So, after reading so many posts where ppl are saying DBs are terrible at high ISO and how they are afraid to shoot at anything over 100, I decided to show what it can look like if you process them correctly. These are shot with a regular P25 (non +) at f5.6 and processed in C1 3.8 with ONLY color noise suppression applied. So there is no luminance noise suppression applied at all. Mind you, this is supposed to be the the worst of all the backs at high ISO.



ISO 200:

Click to view attachment


ISO 400:

Click to view attachment


ISO 800:

Click to view attachment

It would be interesting to see them without massive compression and noise filtering applied. The 800 looks pretty dreadful in the boke areas and on the pots, but I'm not sure if that's the software messing it up or it came out that way.

I see banding and lots of "grain" at 800. 200 looks pretty decent, though.

I expect the Mamiya ZD would be worse.

Dreadful? What a joke. You could print that A3 easy without seeing a single noise grain. As far as compression goes they are 50% size 70 quality JPEG. But whatever you say. The point for me with this post is not really to debate it, but to let ppl see what it can look like and let everyone make up their own mind.

Agreed. That noise might be discernible in a print, but not in a bad way.

The noise in the 800 image would never show on a web press. I really like the noise structure in MFDB files, and frequently try to sharpen the luminance noise so that it will show in a print.

I'm super impressed, I never shoot anything other than ISO 200 on my P21 where there is virtually no noise at all, even on a tripod I wouldn't bother shooting at 100 as there is really no advantage you can see in print. I did do a quick dirty test on 400 the other day and that was really good too. As you've shown with your test, by the time you get to 800 theres plenty of noise but it's a nice noise and it still has that lovely thickness and sparkle to it.

hmm, sounds like a good trick, i'll try that.

The trick is to use C1, ACR does a much worse job with color noise. It's obvious PhaseOne has some kind of custom noise reduction for their backs implemented. And set the color noise slider function to manual and play with the slider till you're happy. For iso100 & 50 I usually use ACR though.

Lightroom is really good at ISO 200 on the P21. The C1 version 4 does do better, it's just I can't control color like I want to on C1 which is a pity because the quality of the conversions is really great.

Another side

Any comparison requires at least two participants. Here are examples from the lowly (but not the "lowliest") Canon 40D (a 10 Mpix camera for $1000), ISO 200, 400, 800 and 1600. Shutter between 1/20s (ISO 200) and 1/160s (ISO 1600). Low color noise suppression on 800 and 1600. The JPEGs are full size, from 2.5 MB to 4 MB (the noise of the 1600 shot is observable on the file size as well).

ISO 200 JPEG

ISO 400 JPEG

ISO 400 JPEG

ISO 1600 JPEG

However, I do not trust JPEGs, they do not reflect the camera's capability, for

a. JPEG reflects already the raw processing,

b. the exposure counts more than the ISO in noise, and that can be seen only on the raw image.

So, here are the raw files as well, 11MB to 13MB.

ISO 200 raw

ISO 400 raw

ISO 400 raw

ISO 1600 raw

As I understand this thread, this is not a comparison. We're talking medium format digital noise here. Nothing to do with 35mm.

Really? The OP wrote



Do you think "ppl are saying" that DBs are terrible on their own, or in comparison to other cameras?

As I posted yesterday, I think that MFD noise looks terrific when properly processed. Very film-like.
This is how my P20 looks at ISO 800 with a bit of NR. Obviously this can be worked out more...



/Samuel

I don't think there's single MFD user here interested in a noise comparison against 35mm at this point.

My vote is still - keep this thread about MFD noise only. Thank you.

There's two types of behavior under different conditions: LotsaLite and BadLite.

Usually LotsaLiteis ok at hi-ISO, because the spectrum is well balanced (sun, flash).

But BadLite is low on blue or spiky and gives strange chroma noise.

Edmund

I want to than AMSP and Samuel for bringing this up. I've just shot some ISO 800 on my P20 today tethered with C1 4.1 and was amazed. I didn't think my p20 could do this well.

In my tests I used a flash, my 150mm TX lens and 1/1000 C1 4.1 does do a slightly better job with the color noise and indeed as ASMP has suggested using higher chroma noise reduction and low luminance works well. I'm impressed!

All my previous tests with the p20 and ISO 800 were with slow shutter so I'm going to do more tests to see if ISO 800 with slow shutter (1/60) are just as good.

