The Naked Sensor
Nikon D800E – Bayer Pattern Sensor, “Optical Low Pass Filter Without Anti-Aliasing Properties”
A little more than a year ago Michael Reichman, Nick Devlin and I were all sitting down to lunch in a cafeteria at the Jacob Javits Center in New York City. We were there for the annual PDN Photo Plus show and the conversation was largely about which new cameras seemed most interesting to us. “For me”, I said, “It’s the Fuji X100. Fuji is now trying to out-Leica Leica.” And what I meant by that was that Fuji had introduced a competitor to Leica’s X1 that not only looked a lot like a Leica M camera – complete with various analog controls – but which also featured a window finder with frame lines. That type of window finder has been integral to the Leica M camera line since it was first introduced and Fuji did a good job of integrating such a finder into the X100 (albeit without a rangefinder). The “M” in “M camera”, of course, refers to “messsucher” – the German word for rangefinder.
That lead us to a conversation about whether or not the X100 had an AA (anti-aliasing) filter and, if so, how strong it was. The Leica M8 and M9, of course, have no such filter. As some may know, I’ve been arguing – over the past few years now – for the advantages of cameras that do not use these filters. In our lunch time brainstorming I suggested that at some point in the not too distant future we were going to be seeing more and more cameras with either very light AA filters or none at all. In my mind, then and now, the AA filter is one of the key elements that “throttles”, so to speak, the potential of a digital camera. I noticed this, emphatically, when I was testing a set of Zeiss ZF SLR lenses on both the Nikon D700 and the Canon 1Ds Mk III (the latter via an adapter). Again and again I saw much higher resolution (from the exact same lens) when it was mounted on the 1Ds Mk III. Now, of course, the files from the Canon were dimensionally larger (given its 21 MP sensor as compared to the 12 MP sensor of the Nikon) but what I came to realize was that what I was really seeing were differences between a camera with a stronger AA filter (the Nikon) and one with a weaker AA filter (the Canon). Like many photographers, I’d long known about this difference but that experience sharpened my sense of how important this factor could be.
The purpose of an AA filter, of course, is to slightly blur certain high frequency detail so that it doesn’t create artifacts. The most notable of these artifacts, of course, is moire (a word apparently derived from a French term for a certain kind of textile). And indeed, textiles with certain fine patterns can very often provoke color moire in a Bayer-sensor digital camera without an AA filter. I know this first hand, having photographed many weddings with Leica M8 and M9 cameras. Kodak experimented with making small format DSLRs (such as the DCS 14n) without AA filters and Leica obviously decided they wanted to nothing to do with a device that would blur detail in their DMR back (for the R8/R9) or, later, their DRF cameras or, later still, the S2.
Leica S2 – Bayer Pattern Sensor, No AA Filter
It’s probably no coincidence that Kodak sensors are in all three of those Leica models (listed just above) and that both companies have chosen to make cameras without AA filters. I’ve never asked Leica about the following but I can imagine some early prototype tests in which an AA filter was included in a DMR test mule. I then imagine the optical design team, in particular, being rather aghast at what that filter had done to the fine resolution delivered by their lenses. Leica lenses, as a rule, are quite expensive and a great amount of time and money goes into tweaking their performance. To then throw some of that performance away, via the AA filter, probably seemed anathema to the company. I can’t say that I disagree with that thinking.
There were already precedents for Leica’s unusual decision, of course. Not only had Kodak tried DSLRs without AA filters but medium-format digital backs – as a rule – did not (and still do not) use them. After all, who wants to spend thousands of dollars on something like a Phase One P65 back only to have part of the lens’ resolution removed – by an AA filter – just before it reaches the sensor.
Ricoh GXR With Mount A12 – Bayer Pattern Sensor, No AA Filter
A look at current cameras without AA filters would be incomplete without some discussion of Ricoh’s Mount A12 unit. This “camera module”, which fits in a Ricoh GXR body, is unusual for two reasons. First, it features not only a Leica M lens mount but also an optical system (including micro-lenses) that is specifically designed to work with rangefinder cameras lenses. The second is that it features no AA filter (something I had been urging the company to do for several years). Recently I have been working on a series of tests to look at how two EFCs (electronic finder cameras), the Ricoh GXR Mount A12 and the Sony NEX-5n, performed with a series of “difficult” rangefinder lenses.
