The Phase One IQ250 CMOS Fully Realized
by Doug Peterson
I’m a technically minded nerd. I’m just as glad to discuss microns and lens design over a beer as sports or politics. But when an entirely new product comes out I like to know more than just the tech specs – I like to hear the story of how it came to be, and the people that helped make it.
If you’re like me, then this article is for you. It’s the story of the IQ250, the world’s first shipping digital back based on a modern CMOS sensor, as told by the team that created it. And it begins much earlier than you might imagine.
The new Phase One IQ250.
Sparks of the Future
On a cold winter day in 2010 a small group huddled around a bench in Phase One’s Copenhagen research and development lab, flanked by several R+D engineers. There was a sense of electricity in the room; so far, it was emanating only from the engineers. For months they had been working on a radically redesigned digital back platform. But until this moment, no one outside the R+D skunk works team had seen their work. This group represented their first guinea pigs, and their guinea pigs looked confused. Strewn on the bench were several components connected, seemingly haphazardly, by hand-twisted interface cables. In the center was a LCD screen with a touch sensitive layer attached by carefully cut duct tape. It was barely recognizable as a digital back.
In retrospect, the guinea pigs must be given some leniency for their initial lack of enthusiasm over this pre-prototype. Most people have a hard time looking at a set of components and envisioning a shipping product – that’s what makes R+D people special. But a moment later, one of the team, firmware engineer Lau Nørgaard, hit the power button, transforming the lifeless LCD into a bright, high resolution image, and the group’s posture changed a bit. He then double tapped the screen. The image jumped to a crisp, 100%view and sparks ignited throughout the group. From that moment on, the R+D team had no doubt their project was the future.
Many months later Phase One would announce the Phase One IQ, to nearly universal fanfare and a long list of pro camera firsts – retina LCD, quadcore primary processor, USB3*, focus mask, adjustable exposure warning, UDMA7 CF card support, and automatic perspective/horizon correction. It had a huge buffer for prolonged fast shooting, a blazing fast touch screen interface, and, notably, it included in camera Live View.
The IQ project was an enormously expensive R+D project. It involved engineers from every area of the company – an entirely new Operating System was created from scratch, every circuit board was entirely custom designed to fit into the required chassis size, and an entirely new set of design tools was created. Up until that that winter night it wasn’t clear that it would all come together. Understandably, the excitement in that room, when the system’s “ready” chirp sounded, was enormous.
When the IQ140, IQ160, and IQ180 shipped, the broader photo community recognized them as three revolutionary backs. But the truth is that it was something more. The IQ was a platform for the future. As Lau put it, “We didn’t want to make a new back. We wanted to make a new way of making backs. The IQ platform was built to be internally modular, modern, and faster to make adaptations and improvements, faster to test and to troubleshoot.”
The harsh reality is, like most digital back manufacturers, Phase One had been building every back on top of the previous back; features were added on top of existing architecture, new components were swapped in and made to work with systems designed for much older components, code was written on top of previous code. The IQ project changed all that – it provided a fully modern, ground up design. In addition to the visible IQ products, Phase One created a new set of in house tools, designed to make the process of developing future backs easier and more robust.
The reason the excitement in that room was so feverish was that the Phase One IQ project positioned Phase One for the fast approaching future. A future that might include wireless, or GPS, or higher resolutions. A future that might include high quality CMOS.
SAAD, Sensor Acronym Avoidance Disorder, effects two out of three photographers
SAS: Sensor Acronym Soup
For those who don’t obsess over acronyms and technical minutia, a broad strokes overview of CCD vs.CMOS is probably in order. In the same way that an electric engine and gasoline engine can both power a car, CCD and CMOS are two types of sensors that can be used in a camera. They are fundamentally different ways of getting to the same end goal – capturing an image. CMOS offers several usability advantages over CCD. For instance, CMOS is typically a bit easier on battery usage. And CMOS sensors are fast, enabling fast frame rates and the kind of movie like Live View you’d find on, for instance, a 5D Mark III. Finally, CMOS can deliver better high ISO results at full resolution.
The 6mp Leaf C MOST featured a 36x24mm CMOS sensor.
It was released in 2002 and met with limited success; CMOS was not ready for medium format.
These usability advantages has been attractive to (and asked for by) medium format users for years. But until the IQ250, with one unusual historical exception (the Leaf CMOST), every digital back that has ever shipped has been based around a CCD sensor. The quality, especially regarding tonality, dynamic range,and color, of CCD sensors had been untouchable in the world of CMOS sensors.
But around the time Phase One began the IQ project it was clear that the huge advantage CCD had overCMOS was not truth written in stone. If and when CMOS ever became a viable option for those who obsess over image quality, Phase One needed to be prepared to take full advantage of it.
Photo by Dan O’Day. Phase One IQ250 at ISO1600, 1/25th @ f/2.8, with Phase One 45mm D lens.
Illuminated by car headlamp after sunset.
How Hard Could it Be...
