phase versus hassleblad

[fredjeang]

It's not just about the DR number, it's "where the DR is" Regarding negative film, it's got a pretty impressive roll off in the highlights, digital is the reverse..it's mostly all in the shadows. Pulling up hard shadows wise, watching the highlights to avoid clipping..I can get not too far from neg film..but it's not really the same, hues and colours shifting can be an issue. Obviously raw and low ISO are a must.

Films vary in DR too, so a blanket statement cannot hold any water. I've serious doubts any DSLR at any price can match some of the b&w films either, FP4 has a simply massive latitude, over and underexposure wise. The only company to seriously attempt good DR via a sensor is Fuji, shame we don't see more of this.
But more to the point..in many cases having a big highlight latitude is more useful. If we added 2-3 stops more highlight end on most DSLR's, you'd hear next to no complaints about burnt out highlights.
I'm glad were talking about DR though, it's an important point, and often takes a back seat to "resolution"

I completely agree with this point Barry. One of my best friend worked a lot with the Fuji and indeed this was one step ahead in DR. May I risk to say also the Foveon solution? But in both case, resolution was in question, so is there a relation between DR and resolution output?
Then it would easily explains in part why the MFD have an advantage over 35mm. But I'm not a scientist neither an expert.

Fred.

[grantleversha]

can anyone out there add to this and/or shed some light on the following. i have the p65 but when used with a Horseman SWD2Pro and a 24xl Schneider Lens i find there is vignetting and purple fringing due the image circle of the lens being too small when using the full frame P65.  i have come to understand through correspondance from Eric Joakimand at Phase One and the guys from Alpa and Teamwork Digital that when using a technical camera, the only wide angle lens that has an image circle large enough is the Rodenstock 23mm HR Digaron. But, you cant use it with shift/tilt. does anyone know when either Schneider or Rodenstock are going to bring out wide angle lenses with large enough image circles to be used effectively with the full frame P65. Thomas at Alpa suggests the following to be released mid 2010 - (Schneider: a new 28mm lens with a large image circle of probably 90mm diameter; Rodenstock: a new 32mm lens also with a large image circle). the HTS 1.5 apparently when used with the 28mm lens causes a crop factor to be introduced and thats only using the 50mp back. what will be the case with the new 60mp Hassleblad sensor. will the tilt/shift be limited and will their still be a crop factora and by how much? also. with Mark's comparison of the two cameras, can they compare to technical cameras in terms of quality. thanks.

[BernardLanguillier]

I completely agree with this point Barry. One of my best friend worked a lot with the Fuji and indeed this was one step ahead in DR. May I risk to say also the Foveon solution? But in both case, resolution was in question, so is there a relation between DR and resolution output?
Then it would easily explains in part why the MFD have an advantage over 35mm. But I'm not a scientist neither an expert.

Fred.

Gentlemen,

All the sensors used in all photographic devices on the market today can be considered to be perfectly linear devices up to the point where they saturate, including the Fuji S5 Pro.

There are only 2 key differences between sensors:
- how the system is calibrated in terms of exposure and in camera histogram (this can give the impression of the ability to recover highlights),
- how clean the shadows are.

In the future we will be getting non linear sensor able to handle infinitely bright illuminations with finite and differentiate RGB values, but we are not there today.

Cheers,
Bernard

[BernardLanguillier]

There we go. It was a topic about comparing 2 MFD cameras and we are now in the war between 35mm FF cmos and MFD ccd  lovers...

Nope, the original article was mostly about comparing 2 backs and a little bit about comparing backs to DSLRs.

This thread is somce the beginning mostly about latter part.

Note that this thread would not have existed had the original article been realistic about the later point.

Cheers,
Bernard

[siba]

Cheers Bernard. Exactly. I started the thread purely because of that one sentence I quoted in the initial post.

The rest of the article was a really great comparison of two digital backs, that most of us wouldn't be able to easily get our hands on and do ourselves.

Fred, as I stated I'm a Phase user, and wouldn't normally care about comparisons between DSLR and MFDB. There was a time, three years ago or so, when I thought there was no comparison. But, nowadays, even though I shoot mostly MFDB, I have no qualms about doing a shoot with my DSLR if the situation calls for it. I have come to see in my professional, real world, handing files over to clients, that there is not that huge a difference.

