new DxOmark test results

[eronald]

In the past I have proposed the following adjustment to DxO Mark scores, it is a great tool to evaluate sensor characteristics but it does not account for resolution. It takes 4 times the pixels to double resolution so why not take the dxo mark score and multiply it by the sqrt of the sensors resolution to get a real world value? 5DII = 79 x sqrt 21 = 362, D800 = 95 x sqrt 36 = 570, IQ180 = 91 x sqrt 80 = 814, this correlates with what I see. I personally use DxO to evaluate whether an upgrade  is worth it or not, if I can't get a 10 point increase in DxO Mark score I don't spend the money. Using this metric it certainly made sense to upgrade my P45+ to a IQ180 and it makes more sense to upgrade my 5DII to a D800 than a 5DIII.
Marc

The problem with DxO is that their tests seem ok, what they mean (S/n vs DR) is less clear, and then the aggregated score is a joke, as you point out.

On the other hand, I have a D4, and find that -as usual- the interesting questions are the focus, metering, ability to take images on the fly, all stuff which MF users of studio flash don't really care about, and shouldn't be paying for, and which DxO is not testing.


Edmund

[torger]

1. The D800, even in accordance with DxO is still a DSLR in terms of color rendition.  They can not touch the quality MF gives for still life. Hasselbad has the 200MS and Phase one the IQ180. I am on the theory that even the 40s will be better than the D800 on this application.

I'm curious about this color rendition thing. Multishot backs will obviously yield better colors on pixel level thanks to complete RGB information in each pixel. But the IQ180? I don't see any large differences in DxO data on color response and color sensitivity, and my (untrained?) eyes don't manage to detect huge differences on the (few) examples I have seen, but perhaps I'm missing something? I was kind of suspecting that color rendition superiority was more about good icc defaults in capture one than the backs actually registering more color information. But I've heard several talk about this superior color rendition so I'm thinking that I may be missing something. Is there perhaps a nice demonstration on the web somewhere?

[yaya]

Just for fun I used the Compare tool to see how my D5100 fares against the more expensive D7000...

Shame I didn't spend more money coz I could have gotten another 0.3 stops of DR...instead I have a camera that can do sports better with 28 more iso's...(but I don't shoot sports)...

Hey but I found out it's got the same DR as the IQ180....yeah, right...

Oh well...

[deejjjaaaa]

I don't think current lens manufacturing technology can deliver the resolving power to the small 135 format (not corner-to-corner at least), but in the larger medium format it is already possible.

really ? you can mount your beloved MF lenses on dSLR camera (even w/o AF) and then 80mp (of IQ180) / 2.5 (IQ180 sensor size vs FF sensor size) = 32mp... that's almost 36mp of FF sensor, but on FF sensor you are using the better central portion of your MF lenses...

[eronald]

Just for fun I used the Compare tool to see how my D5100 fares against the more expensive D7000...

Shame I didn't spend more money coz I could have gotten another 0.3 stops of DR...instead I have a camera that can do sports better with 28 more iso's...(but I don't shoot sports)...

Hey but I found out it's got the same DR as the IQ180....yeah, right...

Oh well...

Picture of Yair looking at his Kodak stock options:

[Michael H. Cothran]

What surprises me is that the likes of Nikon, Canon, and Sony have yet to introduce a new camera & lens system incorporating a "MF" size sensor.

[torger]

really ? you can mount your beloved MF lenses on dSLR camera (even w/o AF) and then 80mp (of IQ180) / 2.5 (IQ180 sensor size vs FF sensor size) = 32mp... that's almost 36mp of FF sensor, but on FF sensor you are using the better central portion of your MF lenses...

I think FF will be ok up to ~40 megapixels, so D800 should do quite fine in most focal lengths. IQ180 is probably quite close to the limit of what MF lenses can do, if corner sharpness is considered important. I recently tested my TS-E 24mm with a the dense 7D sensor, shifted it out to the lens FF corner to see what a 45 megapixel FF corner would look, and well, it is ok but not that great. I see considerably better results with say a P65+ on a Rodenstock wide angle. It is harder to make large image circles with high lp/mm, so you can probably go smaller pixels with DSLR but in total line pairs there's still some gain in MF.

Real well-made tests of MF lenses are not widely available though, so I base this observation of crops and RAWs/JPEGs I've seen in here and there over some time, so I'm not 100% sure this will hold true. I was sure enough to myself invest in a MF tech camera system though...

[torger]

Hey but I found out it's got the same DR as the IQ180....yeah, right...

There's no reason to doubt the measurement, but this is engineering DR of course (at which point signal = noise) which is just an indication of what photographic useful DR might be. Many other factors must be taken into account. One can see that the IQ180 has much less noise at 18% (more than two stops) which actually may a stronger indication of "photographic DR" than engineering DR. Noise quality (blotchiness etc) also matters much, which DxO does not measure...

I have myself when looking at MF raws been impressed by how clean they are so they can be pushed a lot, but if there's an advantage to the best DSLRs it seems to me to be 1-2 stops or so, not 6 stops that I have seen some claim. At some point I guess I need to do a bit more formal study... it is surprising how hard it is to establish what the DR of these things really are... the debate has been on for years it seems.

[jeremypayne]

There's no reason to doubt the measurement, but this is engineering DR of course (at which point signal = noise) which is just an indication of what photographic useful DR might be. Many other factors must be taken into account. One can see that the IQ180 has much less noise at 18% (more than two stops) which actually may a stronger indication of "photographic DR" than engineering DR. Noise quality (blotchiness etc) also matters much, which DxO does not measure...