Building a platform that would be ready for CMOS meant several requirements. Firstly, the back’s internals had to be very, very fast. Internal bus speed had to be on par with a Mac Pro computer (the 50mp Sony sensor can sustain a data flow of 2.3 Gb/s). Likewise, fast CF card read/write speed was a must for fast sustained untethered shooting – otherwise photographers would quickly hit a buffer limit.Other requirements for speed were initially less obvious, and required significant design considerations. For one, there was no off the shelf image processor that could handle the speed of processing required to keep up with the rest of the system, so Phase One custom designed a four-core primary image processor, and several auxiliary processors to handle other tasks so that the user interface wouldn’t be slowed down when the primary image processor was under heavy load. This also allowed for future WiFi features to function even when shooting quickly.
Photo by Dan O’Day. Phase One IQ250 at ISO100, 1/320th @ f/7.1, with Phase One 45mm D lens.
Before and after highlight recovery. The dynamic range of the IQ250 provides excellent pliability in post.
But the most important speed consideration wouldn’t come into play for years. The IQ platform would need enough power to handle the continuous stream of data produced by CMOS Live View, and handle it well. Taking a live feed from a 50mp sensor and down sampling it in real time onto a digital back LCD with retina resolution is no simple task, especially if you want a high quality result. One option is to subsample – simply skipping 90% of all the rows of pixels from the sensor. This considerably reduces the amount of data that needs to be handled, but leads to very poor image quality with false impressions of sharpness and detail, on top of other significant issues. To provide high quality Live View the entire sensor needs to be read out, and then run through sophisticated, and computationally intensive, algorithms to be reduced down to proper viewing resolution. Moreover, this must be done on the fly, and nearly instantly, so that the video does not have any meaningful delay or lag. This requires massive processing power.
To provide this Live View to a computer in a way that provides excellent quality (where a large monitor can allow excellent collaboration between photographers, models, clients, stylists, etc) would require USB3. More than any other feature on the IQ platform, USB3 laid the groundwork to get the most out of a CMOS sensor. The bandwidth of the, then still nascent, USB3 protocol, offered the possibility for large window, high resolution, high frame rate, low latency live view, critical for still life and product shooters who wish to use live view for real time evaluation of composition, focus, and content. It was perhaps serendipity that Phase One’s success in the lucrative Industrial Aerial market took off with the launch of the IQ platform. This very demanding market requires that a high frame rate be sustained continuously for, literally, hours at a time. Meeting that demand meant considerable time was spent fine tuning Phase One’s implementation of USB3 to get the absolute fastest transfer rate possible. This work has paid off in providing the incredible high quality live view on the IQ250.
Photo by Dan O’Day. Phase One IQ250 at ISO800, f/2.8, with Schneider 110LS.
Screw Location – Color. Color. Color.
But Phase One is not a speed and feature obsessed company. They are a company that cares first, second, and third about image quality. So when Sony approached Phase One with an offer to build what would eventually become the IQ250 sensor there was one looming questions, could Phase One tame CMOS color?
Historically, CMOS has not had the best reputation for color rendition. But teasing apart cause and effect has been, up until now, very difficult. CMOS and CCD were being used by very different companies in very different systems. Most CMOS cameras are built for the broadest possible range of applications. They are built by consumer electronics companies with a volume sales business model, where features and price are higher priorities than image quality.
As one example, the selection of a CFA, the color pattern put in front of the sensor, is a choice between quality of color, and ISO performance. If the CFA allows each pixel to see a broader spectrum of color (e.g. for the green pixels to see a bit further into yellow) a camera’s ISO range can be modestly increased. The resulting loss in color quality is subtle – subtle variations in color are missed and a handful of specific colors become difficult to photograph. In a market where a ISO 25,600 camera has a leg up on a ISO12,800 camera, the engineers are under enormous pressure to pick the modestly increased ISO over subtle color quality.
Photo by Tim Kemple. Phase One IQ250 at ISO3200, 1/1250th @ f/2.8, with Schneider 55LS.
Copenhagen, We Are Go / No Go For Color
This sort of mentality is blissfully lacking in the R+D at Phase One. They do not build a system for everyone; they build systems for the photographer that is very discerning about image quality. So where one might have expected an extended debate there was none – there was universal agreement on the Phase One R+D team that they would not produce a back that lowered the bar of image quality, color included. If they could not ensure excellent color, no amount of market pressure, no attractive features, and no competition would convince them to adopt CMOS. Without excellent color, CMOS would not enter a Phase One product.
As the market share leader in medium format, Phase One was at the top of Sony’s list to consult on the features and specifications that would be most important to the success of Sony sensors in their market.The resounding response from Phase One was that color and image quality were far more important than a marginal improvement in ISO. While Phase’s target market would benefit from a very clean ISO1600, and maybe even occasionally shoot at ISO 6400, it was not that important that the sensor perform at crazy high ISOs like 25,600. Sony listened and provided a CFA that emphasizes color fidelity and subtle color discrimination.
