[PB PM]

Compression is just a result of telephoto lenses, you have to learn how to use it to your advantage.

As for VR in body and on lenses. This has been proven to not work, there were some tests done a few years ago when Olympus introduced IS into the body and there were still some 4/3s lenses that had IS in them that it actually made the blur worse than not having stabilization. Modern VR compensates for both vertical and horizontal/rolling shits already. (That is why the lens has normal and active VR settings)

[Matsu]

Oh, I know it's a product of the focal length, not the lens. It's just something I immediately started to think about after looking through my first couple of frames. You could see it's effect very clearly in two ways. On subjects I know well - family - you can see a bit of face flattening happen, and of course you can see busy backgrounds start to look 2D if you don't account for that.

On the VR thing. I don't believe it was tested an rejected from an engineering standpoint, just by curious consumers, post fact. The two 4/3rds makers just took different approaches because the spec doesn't standardize what kind of stabilization ought to be used. To work, it would have to be designed from the ground up to work in sync and apart, maybe needing both new lenses and new bodies.

Just think of all the axes that come into play. Of course there's vertical and horizontal movement of the sensor, but there's also rotational movement and front-to-back movement.

[Matsu]

So, I got out with the 70-200 for a couple of minutes at different times of the day today and last night too. I can't say enough about the value of VR in low light. I'm convinced it would help at short focal lengths as well. Also, the colour rendition and sharpness of the lens helps considerably in low light. The Sigma has a tendency to warm up a bit and combined with a greater tendency to low contrast softening it robs noisier high ISO shots of a bit of detail. The Nikon stays more neutral and sharp. ISO 3200 is not impossible with this lens, especially in RAW, it's like the file has a half stop extra room now... Tells me two things: 1.) the glass is good. 2.) My hands? Not so much...

I wonder if Nikon will update the 24-70 with VR. If they do, I'm getting it. If not, I'll wait to see if Tamron makes a new PZ VC version (surely in the works at this point given the not quite successful update to the 17-50 VC ) Sigma's new 17-50 OS looks very good, maybe there's also an OS 24-70 under wraps somewhere? The current 24-70 HSM is sharp but has somewhat nasty bokeh - though new, they should set about redesigning it to better match the performance of the 17-50 OS.

I deferred the purchase of the 50 and 85 because I needed those dollars for the zoom. I'm less skeptical about screw driven lenses for FX, because it's going to be a long time before the high end loses that function. I am now curious whether Nikon's older AF primes will reveal yet another performance plateau. I don't have the cash for their eye wateringly expensive ASF primes, but may have caught a mild case of lens acquisition syndrome.

Have to stick to plan. Work with the current kit, see what comes out in 2011..

[PB PM]

If the Nikon 16-35mm F4 VR is any indication, Nikon has no fear of putting VR in wide angel lenses.

[Matsu]

I went to test the 85 and 50 f/1.8 afterall, since they belong to an acqaintance who's not in a hurry to get rid of them. I quite like them for their size. The 85 is barely bigger than the latest fast 50's and the 50 is just tiny. I may give myself one last xmas gift this year. They're both bargains, but not quite the right fit for APSC...

[Dorian Gray]

Originally Posted by Matsu So, I got out with the 70-200 for a couple of minutes at different times of the day today and last night too. I can't say enough about the value of VR in low light.

It's pretty impressive to see the viewfinder image stabilise on a long lens. You should try Tamron's 70-300 mm VC lens if you get a chance. Its image stabilisation system sounds a bit agricultural and takes over a second to settle, but when it does its efficiency is phenomenal.

Regarding the old 28 mm f/2.8 AI-S: I bought a used copy and returned it because of an obvious centring defect. I tried another one in another shop and it too had clearly asymmetric performance. My guess is that this lens is very sensitive to manufacturing flaws. It has 8 elements in 8 groups, which probably implies the use of very precise air spaces somewhere in the design. But I've seen online samples that show it can be a very good lens even by today's standards.

[PB PM]

Originally Posted by Matsu I went to test the 85 and 50 f/1.8 afterall, since they belong to an acqaintance who's not in a hurry to get rid of them. I quite like them for their size. The 85 is barely bigger than the latest fast 50's and the 50 is just tiny. I may give myself one last xmas gift this year. They're both bargains, but not quite the right fit for APSC...

I've heard suggestions of using the 60mm F2.8G Micro as a portrait lens for APS-C cameras, although I've not tried one myself. I have the 85mm F3.5G DX micro and I like the focal length, but I don't do portraits very often.

