7D dynamic range?

[John Sheehy]

I have to agree with you somewhat. What I remember were posts showing the "true" base ISO for cams like the 40D and 50D to be ISO 160 and that was where the DR peaked on the chart.

I wrote years ago that the 20D would need an ISO of about 115 to have the same highlight headroom as ISO 200. The reason being is that what the 20D recorded as 4095 was really about 3550, and RAW highlights were stretched to hide this fact.

On a completely different note, I mentioned that the DR was slightly higher at ISO 160 than at 100 on one or more models, because the cameras clipped ISO 100 more than necessary.

These are engineering DRs, which may not correlate with people's own subjective opinion of what "usable minimum signal" is. Camera A may have more DR than camera B, but camera B may still have better SNR at a number of tonal zones below saturation.

--
John

 

[John Sheehy]

Most of us probably all remember that the 20D and even the 40D use numeric processing to achieve the "in between" ISOs.

And thus, if you're shooting RAW, there is little point in using anything but the "full stop" ISO settings (100, 200, 400, etc.).

But does anyone know if the 7D continues this design philosophy, or if it might actually use analog amplification ahead of the A/D to arrive at the intermediate ISOs?

That would be a first. Canon has never used 1/3-stop increments in the main gain stage. When the variable ISO is all analog gain, it is done at a late stage where it is not very useful.

--
John

 

[RedFox88]

I also was pretty surprised when I heard that digital cameras have 9 stops of DR, as I learned negativ film has 7-9 stops DR.

Unfortunatly I can't find the website right now, but I once saw a comparison on a website that showed that film is more like 14 stops DR.

Here's a pretty good test:
http://forums.dpreview.com/forums/read.asp?forum=1031&message=32891438

That's a horrible test. It's not even a test of DR, only highlight range, and it doesn't even test that correctly. Most of a DSLR's range is on the shadow side.

It shows exposure latitude. If you can expose 3 stops over a "normal" exposure your shadows will be well exposed and as you can see Ektar has great highlight retention. If you have problems with the test, take it up with the user on here that is responsible for it. I'm sure he'll give you a good discussion.

I'd rather have a bright image than an image that has most of the light being shadows.

 

[photog7320]

I also was pretty surprised when I heard that digital cameras have 9 stops of DR, as I learned negativ film has 7-9 stops DR.

Unfortunatly I can't find the website right now, but I once saw a comparison on a website that showed that film is more like 14 stops DR.

Here's a pretty good test:
http://forums.dpreview.com/forums/read.asp?forum=1031&message=32891438

That's a horrible test. It's not even a test of DR, only highlight range, and it doesn't even test that correctly. Most of a DSLR's range is on the shadow side.

It shows exposure latitude. If you can expose 3 stops over a "normal" exposure your shadows will be well exposed and as you can see Ektar has great highlight retention.

Exposure latitude is merely medium range - scene range. It's not an independent characteristic. You cannot overexpose by 3 stops if the scene range is as wide or wider than the film range. All this test tells us is that the scene had a short highlight range, so it was possible to overexpose it given a medium with sufficient range on its highlight side. It doesn't tell us the total range of anything. (Nor could it considering the lack of any calibrated points by which to measure.)

You don't need to overexpose digital to get well exposed shadows because that's where most of its dynamic range exists.

 

[RedFox88]

You don't need to overexpose digital to get well exposed shadows because that's where most of its dynamic range exists.

You still aren't getting it. If you can "over" expose Ektar by 3 stops and have a very good looking image, then you can capture a very high contrast scene without losing detail, unlike the digitals shown in that example.

"blown highlights" is a common theme with dSLRs which is not an issue with normal contrast color negative films. But somehow I think you're someone who no matter what evidence is shown is a "digital is better" person even if it means most of the recorded light are shadows/darker light. Good luck with that.

 

[Peter 13]

It is not so black and white. I have scanned many negatives, played with analog gain, etc. While it is hard to see blown highlights with film, getting completely black shadows is common. But wait, the shadows are actually highlights on the negative...

You don't need to overexpose digital to get well exposed shadows because that's where most of its dynamic range exists.

You still aren't getting it. If you can "over" expose Ektar by 3 stops and have a very good looking image, then you can capture a very high contrast scene without losing detail, unlike the digitals shown in that example.

