Best ISO for IQ280

[chrismuc]

I made four test shot with my IQ180.

ISO 035 - 4 seconds
ISO 050 - 4 seconds
ISO 100 - 4 seconds
ISO 200 - 4 seconds

All same aperture. Stupid indoor shot (it's night in China) with not too big dynamic range.
I open the four pictures in raw digger and get 4 times exactly the same histogram (see enclosed)!

Now you guys tell what that means?
Does it mean the back works at ISO 35 at any ISO setting?

I don't see highlight clipping but maybe my motive had too little DR and I exposed not too bright.

Another thing I don't understand is: If I open all pictures in ACR, I have to apply the following exposure adjustments to achieve the +/- same exposure and histogram:

035 ISO: reference
050 ISO: -0,5 stops (ok)
100 ISO: -1,5 stops (ok)
200 ISO: -1,5 stops (not 2,5 stops like the ISO difference would suggest)

??

And: After adjusting the exposure, the grain in brighter or darker grey values is the same in all pictures, the ISO 200 picture is not more grainy than the ISO 35 picture.

Tell me if I should upload the raw pictures for your examination.

(Or maybe I do new ones with wider DR.)

[voidshatter]

I made four test shot with my IQ180.

ISO 035 - 4 seconds
ISO 050 - 4 seconds
ISO 100 - 4 seconds
ISO 200 - 4 seconds

All same aperture. Stupid indoor shot (it's night in China) with not too big dynamic range.
I open the four pictures in raw digger and get 4 times exactly the same histogram (see enclosed)!

Now you guys tell what that means?
Does it mean the back works at ISO 35 at any ISO setting?

I don't see highlight clipping but maybe my motive had too little DR and I exposed not too bright.

Another thing I don't understand is: If I open all pictures in ACR, I have to apply the following exposure adjustments to achieve the +/- same exposure and histogram:

035 ISO: reference
050 ISO: -0,5 stops (ok)
100 ISO: -1,5 stops (ok)
200 ISO: -1,5 stops (not 2,5 stops like the ISO difference would suggest)

??

And: After adjusting the exposure, the grain in brighter or darker grey values is the same in all pictures, the ISO 200 picture is not more grainy than the ISO 35 picture.

Tell me if I should upload the raw pictures for your examination.

(Or maybe I do new ones with wider DR.)

This is interesting. At least for my IQ260 (firmware pre-june-2014) the histogram starts to get more clipping for highlight ever since ISO 200 and higher (which is consistent with dxomark). For your case your IQ180 seems to behave with no additional highlight clipping even for ISO 200, which is against dxomark/sensorgen. What's the date of release of your firmware? Perhaps if you are running the latest firmware then Phase One has changed the implementation and made the back truly ISO-less to avoid any additional highlight clipping beyond ISO 100?

[dchew]

I will check mine. Can't remember the last time I shot it above 100.
 

Dave

[dchew]

Mine matches chrismuc

FW 5.14.29.

Note that I am on a technical camera with a not-exactly-accurate Copal shutter.













Edited to fix normalize scaling in each graph

[voidshatter]

Dave, could you upload larger screenshots? I could barely see the x-axis!

[Ken R]

Hi Ken,

You are both right… or false.

DxO presents two kind of diagrams one is normalised to a small print size, like A4, that is the one that you show. The other one is unnormalised and shows per pixel performance.

Both diagrams are valid. For instance, if you would compare unnormalised DxO data with spec sheet coming from Dalsa you would find a very good match. But, a pixel-wise comparison ignores the number of pixels. The normalisation that DxO does takes the number of pixels into account.

As you see, the effect of Sensor+ is very significant on the "per pixel (Screen)" graph but far less pronounced on the "normalised (print)" graph.

I would say both diagrams are valid, but there is good reason DxO does this normalisation.

Another small point may be that we often show images on screen, but most screens today are just two MP, 4K is eight MP. DxO-s normalisation would be decent for a 4K display. For full image resolution we need to print, but prints struggle to even reach a DR of 7EV.