[Matsu]

The 85 might be perfect on FX, but it's a little so-so on DX. My perfect portait trio for FX would consist of a 35, 50, and 85. Used properly, none of these would unduly distort or flatten a subject. Some faces want a lens that pronounces the features a bit more, some less. I find a little creative framing on a 35-50 focal length does a lot for chubbier faces.

On DX, the 50 becomes a 75, which is nice, and you can get the 35 to become your 50, but there's no fast/cheap 24 to become the 35? Not that there's a fast, cheap, and good 35 for FX either...

Biggest thing to get used to with the 70-200 and DX is the working distance. Too close and any faces wider than an anorexic model won't look their best. Stand back enough and it frames nicely - very good for brides with big noses.

[PB PM]

Yeah the 24mm F2.8D may be the sharpest of their "cheap" wide angle lenses, but it still isn't super cheap, and it still requires a screw driver AF camera.

[Matsu]

The 24-70 is impressively sharp, equals or beats many of the older primes from tests I've been looking at. The new 1.4 primes must be phenomenal. But it's big, costly and doesn't have VR. The size is the main thing here. The price even sort of makes sense when you think about what it might cost to buy 2-4 of the screw drivers, say 24-35-50-85. Even taking the slower f/1.8 variants of the 50 and 85, get all four and you're almost there in total cost. These are all faster apertures, but slower focusing, and you have to switch lenses. Sharpness seem to be a toss up: depends on which prime/aperture/focal length you're comparing. All close enough, however.

I think subject isolation at f/4 on FX should be the same as f/2.8 on DX at equivalent fields of view? Which from one dimension would make Nikon's 24-120 f/4VR a viable upgrade for those moving from an f/2.8 DX standard zoom. Probably quite a nice combination on something like a D700?

Get the 24-120 for versatility and a couple of primes for speed?

[PB PM]

The advantage of the 24mm F2.8D over the 24-70mm F2.8G is that the prime has less barrel distortion, so if you don't have software to fix that it could matter. I agree, from in store testing that I did on a D700 the 24-70mm is a dream lens for FX, but not so much for DX, IMO.

I think the 24-120mm F4VR would be a good general purpose zoom for FX, and not bad for DX if you pair it up with something like a 12-24mm F4. Have that zoom combo, the 35mm F1.8G, 50mm F1.8D or 85mm F1.8D and you're set for general purpose photography.

[Matsu]

The MTF charts for the 24-120 suggest that it's right there with the 24-70 performance. Reviewers seem to agree: it's happy to be used wide open (f/4), and sharp.

If we need more speed for darker environs, the question then becomes, how much speed? I ran a bunch of equivalent focal lengths through DOF master and here's what I got.

The 24-120 f/4 on FX is the equivalent of the 17-55 f/2.8 on DX - at all shared fields of view it has the same light gathering ability and depth of field when each is mounted on it' respective format.

Think of it as a 16-80 f/2.8 DX with VR. If the subject isolation and low light exposure of 2.8 on DX was sufficient, you will get at least the same from f/4 on FX with VR added in as a bonus.

My guess is that a typical FX + f/4 exposure will still keep a slight low light speed advantage over DX + f/2.8. Am I wrong?

Right now, I'll go up to ISO 1600-3200 on the D300. I don't care about the noise at 3200, it'll either be small or black and white in my final PP. A D700 would give me ISO 6400 that I can use, maybe even a smidge more?

[PB PM]

I'm not sure where you got the idea that F2.8 on DX is like F4 on FX. I get the same meter reading with a F90X film camera on any given scene as the D300 gets at the same aperture with the same lens (center weighted average metering). Light is light, that is not changed by DX or FX.

You must have contused the greater light gathering ability of the pixels on the sensor (due to larger size) thus less grain, verse physical light gathering. Technically speaking the 12MP D3/D700 have the same light gathering ability (pixel wise) as the D40/D50/D70/D100 with their 6MP sensors. The ability of the pixel to gather light has nothing to do with the light metering.

[Matsu]

I'm looking at the quality of the image across the entire frame, regardless of pixels, although they play a roll, just assume it's analogue for now, but I'm certain I'm getting lots wrong here, so let me complete my thinking and hopefully you guys can correct/fill-in the blanks.

f/4 transmits half the light per area as f/2.8, but FX has roughly 2X the area, so just about the same amount of light would be gathered under these conditions:

a DX frame at 1/50th of a sec at ISO1600 @ f/2.8 would gather the same light as an FX frame at 1/50th of a second at ISO 1600 @ f/4. However, the FX frame would be darker (half as bright) because the light is less concentrated.

To get the same exposure would require either a 1/25th sec shutter or a boost to ISO 3200 on FX @ f/4

Then, though I didn't make this clear, I guess that:

DX ISO1600, f/2.8, 1/50th sec = FX ISO3200, f/4, 1/50th sec with more or less similar depth of field and field of view.

