Nikon D7100

[ChrisR]

I see that the new Nikon D7100 (DX, 24MP) has done away in some form, with the low-pass filter. I haven't seen any details on how that may differ from the treatment in the D800E.

It could be a good body for hi-res objectives.

[Charles Krebs]

I haven't seen any details on how that may differ from the treatment in the D800E.

It is different.

See this page:
http://www.dpreview.com/reviews/nikon-d800-d800e/3

[Chris S.]

Charlie, could you elaborate a bit on what differences you expect? I understand what Nikon has done with the D800/D800E in regards to low-pass filter vs. non-filtering glass, but have not yet found details on what they will do for the D7100 in this regard. Presumably, it would be a different approach, since they don't have to deal with making two versions (with and without low-pass filtration) of the same body, but it would be interesting to know more specifics.

Based on the rather limited information so far, the D7100 looks more than a little bit interesting for the macro studio.

--Chris

[ChrisR]

That's pretty much my thoughts. Ref the D7100, I doubt they'd have simply left out any sort of filter at all. If a camera's being re-hacked, then removal is the only option so we get the idea that they simply, either have a low pass filter or they don't.

I know some of the other mfrs say they don't have LP filters but don't they all need demosaicing filters? None of them publish frequency response afaik!
Perhaps Nikon's "leaving it out" is a shorthand to say they just didn't go through the rather odd route used in the D800E for response modification.

[Charles Krebs]

I can only speculate...
From a user standpoint I doubt there is any practical difference. It would certainly seem that manufacturing the two versions of the D800 is why they used the "separate and re-combine" filter method shown in that Nikon diagram on the DPReview page. The practical manufacturing reasons DPReview suggests for doing so seem logical.

With a single "version" of the D7100 whatever filters are in front of the microlens and color filter array would be simpler and preferable. (Less expensive, and less room for any "error" in the separating and recombining filters.)

[Craig Gerard]

There are currently a number of discussions appearing on the internet with reference to the video capabilities of the D5200 and D7100; from those discussions I pulled the following quotes:

In reference to the D7100

The other big change is that the camera dispenses with the anti-aliasing filter altogether like the D800E. There’s no version of the camera *with* an anti-aliasing filter. This could be one of the more disturbing aspects for video shooters, however moire & aliasing on DSLRs is largely to do with how the sensor is sampled. On the D800E moire and aliasing wasn’t really any different in video mode to the D800 with an anti-aliasing filter.

The Nikon D7100 is here and appears to be based around the new sensor and image processor of the D5200 which is proving rather good at video so far.

The D5200 is an extraordinary camera. The street price is $700 yet it’s better for video than the flagship full frame Nikon D800, with none of the same moire & aliasing issues.

The Nikon D5200 is the best mid-range Nikon yet for video. What is more of a surprise is the relish in which it takes on the much more expensive 5D Mark III, Panasonic GH3 and Nikon’s own flagship the D800 and D4.

Craig

[Joseph S. Wisniewski]

I'm in essay mode today...

A long, long time ago, most digital cameras had a plain old piece of glass as the "window" on the sensor package. In front of that, was an independent four layer filter pack,

1) a blue-green (Schott BG-34 or similar) absorption filter
2) a LiNbO3 birefringent plate (think Normanski prism in a DIC microscope) to separate the light into two slightly displaced images
3) a 1/4 wave retarder to take those two images having different linear polarization and render them circularly polarized
4) a second LiNbO3 birefringent plate, 90 degrees to the first plate, to turn the double image into a quadruple image.

That quadrupled image makes each ray of light contribute to four focused points of light, and that square "point spread function" kills aliasing.

At a price. Any good macro person knows a thick stack of "clear stuff" in our optical path is "bad". That's why high NA objectives are designed and labeled for particular cover-glass thickness, and some have adjustable compensating collars.

Small sensor cameras weren't much of a problem, the small pixels meant thin plates, because you didn't need as much image shift, and you were more on-axis. FF cameras needed thicker plates, and dealt with more extreme angles, and have more problems. You can't thin the plates, their thickness determines that point spread function of the AA filter.

