Striations in P45+@800 ISO normal ?

[eronald]

There is no such thing as "clipping the blacks", as far as user expsoure is concerned.  A RAW histogram is not a symmetrical entity; the left side is anchored, and the right side is scalable.

However, bad cameras and converters can clip black, but that is independent of exposure.
[a href=\"index.php?act=findpost&pid=155747\"][{POST_SNAPBACK}][/a]

I'm going to bake up some tools myself, and take the Raw file apart - these black values shown by Panopeeper's tool sound preposterous, except if there is some very strong massaging of the data before the Raw file gets written. I'm also going to shoot @ISO 200 and push, just to see what happens.

Edmund

[Panopeeper]

these black values shown by Panopeeper's tool sound preposterous, except if there is some very strong massaging of the data before the Raw file gets written.

1. the black values are unquestionable,

2. the relative values are questionable. I suggested you to shoot a white board with at least four stops overexposure and post the resulting raw; then I can determine the true clipping points (do this with several ISO, because it *may* depend on the ISO, I don't know your camera).

If it turns out, that the clipping points are at 16000, then the underexposure is only one stop away from the right edge. If the clipping points are somewhere else, the "distance to the right" changes.

Adobe DNG converter gets it wrongly with my Canon 40D as well: it always shows 13600, while the actual clipping is between 12700 and 16383, depending on the ISO.

(I mentioned clipping points in plural, because the different "colored" pixels may have different clipping points, though the difference won't be much.)

3. Why don't you shoot a "normal" daylight scenery with ISO 100 as high as you think you can go (and another one with +1 EV) and post those, for verification

[eronald]

Panopeeper,

1. First of all, we have no idea what actually comes off the sensor of a Phase One camera. At least I don't at this point, if we had signed an NDA we wouldn't be talking. All we are seeing is what gets written to file.

2. Then we don't *really* know what is in this file, because, again, we only know what Dcraw is squeezing out of it, and Dave doesn't necessarily have a *complete* map to all the tricks of the format.

3. I do think you should change the x scales so one look would be enough to compare data with the histograms we see in the Raw converters, and this would tell us what the black box Raw converters like C1 are doing to the data.

4. I would really like to see what is going on with the *dark* pixels. Tracing these, maybe via a color map around a thumbnail would be *VERY VERY INTERESTING*

Edmund

1. the black values are unquestionable,

2. the relative values are questionable. I suggested you to shoot a white board with at least four stops overexposure and post the resulting raw; then I can determine the true clipping points (do this with several ISO, because it *may* depend on the ISO, I don't know your camera).

If it turns out, that the clipping points are at 16000, then the underexposure is only one stop away from the right edge. If the clipping points are somewhere else, the "distance to the right" changes.

Adobe DNG converter gets it wrongly with my Canon 40D as well: it always shows 13600, while the actual clipping is between 12700 and 16383, depending on the ISO.

(I mentioned clipping points in plural, because the different "colored" pixels may have different clipping points, though the difference won't be much.)

3. Why don't you shoot a "normal" daylight scenery with ISO 100 as high as you think you can go (and another one with +1 EV) and post those, for verification
[a href=\"index.php?act=findpost&pid=155871\"][{POST_SNAPBACK}][/a]

[Panopeeper]

All we are seeing is what gets written to file

That's right, but isn't it irrrelevant, what the camera is doing with the raw data before writing the file, if we have to live with the file?

Then we don't *really* know what is in this file, because, again, we only know what Dcraw is squeezing out of it

I am not using DCraw. At this point I am working with DNG converted from your images by the Adobe DNG converter, because I don't know the characteristics of the original raw file (it is very messy).

I do think you should change the x scales so one look would be enough to compare data with the histograms we see in the Raw converters

I think you don't understand the problem. If one wants to know, how far the exposure is from the very right (i.e. from clipping), one has to know, where the clipping occurs. This can not be determined from the images you presented, because there is none clipping there.

The fact, that the camera delivers 16bit data does not mean, that the actual values go up to 64K.

Anyway, do it as you see it fit, it's your problem.

[John Sheehy]

I find it quite extraordinary that such professional photographers who spend so much on their equipment should be so ignorant of these basic principles of ETTR.

My mind boggles. Should they all be sacked and sent off to Siberia?
[a href=\"index.php?act=findpost&pid=155534\"][{POST_SNAPBACK}][/a]

Well, how is he supposed to know where the right is?  How do you know where it is?

What kind of feedback is his camera giving?

Where is the clipping point in the P45+ RAW data?

Are the higher ISOs actually unique digitizations or just logistics?

[Ray]

Well, how is he supposed to know where the right is?  How do you know where it is?

What kind of feedback is his camera giving?

