Problems with a H6D 100c and Phocus

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Tangomaniac

Hello,
I'm importing my photos (taken with a H6D 100c) from the CFast-card into Phocus. I used 3.14 and I have run into some problems:
I like to take available light photos, so this let me use ISO 128.000, and even this was at (or beyond) the limit, since the light was really low.
1. The light was so low, that the autofocus didn't focus any more. (OK: I can do this manually, but this was nevertheless disappointing.)
2. All pictures show different brightness in the upper and the lower part, the division is a straight line exactly in the middle. This effect is reproducible with all pictures taken with ISO 128k and low lighting. Its coming out more clearly, if the EV in Phocus is raised. Compare the three attached pictures. It's looking like one color channel (green) is one bit or brightness-step off in the lower half. Therefore the lower half of the picture is always brighter and has in color a magenta hue. (See third photo!)
My question no.1: Can anybody of you reproduce this effect?
Question no. 2: If not, could it be, does this mean: My sensor is faulty or has a problem? Or could this be a software problem of the sensor or Phocus?

3. When I'm using Phocus on these photos, and I'm using "Highlights" or "Clearness", I get an extreme smoothing effect (like a really exaggerated noise reduction). - Compare the first and second photo! -  Only if I put all the sliders back, this effect vanishes, but not totally.
N.B.: I have noise reduction in Phocus switched completely off.
But even when I have the noise reduction switched off, there seems to be a certain amount applied always. (Using the loupe, I see a quite noisy but clear picture, which is smoothed after some seconds.)
Any hints, how to avoid this?

Best regards,
Ralf

rem

Her the same problem. Sometimes more, sometimes less.

24mm no crop t1/350 / f9.

ChrisL

#2
This is a classic tiling/sensor readout issue.
Two things cause it: The sensor in manufacture is exposed to create the circuits in discrete areas so you have a series of rectangles that make up the surface, variations in this exposure result in edge effects where the exposures meet and produce a tiling pattern, secondly the sensor data is read out through two channels for speed which is the split you are seeing. ( I assume you straightened the first frame posted) The calibration of the back at assembly allows the data to be flagged to match the two halves in the final RAW file. This becomes visible when you push the sensor limits, as you have with exposure, or with shifting lenses, or with pushing the contrast particularly in B/W if the match is not good enough. If you aggressively push the sliders, as you state, then the RAW data is pushed beyond the acceptable limits from a stressed file, a "normal" file would cope with the sliders on maximum.
The only fix is a factory re-calibration which should reduce, but is unlikely to remove completely under really adverse conditions, the issue. It's a result of how large surface area sensors are made and how the data is read off then processed and adjusted in the RAW data.
If you push even harder you may see the individual tiles (I think your second B/W frame is beginning to show one top RHS which is darker and looks unlikely to be natural illumination variation) CMOS backs are less prone than the older CCD backs to see this effect, the read out problem remains.
You can reduce the problem, and often eliminate it, by shooting a lens cast correction frame under the same conditions, I use Capture One and am not familiar enough with Phocus to advise.

This is a DALSA chip showing the tiles, which are read off in two sections top and bottom rows, hence your perfectly straight line in the exact middle of the full frame,I do not own the copyright but Google seems to not attribute any photographer and I would claim fair use to illustrate a technical point.


Tangomaniac

#3
Hello ChrisL,
thank you very much for your explanation! This explains the horizontal division in the pictures completely!
(Yes, the first B&W-picture was realigned a bit.)

Could it be, that with the sensor of my camera just one color-channel is not calibrated well enough?
See the attached picture, which shows the effect really clearly. - The lower half of the pictures is never black, but strongly purple.
I'm seeing this strong effect consistently when using ISO 128000, but it is visible (although much weaker) when using 6400 ISO in low light too.

My second question is still open: Could anyone on the forum please try to reproduce this smoothing/blurring effect (like a really very strong noise reduction) seemingly caused by Phocus when adjusting the exposure sliders strongly?
At first I thought, I forgot to switch of the noise reduction, but I've tested it in the meantime at least five times, and the effect is reproducible with switched off noise reduction, especially if using a slightly underexposed picture:
Just crank up EV, clearness and highlight-recovery. (The last one causes the strongest smearing effect.) - Pictures become really unusable, smeared out like with a really strong soft filter.
Can anybody reproduce this effect?
Or am I doing something wrong? - I'm using Phocus only for half a year, so may be I'm still a DAU (dümmster anzunehmender user = silliest imaginable user).

Thank you!