Best regards
Erik

Ok.

So given equal print (or display) size one is always going to be downsizing less (or uprezing less) the IQ180 file vs the Nikon D810 file given the large difference in pixel count.

[dchew]

Dave, could you upload larger screenshots? I could barely see the x-axis!

Don't tell me you need more resolution!! Hehehe.

I modified the post with individual png's.

Dave

[voidshatter]

Ok.

So given equal print (or display) size one is always going to be downsizing less (or uprezing less) the IQ180 file vs the Nikon D810 file given the large difference in pixel count.

When you downsample an IQ180 file to the same pixel count as of D810 you gain ln((80/36)^.5)/ln(2) = 0.576 EV increase of DR for the IQ180 thanks to the increase of SNR.

[voidshatter]

Don't tell me you need more resolution!! Hehehe.

I modified the post with individual png's.

Dave

Thanks! Your screenshots look right. If you force the same range of x-axis for every plot then you should see the same exposure and contrast for ISO 35, 50, 100 and 200, but anything beyond 200 would start to clip more highlight. Your results are the same as chrismuc's - all settings <= ISO 200 are the same, and additional clipping of highlight starts from ISO 400.

The turning point at which the additional clipping of highlight starts is different from what dxomark observes (and also different from my IQ260). Perhaps Phase One made a change with the firmware. Do you know the date of release of your firmware?

[voidshatter]

Looking back to my IQ280 shots (firmware pre-june-2014) I had identical exposure between ISO 35 and ISO 100, though I did not test anything beyond ISO 100.

[dchew]

Perhaps Phase One made a change with the firmware. Do you know the date of release of your firmware?

4 Nov 2014.

[ErikKaffehr]

Hi,

The "print" setting i DxO-mark is compensating for pixel sizes, and I feel that is the right approach. Note that it happens to be default in DxO-mark.

Best regards
Erik

Ok.

So given equal print (or display) size one is always going to be downsizing less (or uprezing less) the IQ180 file vs the Nikon D810 file given the large difference in pixel count.

[eronald]

BJL,

 Us dumb guys we look at the Raws, and then our mind boggles, because basically one can set any ISO one wants and measurebate away, the camera doesn't care.

 So ok, I shouldn't have talked about "fake" ISO, I should have called the 100 setting a "fictitious ISO" because they're all fictitious. But then isn't "reality" just a shared fiction?  Anyway, I'd still call ISO 35 here the "native" ISO but I'm not really sure why any more. Your criticism of my careless wording is probably justified anyway, I am getting dumb and superficial.

 I'll give your post a careful reading when I have time, I promise, you put great effort into it and deserve a careful reading. I'm caught up in a hack and can't spare the concentration.

 BTW, if you have an up to date version of the ISO:12232 standard or draft of same, could you email it to me please? edmundronald at gmail dot com.

Edmund

From persistent falsehoods like this about underexposure and fake ISO values, it seems that some people still do not understand that

---snip---

In summary, all I can see that Phase One is doing giving more highlight-protecting headroom than the ISO recommend minimum, while still using somewhat higher numerical raw levels than most or all cameras due to its greater raw file bit depth.

[voidshatter]

As of firmware v1.05.1 (Jan 2015) the IQ250 only starts to clip more highlight from ISO 400, i.e. ISO 100 and ISO 200 are the same.

Claff PDR attached below (note that the x-axis is manufacturer's claimed ISO, not measured ISO)

[Ken R]

As of firmware v1.05.1 (Jan 2015) the IQ250 only starts to clip more highlight from ISO 400, i.e. ISO 100 and ISO 200 are the same.

Claff PDR attached below (note that the x-axis is manufacturer's claimed ISO, not measured ISO)

So the PhaseOne IQ250 is significantly better than the 645z?