Where is the clipping point in the P45+ RAW data?

Are the higher ISOs actually unique digitizations or just logistics?
[a href=\"index.php?act=findpost&pid=155964\"][{POST_SNAPBACK}][/a]

Perhaps I'm making incorrect assumptions based upon the way 35mm DSLRs operate. I rely upon the flashing highlight warning and the appearance of the histogram to tell me how close I am to full exposure. But first I have the camera's jpeg picture options set to minimum contrast such that a very small area of 'apparent' highlight clipping indicates exposure is a correct ETTR. Without that minimum contrast setting, I'd be underexposing all the time when shooting RAW, which is fine for out-of-the-box jpeg images of course which have to be underexposed to avoid highlight clipping.

There seems to be an unavoidable but small degree of guesswork here. Even with contrast at a minimum, a histogram and review image with no highlight flashing at all is still likely to be underexposed by a small amount, perhaps 1/2 a stop.

One more accurate approach, which unfortunately is a bit slow and cumbersome which is why I don't use it, is to take a spot meter reading of the brightest part of the scene and then increase exposure by 4 stops. But it's so long since I used this approach I can't be sure if it's 4 stops or less. Whatever it was, it worked just fine.

[Panopeeper]

Perhaps I'm making incorrect assumptions based upon the way 35mm DSLRs operate

You certainly are, if you believe, that it has to do with 35mm DSLR.

The issue is, that the histogram the camera displays is based on the resulting JPEG image (in case of raw recording, there is a JPEG image embedded in the raw file).

That JPEG has been created from the raw data based on the camera settings:

- contrast
- saturation
- sharpening
- color tone
but the single biggest factor is the white balance (which is not always a setting, it may have been evaluated by the camera).

I posted some raw histograms on this thread, and later the corresponding RGB histogram without de-mosaicing but after white balancing. You may notice, that the raw green values are over twice as high as the reds, but in the resulting image they end up as high as the green (the brightest part of that image is the white card, i.e. red, green and blue has to be equal at the right end).

There are some enviable cameras (or cameras of some enviable owners), which accept coefficients as setting. Some Nikons do that; in Canons only the very top models are on par.

Otherwise you have to experiment a lot with your camera to find out a WB setting (not only temperature but tint as well), which results in coefficients 1.0, 1.0, 1.0.

I rely upon the flashing highlight warning and the appearance of the histogram to tell me how close I am to full exposure

It is unreliable without the right guessing of WB.

There are some downsides of such a "raw setting":

- the image appears off-color on the camera,
- the thumbnail is off-color,
- the embedded JPEG is useless,
- you can not reasonably combine raw with in-camera JPEG

[Ray]

You certainly are, if you believe, that it has to do with 35mm DSLR.

The issue is, that the histogram the camera displays is based on the resulting JPEG image (in case of raw recording, there is a JPEG image embedded in the raw file).

[a href=\"index.php?act=findpost&pid=155997\"][{POST_SNAPBACK}][/a]

If what has to do with 35mm DSLR?

I realise that the camera's review image is based on a converted jpeg which is why I set the camera's picture modes to minimum contrast, and minimum saturation even though I'm shooting RAW, because I find the flashing highlight warning is then more accurate, but still slightly overstated.

One has to work with the tools one has and experiment to see what approach gives the most accurate result and in what circumstances. When in doubt, autobracket exposure.

What I meant when I said I might be making some incorrect assumtions is that perhaps I've assumed that an MFDB back has a histogram and review image along similar lines to a 35mm DSLR's. If this is not the case and for some reason the exposure feed-back on an MFDB is not as sophisticated as that on a 35mm DSLR, then I take back my comment.

No need to sack these guys and send them off to Siberia   .

[Ray]

Otherwise you have to experiment a lot with your camera to find out a WB setting (not only temperature but tint as well), which results in coefficients 1.0, 1.0, 1.0.
It is unreliable without the right guessing of WB.

There are some downsides of such a "raw setting":

- the image appears off-color on the camera,
- the thumbnail is off-color,
- the embedded JPEG is useless,
- you can not reasonably combine raw with in-camera JPEG
[a href=\"index.php?act=findpost&pid=155997\"][{POST_SNAPBACK}][/a]

If I'm shooting RAW, I don't care about the appearance of the jpeg based image on the camera's LCD screen. I'm only concerned about the correct exposure.

If I were really fanatical about getting the mximum exposure my camera is capable of, short of clipping one or more of the channels, I'd adopt John Sheehy's method of using a particular type of color filter on the lens to equalise the sensitivity of the camera's red, green and blue pixels.