"blown highlights" is a common theme with dSLRs which is not an issue with normal contrast color negative films. But somehow I think you're someone who no matter what evidence is shown is a "digital is better" person even if it means most of the recorded light are shadows/darker light. Good luck with that.

 

[photog7320]

You don't need to overexpose digital to get well exposed shadows because that's where most of its dynamic range exists.

You still aren't getting it. If you can "over" expose Ektar by 3 stops and have a very good looking image, then you can capture a very high contrast scene without losing detail, unlike the digitals shown in that example.

No, you're not getting it. Without showing the shadow side we don't know the tolerance for underexposure of the tested digital camera. When faced with a high contrast scene on film, you overexpose to take advantage of the fact that most of the total range is on the highlight side. When faced with the same scene on digital, you underexpose to take advantage of the fact that most of the total range is on the shadow side.

Total dynamic range is the sum of both. The total range determines which can deal with higher contrast scenes. Which side that range falls on only determines how you expose to take full advantage of the total range.

There are certainly films with more total range than current digital sensors, but from this so called "test" no one would have any idea where Ektar falls.

"blown highlights" is a common theme with dSLRs which is not an issue with normal contrast color negative films.

But if you ever actually shot film you would know that blocked shadows were/are a major problem with many print films. So much so that people routinely bias their camera to give more exposure by manually turning down the ISO, i.e. shooting a 100 film at 64.

But somehow I think you're someone who no matter what evidence is shown is a "digital is better" person even if it means most of the recorded light are shadows/darker light.

No usable evidence has been shown. You appear to be a person who is too ignorant about dynamic range and exposure to realize that. (You throw an insulting accusation, I throw one back. You want to remain civil, then don't start a fight.)

Good luck with that.

I have great luck with that. Underexposing a high contrast scene and then lifting the shadows in RAW works extremely well. It's amazing just how much detail you can pull out of the shadows of a RAW file. For that matter you typically have at least 1 stop or more of highlight recovery from the same file. I haven't yet tested it properly (which means shooting transmission wedges), but I'm guessing developing and blending two images from a single RAW file, one optimized for shadow recovery, one for highlight recovery, might yield an extra 2-3 stops of usable information at ISO 100. Which would put the best DSLRs right up there alongside portrait print films.

Of course the test results would vary based on acceptable noise and the inclusion or exclusion of noise filtering software.

 

[photog7320]

Bingo! Someone who gets it...

It is not so black and white. I have scanned many negatives, played with analog gain, etc. While it is hard to see blown highlights with film, getting completely black shadows is common. But wait, the shadows are actually highlights on the negative...

You don't need to overexpose digital to get well exposed shadows because that's where most of its dynamic range exists.

You still aren't getting it. If you can "over" expose Ektar by 3 stops and have a very good looking image, then you can capture a very high contrast scene without losing detail, unlike the digitals shown in that example.

"blown highlights" is a common theme with dSLRs which is not an issue with normal contrast color negative films. But somehow I think you're someone who no matter what evidence is shown is a "digital is better" person even if it means most of the recorded light are shadows/darker light. Good luck with that.

 

[bronxbombers]

they were based on tests I made on blackframes using IRIS for the read noise

Does not IRIS perform black level correction before calculating the standard deviation?

--
Gabor

http://www.panopeeper.com/panorama/pano.htm

i don't know, it doesn't list avg pixel value as near 0 but near 1024 (or 2048 for the 7D) so I would think not, but I suppose it is possible

 

[bronxbombers]

Most of us probably all remember that the 20D and even the 40D use numeric processing to achieve the "in between" ISOs.

And thus, if you're shooting RAW, there is little point in using anything but the "full stop" ISO settings (100, 200, 400, etc.).

But does anyone know if the 7D continues this design philosophy, or if it might actually use analog amplification ahead of the A/D to arrive at the intermediate ISOs?

--
Jim H.

the onyl raw samples i've seen posted have been at full isos so no clue

i did notice that for the first time ever ISO3200 truly seems to be real though

 

[bronxbombers]

Those are handy graphs and examples.

I think most people want to see hard, cold, objective measurements so that they can make judgments comparing cameras or ISOs, etc., so that they can come up with easy rules for what ISOs to use, etc.

But the problem is that all of this is subjective in the end.

It's the same with audio, too.