The sensitivity advantage of larger format allowing for use of a comparatively slower lens for the same focus characteristics?

[PB PM]

I just did a test with my 50mm F1.8D, first on the D300 then on my F90X 35mm flim camera. The test was done wide open at F1.8 on both cameras, manual focused to infinity, from the same spot, center weighted average metering.

The D300 metered 1/25s at F1.8 in the test (and the D300 is known for being conservative when metering). The F90X metered 1/20s in the same spot, despite the larger metering area! Some might say that the F90X wont meter the same because it is an older camera, but when you use center weighted metering it doesn't matter what camera you use. If the test was done with matrix metering then I'd question the results, but that isn't the case.

[Matsu]

What was the "film speed" setting for each?

[PB PM]

ISO200. Now ISO200 on the D300 is more like ISO250 on film, so that could easily account for the slight difference in shutter speed.

[Matsu]

Originally Posted by PB PM I just did a test with my 50mm F1.8D, first on the D300 then on my F90X 35mm flim camera. The test was done wide open at F1.8 on both cameras, manual focused to infinity, from the same spot, center weighted average metering. The D300 metered 1/25s at F1.8 in the test (and the D300 is known for being conservative when metering). The F90X metered 1/20s in the same spot, despite the larger metering area! Some might say that the F90X wont meter the same because it is an older camera, but when you use center weighted metering it doesn't matter what camera you use. If the test was done with matrix metering then I'd question the results, but that isn't the case.

Sorry, I missed this earlier, you can't do the test that way. You have to use equivalent field of view and depth of field. So basically, the same image produced through two different formats.

Try it this way:

D300 with 50mm set at f/2.8 ISO 100

f90x with 75mm set at f/4 ISO 200

Should give the same shutter speed for the same scene, but that's not all. Should also give the same minimum and maximum acceptable focus and depth of field.

[Dorian Gray]

Originally Posted by Matsu a DX frame at 1/50th of a sec at ISO1600 @ f/2.8 would gather the same light as an FX frame at 1/50th of a second at ISO 1600 @ f/4.

Sure, albeit ignoring the small discrepancy caused by your rounding the FX/DX area ratio to 2 rather than (1.5^2).

Adding ISO values to the statement isn't necessary, though. The sensors will gather the same amount of light irrespective of the ISO values they're set to.

Originally Posted by Matsu DX ISO1600, f/2.8, 1/50th sec = FX ISO3200, f/4, 1/50th sec with more or less similar depth of field and field of view.

Yeah. This line of thinking hints at the lens issues which arise in practice when comparing FX and DX formats. To achieve equal detail and contrast in the final image, the lens on the DX camera must transfer a given contrast ratio at 50% higher spatial resolution (e.g. 60% MTF at 60 line pairs per mm for DX, versus 60% MTF at just 40 l/mm for FX), all while operating at about 1 f-stop lower (e.g. f/2.8 versus f/4), and possibly with a more retrofocus design (e.g. 24 mm versus 35 mm). This is a tall ask.

Carefully applied unsharp masking to the DX image might bring low-contrast fine detail up to FX levels, but sharpening increases noise: possibly robbing us of the advantage we gained by using a one-stop lower ISO to begin with.

Circa 2010 we still have to consider the performance of actual, specific sensors rather than generic DX and FX sensors. Generally speaking, smaller sensors are better on a per-area basis, despite their higher pixel densities. For example the Sony CCD in the Canon G12 and Nikon P7000 has a higher quantum efficiency than almost all CMOS sensors used in SLRs, though the Nikon D3S sensor is roughly its equal. However, the Sony CCD has much lower readout noise than the D3S at low ISO. So if you compare the P7000 at base ISO versus the D3S at base ISO, the P7000 does far better in deep shadows than it should; though still worse than the D3S in absolute terms.

Some cameras, like your D300, have relatively poor quantum efficiency but exceptionally low readout noise at low ISO, resulting in a high maximum dynamic range but also relatively high mid-tone noise. The D300 has a better high-ISO reputation than its performance strictly warrants, probably because it has absolutely no pattern noise, unlike all Canons and all pre-D300 Nikons. Other cameras like the D700 and D3S follow the Canon trend of high quantum efficiency, low readout noise at high ISO, but high readout noise at low ISO. This throttles the sensor's full dynamic range at low ISO, but delivers very clean mid-tones: ideal for aggressive post-production work at base ISO, or just getting a useable image at all at high ISO.

