Harry Clark
Well-known member
Hi guys,
I have had many people bring up issues with Epic at 2K (usually when shooting 300 fps) It's not the noise issue, it's the softness issue.
Below is a paste of an email I sent to my friend who is a very smart director/ editor down under. I wanted the smart folks here to comment, generally and specifically, on my theory.
Basically, I think the OLPF in the Epic is tweaked for the whole sensor/ 5K res (which is a good thing) but that 2K suffers too much when compared with the Red One.
Not a knock on Epic. LOVE IT.
Cheers,
Harry
Hey man. Hope you are well otherwise...
Without actually seeing the footage I can only speculate, but I have some beliefs about 2K cropping in general and Epic specifically. Take them for what they are; information from a hack corporate video DP
First of all, 2K is soft. there is no way around this. When you shoot any cropped resolution with Red or Epic, the apparent sharpness goes down. 3K is pretty acceptable in my experience; 2K is SO different looking that I generally recommend skinnying the shutter a lot and even hardening the lighting somewhat (you're only really resorting to 2K for high speed so the shot will have a somewhat specialized look anyway) The Bayer pattern on the sensor requires a certain amount of oversampling (remember, it's really "inventing" some color resolution in post when you process the raw data) 2K is 2048, really just a bit more than your final resolution of 1920.
Secondly, 35mm glass (counterintuitively) is "lazier" in its resolution that 16mm in general. Without belaboring the arcane theories behind how they design lenses, generally the smaller the frame the sharper the glass needs to be to resolve the same number of line pairs per millimeter. If you take a 25mm lens that's designed for 35mm film cameras, and a 25mm lens that's designed for Super-16, and film a test chart in Super-16, the lens designed for Super-16 will resolve better (it won't fully cover the 35mm frame so you can't benefit by using the sharper lens in 35mm) The rehoused Hassleblad medium format lenses that they use in the Clairmont "Swing And Tilt" and the Arri "Shift And Tilt" lens systems are incredibly soft in 35mm, even more so in Digital 35 formats. All lens designs are a tradeoff in one way or another, but this is a well-known issue. I think it is likely the smallest factor in play in your situation.
Third, the other factor that comes into play (an the factor that I think is most important in this particular situation) is the Optical Low Pass Filter. All digital motion picture cameras feature an OLPF (often it incorporates some sort of IR and/ or UV in the glass) glued directly onto the sensor. It has a secondary feature of protecting the actual sensor surface and keeping contamination off it. Its primary feature is to tweak the image. An OLPF is just a piece of glass with some sort of diffusing pattern on it, usually a "lenslet" or dot pattern, like a Soft FX or Classic Soft filter. Its job is to actually LOWER the incoming image's resolution before it hits the sensor, specifically areas of high frequency or high detail; hence, it's called "Low Pass" because it passes only low-frequency details (You could perhaps call it a "High Block" filter) Without this filter, most images would feature HORRIBLE moire patterns, chroma artifacting and other pattern noise (think 5D moire issues times ten) All manufacturers have their own philosophy, their "secret sauce", when it comes to how much detail to knock off, but all cameras have them. Red cameras are optimized for the use of the entire sensor area, and the OLPF does a nice balancing act of producing an image that has enough resolution for a large theater screen and enough smoothness for an HDTV. Ideally, if you were shooting a cropped area of the sensor, you would have a much lighter OLPF in place. It's not really possible with current sensor block designs to permit a swappable OLPF, so you are stuck with the one that is optimized for the way we shoot 95% of the time.
Lastly, Epic at 300 fps, specifically. I have no insider knowledge and am not accusing Red of anything nefarious. But I think they changed the OLPF design for the Epic. I think they fine-tuned it to accommodate an sensor area that is somewhat larger and higher resolution that the Red MX (even though the sensor is the same, when installed in a Red One it's cropped to be 4K) I think they made it the best compromise they could for 5K, but in the process made it worse for 2K. My own tests with 300 fps at 2K parallel what you are saying. They look much softer than a 2K image from a Red One.
Not sure what the way around this is, my friend. The use of a 16mm lens, harder lighting, more f-stop, and skinny shutter will all help. Last resort, there are other options for 300 fps (Phantom and Weisscam) Cameras designed specifically for high speed photography are really, really good at that application (and really really lousy at 24 fps traditional shooting; I got stuck using a Phantom to do an interview two weeks ago- arrrrggghhh)
Of course I would always rather use my Epic for everything all the time
Cheers,
Harry
On Nov 19, 2011, at 12:04 AM, XXXXXXXXXXXXXXX wrote:
Hi Harry, hope things are well. I have a Red Epic M question and thought you might know. I'm shooting a commercial on the Epic and I did some tests, but they look terrible. I'm shooting at 2k 300fps, T4 on a zeiss ultra prime and Red Prime - both look incredibly soft. I know it's not the lenses, can't seem to figure out what we're doing wrong with the camera. When we shoot at 25fps 5k - it looks fine in an HD sequence, but only at 25% (filling the frame), if i set it to 100% the image looks soft and has a shitty glow around the edges. I tried in after effects and it was the same.