But you can get rid of that one "useless" layer of clear glass on the sensor, replacing it, instead, with the last LiNbO3 plate from the stack. You can't glue the whole stack to the sensor, because the process of doing that is hostile to the adhesives that hold the stack together, but you can do one plate. The Canon 5D was the first camera to have a sensor with a LiNbO3 window, but it's now standard on FF from Canon, Nikon, and Sony.

So, now you have one birefringent plate glued to the sensor. Every variant on a sensor is expensive, you need to rerun testing to make sure a sensor with an LiNbO3 window isn't more physically stressed than one with a glass window, since LiNbO3 and glass have different thermal coefficients of expansion. They outgas different chemicals (not sure if LiNbO3 out gasses, but glass sure does. You can smell it, if you know what you're smelling). So, there's an expensive design validation for each new window, and separate production runs. D800 is low quantity, and that makes no sense to do for the E. It would cost thousands more, like the separate sensor runs in the Leica M monochrome or the Phase One medium format monochrome.

There must be a cheaper way. And, my God, if anyone would figure it out, it would be a company that made cameras and DIC microscopes. Skip the phase retarder, change the angle of the plate in the stack so that it's 180 degrees from the one on the sensor, and viola, the sensor plate undoes the birefringent point spread of the pack plate.

I'm guessing D3200, D5200, D7100, and D9000, like most APS cameras, have glass over the sensor, and both LiNbO3 plates in the stack, so, to kill the AA on D7100 and D9000, you just need a new stack, simply the blue-green glass. Heck, you could put that on the sensor itself, and go naked, no stack at all, except that you need "something" hanging loose in front of the sensor for a "dust shaker" mechanism.

[Joseph S. Wisniewski]

There are currently a number of discussions appearing on the internet with reference to the video capabilities of the D5200 and D7100; from those discussions I pulled the following quotes:

In reference to the D7100

The other big change is that the camera dispenses with the anti-aliasing filter altogether like the D800E. There’s no version of the camera *with* an anti-aliasing filter. This could be one of the more disturbing aspects for video shooters, however moire & aliasing on DSLRs is largely to do with how the sensor is sampled. On the D800E moire and aliasing wasn’t really any different in video mode to the D800 with an anti-aliasing filter.

This keeps appearing. It's a total non-issue.

The still cameras have their AA filters set up for the pixel spacing of a still image of 16-36mp. They somehow decimate that into a 2mp video image (pixel binning, row and/or column skipping, etc). In any case, a 16mp camera needs a PSF (point spread function) 2.8x larger for video than it does for stills. A 36mp camera needs a PSF just over 4x larger. So a D800E, with "essentially" no AA filter, doesn't cut any of the moire in a video, but a D800, with a filter less that 1/4 the PSF size it needs, cuts less than 1/16 of the moire, an invisible contribution, over 4 full stops less than you need.

Shoot a D800 and D800E side by side, and you will not notice a difference in the moire. Shoot a 5DIII and negate one axis of the AA filter by using a linear polarizer (been there, done that) and again, you won't see a speck of difference.

If you want to see a difference, you carefully load a new AA filter into the camera, one with a larger PSF. There are a couple of companies selling these: I use one from Mosaic Engineering.

http://www.mosaicengineering.com/products/vaf/5d2.html

They're expensive, around $400 for FF, $300 for APS, and they're a little tricky to put in. You do it either by lifting the mirror by hand, or doing it on bulb, but it's not that bad. I'm sure, with D7100 and D9000, and more Canon "rebel" class camas getting into more people's hands, we'll be seeing a larger demand for slide in filters, and the Chinese moving in at 1/4 the price of a Mosaic.[/url]

[g4lab]

Joseph W.
Thank you for the explanation of the anti aliasing filter and how it works.
Very interesting. I knew they were there, but had no idea they were so cleverly made. I thought they were microscreens. Maybe they used to be.