Testing 8x10 vs IQ180

[timparkin]

I don't know if anybody saw the test on Luminous Landscape comparing an IQ180 against an 8x10. Suffice it to say it's pretty bad and they only scanned the 8x10 at 745dpi plus they used less depth of field with the 8x10 camera than the Phase.

Anyway - the original tester is rerunning his tests in cooperation with myself and I'm also repeating the tests in the UK (he is in Switzerland). I've sourced an IQ180, Alpa + 40mm & 70mm Digaron W and Linhof Techno and I have access to a Fujinon 240A and 450C. What I would really like to do is test using a 360mm Sironar S and was wondering if anybody has one that would be used in a very good cause? The tests will be done with Joe Cornish and Chris Ireland from Phase with help from Paula (Linhof Studio).

Also, if anybody has suggestions on test protocol I would be interested in hearing your thoughts.. (We have access to an 11,000 dpi Tango and probably a 12,000 dpi ICG for the really hi res checking).

If anybody wants to know what a Sironar S can really do - take a look at this PSD

http://static.timparkin.co.uk/static/tmp/810.psd.zip

The layers are scans from different dpi's, all upscaled to 6000dpi equivalent - there is an inset 200% view on the right hand side. By my calculations, I reckon you lose detail from the 4500 to the 2400 (I think there is also additional detail in the 6000 - the 'slats' definitely look finer in that scan). Anyway - if the real resolution is 4000dpi 'ish then an 8x10 has nigh on a gigapixel (even if it's 2400 dpi it's still 450 megapixel)

The goal of the test is to show how the two media are different not one better than the other necessarily. We'll also be shooting 4x5 and trying to emulate 100% shift movements on the Linhof techno with wideangle lenses. I'll probably test dynamic range by comparing the Phase with Portra 400. I fully expect the Phase to put on a seriously good showing and to have very 'malleable' files. I also expect the 10x8 to resolve more and to be more aesthetically pleasing and expect the 4x5 to show ability for extreme movements. The Phase will also have all the huge advantages of weight, unlimited exposures, etc. In a perfect world I would have a IQ180 and an Arca Swiss F-Line...

Tim

[Charles Twist]

Good stuff, Tim. I am sure you'll apply the customary rigour.
Just one thing. I could easily be wrong because my understanding of the theory is shaky, so please correct me if need be. You say that the best resolution from the film is with scans at around 4 to 6 thousand dpi. Isn't this kind of sampling subject to the Nyquist theorem which says you must scan at more than twice the frequency of the original in order to resolve its detail? So your second figure of 2,400 dpi might be closer to what film has to offer. Which then compares well with what a lot of people say about 4x5 film, ie that it has 120ish megapixels.
Extraordinary to see that digital results must now be compared to 10x8. I remember the good ol' days when...
It'll be interesting to see how the digital back fares with the highlights. I have no doubt that the dynamic range is greater than colour transparency film in any flavour and so it might be that it compares very favourably. In terms of shadow detail and 'malleability', I expect the Phase back will be technically better. But will it have soul?
Best regards,
Charles

[zoikes]

Hi

I think you are always going to run into trouble if you ascribe a megapixel equivalence to film. (you can work out an equivalence in appearance - which is what people are after I suppose)

In principal film is not nyquist limited, because of the random distibution of the silver grains (i.e. no moire!). The scanner is, of course, nyquist limited.
Your scanned image will have a modulation transfer function that is the combination (convolution probably) of the MTFs of the film image and the scanners optical system. And the taking lens, of course - which may make meaningful comparisons a bit tricky

Maybe that is the answer - if you really want to be a geek about it - measure the MTF of the two systems.
Take a picture of a slanted edge (building against sky), and use someting like imageJ to give the MTF curve.

http://rsb.info.nih.gov/ij/
http://rsbweb.nih.gov/ij/plugins/se-mtf/index.html

Ian

[dave_whatever]

Alright Tim.
Been kind of interesting following this debate on the US forum. On one hand it does seem interesting to do a proper test in a kind of bored-and-sat-at-work way, but on the other hand I might be tempted to think we could do without getting bogged down in technical semantics for the sake of it. This is supposed to be art not science afterall, film and digital are different and more importantly than resolution they have different looks, and different ways the fine detail and grain manifest themselves (but I expect you're aware of this as much as anyone). But putting that feeling aside for a moment here's a few thoughts:

• Tests should clearly state the cost of all the clobber used in each mode of capture, which should put any results into context. Also factor in resale value of kit, depeciation etc. It would maybe also be useful to compare a cost-per-banger, assuming the Ansel Adams output rate of one shot of 12 shots you're truly happy with per year, and allowing for whatever the percieved professional lifespan of a MF digital back is (2-3 years before you have to be seen to be shooting the most recent gear?). I don't think this output rate is too far off the mark for fine art work (maybe ask Joe?) - OK you could argue that a professional shooting a digital back needs to produce more than 36 keepers over 3 years, which is undoubtedly true if they're shooting for advertising/corporate/stock, but this area is hardly the mainstay of film shooters anymore so there's got to be some give or take! You've got to make each medium to have to play both home and away in this test. Similarly you could do the maths on shooting 10x8 where you need to churn out 500 hits per day's shooting....
• Total comparable kit weight of all the field gear in each case should be stated.
• Some real life comments on practicality, logistics and time constraints of the peripherals when shooting in the field - i.e. realities of reloading holders, charging batteries, downloading files onto laptops etc.
• Lets have some tests in British weather, mud, wind, pissing rain etc.
• Would be interesting to actually compare some final results, i.e. prints, rather than just pixel peeping. Also get some non-photographers to compare results, some people who don't have a vested interest. Punters off the street, people's mums, fine art print buyers etc. And it doesn't have to be massive prints either, realistic art-print sizes (20x16"?) would also be fine for comparing the "look". If you really want to go to town, throw in a few goog enlarger prints from 18x0 print film, or even a cibachrome or two.
• Might also be intersting to also compare the drumscanned 10x8 and the IQ180 to the results you can get out of a consumer flatbed (i.e. your v750, wetmounted). This might make the comparison a bit more relevant to your average 10x8 landscape shooter, who like most of us will never afford or need a MF digital back but likes to see how their cheapo setup compares.
• Can you get hold of a 4x5 scanning back too?
• I wouldn't get too hung up on shadow detail, afterall images often need some black areas, bit of mystery.
• Try and devise a "gratification" rating and tactile enjoyment factor to evaluate how much fun shooting with each setup it - afterall, we do this for enjoyment don't we?

I do think however that its always going to be a bit of a funny comparison, as I suspect most 10x8 art shooters will never want or afford an IQ180, and most shooters who feel the need for an IQ180 will always find something in their results to justify their purchase (who actually ever admits to having wasted tens of thousands of pounds?). So I suspect that regardless of the results, probably advocates of either mode of capture will claim a win.

Good luck.

[timparkin]

Good stuff, Tim. I am sure you'll apply the customary rigour.
Just one thing. I could easily be wrong because my understanding of the theory is shaky, so please correct me if need be. You say that the best resolution from the film is with scans at around 4 to 6 thousand dpi. Isn't this kind of sampling subject to the Nyquist theorem which says you must scan at more than twice the frequency of the original in order to resolve its detail? So your second figure of 2,400 dpi might be closer to what film has to offer. Which then compares well with what a lot of people say about 4x5 film, ie that it has 120ish megapixels.
Extraordinary to see that digital results must now be compared to 10x8. I remember the good ol' days when...
It'll be interesting to see how the digital back fares with the highlights. I have no doubt that the dynamic range is greater than colour transparency film in any flavour and so it might be that it compares very favourably. In terms of shadow detail and 'malleability', I expect the Phase back will be technically better. But will it have soul?
Best regards,
Charles

You are right about the Nyquist theorom up to a point which is why I think we probably have about 3000dpi of actual detail which means about which equates to 500megapixels. Even if we drop 10x8 right the way down to 1600dpi which I think most people would say is undersampling, then we still have 200 megapixels for 10x8. I think I agree with the roughly 100'ish megapixels for 4x5 although I've taken a couple of shots of Army types where I'm resolving pixel level text with a

The reason I say 'up to a point' is that even scanning at 20% over the limit gives you 90% of the actual (e.g. if you have 2000 dpi of actual detail and you scan at 2400 dpi then you will be getting 1800dpi of raw information).