The point at which you suddenly decide that the SNR is too bad to tolerate depends on a lot of factors and personal taste, etc.

Back in the days of analog tape, there was a lot of personal taste involved in how "hot" to record things.

Record hotter, and the tape noise ends up lower with respect to the average signal level. But you have higher levels of distortion in the loud passages.

Record lower, and the tape noise ends up being higher with respect to the average level, but you have lower distortion in the loud passages.

So different people, using the exact same recording system, ended up with different signal to noise ratios in their final product. But who's to say which "sounded better"? I tended to find tape hiss less objectionable than distortions. So I tended to record fairly low. Others preferred to lower the tape hiss by recording hotter, and could tolerate the distortion in the loud areas in trade for that.

And then we get into the question of "pattern noise" versus random noise. I suspect this ends up being very similar between audio and digital image capture, too.

If the non-signal artifacts in the "recording" have no pattern to them (that is, they're truly random in nature), then our eyes or ears can put up with them at a much higher level than if that non-signal artifact has a discernible pattern to it.

Vertical or horizontal banding in an image is a lot like having a low-level "whine" in the background of a recording versus random noise. The relative power level of that "whine" may be the same as the power level of the random noise, but your ear/brain can pick it out easily, and it'll be annoying. And the relative power level of a bunch of lines in an image might be no higher than another image's random noise, but the image with the lines in it will be very ugly to most people.

Our brains tend to look for or listen for patterns. So reducing pattern noise is paramount IMO.

It appears that the 7D is pretty good in this regard.

To me, therefore, I'd probably tend to find its DR to be greater than a camera with the same "measured" DR but which had a more defined pattern to it's "noise".

If we got a batch of film that was made when the coating machine was having a problem, and it had an uneven coating thickness such that there were horizontal "bands" in it, people hated it. And yet, the intensity of those bands didn't need to be very great for it to be unacceptable. You might even argue that this "banding" was below the "noise level" of the grain. But it still looked awful.

So I think the nature of the "noise" in question can have a huge impact on whether or not it's acceptable.

Two cameras or ISO settings might have the same DR or SNR by some measurement, yet the images may not look the same.

--
Jim H.

yeah that is why i was saying that the effective DR of the 7D may be a little higher than that of the 5D2 and why the 40D may actually be close (but note the SNR at 18% is much better on the 5D2) since the 7D seems to have VERY little pattern banding and the 40D some and the 5D2 a lot.

the 50D has a real lot so even in the less optimistic scenario for the 7D where the measured DR advatage may not hint at anything noticeably different in the realworld, there may well be a noticeable real-world advant for the 7d

 

[pidera]

thanks, Pieter

 

[RedFox88]

No, you're not getting it. Without showing the shadow side we don't know the tolerance for underexposure of the tested digital camera.

Most people prefer to give a good exposure instead of under expose, but hey. Here are more Ektar vs. digital which includes under exposure tests:
http://forums.dpreview.com/forums/read.asp?forum=1018&message=31266941

No usable evidence has been shown. You appear to be a person who is too ignorant about dynamic range and exposure to realize that. (You throw an insulting accusation, I throw one back. You want to remain civil, then don't start a fight.)

I've seen many of your "digital is god's best creation" attitudes on here to know one. If you cannot see how Ektar has outperformed digital in the tests I just included, then I shall say no more since it would not sway your opinion.

 

[tigger183234]

When faced with a high contrast scene on film, you overexpose to take advantage of the fact that most of the total range is on the highlight side. When faced with the same scene on digital, you underexpose to take advantage of the fact that most of the total range is on the shadow side.

Isn't the most tonal range on the highlight side on digital?

Film tends to be "exposed for the shadows", to get more tonal range in the shadows. This can lead to a overexposure of the highlights, but film is very forgiving in this regard, as the exposure curve flattens in the highlights.

Digital on the other hand is "exposed for the highlights", to make sure the highlights don't blow out, as blown out highlights look worse than dark shadows. So on digital, you rather underexpose to minimize blown highlights, but I though the best s/n ratio is in the brighter areas.

This article seems to support my thinking: http://www.luminous-landscape.com/tutorials/expose-right.shtml

 

[GaborSch]

i don't know, it doesn't list avg pixel value as near 0 but near 1024 (or 2048 for the 7D) so I would think not, but I suppose it is possible

Well, you should know that. There is a huge difference between the standard deviations before and after black level correction.

