PhaseOne Sensor Plus Observations and Questions

[ondebanks]

Ray

Thanks for that detailed response.  That does help.  I have always assumed the 60 and 80 MP backs from Phase One did not have micro lenses like the older Kodak chip in the P30+.

Paul

Yes, I originally assumed so too. I discussed this here a couple of months ago: http://www.luminous-landscape.com/forum/index.php?topic=82870.msg679019#msg679019

Ray

[torger]

That link says: "Our measurements show that the ISO Sensitivity settings of the camera are real (that is, all ISO Sensitivity settings are obtained by applying a gain, analog or digital, before RAW delivery) for the entire ISO Latitude range of 50 to 800."

Doesn't that mean that the Aptus 75s needlessly throws away up to 4 stops of overexposure headroom (DR) above base ISO? Why would Leaf have taken that decision?

Huh? How do you mean? Do you mean that since gain is being applied the digital range will clip before full well is reached? Ie it would be better to use fake ISO and just scale up, since noise would be same anyway?

[ondebanks]

Huh? How do you mean? Do you mean that since gain is being applied the digital range will clip before full well is reached? Ie it would be better to use fake ISO and just scale up, since noise would be same anyway?

Yes, exactly. Applying pre-RAW gain with no corresponding noise reduction benefit is worse than pointless; it's damaging to what the sensor if capable of.

Ray

[eronald]

Yes, exactly. Applying pre-RAW gain with no corresponding noise reduction benefit is worse than pointless; it's damaging to what the sensor if capable of.

Ray

Has anyone got a link to the actual S/N curves for the 75s?

Edmund

[torger]

Has anyone got a link to the actual S/N curves for the 75s?

I have an Aptus 75 back, I could do a test with ISO50 and pushing it and comparing it to higher ISOs. It would be good to know for me if I can/should always shoot at ISO50. The problem with that though is that images become hard to review on the display as they will be dark there.

For "long" exposures I already today always use ISO50 and underexpose. The attached picture is an example of a 30 second exposure (max) underexposed 3 stops and then pushed, ie "ISO 400", including a 100% crop (processed in RawTherapee, don't remember the settings but I rarely use noise reduction so I don't think it was enabled, except for hot pixel removal which is important for longer exposures).

The main problem I experience is not noise, the CCD noise is nicely random similar to film grain, but that color fidelity/tonality drops like a stone. The leaves in the example picture are like one saturated color of green or red, much less hue variations than for a normally exposed image. I sometimes exaggerate that effect in post-processing and use this limit of the back as an effect.

[eronald]

Are you using a Matrix profile?
Try balancing the channels with a filter - once one channel dies deconvolution goes haywire.

Edmund

I have an Aptus 75 back, I could do a test with ISO50 and pushing it and comparing it to higher ISOs. It would be good to know for me if I can/should always shoot at ISO50. The problem with that though is that images become hard to review on the display as they will be dark there.

For "long" exposures I already today always use ISO50 and underexpose. The attached picture is an example of a 30 second exposure (max) underexposed 3 stops and then pushed, ie "ISO 400", including a 100% crop (processed in RawTherapee, don't remember the settings but I rarely use noise reduction so I don't think it was enabled, except for hot pixel removal which is important for longer exposures).

The main problem I experience is not noise, the CCD noise is nicely random similar to film grain, but that color fidelity/tonality drops like a stone. The leaves in the example picture are like one saturated color of green or red, much less hue variations than for a normally exposed image. I sometimes exaggerate that effect in post-processing and use this limit of the back as an effect.

[torger]

Yes, exactly. Applying pre-RAW gain with no corresponding noise reduction benefit is worse than pointless; it's damaging to what the sensor if capable of.

Thought more about this, and while you are correct, raw converters and formats are not designed to handle an "ISO-less" camera, ie that the only effect of higher ISO is less exposure. The conversion pipeline will then show a darker image.

In order to show a 1 stop brighter image on the camera display when you go from ISO 50 to 100 and you need to scale the data. It would not be particularly hard to have a format with two clip points, ie if ISO 50 is 0 - 65535, ISO100 would be 0-32768 with "highlight recovery range" in the remaing range up to 65535. However I know of no raw converter or raw format that currently has this concept. There's only one clip point, and then if you do digital scaling (which you would if analog scaling would not give you less noise) you need to throw away that top stop. In practical photography that is normally not a problem, and if you really need to keep a huge highlight range than you just shoot at ISO50 and underexpose.