All of these tests presume digital is getting 100% of the MTF but, as we know from the way the Bayer alogorithm and pixel packing ratios work this is actually untrue. Actual mp is probably only about 60-80% of that shown (for real world colour images, not black and white edges).

Tim

[timparkin]

Alright Tim.
Been kind of interesting following this debate on the US forum. On one hand it does seem interesting to do a proper test in a kind of bored-and-sat-at-work way, but on the other hand I might be tempted to think we could do without getting bogged down in technical semantics for the sake of it. This is supposed to be art not science afterall, film and digital are different and more importantly than resolution they have different looks, and different ways the fine detail and grain manifest themselves (but I expect you're aware of this as much as anyone).

Absolutely - a photographer needs to take into account more than just resolution. However, resolution is a concern for many photographers (rightly or wrongly). For instance, I've got a job to produce a 10m x 5m print which people will be walking right up to to 'pixel peep' a view. This isn't going to happen that often and fortunately they want a new image rather than an existing one. If I was shooting 645 film I would struggle, 4x5 would probably do OK and 10x8 is optimum for this. Knowing whether it would make sense to hire a Phase IQ180 or P45+ would be useful information.

• Tests should clearly state the cost of all the clobber used in each mode of capture, which should put any results into context. Also factor in resale value of kit, depeciation etc. It would maybe also be useful to compare a cost-per-banger, assuming the Ansel Adams output rate of one shot of 12 shots you're truly happy with per year, and allowing for whatever the percieved professional lifespan of a MF digital back is (2-3 years before you have to be seen to be shooting the most recent gear?). I don't think this output rate is too far off the mark for fine art work (maybe ask Joe?) - OK you could argue that a professional shooting a digital back needs to produce more than 36 keepers over 3 years, which is undoubtedly true if they're shooting for advertising/corporate/stock, but this area is hardly the mainstay of film shooters anymore so there's got to be some give or take! You've got to make each medium to have to play both home and away in this test. Similarly you could do the maths on shooting 10x8 where you need to churn out 500 hits per day's shooting....

Agreed - but this is all relative again isn't it. Someone has already mentioned that they used to charge clients for film and developing and so it was never a relevant cost. Digital shooters will always shoot more pictures and hence you can't compare their current output with what they would have shot on film. I'll try to put a few figures down though

• Total comparable kit weight of all the field gear in each case should be stated.

For what cameras? My Toyo isn't the lightest field camera..

• Some real life comments on practicality, logistics and time constraints of the peripherals when shooting in the field - i.e. realities of reloading holders, charging batteries, downloading files onto laptops etc.

Yep - Joe can comment on this as he has used both for some time.

• Lets have some tests in British weather, mud, wind, pissing rain etc.

Again, Joe has produce a video showing the P45+ being used in these weather conditions so we'll get his comments

• Would be interesting to actually compare some final results, i.e. prints, rather than just pixel peeping. Also get some non-photographers to compare results, some people who don't have a vested interest. Punters off the street, people's mums, fine art print buyers etc. And it doesn't have to be massive prints either, realistic art-print sizes (20x16"?) would also be fine for comparing the "look". If you really want to go to town, throw in a few goog enlarger prints from 18x0 print film, or even a cibachrome or two.

Agreed - I would love a print so I'm going to create small prints and then extract sections to print at full resolution and also finally I'll cut out a 35m or MF section from the centre of a print and do a darkroom enlargement. I've also bought a microscope to actually check what is on the film itself.

• Might also be intersting to also compare the drumscanned 10x8 and the IQ180 to the results you can get out of a consumer flatbed (i.e. your v750, wetmounted). This might make the comparison a bit more relevant to your average 10x8 landscape shooter, who like most of us will never afford or need a MF digital back but likes to see how their cheapo setup compares.

• Can you get hold of a 4x5 scanning back too?

Sadly not but I think the real limitations of the scanning back make their use all but studio work moot.

• I wouldn't get too hung up on shadow detail, afterall images often need some black areas, bit of mystery.

