/Tv, I still prefer the main ISO's and tweaking the Av/Tv a third stop to get the proper exposure.  At least, this is what I do up to about ISO 1000-1250 on my 1D3, at which point the read noise is controlled by the sensor readout and it doesn't matter whether the secondary amplifier is kicking in or not, and one can use the intermediate ISO's without penalty.  

This is all due to Canon's bizarre double stage amplification mechanism for providing these intermediate ISO's.  Nikon just uses a single variable gain amplifier for all ISO's including the intermediate ones, and the read noise just smoothly increases with ISO rather than bouncing around like it does for Canon.

Oh, and for those who shake their heads at this sort of "science experiment", it's a good example how such experiments inform exposure decisions in order to maximize image quality.

[rethmeier]

[Dustbak]

Oh, and for those who shake their heads at this sort of "science experiment", it's a good example how such experiments inform exposure decisions in order to maximize image quality.
[a href=\"index.php?act=findpost&pid=202974\"][{POST_SNAPBACK}][/a]

No no no, go ahead. If scientifically proven what the best practice is per body/sensor please let us know so I can use it to my benefit. No problem in people sorting out all the scientific details for us to benefit from. At least that is my opinion. Trial & error does work as well but takes longer I guess.

It just sounds a bit like you are kind of missing out on the really fun part, taking or making images.

[rethmeier]

"It just sounds a bit like you are kind of missing out on the really fun part, taking or making images."


You hit the nail on the head!


Jammer dat Holland niet zo gelukkig was!

Guus Hiddink did wonders for the Socceroos too!

[Dale Allyn]

Finally, pls post a raw file of whatever with whatever setting, except a black frame, only as technical reference for me. I am hunting for the error, which I noticed in the raw file of the P30+ posted above by 203.
[{POST_SNAPBACK}][/a]

Gabor,

Here is a RAW file which has no photographic merit. It's simply a shot which includes data from slightly clipping highlights to clipping shadows. I was outside today and was curious about how the P25+ would handle this ridiculous range. If it's not useful for your purposes I can upload something different.

Dale

[a href=\"https://download.yousendit.com/138ED46D1AAA8B09]https://download.yousendit.com/138ED46D1AAA8B09[/url]

[rethmeier]

For the Ruski's I mean!

[Dustbak]

"It just sounds a bit like you are kind of missing out on the really fun part, taking or making images."
You hit the nail on the head!
Jammer dat Holland niet zo gelukkig was!

Guus Hiddink did wonders for the Socceroos too!
[a href=\"index.php?act=findpost&pid=202988\"][{POST_SNAPBACK}][/a]

Staat er in ieder geval 1 Hollander in de 1/2 finale  

Guus has proven to be a top trainer more than once. This does include the Socceroos indeed.

[rethmeier]

"Guus has proven to be a top trainer more than once. This does include the Socceroos indeed."
  $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$

[ejmartin]

It just sounds a bit like you are kind of missing out on the really fun part, taking or making images.
[{POST_SNAPBACK}][/a]

What makes you think I do?

[a href=\"http://theory.uchicago.edu/~ejm/pix/20d/Trips/Homer_Alaska_2008/IMG_6318-web.jpg]http://theory.uchicago.edu/~ejm/pix/20d/Tr...MG_6318-web.jpg[/url]

[Dustbak]

What makes you think I do?

http://theory.uchicago.edu/~ejm/pix/20d/Tr...MG_6318-web.jpg
[a href=\"index.php?act=findpost&pid=203039\"][{POST_SNAPBACK}][/a]

Now that's what I like looking at much more