[i myself, would like to know Dalsa's true resolution. Why won't you say anything about it?]
Exactly what is "true resolution?" System MTF? The number of pixels on the sensor? The resolution of a chart shot through a 1960's Angenieux zoom or the same chart shot with an Arri Master Prime? The 'resolving power' of the entire imaging system? I can give you an answer for all of those variables, and still it will tell you nothing about "true resolution."
Why won't I say anything about it? Because this is a complex question, and indeed as David pointed out in this case, a loaded one. I can however give you the following facts coupled with some personal observations:
1. The Origin sensor has 4096x2048 pixels in a Bayer pattern on the CCD. Of those 4096 pixels 50 (on the far left side) are 'dark' and used as a black reference so the actual captured imaging pixel count in the horizontal direction is 4046.
2. The sensor is larger than a 35mm Academy frame. In fact the Origin sensor is the same size as an entire 35mm frame from side to side. The aspect ratio is actually 2:1 (1.9755859375:1 if you take into account the 50 dark). Now, we disclosed this on our web-site from the start and yet people still cried: "There, you see...it's not 4K!" or, my favorite: "Your mileage may vary." Right-e-oh. We were accused of hiding the fact of lens coverage and its relationship to 'resolution.' Let me state this as plainly as possible due to the fact that our picture on the web site of a 35mm motion picture frame laid over a picture of our sensor did not seem to get the point across. When using some standard 35mm lenses with the Origin the lenses will not cover the entire sensor, so one must extract the useable area (full Academy) from the captured pixels. Now, we have purchased a number of different lenses for rental (you can use any PL mount lens you want on the Origin) and based on the lens coverage we have a different ground glass for each format (1.85, 2.35, 16x9, etc). For example if you are using the Cooke s4 series, or the Arri Ultra Primes they were designed to cover only the Academy aperture, so you tell us: "I want to shoot with the Cooke s4 lenses at 1.85." We then put in the ground glass that has 1.85:1 markings imaged *inside* the Academy frame and you shoot some test material. We then extract the Academy area from the 4046 pixels and give you the frame (I believe the extraction becomes 3858, but I am not at work right now and don't remember the exact number off the top of my head) or we give you the entire frame and let your post people crop as they may. If you like what you see you shoot with those lenses and that ground glass for your show. However, we also have purchased a number of re-barreled and PL mounted still camera lenses (Leica, Zeiss, etc) like we use in Vista-Vision VFX photography that DO cover the entire range of the sensor. Then you get a ground glass that matches your chosen aspect ratio, but covers the full 2:1 sensor size and you shoot with that. Honestly, I never expected this to create this much controversy. Again, it is really up to personal taste: the Cooke s4 series have the smallest coverage (indeed they fall off RIGHT at Academy) but they are such beautiful lenses with such a 'warm' and wonderful look that many people (including my team) really like to shoot with them. On the other side the Leica's are incredible lenses as well; they cover the entire sensor, they are very sharp and they have a unique look that is hard to beat. We tested the Arri Master Primes and we really liked them too, so we also ordered some sets of those and they seem to voer the entire sensor as well. Indeed it really isn't a question of 'resolution' but one of how an entire imaging system combines to create a look that works (or does not work) for your project.
3. We have a number of very high level image processing algorithms for color reconstruction working on the Bayer data. However the last time I even uttered the word 'algorithm' in public I was misquoted on the internet and for a year I had to deal with fall-out and totally ignorant BS. So I am going to leave it at this: if you want to know about our image processing software please go to the DALSA website and read some of the white papers about our imaging kernel and how the algorithms reconstruct color from a Bayer pattern. If you wish to know more contact me privately.
4. Coupled with the issue of 'resolution' is that of 'color depth.' The Origin captures data in 16 bit linear space. We claim 11+ stops of dynamic range, but again please come in and test it for yourself. I know what I have seen come out of the Origin, but you should test for yourself. One point I do wish to make is this: once rendered the 'native' space we like to deliver the imagery in is a 16 bit High Dynamic Range SMPTE DPX file in un-mapped color space. With that format you have the entire 65,536 DN value range with no gamma applied as a starting point. This gives the colorist and VFX people the most 'meat' to work with in post. However, we learned the hard way that 95% of the people we give Origin data to simply flip out when they see unmapped HDRI imagery. Visual effects people who have been dealing with HDRI and film scans get it immediately, but most people start to have tantrums: "I can't see anything, its all dark!" I think you may be starting to see why we are so careful about what we 'say' about resolution.
More to the point, you say the following:
"Basically, it was proven to me that some of your claims were not true."
Ok, here we go, sure I'll take the bait: what 'claims' were proven not to be true and who 'basically' proved that "some of our claims were not true?" Please be more specific, this is just more internet posturing if we leave it as it stands. We can start a useless flame war or we can get to the bottom of it now. You choose.
Let me ask this another way. What do you think the technical people at Arriflex and Panavision claim the 'resolution' of the D-20 and Genesis are? I have seen both of those tools up close and I will say without question that they are both excellent cameras; truly remarkable pieces of engineering. I have the utmost respect for both of their tech teams as well. Now, do you think they would say that the 'resolution' of their respective cameras was 1920x1080 just because they are currently recording out to HDCAM-SR, or indeed must we look deeper at the underlying sensor technology? Soon both of those cameras will also support recording DATA directly out of the pipe (as we do at DALSA), and we will see 'resolution' based more on the native 'resolution' of their sensors. Let's put this another way: what do you think PIXAR considers the 'resolution' of MONSTERS, INC? We all know that PIXAR renders their material at a bit less than HD, but because they have no "imaging system" to go through (no lenses, film-stocks, printing, etc) they end up with imagery so sharp that in many cases they have to soften it before film-out. What would the 'resolution' of imagery coming out of RENDERMAN at 1600x900 be?
Alan Lasky
DALSA Digital Cinema