The new Sony sensor in the Nikon D7000 and Pentax K-5 combines, for the first time in a large sensor, state-of-the-art quantum efficiency and very low readout noise at low ISO. So it performs better than its sensor size would indicate in comparisons with FX sensors, especially at ISO 100. Future cameras from Nikon and Canon will have to follow. It would help if influential reviewers like DPReview would understand and explain what this means. It means, for example, that an ISO 100 image underexposed by 5 stops and subsequently pushed by 5 stops has the same noise as a properly exposed one at ISO 3200, but 5 stops more highlight range. This gives great shooting flexibility.

[PB PM]

Originally Posted by Matsu Sorry, I missed this earlier, you can't do the test that way. You have to use equivalent field of view and depth of field. So basically, the same image produced through two different formats. Try it this way: D300 with 50mm set at f/2.8 ISO 100 f90x with 75mm set at f/4 ISO 200 Should give the same shutter speed for the same scene, but that's not all. Should also give the same minimum and maximum acceptable focus and depth of field.

I did later. 35mm F1.8 on D300 and 50mm F1.8 on F90X, same result. Again ISO200, same spot etc. This time, darker subject. Both at F2.8 (sorry your theory is wrong), get similar shutter speeds. F90X 2/s, D300 3/s. The depth of field between FX and DX is the same, sorry. People talk about this over and over again, tests have shown that the depth of field does not change enough between DX and FX to be measured in terms of F2.8 vs F4. Its more like F2.8 verse F2.9. I can tell you this because I used the depth of field preview buttons on both cameras to check.

[Kyros]

No, Matsu is correct, but you aren't understanding what he's saying, PB PM. Think of it this way: let's try to take a similar photo on both FX and DX.

Take two cameras, one DX, one FX, with the same number of pixels, and assume that they are using similar noise reduction technology. First of all, the noise characteristics of FX at ISO 200 should roughly equal that of DX at ISO 100. This is because of the higher pixel density on the DX camera.

We want to get the same field of view standing from the same spot, so we use, say, a 75mm lens on FX and 50mm on DX.

Now for the aperture: since the ISO is set to 200 on FX, the aperture needs to be set a stop slower than on the DX camera. Thus, if we use f/2.8 on the DX camera, we need f/4 on the FX camera. So, your 50mm f/2.8 lens on your DX camera is basically equivalent to a 75mm f/4 lens on FX, if you want to keep noise and depth of field roughly the same. You could also take a shot at 75mm f/2.8 on the FX camera and keep ISO at 100 (which is basically what you are talking about doing), but then the shot is different. There is a shallower depth of field and there is less noise.

Using a crop sensor is even more like using a teleconverter than people think. You are basically losing a stop of aperture as well as gaining more reach. People say that putting a 200 f/2 on your crop camera is like having a 300 f/2, but it's really more like having a 300 f/2.8. You gain some reach, but you don't have the shallow depth of field and lower noise that you would with a 300 f/2 on FX.

I hope this makes sense, I really need some sleep.

[PB PM]

We are talking about two totally different things. I'm talking about light metering, not depth of field, contrast... etc. F2.8 will meter the same on a FX, DX or an iPhone, it doesn't mater. Light is light.

Also the theory that ISO200 on a DX body is more noise than ISO200 on a 35mm frame body isn't correct either. At ISO200 the D300 and D700 have the same noise characteristics. The D700 has slightly more dynamic range (see DXO marks website), but in terms of pure signal to noise ratio that isn't true. That starts to become true north of ISO1600, but until then output is very similar, again in that range the D700 has more dynamic range, but that is a side issues to noise. Now take a more modern camera like the D7000 or D3100 and you'll see that those DX bodies have better performance, despite higher resolution, than the D300, coming close to matching the aging D700.

[Matsu]

Of course we run into trouble comparing different generations/brands of sensor. We can't completely take that out of the equation but can minimize it by comparing platforms within the same generation.

If, for the purposes of lens-platform comparisons, we assume that -- per unit area -- any given sensor will be more-or-less capable of measuring light with the same efficiency -- think of a sheet of the same film cut into different sized formats -- then we can figure out what sort of lens covering what sort of image circle will give the same look.

For APSC to 35mm (DX to FX, EF-S to EF) focus distances and depth of field measurements indicate that f/4 on 35mm gives the same subject isolation as f/2.8 on APSC, but not the same exposure. For that we either need more time or more sensitivity - no longer the same "film".

Now digital is not film, though for comparison I just wrote that we should treat them all the same, which is what PB PM (I think in part) was trying to answer.

I think I understand the source of disagreement:

Use the film cut into pieces analogy. It performs exactly the same per unit area. For digital, any DX pixel would be no different from an FX one if you apply our strict film analogy, it's just that you have more of them.