Therefore, at 1 in the morning, when I was happily reading the article - happy to have a new article by Mark Dubovoy, and a glass of wine, the sentence about being able to see the difference in DR of a DSLR small print versus a MFDB small print from 30 feet away jumped out of the page, and it seemed like bar talk. If a photographer mate of mine would have made the same sweeping comment I would have immediately reacted. So, I couldn't help myself and started the thread.

And Erik, you have made the same sort of sweeping comment that has made me come in once again. According to you a "blind man" can tell the difference between two small prints at thirty feet. That the issue is just being able to discern between black and white in the two prints.

I quote Mark again:
"The dynamic range of a typical high-end professional 35 mm DSLR is around 7 F/stops. Medium Format cameras are closer to 13 F/stops of dynamic range. You can see the difference from 30 feet away in a small print"

This sentence implies that the dynamic range of DSLRs and MFDBs is so astronomically different that it is immediately apparent from thirty feet away. Erik, you are agreeing with this. I just can't agree. If you take a photo of the same scene under the same conditions and take the same care with printing, then from thirty feet away I don't believe anyone could tell the difference, from two 8x10 prints, which was DSLR and which was MFDB. Not with any certainty at all. If I was just talking to mates in a bar I would be more adamant about this, but because we are in a forum, I am just politely suggesting that this one sentence is cause for discussion. Even though we've obviously been there before.

Now if it was just a question of semantics then I don't think I would have bothered reacting. But this is fighting talk  . And I'm usually on the side of the Digital backs. Not so in this case

cheers all,

Stefan

[fredjeang]

Nope, the original article was mostly about comparing 2 backs and a little bit about comparing backs to DSLRs.

This thread is somce the beginning mostly about latter part.

Note that this thread would not have existed had the original article been realistic about the later point.

Cheers,
Bernard

I agree Bernard.
I read your posts carrefuly, and you certainly point some interesting details. But in another topic (about the pentax MFD), you assumed that the 645D could be a game changer. Maybe I did not understood in what way or context you pointed that fact, and how could it be a game changer in the current panorama.  
I simply think that in these kind of debates, there is fast much more writings and concepts that images, in a website dedicated to...image. So what happen is that we have an incredible amount of arguments but very little real life images in order to endorse the arguments. Yes, a lot of graphics.
At the end, we all looking and participating for better information, and the overall sensation that I get is that there is more confusion.

Best regards,

Fred.

[Graeme Nattress]

There are numerous backlit dynamic range charts - all we need to do is point a range of cameras at them and publish the images. This discussion on 35mm v MF DR comes up time and again, and nice though DxO numbers are, I think we all need to see the test images.

Another point - "highlight DR" is just about where you set your mid grey. The less noisy the sensor, the lower you can set the mid grey and the more highlight protection you can build into the development gamma curve. That is why it's important to take the DR test measurement back to linear light to compare cameras properly. Once you know how many stops your camera protects in the highlights - where it biasses mid grey effectively, you can use exposure compensation to adjust accordingly, or even on a scene by scene basis. In the end, it's just another way of looking at ETTR.

Graeme

[michael]

Please see my Editor's Note just added to Mark's review. It will hopefully shed some light on the DR question.

Michael

[bjanes]

Please see my Editor's Note just added to Mark's review. It will hopefully shed some light on the DR question.

Michael

I think your "standard definition" of DR, "The standard definition of dynamic range in the industry has always been how many F stops above and below middle gray can be recorded while delivering full texture and detail", is from the film era. Digital is linear and does not have any inflection point in the mid range. Middle gray has no special significance in a linear digital characteristic curve and we don't meter for middle gray but rather for the highlights. In any case, we are more interested in the total DR. The standard definition of DR for digital sensors is the full well capacity divided by the read noise. This engineering definition sets the noise floor quite low and photographers might set the floor at a higher value. For objectivity, a signal:noise cutoff should be used in this case.

This is the approach taken by Norman Koren (who was recently featured in a LuLu video review) in Imatest. For example, with my Nikon D3 at base ISO and a Stouffer wedge, I get the shown results when rendering into a 16 bit TIFF with ACR set at defaults but with a black point of 0 (using the default black clips the shadows and gives a DR of 7 stops).