But a good sensor, with a color quality oriented CFA, are only raw ingredients to cooking up good color. As Niels says, “Having a sensor with a good CFA is only the start of the process to making a back with great color.”
Fortunately for Niels, and for Phase One, he had been, unknowingly, preparing for this task for two decades. Niels has the informal title “Image Professor,” and is best known inside Phase One as the dark wizard of color. He is responsible for the fine tuned profiles made for all the cameras that Capture One supports.
This has meant that for over a decade he has produced color profiles for every major CMOS camera. He has come to know CMOS color from dSLRS like a friend or perhaps like a frenemy. As he describes it, “I am always fighting with profiling CMOS dSLRs to control accuracy while allowing for subtlety and robustness. Every time I fix one color another color jumps out and bites. These cameras’ color response is brittle.”
As anyone who has compared Capture One (which uses Niel’s profiles) to LightRoom for processing a 5D3 or D800 file knows, Niels’ profiles represent the very best that can be done with these cameras. But without any input or control over the sensor, CFA, and all the supporting electronics and components, and often without access to the underlying sensor specifications to get some sense for what’s going on under the hood, he can only do so much.
“Getting involved in the entire imaging chain of a CMOS camera for the first time, and not just having to make the best of what came out of someone else’s CMOS camera was like a breath of fresh air after being stuck in a smoke filled room,” Niels said. The initial work involved working with a hardware team for rapid prototyping profiles to aid with the selection of system components. Much of this process was lab based, and drew heavily on the team’s experience with high quality sensors built with color oriented CFAs.
With the hardware finalized, and a very rough profile generated, the real work was about to begin. The success or failure of the IQ250 project would depend on Niels’ ability to generate real world profiles.
Photo by Tim Kemple. Phase One IQ250 at ISO800, 1/1600th @ f/5.6, with Schneider 28LS.
The Proof is in the Pictures
“At Phase One we don’t use test charts very much for profiling. Especially when it comes to optimizing for skin tones. Using test charts to generate profiles ensures profiles that make test charts look great; whether anything else looks great is a gamble,” Niels says of his approach to profiling. “We use test charts and other lab based tests primarily at the beginning of the profiling process and rapid prototyping stages, but such profiles are the start, not the end of the process.”
The downside to this approach is that it requires enormous experience, what Niels calls “mileage” – you need a large number of sample files of different subject matter shot under a variety of lighting conditions. In the case of the Phase One IQ250, several professional photographers were given prototype IQ250 backs under strict NDAs (see IQ250 shots from Justine Ungara for instance). Images from these photographers were added to the R+D team’s own real world and lab test images and Niels got to work.
Photo by Justine Ungaro. Phase One IQ250 at ISO1600, 1/320th @ f/4, with Schneider 80LS.
Does This Skin Tone Need More Salt?
The process of crafting a good profile is a lot like a chef making a sauce. There are some good rules of thumb, and many years of experience speed the process of deciding what the profile needs more and less of, but it’s still necessary to constantly taste the result. With thousands of example files representing a broad range of uses, Niels could make a change to a profile and quickly see its effects on many real world images. The tools are different (mostly custom in house tools built from scratch for Niels’ preferred manner of adjusting profiles), but the process is not unrecognizable to any photographer who has worked on tweaking batches of images.
When I asked the head of the Product Management Team, Franck Rasmussen, about Niels’ work on the CMOS color profiling, he laughed and recalled, “He was more or less locked in the basement for a month straight. He only emerged sporadically to demand more sample files of this subject matter or in that lighting.”
The hard work, however, has clearly paid off. The sample IQ250 files we’ve seen sing with color.
Photo by Justine Ungaro. Phase One IQ250 at ISO400, 1/200th @ f/5, with Schneider 80LS.
At Journey’s End
After several hours of talking about technical details I asked Niels what he has concluded from all of this work. He spoke like a proud father: “I have fought with color from CMOS cameras for many, many years. I’ve always assumed that CMOS itself was not the issue, but rather the issue was the priorities of the companies using CMOS sensors. Getting to test that thesis was very satisfying. When we first started this project we were not sure we could take a CMOS sensor and craft the color our customers expect from Phase One.” His gaze drifted, as if taking in the totality of the journey he has been through – or perhaps his eyes are just still adjusting to bright lights after entire days spent in front of a carefully calibrated Eizo monitor. His focus returned and, with a reassuring nod, he concluded, “But now we know: yes we can. We can make CMOS sing.”
About the Author
Doug Peterson is on the technical team at Digital Transitions, the Phase One Dealer. He has worked in medium format for six years, and is also a wedding photographer on the weekends, and brewmaster of Digital Transitions’ limited edition home brew ale series. Digital Transitions specializes in support, sales, rental, and training of medium format equipment.
*USB3 ended up taking a lot longer than expected to get up and running, largely because USB3 itself was still in the “wildwest” stage of industry adoption. As frustrating as that process was, USB3 became a vital part of the IQ250 project.