Ralf

ChrisL

Quote from: Tangomaniac on June 05, 2017, 08:37:23 PM
Hello ChrisL,
Could it be, that with the sensor of my camera just one color-channel is not calibrated well enough?
See the attached picture, which shows the effect really clearly. - The lower half of the pictures is never black, but strongly purple.
I'm seeing this strong effect consistently when using ISO 128000, but it is visible (although much weaker) when using 6400 ISO in low light too.

Thank you!

Ralf

Hi Ralph
I don't know enough about the calibration to say if it's one colour channel but really it doesn't affect what you need to do about it, it can't be "fixed" in post processing it needs re-calibration. I would send Hasselblad some files with all the details you have posted about sliders and settings and seek their advice, if this is too far from expected behaviour. I know in the past they, and Phase One, have said if you push enough you can still see some effect even with calibration as good as they can get it. Certainly for a CMOS back I would expect better but you are at 128000 and low light so the internal circuits are struggling to produce a signal high enough above the noise floor of the electronics for a  medium format sensor. Obviously there are full frame 35mm sensors that are able to cope more effectively with low light shooting, you are pushing the limits of this sensor but I still think it should be better than you are seeing, but ask Hasselblad.
On Phocus I cannot sensibly comment so I won't.


rem

This is the answer we received from Hasselblad to this problem. I'm not very satisfied. What you think?

"The system should be free of "split" up to 1600 ISO, it may be possible to provoke at 6400 ISO, although not clearly visible with standard adjustments.
It is impossible to prevent at 12800 ISO. The problems occur due the physical sensor layout from Sony and we have confirmed that PhaseOne suffer the exact same problem at high ISO.

The sample you've provided seems quite severe and i would need the RAW file for proper analysis."

Tangomaniac

Hello rem,
thank you very much for getting this answer from Hasselblad.
I've contacted Hasselblad myself and sent them three RAW-files. I'm now waiting for a reaction; possibly I'll have to send in my camera.
I'm really astonished: If this "split is unavoidable at 12500 ISO", then 12500 ISO is one ISO step to high, one more than the H6D delivers, and 6400 ISO should have been the highest value possible. (The difference between advertising and delivering seems quite obvious here.)
And this is truly disappointing.

I'll let you know, whether there will be an improvement after recalibration. (In my eyes, an improvement should be possible.)

Thanks again!

rem

Tangomaniac, I think the same like you! That's really the wrong answer! Because I like sometimes to use the 12'800. For me are the higher ISO's the main reason to buy a H6D to the H5D. Marketing and reality... Split free til 1600!??? I do expect more from Hasselblad. Mine is not yet here. Have a great day! :-D

Tangomaniac

I only saw the split in one (underexposed) picture taken with 6400 ISO.
Not at all with correctly exposed photos with ISO 6400.
I'll have to recheck the 12500 ISO pictures. I hope: If you expose correctly, you won't see the split.
So - may be - the problem is a bit less severe, than it sounds right now.

But to be honest: I am  disappointed. I was using a CFV-50c before, which is really good. So good, that I started again with Hasselblad and found all the joy I had 30 years ago in taking an making pictures again. And I hoped to get one EV value more without discussions or caveats with the H6D.

So far, my experiences in really low light situations are not so great: I have a non-working autofocus and the split problem. (OK: I expected the autofocus-problems because of the published technical data, so no complaints in this regard from my side. But I would like to ask Hasselblad: Could you make an H7D, where autofocus works with two or three EV values less? - This would be really great.)
And I have this split-problem, which was documented nowhere.

I would much prefer, if Hasselblad would publish the limits of the shooting envelope of all its cameras more clearly. (And what happens, if you transgress them.) Mustn't necessarily be very open, but at least somewhere in the small print in the technical data, so you can find it.

Hasselblad's H cameras (at least the newest models) are primarily used as tools by professionals, and it can really become very expensive to run into a hard limit, the existence of which you didn't even guess. 
Especially after you payed the price of a quite nice car for an new H6D.
I'm definitely not a pro, but the disappointment is the same.
And the disappointment(s) that neither the CF-converter nor the GIL work with the H6D as they are supposed, are adding up...

stephanbruehl

Quote from: Tangomaniac on June 14, 2017, 09:31:25 PM

But to be honest: I am  disappointed. I was using a CFV-50c before, which is really good.
...

I had the same problem with a H4D-60 (CCD) three years ago. And it appeared even under low ISO settings (100) if I tweaked the contrast very much. I had my camera recalibrated and it was better then but hadn't disappeared totally. As explained by ChrisL the effect is due to the various sensor partitions.

This was one reason for me to switch back to the smaller sensor area and exchange the H4D-60 for a H5D-50c. My new camera as well as your previous CFV-50c has a single piece sensor with no partitions so that edge effects cannot occur....

Regards, Stephan