[attachment=20738:Stouffer...0_Step_2.png]

According to Roger Clark's data, the D3 has a full well of 65500 electrons and a read noise of 17.6 electrons at base ISO, giving a calculated DR of 11.8 stops. DXO gives the D3 a screen DR (per pixel) of 11.92 stops. That compares with the low quality DR of 11.9 stops in my tests. For high quality results, the DR is only 7.82 stops.

To get a better idea of how MFDBs compare to DSLRs, why doesn't someone with a P65+ post some data?

[bjanes]

There are numerous backlit dynamic range charts - all we need to do is point a range of cameras at them and publish the images. This discussion on 35mm v MF DR comes up time and again, and nice though DxO numbers are, I think we all need to see the test images.

Another point - "highlight DR" is just about where you set your mid grey. The less noisy the sensor, the lower you can set the mid grey and the more highlight protection you can build into the development gamma curve. That is why it's important to take the DR test measurement back to linear light to compare cameras properly. Once you know how many stops your camera protects in the highlights - where it biasses mid grey effectively, you can use exposure compensation to adjust accordingly, or even on a scene by scene basis. In the end, it's just another way of looking at ETTR.

Graeme

+1. See my previous post.
Bill

[Jeremy Payne]

Please see my Editor's Note just added to Mark's review. It will hopefully shed some light on the DR question.

Michael

With all due respect, you don't need to believe in DxO or have access to a MFD back to have an view on the dynamic range - by your definition - of your own DSLR.

Right?

If we all agree that good slide film could deliver 8 stops, then I think I get at least 10 from my D700.

[ErikKaffehr]

Hi,

Thanks for the note!

Just a few comments. I may feel it unjust to blame DxO for using the SNR equals one definition of DR, AFAIK it is the standard (textbook) definition. I do agree that SNR (Signal to Noise Ratio) of one is pretty useless, but that is another question. The way I see it DxO makes measurements of the sensors they need to make their DxO raw converter. They publish their results, which I see as a great service. The one thing I don't like with DxO is putting a single figure of merit on their measurements.

Another point is that much of the characteristics of a sensor is decided by physics. DR is definitively only affected by the number of free electrons generated by photons and the read noise. What Michael essentially says that the normal definition of DR is not relevant to photography and I can agree with that. On the other hand what is probably important to photography is what is known as shot noise a purely statistical variation of the number of photons reaching each sensel. Increasing the sensor size reduces noise, doubling the area of the sensor reduces noise with a factor of 1.41 (square root of two) to my understanding.

From a standpoint of physics I cannot see how a 4-5 stop advantage of MFDBs over DSLRs would be feasible. Would MFDBs really have much less noise than DSLRs they would also be able to use much higher ISO as they would be able to take like 4-5 steps of underexposure.

I have great respect for the knowledge and experience of both Marks and Michael, but I would really like to see some analysis which explains the differences in a way that makes sense to engineers.

The most comprehensive analysis I have seen is here:

http://www.clarkvision.com/articles/digita...mary/index.html

Unfortunately, Dr. Clark does not discuss MFDBs, as far as I know,

Best regards
Erik

Please see my Editor's Note just added to Mark's review. It will hopefully shed some light on the DR question.

Michael

[barryfitzgerald]

The "quality" aspect seems to be important. Looking at the DR results on imaging resource, they use imatest, but have quality from low to high, jpeg and ACR best. Obviously there will be a point you can pull any image apart, esp shadow end..but if you get unprintable results, it's of no real value in the field in such cases, purely technical.

I guess someone is going to have to do this "Clint Eastwood" style, take a phase one, Hassie, and a respected DSLR out, take some real world shots, and if there really is a 5 stop gap DR wise, that would show up very quickly indeed. I'd be surprised if the MF digital's can get 13 stops of high quality DR, but I'd be curious to see that. I would not expect the MF ones to blow out like that if they did have huge DR, but I won't bar stool this one, I've never even seen either MF digital, let alone touched one ;-) I bet 13 stops has more to do with a technical test, rather than a high quality printable image. Ditto on DxO and Clarkvision, in particular some compacts seem to have a DR score that is not really believable for actual usable images.

Still, the discussion is useful, and interesting..and so far on this thread at least..seems civilised and respectful.

[BernardLanguillier]

Please see my Editor's Note just added to Mark's review. It will hopefully shed some light on the DR question.

Michael,

I was not aware that you had other problems with the relevance of DxO results besides them not factoring in resolution enough.