I'll go for highlight detail instead

• Try and devise a "gratification" rating and tactile enjoyment factor to evaluate how much fun shooting with each setup it - afterall, we do this for enjoyment don't we?

heh.. Well that is all opinion isn't it. Some people will enjoy the IQ180 best as they get to feel rich and special because they have bought something no-one else can afford. Other people may feel morally superior for having carried a 30kg camera kit into the mountains or because they have saved a stupendous amount of money. etc.

I do think however that its always going to be a bit of a funny comparison, as I suspect most 10x8 art shooters will never want or afford an IQ180, and most shooters who feel the need for an IQ180 will always find something in their results to justify their purchase (who actually ever admits to having wasted tens of thousands of pounds?). So I suspect that regardless of the results, probably advocates of either mode of capture will claim a win.

Yes they very well may do but the test is aimed at people who haven't already made up their minds and are wondering 'what the state of the art' is and allow people to make an informed decision about the differences between LF film capture of MFDB capture.

Tim

[timparkin]

Hi

I think you are always going to run into trouble if you ascribe a megapixel equivalence to film. (you can work out an equivalence in appearance - which is what people are after I suppose)

In principal film is not nyquist limited, because of the random distibution of the silver grains (i.e. no moire!). The scanner is, of course, nyquist limited.
Your scanned image will have a modulation transfer function that is the combination (convolution probably) of the MTFs of the film image and the scanners optical system. And the taking lens, of course - which may make meaningful comparisons a bit tricky

Maybe that is the answer - if you really want to be a geek about it - measure the MTF of the two systems.
Take a picture of a slanted edge (building against sky), and use someting like imageJ to give the MTF curve.

http://rsb.info.nih.gov/ij/
http://rsbweb.nih.gov/ij/plugins/se-mtf/index.html

Ian

Easiest way to do this is compare details in the final resultant scan. If there is more detail in one that the other, we have a comparitive measure. We can definitely get close to a result by doing this (and it won't be mathenatically absolute but comparative is all we need)

[zoikes]

Hi Tim

Apologies for continuing the physics approach...

After looking at Norman Koren's website, I'd tend to agree - it may be impossible to compare MTF fully. The film manufacturers don't give MTF beyond 50% (which determines sharpness). Limiting resolution is around 2 - 5% MTF usually.

However, the quoted figures for film are interesting (to me!)
velvia is about 45 line pairs/mm at 50% - or 11430 lp along the long edge of 10x8
TMax 100 is about 125 line pairs/mm at 50% or 31750 lp along the long edge of 10x8

The iq180 sensor is 10328 x 7760 - ie 5164 line pairs along the long edge. 50% MTF will depend on post processing but will be less than this.

So - no contest! (in theory). This of course is why the new lenses have to be so good for digital

It would be worth trying with Tmax (or delta) 100 if possible. For B+W workers the difference would appear to be much larger.

Ian

[timparkin]

Hi Tim

Apologies for continuing the physics approach...

After looking at Norman Koren's website, I'd tend to agree - it may be impossible to compare MTF fully. The film manufacturers don't give MTF beyond 50% (which determines sharpness). Limiting resolution is around 2 - 5% MTF usually.

However, the quoted figures for film are interesting (to me!)
velvia is about 45 line pairs/mm at 50% - or 11430 lp along the long edge of 10x8
TMax 100 is about 125 line pairs/mm at 50% or 31750 lp along the long edge of 10x8

The iq180 sensor is 10328 x 7760 - ie 5164 line pairs along the long edge. 50% MTF will depend on post processing but will be less than this.

So - no contest! (in theory). This of course is why the new lenses have to be so good for digital

It would be worth trying with Tmax (or delta) 100 if possible. For B+W workers the difference would appear to be much larger.

Ian

Forget about T-Max, we were thinking Adox for about 180 line pairs although this will obviously be lens limited but the sironar s should get 80 lp I imagine for maybe 70 on film giving

Obviously there is still 10% MTF for things like fine branches against the sky etc which will render up to the film limit I imagine - who knows?

Whatever the figure, the potential is there for a very good result for film..

