After held a 645D a few days before and checked for its street price, I have some thoughts to wonder what cost that much for a MF DSLR? And, are those "differentiators" of the 645D really worth for paying that significant extra premium? How long will take for those advantages to vanish, or at least to be less significant? (in nowadays when the technologies are growing so fast..)
To begin with the investigation, let's go back to the sensor size/pixel width talk I made last time for some MF DSLRs and others. Recently, after completed my in-depth full review on the Canon 550D and also brief tested the S0ny NEX-5 and purchased the NEX-3 myself, I found that in fact the Image Quality and Noise Control of those latest DSLRs and sensors (and their image processing as well) are actually not bad. In fact, I found that both models/sensors are not worse than the noise performnace of the K-x, if they are not better, even though the K-x has already been highly praised by Pentaxians for its "superior" noise performance at higher ISO speeds. (Well, Pentax people are amazed for what the K-x can give for the results just because there is a K-7 in comparison! Right? :-o)
The 645D has 40 Megapixels with a (square) pixel pitch width of 6.1 um. The 550D has a resolution of 5184 x 3456 pixels on a 22.3 x 14.9 mm sensor. As such, the pixel width of its sensor is of 4.3 um only, which is only 50% in area per pixel when it is compared to the 645D sensor. If that's what the current and latest technologies, no matter for the manufacturing process of the sensors as well as the associated image processing techniques (for both the hardware and software), could give acceptable or even good results, then we can calculate the maximum pixels and count that can be put into a 135 Full Frame DSLR, which is yet considered as the mainstream product for most (for advanced users/"richer" amateurs anyway).
Reverting the maths and calculating packing 4.3 um sized pixels on a 36 x 24 mm full frame sensor will give an overall resolution of 8372 x 5581, which gives an aggregate pixel count of 46.7 Megapixels, which is just more than the pixel count of the current 645D actually.
But of course, with the same sensor size of the 645D, we can have a DSLR with 10232 x 7674 = 78.5 Megapixels, which is just huge in number! But do we need that anyway??
So, my quick conclusion is that the Full Frame sensors will eventually reach and exceed the pixel count of that of the current 645D, even with the current proven/mature technologies that are already adopted and used in the latest APS-C sensors of Canon and Sony in their latest "low-end" (but actually high-end in real!) DSLR models. As there is still no popular 135 FF sensor that is more than 25 MPs at this moment, we shall still need to wait and see if any of the sensor vendors can make it later on - but possibly the day will come sooner than we would expect - as the technologies are already there. One more thing to note is that those 21 and 24M FF sensors of Canon and Sony were designed and put into mass production at least about two years ago. So, I do believe that if they are willing to make one, they could!
So, with the above conclusion and prediction in mind, the high resolution and high pixel count advantage of the 645D will soon vanish in a year or so as soon as when 135 FF DSLRs and sensors are catching up in this regard. As such, it should be wise to stay away from the MF even if you like it a lot, considering the huge investment probably cannot be safeguarded for just a bit longer. My believe is that the 135 FF will become the MF of the old film days, as IQ wise they will be more than adequate for 99% of use for people on this planet, no matter they use their cameras professionally or just amateurishly.
For other general shooting purposes and when smaller cameras and systems are desired, the APS-C will be the way to go. So, how about micro-4/3 then? I guess they will be killed by the APS-C ILDCs/EVILs eventually, as people don't need cameras that are not as small as and as "versatile" as compact P&Ses but without the minimum level of IQ that APS-C systems would give. (Honestly speaking, I am not using my GF-1 anymore with the purchase of my new NEX, as the IQ difference is just huge! Even if when I haven't got my NEX, I have already rarely to pick up my Girl Friend and I would just use my "huge" K-x and Pentax APS-C system instead.)
Monday, July 26, 2010
Logging you in...
Comments by IntenseDebate
Subscribe to:
Post Comments (Atom)
Alex · 763 weeks ago
I don't have to make any "calculations". I've saw what MF is capable of (true, not 645D prints - but eg. Hassie, with similar if not the same sensor). Telling that I was impressed is an understatement. The sheer amount of detail, the quality - even if close looking at large prints - the MF *is* something special.
And yes, I've also saw prints made with smaller formats. Even with APS-C one can get very good results; but reaching MF's level is another story. FYI, just cramming more pixels won't be enough ;)
Btw, Pentax can upgrade the 645D to the readily available 50MP Kodak sensor. In contrast, there is no 40MP+ "FF" (the correct name is: small format) sensor. Yet you're telling us that, in an year or two, such a huge advantage will simply vanish? Ridiculous.
2 cents · 763 weeks ago
Maybe Pentax should offer a job to RiceHigh, coz at least he is being critical and not pretending that everything is all right and everything will be fine one day. I'm really wondering how come Pentax is still in business when they have a bunch of retards in their top management.
RiceHigh for CEO of Pentax! :)
Michael · 763 weeks ago
Read your MTF carefully and you'll find out.
Zebooka · 763 weeks ago
645D has excellent performance in DR, noise, tones and resolution.
