PureRaw + Fuji X-T5 + Lightroom = purple in highlight

DR100/200/400 on fuji is a very impressive funcion. It allows to shoot with an highest ISO value (so the camera compensate with SS and/or aperture), and just before creating the RAW file the camera decrease ISO ONLY in hightlight.
So Ihave almost 2 more stops of dynamic range.
Of course, there is theorically more noise in shadows, but in DR400 the minISO is only 500 on the X-T5 and with DEEPPRIME it’s almost invisible.

One more time, the main question is not “how to preserve highlight ?”. I think I know how.

Oh i didn’t imply that :blush:, i was just interested if the fuij had the same kind of feature.
Often they do, and often it’s function isn’t completely written down in the manuals.
This was side tracking about how camera’s work in the highlight section of a image.
General interest in technical stuf. You own one i don’t so i can’t test this myself. :slightly_smiling_face:
(At the same time this can be in general manner interesting. Because of how this is flagged in the rawfile. Does it alter capture values or does it only alter the actual exposure moment?
And (how) does this influence the demosiacing algorithm in dxopl?

I don’t examin or question your skills nor knowledge about photography.
I don’t know you so i have no clue which level you are skilled.
Apology if my writing implied that thought.
:relaxed:

No problemo :smiley:

If you are interrested, here 2 RAWs. One in normal, the second with DR400.
DSCF5534.RAF (35,0 Mo)
DSCF5532.RAF (36,0 Mo)

Reduce exposure to -5 in LR on both and watch the magic (in any case for me it’s impressive)

1 Like

This is for others you know already :wink:

Interresting, it looks like it uses the 1/3 stop exposure algorithm of 500iso( less exposuretime to compensate 1/3 isostop) for 400iso as in sensor sensitivity, and we know a sensor doesn’t change much in sensitivity only is a bit optimised for low light when you crank up de iso value so that’s not the case. But the shuttertime is the same. aka same amount of exposure.
the 5532 seems to me the normal:( in -3ev it’s still overexposed in outside view.) but has the least amount of under exposed area’s less then 2%/channel)

the 5534 has much more “under exposure” signs 42%/13%39%
but at the same time the r,g,b channels are much earlier and more raised in the histogram indicating more details to see indicating the shadowpart of the image is boosted or more likely the highlight is dropped so the shadow looks more exposed. confusing.


inside dxopl:

highlight preservation in the 5534 file.

i see a bit more “blue” in the shadow around bonheur and windowdraping which shows they are not the same file.
So any issue in noise? ( screen dumps of dxopl preview so no deepprime)
los of detail in the shadow?

nearly no difference!
impressive.
based on this 5534 is the DR400 file right?.

Tomorrow when the sun is shinning i try the same with my good old G80…
multi meatering right?

It’s in Dutch. Just search on fuji dr400.

George

1 Like

wel they say 2 things: (note: they(DxO) are cooking a cure so lets entertain our selfs in the maintime :slight_smile: )

  • underexposure of max 2 stops by 400DR and raise shadows and highlights again wile fiddling with the tonality/curvetone.
  • And it has no value for RAW shooters because you effectively underexpose max 2 stops.
    which is strange with these two images provided by @Merlinx63 with the same shuttertime and Aperture aka exact the same amount of exposure. (only difference is the 1/3 stop isovalue which is part of the exposure triangle ( i don’t start up the old discusion about this by the way.))
    So it does something with the mapping from charge to r,g,b data i think.

Against idyn which doesn’t need to raise iso value ( 200 iso is base)
there it seems to be shuttertime which changed.
( i have to check tomorrow one active one off.)
how effective that is.

That Super CCD is great the G9 has HighRes which uses multi exposure to combine two images (sensor shifting) for higher resolution so that system could be used for HDR-rawfile. (combination of electronic curtain and mechanical.) :thinking:
open mechanical and electronic, readout electronic for highlights( accoording to amount of stops needed),close mechanical shutter for shadows. Then combine those rawfiles in one HDR one.
:slightly_smiling_face:

and i am off to file this idea :stuck_out_tongue_winking_eye:

i was looking for the article about panasonics idyn and stumbled on this:

whats iso about part one what it’s place is in the exposure till exported jpeg is and rawfile.
if you follow the words you get to this part: DR-modes
explains the technique behind it.
This why the image has no noise degradation in the shadowparts:

Because there’s no connection between ISO and amplification, and because the sensors it uses are highly ISO invariant, Fujifilm is able to offer a series different ISO modes with the sensor’s amplifier in its ‘base ISO’ state.

and also interresting:

This can be useful for Raw-shooting photographers: the DR200 and DR400 modes essentially let you expose one or two stops to the right (ie shifting exposure to include highlights that would otherwise be lost), while maintaining a comprehensible preview image. Snapping a quick DR400 mode image lets you check which additional highlights will be captured, without the rest of the image becoming too dark to interpret.