[ErikKaffehr]

Hi,

Interesting to note that the DPR curve of the IQ-250 differs so much from the Pentax 645Z, as bot use exactly the same sensor. As a hint, BClaff indicates that the ISO-s on the X-axis are rated and not measured. So my guess is that ISO on the IQ-250 is a bit like IQ-180 and on the P645+ it is in accordance with the saturation based ISO standard.

Best regards
Erik

As of firmware v1.05.1 (Jan 2015) the IQ250 only starts to clip more highlight from ISO 400, i.e. ISO 100 and ISO 200 are the same.

Claff PDR attached below (note that the x-axis is manufacturer's claimed ISO, not measured ISO)

[BJL]

So you've reached a conclusion that Sony is falsely advertising the native base ISO for the A99? They advertise for 100 but apparently they are deceiving and they should modify it to 50, is this correct?

I see what you are referring to now: the Sony A99 offers exposure index settings ranging from 100 up, whereas (if I am reading DXO's results correctly) the ISO SSat recommendation for the MINIMUM safe exposure index of the sensor is about 48. So Sony is giving four stops of highlight headroom above the midtones placement when auto-exposure is used instead of the three stops suggested as the bare minimum by the ISO standard with its SSat measure.

So I see nothing "false" and no "deception" there, just a perfectly reasonable and allowable design decision to offer more than the bare minimum highlight headroom on a sensor that has plenty of dynamic range to spare.

To repeat yet again: the ISO SSat is intended as a bare minimum Ei (maximum advisable among of exposure to the sensor) not an ideal exposure placement!  


Aside. There is an irony here: one common if misguided criticism of digital cameras is that they have more problems with blown highlights (while handling shadows far better.)  But when a camera maker sets AE in a way to improve highlight handling (while still having far better shadow handling than film), some people criticize this as deception or fakery.

[BJL]

So ok, I shouldn't have talked about "fake" ISO, I should have called the 100 setting a "fictitious ISO" because they're all fictitious.

So long as we thing of "ISO" meaning what it meant for film, yes the ISO settings on camera and the ISO SSat (aka "DXO ISO") are all fictitious; the measure completely different things.  ISO:12232 recommend using the noise-based S40:1 measure as the "ISO sensitivity" but for whatever reason I have not seen any camera maker do that!   It might be the case that when some cameras have "low" and "high" ISO exposure index setting outside the normal EI setting range, the former are ones low SSat and the latter are ones above some noise-based threshold like S40:1 or S10:1.

BTW, if you have an up to date version of the ISO:12232 standard or draft of same, could you email it to me please? edmundronald at gmail dot com.

Unfortunately not, since the ISO charges way too much for it.  I have gathered multiple summaries from other sources (starting with the Wikipedia article that I often cite).  I do have a link to the Japanese CIPA DC-004 standard that specifies how "ISO" settings on cameras must be calibrated, at least on Japanese cameras: it's at this page: http://www.cipa.jp/std/std-sec/std-list_e.html
This originated the "Standout Output Sensitivity" part added in the 2006 revision of ISO 12232, and gives some rationale for that addition.

[voidshatter]

So the PhaseOne IQ250 is significantly better than the 645z?

Of course no. The IQ250 is exactly the same as the 645Z. It's just Phase One's "clever wrap" to make the plot "look better". If Phase One follows Canon/Nikon/Sony rules then the IQ250 would overlap with the 645Z and the IQ260 would "look worse".

But of course it's good to know that I don't need to shoot the IQ250 at ISO higher than 200 if I don't need to check focus/composition via playback - digital amplification wouldn't squeeze any more DR out of a Sony CMOS (in contrast with a Canon CMOS with 500nm fab process and external ADC presumably made in China).

[ErikKaffehr]

Hi,

Both cameras use the same sensor.

Please note the statemnet: (note that the x-axis is manufacturer's claimed ISO, not measured ISO)

The IQ-250 is probably overrating ISO like the IQ-180 discussed before. As discussed before, this can be seen as a way of protecting highlights.

Best regards
Erik

So the PhaseOne IQ250 is significantly better than the 645z?