I have myself when looking at MF raws been impressed by how clean they are so they can be pushed a lot, but if there's an advantage to the best DSLRs it seems to me to be 1-2 stops or so, not 6 stops that I have seen some claim. At some point I guess I need to do a bit more formal study... it is surprising how hard it is to establish what the DR of these things really are... the debate has been on for years it seems.

BCLAFF ... a forum contributor ... has done excellent work on measuring what he calls PDR - Photographic Dynamic Range.

http://home.comcast.net/~NikonD70/Charts/PDR.htm

Unfortunately, it doesn't have any MFDBs but has a nice selection of DSLRs.

[yaya]

Kodak stock options

What's that???

[jeremypayne]

Just for fun I used the Compare tool to see how my D5100 fares against the more expensive D7000...

Same sensor ...

http://www.dxomark.com/index.php/Publications/DxOMark-Reviews/Nikon-D5100-DxOMark-Review/Nikon-D5100-vs-Nikon-D7000-same-sensor-same-results

[Dan Wells]

      I've seen some semi-real-world results (deliberate test subjects) showing the D800 pretty close to the 40 mp generation MFDBs (which are not only lower resolution than the IQ180, but also older sensor generations). They are close enough that I wouldn't be surprised if the D800e reaches P45+ quality levels. I'll be very interested to see images that are not just tests, and to examine large prints - pixel peeping any of this stuff on screen is crazy because no monitor displays more than (rarely) 2560x1600 pixels!
     The D3x was a close competitor to 20+ mp backs (another generation older), suggesting that  FF sensors may just about keep up with MF sensors at the same resolution (obviously, except for depth of field and other direct effects of sensor size) because the extra generations of sensor development in the years it takes FF to reach a given resolution more or less cancel out the effect of pixel size? The newest generation MF sensors will always be significantly ahead, but at a high price.  We seem to have the same generational effect with the (APS-C) NEX-7 possibly catching the D3x (or coming closer than one might think) - the 3 years newer sensor makes up for the smaller size... Of course, the NEX-7 has lens issues - Sony's built a camera their lenses don't always keep up with... They aren't alone, hasn't Phase had that problem with some of the older Mamiya lenses?
        I wonder why no MFDB has gone CMOS? A double-size D800E sensor might be much cheaper to produce than the IQ180 sensor, and perhaps as good?

[yaya]

Same sensor ...

http://www.dxomark.com/index.php/Publications/DxOMark-Reviews/Nikon-D5100-DxOMark-Review/Nikon-D5100-vs-Nikon-D7000-same-sensor-same-results

Yup, but I've got 28 more iso's and 0.3 less f-stops....outrageous in'it

[eronald]

Yup, but I've got 28 more iso's and 0.3 less f-stops....outrageous in'it

You're just field, not policy - if you were policy, maybe the policy would be saner?

Edmund

[itsskin]

2012 vs 2004 
http://dl.dropbox.com/u/17452116/DynamicRange_5DMIIIvsAptus22.jpg

[yaya]

You're just field, not policy - if you were policy, maybe the policy would be saner?

Edmund

So I googled Field-Policy and found this article...

Their circles theory makes DXO's measurements look like kids math...

[Peter Devos]

2012 vs 2004 
http://dl.dropbox.com/u/17452116/DynamicRange_5DMIIIvsAptus22.jpg

And imediately i now remember why i love my MF backs, from the 6Mp CantareXY to the Hasselblad 40Mp...... Just take a look at the shadows of the 5d. These are as ugly as all dark parts with my 5DMk2. 

[yaya]

This might or might not be relevant:

Link 1 and Link 2

But it'll surely add to this already all-over-the-place thread....

[eronald]

This might or might not be relevant:

Link 1 and Link 2

But it'll surely add to this already all-over-the-place thread....

Kudos for providing the link. Maybe Nikon needs a Chuck Westfall clone, you could apply.

In the mean time, I went the other way, and got a D4. It has an interesting texture and look for portraits, see the crying baby above.
You will remember that a lot of people here said they wanted fewer better pixels, in a certain sense the D4 is providing that.
But then I think Saint Augustine said, O Lord, make me chaste and poor, but please wait a bit before doing it ...


Edmund

[bjanes]

There's no reason to doubt the measurement, but this is engineering DR of course (at which point signal = noise) which is just an indication of what photographic useful DR might be. Many other factors must be taken into account. One can see that the IQ180 has much less noise at 18% (more than two stops) which actually may a stronger indication of "photographic DR" than engineering DR. Noise quality (blotchiness etc) also matters much, which DxO does not measure...

I have myself when looking at MF raws been impressed by how clean they are so they can be pushed a lot, but if there's an advantage to the best DSLRs it seems to me to be 1-2 stops or so, not 6 stops that I have seen some claim. At some point I guess I need to do a bit more formal study... it is surprising how hard it is to establish what the DR of these things really are... the debate has been on for years it seems.

Perhaps you should recheck your math. The print DXO SNR at 18% is 48.7 dB for the IQ180 at a measured ISO of 29, and the corresponding figures for the D800 are 45 dB at a measured ISO of 74. 48.7 dB = 272.1:1, for a DR of 8.1 stops. 38.4 dB = 177.8:1 or 7.5 stops.

48.7 dB is rather high for a noise floor used to determine photographic DR. One can use the full SNR plots by DXO to determine the DR for any noise floor (see Emil Martinec). For a noise floor of 18 dB (around 25:1), I get a normalized (DXO method) DR of 10.5 stops for the D800 and 10.2 for the IQ 180.

Regards,

Bill