When Fujifilm announced the S3 Pro-UVIR back in August, it was very exciting to me, because I’ve been shooting digital infrared photos for about 15 months now, starting with a converted Nikon D50. Here was an official manufacturer-supported camera capable of shooting everything from ultraviolet to visible light to infrared—wow!
First of its kind (and not the last we hope)
The Fuji S3 Pro UVIR is the first of its kind—no manufacturer has ever offered a full-spectrum camera before. Fuji is to be commended for taking this brave first step. We can only hope that Canon and Nikon will follow suit, embedding true monochrome, Bayer-Pattern-free 20 megapixels imagers in an compelling camera.
Canon did offer (and might still offer) a special version of its 20D (the 20Da with mirror-up focusing ability) for astrophotography, but its spectral response was limited. It might not been available in the USA however.
Before reading this article, you might want to first read the diglloyd.com Digital Infrared article to get an idea of what to expect when shooting a digital camera using infrared light.
This article is a brief review of three cameras capable of taking pure infrared pictures: the Fujifilm S3 Pro UVIR, the Nikon D70-IR, and the Canon EOS 5D-IR. My interest in these cameras is for portrait and landscape photography. If you live in the San Francisco Bay Area, you might want to attend my Digital Infrared class in late October 2006 (if the flyer is hard to read, you’re probably running Internet Explorer as your browser).
Fujifilm loaned me the S3 UVIR for a few days—I did not purchase one. Due to other commitments, I was not able to do any field shooting with the camera, and so an up-front proviso is that the test images are rather boring—in my paid reviews, I seek out much more interesting subjects, but such reviews take many weeks to prepare. Still, the “boring shots” are more than adequate for comparing the cameras, and you will learn a lot from this review.
Announced in late August 2006, the Fuji S3 UVIR is a full-spectrum camera, capable of recording everything from about 380 to 1000 nanometers (according to Fuji). Fuji modified 500 S3 Pro camera bodies, removing the normal sensor glass (which blocks infrared and ultraviolet), replacing it with clear glass which lets ultraviolet, visible light and infrared through to the sensor.
Whereas the Nikon D70-IR and Canon EOS 5D-IR sensors appear black to the eye, the Fuji sensor looks very colorful (see below). The Fuji sales representative wasn’t sure whether the pre-production model actually had glass over the sensor or not. It appeared to me by visual inspection that there was glass over the sensor, but I did not verify this.
According to my Fujifilm sales contact, the production cameras have cover-glass over the sensor which matches the optical characteristics of the original sensor glass.
Fuji targets the S3 Pro UVIR at the forensic (criminal investigation) market. This is not a large market, but targeted customers (such as the government) need a camera that is “official” in its capabilities; a modified camera won’t do for evidence purposes, etc. There are also the warranty and support issues. So if you work for a law-enforcement or criminal investigation organization, your only credible choice is the Fuji S3 Pro, as opposed to modifying an off-the-shelf digital SLR.
Because Fujifilm sells the camera as a supported product, it comes with the usual one-year warranty, whereas modifying a Nikon or Canon body presumably voids the warranty, since the camera must be disassembled and reassembled. I haven’t had to put this theory to the test, as neither my D70-IR or 5D-IR have had any problems whatsoever.
The top marketing points (claimed) of the Fuji S3 Pro UVIR are:
- Superb image quality with high dynamic range;
- Imaging sensor response from approximately 350 nanometers to 1000 nanometers (“Generally, you can expect to easily capture acceptable images between 350nm and 1000nm”);
- A unique (for a DSLR) “live preview” mode in which the mirror is raised, and a live through-the-lens preview is displayed on the 2.0-inch, 230,000-pixel color rear LCD;
- A high dynamic range through the use of dual photosites, one with extended range.
My first pure infrared camera was the Nikon D70, modified by irdigital.net for approximately $450 (I paid $100 extra for a “hardened” filter, which is much less likely to scratch).
The filter over the sensor has a 50% cutoff at 800 nanometers, so it passes no visible light. I still use this camera, though for most shooting I now use the EOS 5D-IR.
The Canon EOS 5D-IR is now my preferred infrared camera. It is a regular EOS 5D modified by maxmax.com (see Digital Infrared). Look for a complete report for why this is so in a future diglloyd.com article. I will also be comparing the 5D directly to a converted Nikon D200, which is now in-hand (Sept 12, 2006). I likely will not be adding a Nikon D80 to my outfit, though had it been announced earlier, I might well have gone that route, though mirror lockup is a feature sorely lacking on the D80.