With current technologies, some new 35mm FF sensor with same resolution will loose battle because of noise, worse DR and tones. Yes resolution will be same, but people buy MF cameras not only for resolution. Also this segment is not wide spread. Especially WR MF. It is small and specific - like Leica segment. As for me - I would never buy MF because FF provides shallower DOF because of faster lenses and gear is lighter and smaller too, and tele lens are not "portrait".
As for low end cameras - Pentax here is in deep trouble now. This segment may die very soon, like prosumers did (in two years). People will prefer small EVIL cameras instead of larger DSLRs. And Pentax has no hi-end range cameras, only mid-range K-7 with advanced amateur features, but lacks really needed pro ones. (flash performance and AF - no 2.8 sensor, AF-C is terrible).
PS: It appears that it's a 70% change now that Pentax will loose two Pentaxians next year (for Canon 50D/60D and 5DmkII). Just because we are tired of bad AF performance, terrible not working flash system, and very high prices in Russia. Only affordable FF (till spring 2011) will save situation.
Doc_Dick · 763 weeks ago
Hope that adds something to the discussion...
David · 763 weeks ago
Really, you are comparing an entry level APS-C sensor Canon to a medium format camera? And you think the MF format will be surpassed by FF (and APS by your mention with the 550)
Well, Hasselband, Maimya, Leaf - shut it down boys, your days are numbered. Better get into entry level SLRs before you get left in the trashheap of history.
This from a guy who did not even get to touch a 645 that was turned on? Really.
I want to be nice, but you just don't get it, do you. Yes, you will never buy a 645D. I will never buy a 645D (I am like Doc_Dick, still rocking a K10 while debating the K7 but talking my girlfriend into a KX). We're not the market for this camera. But you know who will be in the market for this camera? Every portrait photographer that now rocks a full-framer who could never afford Hasselband's $25k camera or Leica's new $25k S series, but has wanted something more. Its the fashion photographer who could never part with his/her 645 system and lenses despite the world moving to digital. Its the big outfit photography studio that sees it can buy a 645D and a bunch of lenses for the same price as a second HB. Its for that diehard film photographer that has been buying all the 645 lenses dirt cheap and can afford $10k, not $25k for the other guys.
Its for the wildlife photographer who is using lenses so large a heavier camera is no big deal.
This 645D is a niche product, but it is in high demand from those who are devoted 645 fans. Pentax has a large, devoted group, if you take into account the 6X7 crew, and thankfully they made an adapter.
THis medium format camera can do things the NEX, m4/3, APS, or even FF can do.This is for billboards, this is for super posters. Are those other formats closing the gap? Yes, but they are not surpassing MF anytime soon.
Stop fighting the megapixel race, because counting is really a crock after 10 MP. Its sensor size, sensor size, sensor size. Nikon proved that with their 12MP D3s. You said it yourself in this post with the m4/3 versus NEX and Kx. A MF sensor just can do things an APS or FF cannot do.
Like I said, this camera is not for me or you. As much as I drool over it, the FPS is weak for what I do and my back would be in the shape of a C carrying this thing. But its not designed for me, and I don't think its a bad idea. Its not for you - someone who never owned any of the Pentax DSLR flagship models - either, but its selling well, which is why they have had such a slow rollout.
It would be interesting how many Hoya has to sell to break even? I am sure they reached that in the Japan market alone.....
David · 763 weeks ago
tom stermitz · 763 weeks ago
Pixel size is like aperture: factor of 1.4x times diameter, leads to area increase of 2x. This means 5.6 micron (Nikon D300) has half the light and twice the noise of 8 micron (D700).
DIFFRACTION (RESOLUTION) LIMITED:
Resolving smaller pixels requires larger apertures. Specifically, f/8 resolves about 5.6 micron, while f/11 resolves 8 micron. See Clark Vision for extensive discussion of noise, diffraction and pixel effects.
YIELD (PRICE) LIMITED:
MF is twice the area of FF, which is twice the area of APS. Yields drop dramatically with size. Note Tom Hogan's (July 26 2010) suggestion that Sony is dropping out of FF sensors. Tom estimates the FF sensor costs $500 whereas APS sensor costs $50. Manufacturing rule of thumb is retail price equals 5x wholesale parts price.
CONCLUSION:
Nikon, Sony and Canon have 21 Mpixel FF cameras with about 6 micron sensors. Leica M9 has 18 Mpixels at 6.8; Pentax 645D has 40 Mpixels at 6 microns. P645D and M9 systems eliminate the anti-aliasing filter. All these cameras can resolve without diffraction limitations at f/8. Nikon D3 with 8 microns has extremely good ISO. In other words, 6 micron is a sweet spot for noise and diffraction. That size leads to 12 Meg APS, 21 Meg FF, and 40 Meg "MF".
Maybe you can squeeze out 40 Mpixels FF at 4 microns, but your tradeoff will be more noise (worse ISO), more diffraction (less resolution) unless you shoot at f/4 (worse DOF), higher costs due to lower yields at the larger silicon area. Will pro photographers (fashion, landscape, advertising) choose your 40 Mpixel FF camera with the compromises I suggest, or will they stay with Leica S2, Hasselblad, etc?