Update: The quick answer is no they are not. And the longer the exposure the bigger the differential between base ISO and ISO 800 gets. So for me I get the quality ISO 800 files that ASMP and Samuel posted if I am using fast shutter and small aperture.

Another interesting thing I think could be explored more is noise in OOF regions. It appears that this is where the noise gets out of control. If true then why?

What Edmund said is key to the "useability at high ISO" question:

- Shooting with plenty of light will look good on most cameras, low or high ISO; high ISO well exposed can look terrific. This is clearly seen here in amps example.

- Once you get into low-light levels, particularly tungsten-based lighting, the high-ISO challenge hits you hard and fast. This kind of scenario is the proof-of-the-pudding. Chroma noise can get very ugly. Sometimes black-and-white conversion is the last resort.

Peter

Ah. So it is the software getting in the way as I suspected. Fair enough.

I do like how the artifacts and noise are very evenly spread out, very much like pushed film. This is nothing like you get from a typical DSLR. Probably because it has no AA filter on it and so any AA or NR you do is a one-time thing when you process it.

Does it look "good" at 800? No. 200 plainly looks better. Does it look tons better than any DSLR that I've seen at 800? Without a doubt.

C1 3.6.x and later are the king of chroma noise removal, IMO. I think version C14 is terrible with color noise, and I have to use other software to remove it from the converted 16 bit TIFFs, if I were to use version 4.

I think Lightroom is tied, or second best with color noise removal and control. Plus, it just gives you clean files, no matter what file you're processing, however, sometimes you need to add in a fine edged sharpen after converted from LR, where in C14, it gives you the control for the radius, amount, threshold, etc, more so than LR does.

Absolutely right, Edmund: there is a huge difference at high ISOs when LotsaLite or BadLite.

I refer to this situation, taken at ISO 800 with the eMotion 75:

http://forum.getdpi.com/forum/showthread.php?t=1683&page=7

There was also light in this situation, but not "Lots", as it is shot in the early morning during sunrise.

Thierry

Really! MFDB users pay amazing sums of money for their equipment because they want the best possible image quality, whatever the ISO used.

The fact that 35mm DSLRs seem to do a better job at high ISO (albeit starting off from a lower quality base) must be of concern to users of MFDB.

From reports I've read from Edmund, Phase DB's produce equally good results, if not better results, when underexposed at base ISO instead of the option of using the same shutter speed at a higher ISO.

Canon DSLRs, on the other hand, produce significantly better results at a higher ISO, compared to the same exposure at a lower ISO. That's the key issue here, for me and I would think for anyone interested in image quality above base ISO.

Why aren't DB manufacturers using the same technology that 35mm manufacturers use to boost the analog signal before A/D conversion? My guess is, because the CMOS sensor makes this possible and most (if not all) DBs are CCD.

My impression is that the actual sensor used are pretty good with respect to noise at high ISO and should yield good images when scaled down a bit, due to their large number of pixels. However, the camera makers seem to be using chips which are out of spec (cheaper), and correcting this by software, but the software corrections don't work well at high-iso.

Edmund

Please don't feed the troll's.

Here is the same iso800 image with Neat Image noise removal applied...

Click to view attachment

I prefer it with a bit of noise I think. I always add noise to files anyway.

Yeah, I sometimes remove the digital noise if I think it's ugly and then add fake film grain. It both hides the noise reduction and adds more pleasant grain.

This can not be stated so generally. For example the Sinar eM22 has only one ISO, which is I guess about 40; everything else is software.

Thanks for the test. Please are you talking ISO or are you talking exposure? Its a significant difference. Do you expose to the right or do you let the meter decide how you expose.
ISO is a standart for film/sensor sensitivity. Using the "expose to the right" technique would eventually result in an overexposure of the sensor or in other words by ISO standart, you are using a lower speed than what your back tells you.

It is also difficult to judge correctly on a downscaled image and as has been said its a good light shot. You can certainly print it to A3, would it look different from an image taken with a APS camera? Probably not!
Please keep going with your testing, I am not yet confinced that MF backs are able to deliver the image quality of a top notch APS camera at high ISO

Woof, that is really nice looking..
I will have to go back through the thread to see but I think your camera was the P25 non+ series, correct?
I would never expect that at 800.
I have a P30 and it is suppose to be even better at higher iso/asa than the P20 or P25, as far as I have read.
Looks good
Snook

It's actually asmp's test.