The lenses were mounted on the Sony using a Voigtlander adapter. I was especially interested in how each camera would perform in the outer zones (with respect to resolution, luminance vignetting and color drift) and it turned out that there were indeed notable differences between them. But what I also discovered (in the course of doing careful, controlled, focus-bracketed studio tests of resolution) was that – with every tested lens – the Ricoh out-resolved the Sony on center. The key difference between them, in this respect of course, is that the Sony uses an AA filter and the Ricoh does not. Of course, if one has to up-res. the 12 MP Ricoh file to match the dimensions of the 16-MP Sony file the former’s resolution advantage is lost. But looking at actual size files without interpolation (made using the same lens) the resolution advantage of the “non-AA filter” camera was undeniable.
Leica M8 File Converted in C1 DSLR Pro (Left version as converted, right version spot corrected for moire.)
The higher end of digital photography has long done without AA filters. Kodak brought this approach to DSLRs and Leica brought it to their DMR back, their DRFs and their S2. But this strategy comes at a price, of course – the photographer sometimes has to deal with correcting color moire. I’m sure that many professional photographers like myself know the sinking feeling of being under deadline for a client or art director and having to spot-correct moire in dozens or more pictures from a shoot. I recall once having to do this for about fifty pictures made during a wedding shoot (done mostly with my trusty M8.2). The challenge of life without an AA filter is real and it can sometimes create a lot of post-processing work. I should note, however, that color moire certainly can be *spot* corrected. One doesn’t need to, and I would say shouldn’t, apply the corrections to the whole picture. Not all photographers will want to take on the correction of moire in exchange for a gain in resolution (a fact that Nikon has implicitly acknowledged in making two versions of the D800) but I’m glad to see more options appearing for those of us who are.
Sigma SD1/SD1 Merrill – Foveon Sensor, No AA Filter
But this, of course, assumes that one is using a camera with a Bayer pattern sensor. What if a sensor was designed so that it was not prone to color moire in the first place? That, of course, was part of the thinking behind the creation of the Foveon sensor. At this point, many of us are well aware that Foveon sensors have some weaknesses. But what some may forget is that Foveon cameras not only do not use AA filters but also aren’t subject to color moire. With only one exception (to be discussed below) I can’t think of a single non-Foveon camera that even claims to combine those two advantages. To date, the Foveon cameras I’ve tested have all had some problems with color accuracy and high ISO rendering. But I must say that I have often been deeply impressed with the ways in which these cameras render color, tone and detail. So while there are some flies in the Foveon ointment I still think the technology has undeniable strengths. I’ve looked at certain pictures made with these cameras and just whistled through my teeth at how beautiful their rendering was. It’s a subjective response…certainly…but has been a strong one at times.
Sigma SD1/SD1 Merrill – Foveon Sensor, No AA Filter
With all this mind, I was quite excited when I first heard about Nikon’s D800E which – of course – claims to use, in Nikon’s words, an “Optical Low Pass Filter Without Anti-Aliasing Properties”. It’s a seemingly strange approach (no doubt chosen for some practical reasons) but if it works, it works. I won’t know, until I test the camera, but if the D800E renders like a camera with no AA filter its output may be quite remarkable. That said, Nikon is very wisely cautioning photographers to expect moire from this camera (just as we’d expect moire from an M9 or a medium format back). That’s prudent and Nikon is going one step further by offering photographers a choice. One can have this D800 with or without anti-aliasing of the image. So one can choose higher resolution with some risk of moire or slightly lower resolution and less (or no) risk of moire. What more could one ask for in that respect? It’s a novel approach to offering camera choice and I applaud it.
That said, it would be a mistake to assume that the Nikon 800E has made the (newly affordable) Sigma SD1 irrelevant. The SD1, which I have tested, delivers the high resolution of a camera with no AA filter coupled with a lack of color moire. The D800E, according to Nikon’s own statements on the subject, cannot offer this. In fact, no Bayer Pattern camera, to my knowledge, can offer it. Based on my experience with the SD1 and my experience with the D700, I imagine the D800E may have several practical advantages over the Sigma (faster operational speed, better overall color accuracy, better high ISO performance, video functionality, more choices for RAW conversion software, access to a wider range of lenses, etc.) but some of those are just guesses since, again, I have not yet tested the Nikon. Still, I think it’s safe to say that the two cameras will not render in the same way (even apart from the moire issue) and photographers will have different preferences when looking at those different renderings. How important is the way a certain camera/lens pairing renders? For some photographers it is paramount. In other words, whatever limitations the SD1 has (and it does have some) there will be certain photographers who embrace it because of the way pictures made with it *look*. And now that this camera is not priced in the stratosphere more people may decide to find out how much they do or do not like that look. Since Sigma is no doubt reading this essay let me again strongly urge them to sort out the SD1’s color accuracy and work with Adobe to get the camera’s files supported in Lightroom (ideally also working with Phase One to get them supported in Capture One DSLR).