--
Gabor

http://www.panopeeper.com/panorama/pano.htm

 

[bronxbombers]

i don't know, it doesn't list avg pixel value as near 0 but near 1024 (or 2048 for the 7D) so I would think not, but I suppose it is possible

Well, you should know that. There is a huge difference between the standard deviations before and after black level correction.

--

well considering that it doesn't list in the statistics means/avg/high for values around 0 but around 1024 wouldn't that suggest that they don't do the stddev after black level correction? I maybe they would. I've never head anyone bring this up before or say that you can't used IRIS for this. Do you have any documentation stating when they apply it?

 

[John Sheehy]

well considering that it doesn't list in the statistics means/avg/high for values around 0 but around 1024 wouldn't that suggest that they don't do the stddev after black level correction? I maybe they would. I've never head anyone bring this up before or say that you can't used IRIS for this. Do you have any documentation stating when they apply it?

This all depends on the software. Iris doesn't do anything to the RAW data, until you do things yourself. Black is about 128, 256, 1024, or 2048 (new for the 7D), and you have to subtract these numbers before black is zero. In Iris, the standard deviation does not change for black, because it uses negative numbers and does not clip at 0 unless you tell it to, or it writes a file.

--
John

 

[GaborSch]

I don't use IRIS, I have no idea how and what it is doing.

The point is, that the standard deviation is much greater with "raw" raw than with the black level corrected data. Example: my 40D with ISO 100, 1/30sec, shows 5.95-6.00 on a black frame shot, equal on the masked pixels and image area, without black level correction . However, after black level correction, it is only 3.65-3.90, in the very same shot.

This fact makes comparison of the DR this way between Canons and most other cameras impossible - another reason I prefer to measure the relative noise on uniform patches of non-black frame shots. When doing so, I am watching for the proportion of "black clipping", because when you measure the noise on such deep shedows, which contain a lot of black clipped pixels, the noise becomes "moderated" (lowered). This is the source of the funny changes in the curves of ignorant measurements.

--
Gabor

http://www.panopeeper.com/panorama/pano.htm

 

[ejmartin]

I don't use IRIS, I have no idea how and what it is doing.

The point is, that the standard deviation is much greater with "raw" raw than with the black level corrected data. Example: my 40D with ISO 100, 1/30sec, shows 5.95-6.00 on a black frame shot, equal on the masked pixels and image area, without black level correction . However, after black level correction, it is only 3.65-3.90, in the very same shot.

This fact makes comparison of the DR this way between Canons and most other cameras impossible - another reason I prefer to measure the relative noise on uniform patches of non-black frame shots. When doing so, I am watching for the proportion of "black clipping", because when you measure the noise on such deep shedows, which contain a lot of black clipped pixels, the noise becomes "moderated" (lowered). This is the source of the funny changes in the curves of ignorant measurements.

--

What do you mean by "black level correction"? I can imagine a couple of things:

1) Clipping the data at a reference black point, setting all the RAW values less than this equal to the reference black point. This will substantially reduce the std dev of the data. IRIS does not do this (though you can do it manually using IRIS if you want, nothing is done to the RAW values by default by IRIS other than read them in, and trim the borders of the frame to for instance exclude the masked pixels).

2) Some attempt to eliminate pattern noise by individually adjusting the average of each row/column of the RAW data. IRIS does not do this either.

Did you mean either of these, or something else?

I didn't code IRIS, my understanding (which may be incorrect) is that it uses that part of dcraw code that reads in the RAW file, then displays that data without further manipulation (in particular without passing it through the rest of the dcraw data pipeline).

--
emil
--



http://theory.uchicago.edu/~ejm/pix/20d/

 

[GaborSch]

off-setting the original pixel values, so that the pixel value 0 represents "black". The corrected corrected pixel values are theoretically linear.

The way to achieve that depends on the camera; some cameras (notably, the Canons) allow for different approaches. Rawnalyze can perform the correction based on a global average of the masked pixels, based on column and/or row aveages, or with a constant number (or ignore it all together). The appearance of the noise on the very darkest areas does change with the methods, but the standard deviation changes only negligably.

--
Gabor

http://www.panopeeper.com/panorama/pano.htm