No offense, but I'm not even remotely interested in convincing you of anything. This post is a nudge to DB owners, who have been afraid to shoot at anything but iso100 in the past, to go try it themselves. If you can't judge from a 50% of a 20MP image that's your problem. And if you think these images would print the same as an APS sensor at A3 you are being quite ridiculous and obviously have no real experience on the matter. For all you pixel-peepers and whiners, if you are not happy with this test go do your own, it's that simple. Thank you.

Word.

Come on! I made the test some time ago with the P30. A 5D had the better IQ at iso 800 than the DB. The + is certainly better though.
And sorry I dont see why you name your thread a test and show us a downscaled image, but then again expect to read comments about the full size image quality.

I would be far more appropriate to answer to questions than to ride personal attack against posters. Have fun

At least his showing images. You've got nothing more than words...
It would be nice to see your tests...

I do much the same. But I also tend to add a touch of grain as I'm not so keen on digital's smooth video look.

Mr. Horrible,

I'd like to see your test images as well not because I don't believe you but because IQ is subjective. Some people like sterile, (over) saturated images, some like color accuracy, some people hate colour noise. I can't stand color noise but love luminance noise on a high pixel image, kinda like pointalism in the OOF areas, and I require good color. The main advantage of a DB is in color. Even pushed images at 1600 on a DB have much better colour accuracy than any DSLR. Any colour noise cleans right up with a little colour NR, leaving the luminance noise to add to acutance. I know the 5D well and can say that 800 is great noise wise, but colorwise is just OK. This is my experience, I'd like to see your pics to see what you are seeing.

*Yawn* This NOT a DB vs. X thread. Why is that so hard for you to understand? This is just a couple shots to show ppl that DBs aren't necessarily as bad as you might think. I don't care if this test is not up to your standards, am I being clear enough? The fact that a couple of DB owners have already expressed their surprise at the results tell me that my intention with this post got through. But feel free to go start your own P30 vs. 5D thread, just stop posting your nonsense on mine.

Yes I know, I did not bring in the 40d or whatever, I just shared my experience in response to your complaint.
Its about high ISO, and the question was if the light were metered acording to ISO specs or if you simply exposed to the right.
I dont understand what so offending about this question. As you name your pics a test, it may be allowed to ask, otherwise it could give the impression that you want to proof something which isnt there.

Here's a shot from this afternoon with the Rollei 6008 / p20 handheld at ISO 800 forget the shutter but around 1/125. I made no adjustments to the file other than changing the color noise to 71 and the luminace nose slider to 6 (defaults are 58 and 25) in C1 4.1

Honestly I am really surprised! I had been holding off from even ISO 200 before seeing this post, and getting blurry images because I was afraid to bump the ISO.

Looking great! I'm glad my post could inspire you to start experimenting. Did you try and develop it in C1 3.8 too? If not, you should try it and see which one you like best. I played around with your photo a little, hope you don't mind. It's quite a nice capture, I especially like it in B/W.

Click to view attachment

Click to view attachment

Please show me the DSLR which produces better IQ than this ISO 800 sample from a MFDB.

This is at 100% which is not a true comparison, because the image size of the MFDB is so much larger:

Click to view attachment

And this is a 100% crop when reduced to 16MP (more like a high end DSLR):

Click to view attachment

NO noise reduction used in either sample.

Honestly I couldn't ask for more at ISO 800 and I hope this kills the prevailing "MFDB can't compare to DSLR at high ISO" myth. It used to be true but not any more.

I just got back from doing my workshop in San Juan and Lance from Capture Intergration was there with all the backs for me to play with and our attendees. These are not hard core tests but i can assure you the P30 Plus can do high ISO' very well , the whole fashion shoot we did I shot at ISO 400 and there very clean plus some other casual testing. Read this thread

http://forum.getdpi.com/forum/showthread.php?t=1896


I also got the chance to shoot the P25 NON plus and again ISO 400 was pretty darn good and I expect better from the PLus back which BTW i am sitting here waiting to be delivered . I bought the P25 plus with the new Phase Body and Lens plus the Horseman SWDII with 35mm digitar. Feel like a kid in a candy store but i was really impressed with the high ISO's even from the P25 and I want , need a back that can at least give me a clean ISO 400 but the P30 plus was even pretty clean at 1600 and maybe with some NN software it would clean up better or better with working with C1 with the correct settings. I was more guessing on what those maybe. I will do more formal testing when it comes . It was hard doing hard tests when your teaching a workshop but it was enough for me to buy.