Fuji X-Pro 1 – “X-Trans” Sensor, No AA Filter
And then, of course, there’s the very interesting Fuji X-Pro 1. I have not yet tested this camera so I don’t know what its files look like. But Fuji claims to have done something rather remarkable. They tell us that this camera uses a unique type of “X-Trans” color filter array (not a standard Bayer pattern) that is not prone to color artifacts such as moire and thus, they claim, needs no anti-aliasing filter. If this works, they’ve essentially matched the one-two punch of the Foveon sensor (no AA filter and no color moire). That’s not to say that the Fuji will render like a Foveon camera but rather that the company has taken another path towards accomplishing those two goals. Needless to say, I am very curious to see how well this new Fuji system really works in practice. I congratulate the company on trying a new approach that could be very promising.
Leica M9 – Bayer Pattern Sensor, No AA Filter
Not long ago, the only small format digital cameras being produced without AA filters were the Sigmas and the Leica M8/M9 DRFs. The Kodak DSLRs and the Leica DMR back, unfortunately, are no longer made. But now, in a fairly short period of time, we’ve seen the announcement or introduction of several new cameras without AA filters (or the equivalent): the Ricoh Mount A12, Sigma SD1/SD1 Merrill, Sigma DP1 Merrill, Sigma DP2 Merrill, Fuji X-Pro 1 and the Nikon D800E. Add the existing Leica M9/M9P into the set and we have a group of seven. Of course, that’s only seven cameras but they are seven very interesting cameras and I think they show us an important direction that at least some small format digital cameras may be headed towards in the future. My “Naked Sensor” term is a bit of hyperbole, of course. These camera still have – and need – micro-lenses, infrared filters, color filters, etc. Their sensors are far from naked. But by stripping away the AA filter (or, in Nikon’s case, using a system that is supposed to do the equivalent) they’re getting closer to being able to render a lens’ drawing without degrading it. In other words, these cameras are moving towards a greater fidelity to the image cast by the lens. And that, in my mind, is important progress in a worthwhile direction.
Added February 15, 2012: I heard from a reader, not long after this article was published, who has found that certain Sigma cameras, though not subject to color moire, are subject to luminance aliasing. I haven’t experienced that, or perhaps haven’t noticed it, but it certainly is possible. His comment is a fair reminder, though, that color moire is not the only potential artifact that an AA filter tries to counter. That said, however, I’ve been working with Leica M8 and M9 cameras for over five years now and I can’t recall ever having a single picture ruined by luminance aliasing. I’m sure it’s possible but it hasn’t yet been a problem in my personal or professional work. I do, however, as I noted above, sometimes run into color moire that I have to correct for. So in my own experience with the Leica DRFs, color moire has been a more important issue than luminance moire – and even the latter hasn’t been a frequent problem in my pictures. Moreover, what I notice most often in the M8/M9 files are not artifacts but high levels of detail (at least at lower ISO levels). Others’ experiences with this, of course, may be different.
While it’s true that eliminating the AA filter is not a panacea, the advantages – for my needs at least – tend to outweigh the disadvantages. So perhaps the fairest way to look at this is to say that there are pros and cons to AA filtering and that a photographer, ultimately, has to decide which of those pros and which cons matter most to him or her. That makes Nikon’s decision to offer the D800/D800E especially interesting because in essence the company is saying to the photographer: “You can have this camera either way. You decide.” I, personally, lean towards cameras without AA filters but I hold that preference with the full knowledge that such cameras can sometimes trigger artifacts that require time to correct in post (or which, sometimes, cannot be corrected).
This article was first published in February of 2012 on Luminous Landscape.
Many of Sean Reid's reviews, essays and other articles can be found in the subscriber section of Reid Reviews. They include an extensive review of the Leica SL and tests of various rangefinder and Leica R lenses on both the SL and the Leica M-240.
Reid Reviews accepts no advertising and is paid for entirely by subscribers. Written primarily for professional and serious amateur photographers the site has become known for its in-depth analysis based on both field and studio testing.