I always use the camera's auto WB. If for some reason, as is sometimes the case, that doesn't appear quite right in the ACR window, I'll make adjustments before converting. Sometimes simply selecting the 'flash' WB instead of the 'As shot' WB in ACR does the trick, if the image was shot using flash. Sometimes I mess around with the temperature and tint slider or try a gray balance.

I bought a gray card once, but never used it. I had some idea I would take 2 shots, the first with a big gray card in the scene and the second shot without it. But it's not really my style of shooting, waving a gray card around   .

[Panopeeper]

If I were really fanatical about getting the mximum exposure my camera is capable of, short of clipping one or more of the channels, I'd adopt John Sheehy's method of using a particular type of color filter on the lens to equalise the sensitivity of the camera's red, green and blue pixels

I did not know about that, but I am waiting for the arrival of a magenta filter (delivery time: 6-10 weeks), just for this purpose.

The wast majority of my shots (landscape panos) are "green limited. I hope to gain almost half stop this way.

[Jonathan Wienke]

Well, how is he supposed to know where the right is?  How do you know where it is?

What kind of feedback is his camera giving?

Where is the clipping point in the P45+ RAW data?

Are the higher ISOs actually unique digitizations or just logistics?

It's pretty much a given that the camera histogram is bullshit, since no camera histogram yet is based on actual RAW data, but some conversion/interpretation of that data instead. To really know what the camera histogram is telling you, you need to conduct the test I outline toward the end of this article. Shoot RAWs of a Color Checker or WhiBal card along with some of your favorite subject matter under your typical lighting conditions, starting about 2 stops under meter and increasing exposure in 1/3-stop increments until you're about 2 stops over the meter. Make a note of where the camera histogram first indicates clipping. Then look at the RAWs in ACR or the CryptoBola analyzer (or better yet, both) and see where clipping actually occurs in the RAW data.

Once you've done that, you'll know what your camera histogram really means, you'll know what the exposure interval between camera histogram clipping and RAW clipping really is, and you'll be able to judge exposure much more intelligently.

[John Sheehy]

I did not know about that, but I am waiting for the arrival of a magenta filter (delivery time: 6-10 weeks), just for this purpose.

The wast majority of my shots (landscape panos) are "green limited. I hope to gain almost half stop this way.
[a href=\"index.php?act=findpost&pid=156003\"][{POST_SNAPBACK}][/a]

Using a magenta filter in daylight will increase the DR of full-color response.  It will not help with those cases where you can retrieve greyscale highlights from the red channel, however.  You can exaggerate that benefit by using a green filter.

The magenta filter in daylight will make the shadow noise less chromatic.  Same for tungsten; except you need a filter that cuts red and green about two stops relative to blue.

This all assumes you have enough light to get a good exposure with the filters.  They don't really help if they force bad exposures.

[godtfred]

Axel,

 Seeing you have a P45+, could you post some underexposed shots at hi iso please ?

Edmund
[{POST_SNAPBACK}][/a]

Sure, no problem...

Here are some bracketed exposures at ISO 800 and ISO 400:

[a href=\"http://rapidshare.com/files/72455887/ISO800_Tests.zip]ISO 800[/url]

ISO 400

The images are taken with a Profoto B600R flash as the only light source, bounced on a umbrella (white from above.) They are the untouched RAW files from the Phase One P45+ back on a H2 camera using the HC100 lens.

If you push and pull the file a lot in PS, you will notice some lines in the black bag, but the adjustments have to be way over the top like in this screenshot:

[attachment=4035:attachment]

-axel

[Dustbak]

These are less underexposed than Edmund's examples. The one with the lowest exposure of the ISO400 samples is only about 1 stop under. At least that is what ACR is telling me.

Pushing it towards the point of clipping does show the bag go mildly to magenta & greenish.

I bet when you would underexpose that shot another stop, with which you will come close to Edmunds example, there will be a lot more garbage in the bag.

I am currently downloading your ISO800 examples...

Yikes! The ISO800 examples look horrible. These are going from being exposed right on the point of clipping to being 3/4 of a stop over? Even than there is a lot of Green/magenta rubbish in the darker parts.

[godtfred]

These are less underexposed than Edmund's examples. The one with the lowest exposure of the ISO400 samples is only about 1 stop under.
[a href=\"index.php?act=findpost&pid=156155\"][{POST_SNAPBACK}][/a]

This is correct, there should be examples of both underexposure by between 2/3s and a full stop, as well as one correctly exposed, and one with highlights clipping in both of the downloads (ISO400/800).

If you take the correctly exposed 400 and 800 images, and pull back the midtones a lot, while clipping the blacks around 16-18 from zero, you can see striations in the 800 image, but not the 400. Anyone with knowledge as to why this would happen?