here is the P25 thread

http://forum.getdpi.com/forum/showthread.php?t=1921

As I'm sure everybody else did too, I did a lot of research before buying a digital back, and based on what I had read, I felt that I had been led overwhelmingly to believe that ISO 400 was so horrendous (see also the ISO 400 shot in bright daylight in the Luminous Landscape review of the older Ixpress 96 back) one might only use it in an emergency. So I braced myself to not be able to use ISO 400 unless absolutely necessary. However, when I got the 96C version of the back finally, and did my own testing, I have found the results to be quite usable and not at all as bad as I imagined it would be. Presumably, between the writing of that review and now, there have been numerous firmware updates, which have undoubtedly improved IQ. Here's a 100% crop taken at ISO 400. The light was window light only, and I was forced to shoot at lower shutter speeds and higher ISO than I would have liked, but I am pleased with the result.

Bronica Zenzanon-S 150mm f/3.5 lens & Ixpress V96C @ ISO 400, 1/60th sec, f/3.5, mounted on a tripod. It was a little tricky as he was talking while I was shooting. Sorry for the rather stern look on his face and the nose hair is certainly not as appealing as Frank Doorhof's models, but the stern look seems to capture this guy's essence...
Click to view attachment

(Edit:) I forgot to mention that I didn't do anything to the shot other than adjust black point slightly, output to JPEG and crop then post. No noise reduction or anything. FlexColor 4.8.4

Actually, ISO is an organization (the International Organization for Standardization?) that produces a great many standards, including
1) A scale of units for describing Exposure Index, used for example for both items (2) and (3): roughly a measure of what shutter speed goes with a particular combination of light level and aperture ratio to get appropriate mid-tone handling.
2) A film speed measurement (adapted from ASA/ANSI standard) based on adequate _shadow_ handling: roughly the _highest_ exposure index at which shadows are handled well enough.
3) a "base exposure index" definition for sensors, based on adequate _highlight_ handling: roughly the _lowest_ exposure index at which highlights are handled well enough.

These days, when people talk about "ISO" for digital cameras they are most often referring to the exposure index scale (1), as used in a camera's "ISO" settings, without necessarily making any claims about how well shadows or highlights or noise are handled at the various exposure index settings. This may be confusing to us old timers who think of ISO (or even ASA) as a measure of film speed based on shadow handling and such, but it is not totally incorrect.

That is why it is useful to have tests of cameras at various exposure index levels, like the ones that started this thread, and using "best development practice" rather than "standard development" makes sense for high end cameras.
So my thanks to the OP.

I think some misunderstandings need to be eliminated, for they are causing some confusions.

The ISO standard referred to often on these forums does not represent any sensitivity regarding digital imaging, but gain. BJL has posted this just above without spelling it out in cleartext.

Sensor gain is not identical to "pushing the exposure" in post processing. Higher gain is supposed to add details in the dark areas compared to a lower gain with equal exposure.

Therefor underexposing a shot by 3 EV and adjusting the lightness in the raw processing does not make ISO 800. It speaks a lot for some MFDBs, that they can produce high quality images with low exposure even though they do not offer higher ISO; the Sinar back is one such. (Note, that the PhaseOne backs do have higher gain, at least those I have analyzed.)

My observation is, that many MFDB photographers do not know the characteristics of their camera enough.

For example they believe, that they expose a shot properly, while heavily underexposing it; unfortunately some raw processors actively contribute to such misunderstandings. Or the photographer exposes lower than it would be ideal, because (s)he mistakenly believes, that the ISO gain will make up for that.

A further source of misconception is the belief, that no sharpening took place yet. For example the image posted just above by Graham has undergone noise reduction, but (I guess) the software did not tell about that.

Unfortunately, paternalizing and misleading the user is a quite common habit of raw processors.

Guy,
I hope you will post some examples from the Horseman either here or on your site!

Thanks
Paul C

No don't mind at all and thanks! Would not have tried this were it not for your post.

I did just now try the same image in C1 3.8 DB The B&W conversion using the B&W film looks and also using the panachromatic color profile looks good, but I think now seeing both I can say the noise handing of C1 4.1 is better. There is are a few vertical bands in the image in 3.78 DB that I didn't notice in 4.1

amsp, which do you prefer 3.78 or 4.1?

Graham,
The new king of the block with regard to high ISO performance is the Nikon D3 with an effective, analog-gain based ISO up to ISO 6400 plus a couple of more 'false' ISO settings up to 25,600, which are basically ISO 6400 underesposed by one or two stops.