(stil not found the panasonic idyn article.)
update found that article also. :slight_smile:

somewhere down the stream of words:

Panasonic’s iDynamic and Olympus’ Auto Gradation modes will both reduce exposure by a 1/3EV or so to capture some additional highlights, then use an adaptive tone curve to brighten the shadows and as well as adding highlights.

has it anything to do with the topic starting subject? Nope. But still if you know how it works under the hood you can benefit from that when you are taking photo’s. :slight_smile:

even DxO acknowledged the bug in the code for X-T5 …" but why you would assume it is PL at fault" :joy:

you need to take X-T5 raw file with an area where all channels are clipped in raw, convert it into linear DNG file ( w/ only optics correction and NR applied ) in that DxO PL6 version with a bug, open it in rawdigger and there you will see that DxO maps clipping too far away (57837) from declared ( through a tag ) whitepoint (65535) for ACR/LR to start using their “magenta skies” prevention when brightness correction is pulled back … ACR/LR rightfully see that there is NO clipping channels in linear DNG - the bug as acknowledged by DxO is on their side

that is why people use tools like rawdigger to find the bugs …

So maybe DxO should include a rawdigger licence with photolab to help users :wink:
Would often be usefull :wink:

no, please refer to DNG Specifications… software producing linear DNGs must scale data & tag linear DNG according to the standard, period … DxO PL6 did not, DxO acknowledged the error with how they deal with X-T5 raw files

no, DxO shall pay more attention to basics of software development and testing … it is hard to believe that after so many years ( this is not something invented in a recent DNG v1.7 specs ) the piece of code does not check data scaling vs a specific white point tag for linear DNG … I mean this is just code it once following the formula clearly provided by Adobe and then just call that code everytime

as I told you I never said there was no bug.
I just said that scaling datas inside photolab with curve tool to get values in DNG between 0 and 65536 was probably the best fix for it.

@noname
probably because we’re talking here about values (brightness), but not about colors.

no, the best fix is for DxO to fix the code… and using “curve tool” as a fix works only if the topic poster agrees to deal with linear DNGs with other things baked in ( and curve tool corrections will NOT be the only thing baked in - color transform /that is a camera profile/ and white balance will be baked in, etc ) … it is his call but one can assume he wants linear DNGs only with optics corrections and NR for a reason

Of course …
curve tool does not fix code. Seems so obvious.

But when you’re in production and have a problem and go to a forum (as I often do when things goes wrong), you generally need a quick fix (or workaround) against your bug.

I assume you mean “DR” by “dynamic” ?

it is not… you have less than 16 bit values from original raw (12, 13, 14 bit - whatever it is ) scaled to 16bit… so it does not materially matter for precision if max is 57837 or 65535 ( you are not losing anything from the original data ) … but it matters ( along with the relevant white point tag ) for ACR/LR to see that there was a clipping in raw and engage “magenta skies” prevention when brightness is reduced by operator in ACR/LR

still do after DxO acknowledged this error ? just curious …

you’re right.
Indeed 14 bits is generally the max sensor dynamic (the bests in photography cameras) - and even not fully used when not at base iso (except for some high-end sensors - used in cinematographic production only for now i think - unless mnufacturers are lying - I have to carefully choose my words with you :wink:).
So yes 57837 is enough to keep full sensor dynamic.

you do NOT understand the basics of raw converters operations for start ( as in TOTALLY NOT ) - ACR/LR do not change /but read further what happens/ “relation between the channels” at all ( because the linear DNG presented to ACR/LR does not have any clipping as written by DxO PL6 and so ACR/LR does not engage in special prevention of “magenta skies” effect )… but ACR/LR do apply white balance ( now try to engage some grey cells you know where ) … and when ACR/LR has equal values (57837, 57837, 57837) in all 3 channels they are scaled differently through WB application … daylight-ish WB means BLUE and RED channels are multiplied by some coeff = 1.x ( while GREEN is not multiplied at all )… and what BLUE + RED is ? MAGENTA… that is why “magenta skies” effect appears… but when ACR/LR see clipping they engage additional correction to prevent WB application to result in that effect ( and this is why a bug in DxO PL6 prevents ACR/LR to properly operate when brightness is reduced )

capisce ( capiche, capeesh, capisch, capische, capishe, coppish, kabish ) ?