So I do not think it's a strange design, although the "two clip point" design and no scaling would be better for truly ISO-less cameras.

In the future I guess we will see ISO-less raw formats when the ISO is nothing more than a tag to hint the raw converter how to render the data, like white balance is today. But it's not how it's done with current and past formats, as far as I know. Hmm... maybe could be done in DNG with the BaseLineExposure tag, but I don't think anyone's doing it.

[torger]

Are you using a Matrix profile?
Try balancing the channels with a filter - once one channel dies deconvolution goes haywire.

Found the original raw file with processing sidecar file. I'm using a matrix profile yes, I usually do when the exposure or lighting conditions are such that accurate color is not possible anyway. Red and green has been somewhat exaggerated for effect. 30 second ISO 50 exposure pushed about 3 stops, with a custom contrast curve. Unsharp mask is used, not deconvolution for sharpening. Hot/dead pixel filter, otherwise no noise reduction. RawTherapee can display the raw histogram, and it's healthy (ie information in blue red and green channels), but underexposed.

(I'm assuming that applying a LUT-based ICC v2 profile on heavily underexposed data is bad news due to quantization issues, haven't tried that on this file.)

Oboy, I just love to get way off-topic!  

[ondebanks]

In the future I guess we will see ISO-less raw formats when the ISO is nothing more than a tag to hint the raw converter how to render the data, like white balance is today. But it's not how it's done with current and past formats, as far as I know. Hmm... maybe could be done in DNG with the BaseLineExposure tag, but I don't think anyone's doing it.

"ISO is nothing more than a tag to hint the raw converter how to render the data" has actually been done before - the one example I know of is the Kodak .DCR raw format, which was used in two cameras I have/had (DCS 720x,  DCS 645M Proback). To verify this, you have to pull out the RAW data into a format which ignores the ISO tag...I use FITS format, which also has the benefit of permitting floating point precision, facilitating non-quantized (non-integer) calculations, and arbitrary extensions of the intensity range above 65535 and below 0.

Thought more about this, and while you are correct, raw converters and formats are not designed to handle an "ISO-less" camera, ie that the only effect of higher ISO is less exposure. The conversion pipeline will then show a darker image.

In order to show a 1 stop brighter image on the camera display when you go from ISO 50 to 100 and you need to scale the data. It would not be particularly hard to have a format with two clip points, ie if ISO 50 is 0 - 65535, ISO100 would be 0-32768 with "highlight recovery range" in the remaing range up to 65535. However I know of no raw converter or raw format that currently has this concept. There's only one clip point, and then if you do digital scaling (which you would if analog scaling would not give you less noise) you need to throw away that top stop. In practical photography that is normally not a problem, and if you really need to keep a huge highlight range than you just shoot at ISO50 and underexpose.

So I do not think it's a strange design, although the "two clip point" design and no scaling would be better for truly ISO-less cameras.

That sounds reasonable. Or perhaps you could leave the single clip point at 65535 to retain all the highlights, and use curves or single-file-HDR to bring the underexposed rest of the scene up to normal display brightness. That would be my preference. Of course if there's no signal recorded near 65535, you can drop the clip point accordingly, which brightens everything anyway.

Ray

[Daf]

I have an Aptus 75 back, I could do a test with ISO50 and pushing it and comparing it to higher ISOs. It would be good to know for me if I can/should always shoot at ISO50. The problem with that though is that images become hard to review on the display as they will be dark there.

I've run some test month ago with my old Aptus22, and using C1 i get better result by using base iso until 30s...
I mean for exemple 30 sec @ Iso 25 is better than 15s@iso5O and better than 8s @iso 100

But then "past"  30secondes best results are @ 100iso
I mean 30s @iso 100 is better than 30s @iso25+2stop push in c1
or even 30s@iso 100 push by 2 stop in C1 is better than iso25 push by 4 stop.
And also beter than 200 or 400 : mean 30s iso100push 2 stop in C1 is better than 30s @iso400
I don't know why but it's better with and without noise correction.

I haven't tryed with my aptus 7II but i guess it's the same.
So for me until 30s i shoot at base 25 iso then if i need more i shoot @100