It would be interesting to compare the image quality of identical scenes with identical exposures, using a DB and the Nikon D3, because that's the only way one could get a meaningful comparison.

As one moves up the ISO scale, deterioration of image quality is always first noticeable in the shadows. If you choose your subject well, have lighting which is even, no deep shadows or bright highlights, it's often possible to get surprisingly good results at high ISO, with any camera.

Another issue, which I alluded to earlier, is the fact that at base ISO the DB produces images with greater dynamic range than any 35mm DSLR. If we consider that the DR of the Nikon D3 is 9 to 10 stops, then the DR of a Phase DB is probably 11 to 12 stops using the same measurement standard, ie. about 2 stops more.

As a consequence of this fact, it is reasonable to suppose that one should be able to raise ISO on a DB by 2 stops above base ISO and still get image quality which is no worse than that from a 35mm DSLR used at base ISO, with regard to tonality and shadow noise. If the DB has substantially more pixels than the DSLR (which is usually the case), then resolution in the brighter parts of the scene should still be better than the DSLR.

The issue here, as I see it (let's pretend I know something), is how the camera handles that underexposure. From my own experience with cameras such as the Canon 20D and 5D, it is clear that image quality is significantly better at high ISO than the same exposure used at low ISO. Ie. Image quality at ISO 800 is significantly better than at ISO 100 using the same exposure, which means underexposure by 3 stops.

As I understand it, this improvement at high ISO (in Canon DSLRs) is not due merely to software manipulation, but is hard-wired into the sensor's electronics. I get the impression that such technology is possible (or at least easier to implement) in the CMOS type sensor rather than the CCD sensor of the DB.

If you want to do a meaningful experiment, try comparing an image from your DB underexposed at base ISO, with the same scene at the same exposure but correctly exposed at a higher ISO. Is there any improvement in image quality at the higher ISO setting after appropriate post processing? Edmund found there wasn't.

Dear Panopeeper,

this is absolutely not true, and I wish to insist on it: there is no NR applied in the Brumbaer "eMotion DNG Converter" tool: we have had this discussion on another forum and I have given you the explanation(s) why it does not happen. Please refer to my post:

http://forum.getdpi.com/forum/showthread.php?t=1846

But I will let Stefan Hess together with Rainer Viertlböck speak about this, who are both members on this forum, if they wish to intervene, since they are the ones having created this tool together, as such much more able to speak about it.

Also, your second last sentence does suggest that sharpening is taking place somewhere during DNG conversion, if I understand you right. No, that's not the case at all, there is no sharpening done in either Brumbaer's DNG Converter nor in Sinar eXposure.

I wish also to make something very clear here, since it seems to be misunderstood by you:

Although the Brumbaer "eMotion DNG Converter" tool does read, handle and convert eMotion and other Sinarbacks raw files, although I have myself and in many occasions praised and still am praising the quality of the DNGs produced by the Brumbaer, as well as the simple, and most fast and efficient workflow I have ever seen from ANY application in the market, I have to emphasize here that Sinar DOES NOT by any means SUPPORT this tool and it is NOT a Sinar application.

The Brumbaer "eMotion DNG Converter" from Stefan Hess is an application standing on its "own", developed and written by Stefan Hess and Rainer Viertlböck, both members on this forum, for the purpose to be able to handle eMotion files (and now as well other Sinarback files) with the best possible IQ quality and fastest possible workflow when needing to apply "white shadings" (very useful for architecture photography). It is however, though not supported by Sinar, still today the most powerful tool I know, praised by all who are using it.

As a consequence, and even though this Brumbaer toold DOES NOT apply any NR during conversion into DNG, I feel the sentence you have used, "Unfortunately, paternalizing and misleading the user is a quite common habit of raw processors" absolutely wrong and misleading in itself.

I would expect it to be taken away, if aimed at Sinar, for the rightness and truth. And I believe Stefan Hess, as well as Rainer Viertlböck might wish to jump in here to give their own comments about it.

Where I do absolutely agree with you, is when you are saying that many users do not know (without meaning this negatively) the characteristics of the camera/back fully.
Please do not use terms as "paternalizing" and "misleading" when you do not have the full details.

Thanks and best regards,
Thierry

PS: there is much more to say than what is presented here in a very "summarized" explanation, concerning "gain" and "sensitivity", in respect of their respective advantages AND disadvantages and the differences. BOTH have advantages BUT BOTH have disadvantages also.