If follow up questions are allowed, how is the DxO definition of DR not taking the linearity of sensors into account? My view is that their definition only makes sense for the very reason that sensors are linear. If they were not, highlight behavior would be less clear, and there would be no way to base DR computation from a fixed reference in the highlights.

In fact, both definitions are very similar.

- On your side you use a film days concept called mid-grey and add the amount of stops on both sides to compute DR... DxO doesn't split the range.
- Other than that their definition and yours would be identical if "delivering full texture and detail" (the industry generally uses "perceivable detail" rather than "full detail") were equal to "signal to noise ratio equal to 0DB". This is indeed not the same, Full detail would correspond to a signal to noise ratio significantly higher than 0DB, and the resulting dynamic range would be lower than the value provided by DxO... but it would be lower for both the DSLR and the MFDB wouldn't it?

The question then becomes, why would DSLRs be impacted more by a more severe definition of noise floor compared to MFDBs?

Cheers,
Bernard

[fredjeang]

Please see my Editor's Note just added to Mark's review. It will hopefully shed some light on the DR question.

Michael

Thank you for this Editor's Note.
Michael pointed exactly some concepts that I tried clumsily to express with my limited english.
It is also nice to see that the debate here in Lu-La is animated and sometimes hot but did not fall into uncontroled manners.

As a non-expert and in a learning process, I'm always interested to read everyone's point of view. But after a while, what I find is that there is a lot of presumed knowledge that seems to be given without real experience with these tools. Michael pointed this in the note and I think it is sadly often the case.

So, do we get a trustable or better information?  I think we are all looking for that, no? But it turns into a passionate and conceptual answers game, like in a football stadium where each one supports his team, even if it means deny clearness.
For a not so technicaly knowledgable members like me, it is very frustrating and confusing because I do not know at the end who and what I can trust, and the only answer I got so far is I won't know it till I buy a MFD, work some time with it and see by myself.
I certainly can trust Michael because I know that he uses any kind of gear, from 4/3 to MFD and LF and I have no suspicions that he could be integrist of a system or another so his post was information to me. But in general I do not see very much, arguments that are served with images that illustrate them, so I tend to be carreful and keep an helphy distance. The question is, when you expose these physical and technical posts about DR etc...do you actually own a MFD ? or is it just DxO reading on the web? This point is important I think.
I also agree Siba, FF has changed and evolved, but till wich point?

When I pointed my desire of seeing more images and less theories, is just because there is a moment where images should also speaks.
Why not having for example a fashion session (controled light), like the one they did for the Leica S2, with 2 photographers experts in their gears?
1 with a Nikon D3s and one with a Phase or Hasselblad or Sinar m. Then putting the original files downlable, and see the results.
I think it would be really helpfull.

Cheers,

Fred.

[AveryRagan]

"For high quality results, the DR is only 7.82 stops."


Based on my experience with linear detectors in my day job (univ.prof. mass spectrometry) looking roughly at your data I would say you nailed it.

[JeffKohn]

Please see my Editor's Note just added to Mark's review. It will hopefully shed some light on the DR question.

I don't think this sheds any light on anything at all. If you want to argue that the absolute numbers produced by DxO's DR test are optimistic, and that a more real-world definition of photographically useful DR would result in smaller absolute numbers, that's fine.  

But the real value of the DxO tests is that DxO is consistent in their testing methodology. So as long as you understand what the various tests are measuring, comparing the test results from two cameras will indeed give a meaningful indication of their relative performance. If we look at DxO's testing, there is nowhere near a 6-stop difference in performance between today's best DSLR's and MFDB's.  And I'm sorry, but you cannot dismiss the DxO test as invalid with a flippant comment that "it does not take into account the linear nature of sensors." This statement is, to put it bluntly, absurd.

I will agree that today's latest MFDB's have some DR advantage over most DSLR's. But to claim that it is a 6-stop advantage cannot be taken seriously. If you have tests showing this, I think a lot of people would be very interested in seeing them. You seem to place high value on personal experience and real-world testing. Yet by his own admission, Mark used neither in coming up with his claims; rather, they are based on subjective impressions and anecdotes from disparate sources. IMHO you need better arguments that that if you're going to claim that

- Slide film has 1 stop more DR than DSLR's
- MFDB's have 6 stops more DR than DSLR's.
- The difference in DR between MFDB's and DSLR's can be distinguished in a small print from 30 feet away.