Interestingly I shot some 6x17 of a recent job at f/16 and it's showing pixel level detail at 4000dpi for a 24,000 pixel wide file (200mp of data possibly? Obviously the pixel level detail is low contrast)

It would be great to have 5 or 10% MTF figures for lenses and film

Tim

[timparkin]

velvia is about 45 line pairs/mm at 50% - or 11430 lp along the long edge of 10x8
TMax 100 is about 125 line pairs/mm at 50% or 31750 lp along the long edge of 10x8

What is amazing on that NK page is that the 10% figure for Velvia is upwards of 130lp

Interestingly, and something that is backed up by my experience, the MTF figures for negative film show a higher figure for 10% than velvia, putting paid to the old 'velvia is the highest resolution' myth.

Finally, Chris Perez recorded 78lp onto black and white film (210 Sironar S at f/16) and, if NK is to be believed, we still get that resolution on colour film but with less contrast. 80lp = 20,000 pixels which is about 2000 dpi which fits with my experience and equates to about 350mp

Tim

[dave_whatever]

The velvia "myth" is probably still true given it, as far as I know, was only touted as the sharpest slide film.

[timparkin]

The velvia "myth" is probably still true given it, as far as I know, was only touted as the sharpest slide film.

Ah OK - That makes a little more sense.. However according to the article, E100VS has higher resolution when you get to 60lp but it's the 10lp that has the significant edge effect for Velvia.

Tim

[michaelfinch]

I was told many years ago that not only were LF lenses not designed to resolve as much detail as lenses for smaller formats (makes sense given the far larger film size; why build higher costs into manufacturing) but that LF emulsions resolved less detail than smaller formats (although I have always had serious doubts about that). Assuming that a given emulsion is equal across all formats, would there be any real benefit in having LF format lenses that resolved as well as the best 35mm primes? Just a thought.

[timparkin]

I was told many years ago that not only were LF lenses not designed to resolve as much detail as lenses for smaller formats (makes sense given the far larger film size; why build higher costs into manufacturing) but that LF emulsions resolved less detail than smaller formats (although I have always had serious doubts about that). Assuming that a given emulsion is equal across all formats, would there be any real benefit in having LF format lenses that resolved as well as the best 35mm primes? Just a thought.

Hmm... yes you are right and it's all to do with lens coverage I think. You have a choice to spread some of the sharpness across a certain angle and hence with small formats you only need to cover the diagonal and with larger formats you need to cover lots more. However, tolerances with large format can be substantially less so these sort of trade off each other so you don't lose 4x the resolution with large format lenses (although you may lose about 30% of the resolution).

My Mamiya 7 lenses can resolve 120 line pairs per mm on T-Max at 50% MTF (http://www.hevanet.com/cperez/MF_testing.html) (which suggests their capability is much higher).

The best LF lens in Chris' list are resolving about 80 lp hence losing 1/3 of the resolution but some of these lenses can cover 10x8.. so 4x the linear dimension but 66% of the resolution.

Interestingly the Mamiya 7 lenses seem to outresolve nearly all 35mm lenses..

If you want to see how much you can get on film, take a look at this..

http://www.imx.nl/photo/Film/page169/page169.html

over 200mp on 35mm!

Tim

[Neil Barnes]

As someone who's spent most of his career in imaging and signal processing of one sort and another (usually at much lower resolutions - broadcast TV) I'm watching this with interest... one thing which hasn't been discussed - and which the somewhat vague datasheet for the IQ180 doesn't specify - is exactly what is meant by '80 megapixels'. In pretty much every sensor of which I'm aware, the pixels for a colour 'block' are actually usually a variation on four pixels - two sensitive to green, one to red, and one to blue. Together that makes a colour sensitive (or monochrome, if you prefer) block but with only half the stated resolution - in both directions - in real terms.

Add to that the issue of both Nyquist limits and the comb filtering that occurs as you approach it if an optical low-pass filter is not used and it looks quite likely that the usable resolution of an 11000 wide sensor is going to be on the order of 'only' 2500 line pairs or so. On a 125mm wide 4*5 image that's roughly equivalent to 20 lines a millimetre...

As an aside - at the IBC exhibition earlier this month I got to see the NHK ultra-high resolution screen: 7680 x 4320 rgb pixels on an 85" diagonal. It is absolutely *stunning*... the camera (courtesy Ikegami) uses a colour splitter with 4000*4000 sensors - two for the green, offset half a pixel, and one each for red and blue. The system uses a 22+2 sound system (which strikes me as somewhat over the top, but which couldn't be adequately auditioned in the space provided).

Neil