The 5D-IR uses a 715 nanometer filter glass (50% transmission at 750nm), so it passes 1-2% visible light (deep red).
While the EOS 5D is a 12.8 megapixel camera, its infrared performance in prints (detail rendition) is only slightly better than the Nikon D70-IR. It’s just enough to see in a 24 X 16 print when compared to the Nikon D70-IR. These comments are based on making a single 24 X 16 test print. With other lenses and other subjects, the differences might be more striking.
The 5D noise characteristics are somewhat better than the D70-IR, though not enough to be compelling (and there are other negative image characteristics to the 5D images when it comes to sharpening which were observed using an unmodified camera in color in my paid reviews). A future review will cover this topic in detail, along with the D200-IR.
The 5D is a $3000 camera, and the D70s is a $600 camera. Add about $450 for conversion and it’s a question of spending $1000 or $3500. That will make the choice easy for most users. There are of course other choices: the Nikon D80 and D200, and any of the other Canon 10D/20D/30D and Digital Rebel (see ) models. I converted the 5D to infrared because I greatly prefer the Canon EOS 1Ds Mark II operationally and in terms of image rendition. See the comparative (paid) review: D2x vs EOS.
I received my modified Nikon D200 too late to include in this comparison. Look for a future comparative review on the Reviews page in the late October/early November 2006 time frame. Those living in the San Francisco Bay Area might also want to consider my Digital Infrared class.
Infrared is the only spectral band in which we can compare the 3 cameras, since my Nikon D70-IR and Canon EOS 5D-IR “see” only infrared light. For details on those cameras, click the links in the previous sentence.
|Camera or Filter||50% cutoff (nanometers)|
|Nikon D70-IR (no spectral chart available)||800*|
|Canon EOS 5D-IR ( maxmax.com XNite filter)||715|
|Fuji S3 Pro UVIR||> 1000|
|B+W 091 (deep red) [graph, pg 6 ]||630|
|B+W 092 [graph, pg 6 ]||695|
|B+W 093 [graph, pg 6 ]||830|
In theory, adding the 695 nanometer B+W 092 filter to either the Nikon D70-IR or the Canon EOS 5D-IR should have no effect because it achieves nearly 98% transmission by 715 nanometers.
In practice, the addition of the B+W 092 filter reduces the light reaching each photosite, and causes a shift in the absolute and relative brightness recorded by each monochrome photosite—remember that these sensors use red, green and blue dyes over their photosites, and thus each photosite is a monochrome device, “seeing” only a limited range of the spectral band.
Odd things begin to happen as light passes further into the infrared band, with the color dyes (red/green/blue) apparently becoming increasingly transparent to infrared at longer wavelengths. This is suggested by the increasingly monochrome look to the resulting images (eg the red/green/blue color channels see close to equal exposure). A logical theory is that the image becomes monochrome because the red, green and blue photosites all begin transmitting about the same amount of radiation (infrared light).
The Fuji S3 Pro has been refashioned from a somewhat dated color camera into a ground-breaking full spectrum camera via the replacement of the usual sensor glass with clear glass passing the entire electromagnetic spectrum. It hits the mark for a small forensic customer base (which requires official manufacturer support), and other specialized uses. The Fujifilm-supplied Hyper-Utility-2 software is an exercise in frustration—skip it and use one of the 3rd-party raw-file converters.
Even filtering for visible light photography (using the B+W 486 UV-IR cut filter), color rendering of the Fuji S3 Pro UVIR does not appear to match what an unmodified camera might produce (but see the next paragraph). Though it appears correctable with proper white balance (not tested), the expense and hassle of needing to use the B+W 489 filter is problematic, especially considering that some users might need multiple filter sizes. At $200 just for a 77mm version, the costs add up—and the filter must be treated with care to avoid damage.
Fujifilm sales responded to the above comments on the B+W 486 by informing me that “This filter will not cut enough IR light…. Kenko makes a filter that works better, (DR-655) ”. Sounds good—certainly worth trying if you have the camera.
Users interested in general infrared shooting (especially handheld) will likely be more satisfied with a conversion of their Nikon or Canon camera body (or other brand) to either full-spectrum or infrared-only use.
Combined with the dated ergonomics, AA batteries and other eccentricities, the S3Pro UVIR feels very much passé, rather dated compared to today’s advanced Nikon and Canon bodies. Its main issues are very slow performance when writing extended-dynamic-range raw files, 25MB images (no compression), and an unnecessarily awkward means of controlling the “live view” function. Fujifilm tells me that in production models the “Live View” item is first in the menu, but such a key feature should have its own dedicated button which doesn’t require using a menu.