None of these can be taken seriously. And while Mark's anecdote relating to the last point is amusing, it's hardly relevant. For one thing, there was no comparison; they only saw an MF print from a distance. Furthermore, as another poster pointed out, the DR of DSLR's and MFDB's both exceed the DR of a physical print.

[fredjeang]

I don't think this sheds any light on anything at all. If you want to argue that the absolute numbers produced by DxO's DR test are optimistic, and that a more real-world definition of photographically useful DR would result in smaller absolute numbers, that's fine.  

But the real value of the DxO tests is that DxO is consistent in their testing methodology. So as long as you understand what the various tests are measuring, comparing the test results from two cameras will indeed give a meaningful indication of their relative performance. If we look at DxO's testing, there is nowhere near a 6-stop difference in performance between today's best DSLR's and MFDB's.  And I'm sorry, but you cannot dismiss the DxO test as invalid with a flippant comment that "it does not take into account the linear nature of sensors." This statement is, to put it bluntly, absurd.

I will agree that today's latest MFDB's have some DR advantage over most DSLR's. But to claim that it is a 6-stop advantage cannot be taken seriously. If you have tests showing this, I think a lot of people would be very interested in seeing them. You seem to place high value on personal experience and real-world testing. Yet by his own admission, Mark used neither in coming up with his claims; rather, they are based on subjective impressions and anecdotes from disparate sources. IMHO you need better arguments that that if you're going to claim that

- Slide film has 1 stop more DR than DSLR's
- MFDB's have 6 stops more DR than DSLR's.
- The difference in DR between MFDB's and DSLR's can be distinguished in a small print from 30 feet away.

None of these can be taken seriously. And while Mark's anecdote relating to the last point is amusing, it's hardly relevant. For one thing, there was no comparison; they only saw an MF print from a distance. Furthermore, as another poster pointed out, the DR of DSLR's and MFDB's both exceed the DR of a physical print.

Jeff,
I really like your landscape photographs and I think you are a knowledgable photographer. That is why your post surprises me. How do you give that much credit to DxO and not the same credit to photographers who works daily with MFD ?  
But then I also see that you work with D3x, so is it that you are really serious in your post or in a way you defend your gear?
Because then, how can you explains the enormous amount of top photographers that are working in fashion, landscape, arquitecture and fine arts who only work with MFD or LF ?
I'd like to see a real world comparaison between the Nikon and some MFD, with texture clothes, complex patterns, difficult light etc...to see if this DxO is trustable in the real world. Numbers are one thing, and DxO is NOT absolute mathematics calculations but just a way of testing, reality is another story.
In my understanding of course and with all my respect.

Cheers.

Fred.

[JeffKohn]

How do you give that much credit to DxO and not the same credit to photographers who works daily with MFD ?

I don't think DxO tests are the only meaningful measurement of camera performance. Having said that, I also do not think you can dismiss the DxO tests out of hand, just because you don't like the results or the results disagree with previously-held beliefs.

But then I also see that you work with D3x, so is it that you are really serious in your post or in a way you defend your gear?

I'm under no delusions that my D3x is better than the P65. There are certainly aspects of image quality where such a back far exceeds any DSLR, with resolution being an obvious one, but also color depth. I don't think DSLR's are just as good as MFDB's, and never said that. If the P65 cost under 10 grand and had in-back live-view, I would be shooting one.

The three specific statements I took exception with were:

- Slide film has 1 stop more DR than DSLR's
- MFDB's have 6 stops more DR than DSLR's.
- The difference in DR between MFDB's and DSLR's can be distinguished in a small print from 30 feet away.

I may not think my D3x is as good as a P65, but I still disagree with those statements, especially with no factual basis or even personal experience from the author of those statements to back them up.

Because then, how can you explains the enormous amount of top photographers that are working in fashion, landscape, arquitecture and fine arts who only work with MFD or LF ?

There are plenty of reasons to choose MFDB: leaf shutter for fast sync with strobes, ability to use a view camera with the full range of movements, high resolution for retouching or large prints (DSLR stitching is sometimes a good approach for the latter, but not always), etc. Again, I never said there's no reason to use MF; a 6-stop DR advantage just isn't  one of them.

[fredjeang]

Thanks for these precisions Jeff.

Cheers,

Fred.