Maria · 763 weeks ago
Therefore the question of MF or FF will not be made on basis of pixel count.
But the 645D is a stupid move. The development is far to expensive for a niche market! It's not a manual MF, it is digital. That is 2 orders of magnitude more complex. In result this means that very high development costs have to be taken back from a small market. And since in this price range few are going to buy Pentax, they will lose with this camera. If you spend a few grand on a MF system, you want lenses, you want reliability, you want a proven and working system that integrates into the workflow and you want security for the future. All these things are much easier to get with the established manufacturers. Buying MF is not like buying a Kx, that you can toss if you find out that it sucks.
Joseph Leung · 763 weeks ago
I thought RH always prefer FF over APS-C, now you are saying that APS-C noise level is acceptable for FF. What if technology improves allowing APS-C to reach 20+MP, would you say the current FF advantage vanish? Sounds like you are supporting Pentax to have APS-C format only
Davidd · 763 weeks ago
Why be against the 645D? First of all, people have clamored for this for years. I can remember way back when I got my K10 people were calling for a digital 645. Way before the FF march came...
In addition, a 645D may help in the development of a FF as it gives Pentax experience with larger sensors.
Thirdly, if you recall, one of your earlier posts, the top selling cameras were all under $1000, meanings lot more people will buy the Kx. If Sony' heralded 850 & 900 can't make it, I don't see how Pentax making a FF instead is a moneymaker, especially if they stay under $2k to undercut the competion. Nor how can the 645 development cost more than developing an Evil system.
ray · 763 weeks ago
Ben · 763 weeks ago
KeithJ · 763 weeks ago
It's not true to say that the larger sensors have greater DOF - in fact, for a given aperture, the opposite is true. You can find out more on this and other related subjects in, for instance, tutorials by cambridgeincolour.com.
If we're interested in getting the maximum DOF, we need to take into account the effect of diffraction, which will then give us a minimum working aperture.
For a given pixel-count and lens angle of view, we find that the maximum DOF (i.e. that obtained by using the minimum working aperture) is the SAME for ALL sensor sizes. For instance, at 40 megapixels, a normal lens and a 40-inch print, we have about 2ft total DOF at 10ft subject-distance, and a hyperfocal distance of somewhat greater than 100 feet. It is not possible to achieve greater DOF without reducing resolution, and therefore blurring the image.
Regarding noise, the relationship between pixel area and noise is as follows: when we double pixel area, we reduce noise by root-2. This is because when we double the area, it is like summing the output of 2 pixels: the signal simply doubles, but the two (uncorrelated random) noise sources have to be summed as the square root of the sum of the squares (giving a factor of root-2).
This seems to be borne out in Clark's tutorial. I also looked at the Wikipedia entry on this subject, and the author stated that a doubling of area would double noise, but I think this is a mis-statement. He also came up with another odd statement - to the effect that noise is dependent only on sensor size, and not on how it is divided into pixels!
So, while a 645 sensor will give us lower noise (for a given pixel-count), it won't give us greater DOF. Just to illustrate the cost of this lower noise, I've made a comparison of noise and lens weight (purely theoretical, I have to say!). The lens maximum aperture is assumed to be set at the minimum working aperture (for diffraction not to be a factor), for a sensor of 40 megapixels:
1. So-called "1/2"" sensor (actually 5.4mm) found in some P&S cameras: noise 11 units, f1.65 lens weight 1 unit.
2. APS-C sensor: noise 2.5 units, f5.75 lens weight 88 units.
3. "Full-frame" sensor: noise 1.7 units, f8.7 lens weight 296 units.
4. 645 sensor: noise 1 unit, f14.5 lens weight 1,372 units.
(The lens weight is assumed to increase as the cube of the sensor ratio, since the glass is 3-dimensional).
To make the above comparison, we are assuming that we can actually manufacture good quality tiny lenses and large lenses, that pixel sizes scale exactly with the sensor sizes, and that it's actually possible to for pixel size to approach the wavelength of light! (In the case of the P&S).
If I may draw some conclusions:
The huge weight (and cost) of glass for the 645 seem to me to be a ridiculous price to pay for the improvement in noise - and we also have to suffer lower shutter speeds and/or reduced DOF relative to the smaller formats.
IMO, the main reason medium format cameras were so popular in the days of film was that they were the only way of getting increased resolution (since film grain size is a constant). Nowadays, we have the freedom to vary pixel size (albeit with potential noise penalties). So I just don't see why you'd want a 645D. RH is right, as usual!
Finally, my thanks to RH, for such an interesting, informative and thought-provoking blog!
David · 763 weeks ago
This comment really seems to have 2 parts - one, an articulate look into the numbers to sort of sell the point of MF and lower noise. The second portion, well, it basically said the 645 is too big and with what current slrs do, it's uneeded. I like the first, but the second seems to be a copout. Couldn't I flip the pixel argument back and say that's why Pentax doesn't need FF? Obviously people will argue it's more about sensor size and viewing, which is the same thing that can be said for MF
mrgreycard · 750 weeks ago
Film grain is not constant between iso's.......