If you turn up the sensitivity and double the noise at the pixel level, at the same time you have twice as many pixels over an area twice as large. Because the reproduction ratio for 35mm is half that of ASPC for the same size print, apparent noise would be no worse for it, but you would have an ability to print larger before it became objectionable.

None of this even gets into the effects of discrete sampling that the pixel structure of a modern sensor introduces, less but larger pixels, more but smaller, more efficient etc etc...

What we do know, is that the diffraction limits of the lens put a stopper on ultimate resolution available. Past a certain point the only way to get a bigger 'print' is to make a bigger 'negative'.

[PB PM]

I wasn't talking about digital vs film at all, but that is an interesting thought.

Diffraction can be an issue, but unless you are shooting landscapes or macro how many people stop down past F10 anyway? Of course, as we continue to push more resolution into our cameras sensors diffraction becomes more of an issue, because that limit becomes more apparent at lower F stops.

It would be great if we all had the money, and muscle to drag around medium format cameras to get around that, but who wants to do that?

[Matsu]

Love this tutorial

Diffraction Limited Photography

The mid page interactive widget which shows pixel pitch for a few different cameras relative to airy disk sizes is instructive for a whole other reason as well. The discrete sampling of a digital sensor grid and how it interacts with the light projected by the lens. A 12MP APSC starts to show some diffraction effects by f/11, but that doesn't mean that higher pixel densities couldn't improve resolution even at smaller apertures. The light draws continuous lines across the frame, not discrete points that either fit or don't. Eventually, we'll have so many pixels that they are no longer the measure of resolution. That will be a function of the lens in use.

[Matsu]

Hey guys. The 50 and 85 1.8s ended up being too cheap to pass up, so I took them as a package. I might flip the 50 for whatever I can get, even though I prefer the way it frames on DX, and keep the 85 just in case the market for used FX cameras improves in the new year...

With either of these mounted the camera is completely different. It's super light compared to my zooms. I need faster shutters when using the 85, about 1/100 to get sharp, and it's not too hard to hold there.

So, I turn my attention to flashes.

Would you go for a single SB-900 or dual SB-600s ?

I have some stuff lying around here ready for some experiments on a ring flash adaptor design that I drew out. The best store bought stuff reports between one and four stops of light loss, not ideal. I'm don't think I can do better than their level of efficiency because I'm not using computer aided design and don't know enough about light. That said, I have a couple of ideas to improve the handling and focus beam interference of the Orbis and Ray Ring respectively.

Getting it to light up evenly is another challenge. None of the adaptor based options really do. Two flashes could help here.

I'm curious: is anyone here using one of these circular light modifiers? Do you find that the zoom or wide settings are best for getting an even light distribution?



I wouldn't do too much complex off camera lighting at first, but I imagine I could get a lot of great effects with two flashes plus the camera's commander mode.

I can make my own custom flash brackets and such

[PB PM]

If you aren't going to do a lot of flash work, go for one flash. I have the SB-800 (sb-900's predecessor) and don't regret having it rather than two SB-600s. For one thing the SB-800/900 can act as a command unit, so if you do add other flashes later then it doesn't matter what camera you have. Also take a look at the new SB-700 (which can also act as a commander) if you want to go the cheaper route. The SB-600 is an old flash, and may not even be supported by the next generation of cameras.

For macro work take a look at two or more SB-R200s rather than standard speedlights. They are only $200 so you could get two or three for the price of the SB-900, depending on where you are.

[GSpotter]

Originally Posted by Matsu I wouldn't do too much complex off camera lighting at first, but I imagine I could get a lot of great effects with two flashes plus the camera's commander

I use two SB800s (I bought the 2nd flash when I had my D2Hs, which didn't have an internal flash). Having 2 remote flashes helps quite a bit with lighting, e.g. you can use the 2nd one as backlight to enhance the outline of the subject...

[Matsu]

The SB-R200 is too much of a specialized tool for what I would want to do with it. One look at the guide number tells you that it only has a small fraction of the power of the rest of Nikon's line-up.

For command duties it looks like either of the SB-800 or 900 are a good choice because they control an additional group. SB 700 can only control 2.

Depending on price and application, multiple SB 600s could be the better choice for slaves - they're spec'd to work with all Nikons wireless remote modes - just not as commanders.

[evan]

Got an 85mm 1.8 for early xmas present Still on vacation with the family but I'll upload some shots I took with it this week into the photo sharing thread.

but all this talk about flashes got me thinking... what's sorta a good "intro" into flashes? I don't think I've taken a single picture (out of several thousand) with flash since I got my DSLR this summer.