I'm content with the image quality as of yet for my usual ISO50-200 shots, but I'm seeing some magenta and green casts in some of my images shot against white product tables. I'll be looking closer at image quality as I get to know the back and software more.

-axel

EDIT:

Having seen over the files in ACR, zeroing everything, I now don't trust the histogram in C1 at all, they are completely different, with ACR pulling out a lot more highlight detail and C1 pushing midtones as well as highlights a lot, even using the "Linear" tone curve provided.

The image provided shows the differences. If somebody can show me a "secret button" in C1, it'd be greatly appreciated:

[attachment=4038:attachment]

EDIT II:

There are much larger differences between the ISO 400 and ISO 800 shots in ACR than in C1, I'm not sure what to trust... check out these histograms:

[attachment=4039:attachment]

[godtfred]

Last test of the day...

I did a quick test pushing the underexposed ISO 400 image up 0,8 on the exposure slider in C1. Exported the pushed underexposed ISO 400 image as well as a properly exposed ISO 800 image to 16 bit tiffs and opened both up in PS. I did an identical Levels adjustment on both to bring up any problems in the black bag. Check out the trouble in the ISO 800 image, whereas the ISO 400 image holds up much better and is by far preferable to the ISO 800 image (neither is great though....)


ISO 400 pushed +0,8 on top, ISO 800 correctly exposed on bottom. Levels adjustment shown as well:

[attachment=4040:attachment]

[eronald]

Axel,

Thank you very much for posting these tests which are much more demonstrative than my own.

This raises the possibility that the issues with our P45+ might be a feature of every back of this type and not just ours.

To recap, the ISO 800 image shows some strong striping, similar to the issues I have seen in some of my own shots which I consider well exposed. They are there also in ISO 400 but more restrained. As for the color shifts, for street/lifestyle I think these might be dealt with by a special profile or color adjustments, or moving to black and white, but the stripes make a real mess of the image.

As you remark, C1 seems to be applying a strong curve, it might be interesting to try and determine what the "real" ISO of the back is at each setting.

Edmund
 

Last test of the day...

I did a quick test pushing the underexposed ISO 400 image up 0,8 on the exposure slider in C1. Exported the pushed underexposed ISO 400 image as well as a properly exposed ISO 800 image to 16 bit tiffs and opened both up in PS. I did an identical Levels adjustment on both to bring up any problems in the black bag. Check out the trouble in the ISO 800 image, whereas the ISO 400 image holds up much better and is by far preferable to the ISO 800 image (neither is great though....)
ISO 400 pushed +0,8 on top, ISO 800 correctly exposed on bottom. Levels adjustment shown as well:

[attachment=4040:attachment]
[a href=\"index.php?act=findpost&pid=156223\"][{POST_SNAPBACK}][/a]

[Panopeeper]

This is correct, there should be examples of both underexposure by between 2/3s and a full stop, as well as one correctly exposed, and one with highlights clipping in both of the downloads (ISO400/800).

Well, the raw files make quite different impressions; left side: ISO 400, right side: ISO 800 (the clipping point is still assumed to be 65535. The Phase One site states:

full 16 bit image data and Color depth: 16 bit per color


There is nothing even close to clipping, except if the bit depth is in fact 14 - but that's not probable.

Histogram comparison

However, the underexposure is very low here.

[eronald]

Panopeeper,

 I am getting more and more intrigued by the relationship between the strongly underexposed histograms you show and the photographic reality of decently exposed images we are also seeing. There is an obvious disparity here, and  it might be worth trying to figure out what's going on.


Edmund

Well, the raw files make quite different impressions; left side: ISO 400, right side: ISO 800 (the clipping point is still assumed to be 65535. The Phase One site states:

full 16 bit image data and Color depth: 16 bit per color
There is nothing even close to clipping, except if the bit depth is in fact 14 - but that's not probable.

Histogram comparison

However, the underexposure is very low here.
[a href=\"index.php?act=findpost&pid=156270\"][{POST_SNAPBACK}][/a]

[Panopeeper]

There is an obvious disparity here, and  it might be worth trying to figure out what's going on.

Well, Edmund, I have been telling you for days to make a *totally* overexposed shot (white board with +4EV), with different ISOs. This is important to see, if there is a non-linear range before the "ultimate clipping", or if a channel clips at once (there are examples for both).

Furthermore, you should shoot some well-lit scenery with a series of exposures 1/3 apart, up to +3EV compared to what you would do.

You can upload them with "yousendit". You can download Rawnalyze from from here and take a look at them yourself, but the manual is quite old; a new one will be ready in perhaps two weeks.

It is really a strange situation, that you (and not only you) have such an expensive equipment and don't know its limits.