I've searched Reduser but found nothing too helpful. Any ideas?
Cheers,
XXXXXXXXXXXXXXXXX
I have had many people bring up issues with Epic at 2K (usually when shooting 300 fps) It's not the noise issue, it's the softness issue.
Below is a paste of an email I sent to my friend who is a very smart director/ editor down under. I wanted the smart folks here to comment, generally and specifically, on my theory.
Basically, I think the OLPF in the Epic is tweaked for the whole sensor/ 5K res (which is a good thing) but that 2K suffers too much when compared with the Red One.
Not a knock on Epic. LOVE IT.
Cheers,
Harry
Hey man. Hope you are well otherwise...
Without actually seeing the footage I can only speculate, but I have some beliefs about 2K cropping in general and Epic specifically. Take them for what they are; information from a hack corporate video DP
First of all, 2K is soft. there is no way around this. When you shoot any cropped resolution with Red or Epic, the apparent sharpness goes down. 3K is pretty acceptable in my experience; 2K is SO different looking that I generally recommend skinnying the shutter a lot and even hardening the lighting somewhat (you're only really resorting to 2K for high speed so the shot will have a somewhat specialized look anyway) The Bayer pattern on the sensor requires a certain amount of oversampling (remember, it's really "inventing" some color resolution in post when you process the raw data) 2K is 2048, really just a bit more than your final resolution of 1920.
Secondly, 35mm glass (counterintuitively) is "lazier" in its resolution that 16mm in general. Without belaboring the arcane theories behind how they design lenses, generally the smaller the frame the sharper the glass needs to be to resolve the same number of line pairs per millimeter. If you take a 25mm lens that's designed for 35mm film cameras, and a 25mm lens that's designed for Super-16, and film a test chart in Super-16, the lens designed for Super-16 will resolve better (it won't fully cover the 35mm frame so you can't benefit by using the sharper lens in 35mm) The rehoused Hassleblad medium format lenses that they use in the Clairmont "Swing And Tilt" and the Arri "Shift And Tilt" lens systems are incredibly soft in 35mm, even more so in Digital 35 formats. All lens designs are a tradeoff in one way or another, but this is a well-known issue. I think it is likely the smallest factor in play in your situation.
Third, the other factor that comes into play (an the factor that I think is most important in this particular situation) is the Optical Low Pass Filter. All digital motion picture cameras feature an OLPF (often it incorporates some sort of IR and/ or UV in the glass) glued directly onto the sensor. It has a secondary feature of protecting the actual sensor surface and keeping contamination off it. Its primary feature is to tweak the image. An OLPF is just a piece of glass with some sort of diffusing pattern on it, usually a "lenslet" or dot pattern, like a Soft FX or Classic Soft filter. Its job is to actually LOWER the incoming image's resolution before it hits the sensor, specifically areas of high frequency or high detail; hence, it's called "Low Pass" because it passes only low-frequency details (You could perhaps call it a "High Block" filter) Without this filter, most images would feature HORRIBLE moire patterns, chroma artifacting and other pattern noise (think 5D moire issues times ten) All manufacturers have their own philosophy, their "secret sauce", when it comes to how much detail to knock off, but all cameras have them. Red cameras are optimized for the use of the entire sensor area, and the OLPF does a nice balancing act of producing an image that has enough resolution for a large theater screen and enough smoothness for an HDTV. Ideally, if you were shooting a cropped area of the sensor, you would have a much lighter OLPF in place. It's not really possible with current sensor block designs to permit a swappable OLPF, so you are stuck with the one that is optimized for the way we shoot 95% of the time.
Lastly, Epic at 300 fps, specifically. I have no insider knowledge and am not accusing Red of anything nefarious. But I think they changed the OLPF design for the Epic. I think they fine-tuned it to accommodate an sensor area that is somewhat larger and higher resolution that the Red MX (even though the sensor is the same, when installed in a Red One it's cropped to be 4K) I think they made it the best compromise they could for 5K, but in the process made it worse for 2K. My own tests with 300 fps at 2K parallel what you are saying. They look much softer than a 2K image from a Red One.
Not sure what the way around this is, my friend. The use of a 16mm lens, harder lighting, more f-stop, and skinny shutter will all help. Last resort, there are other options for 300 fps (Phantom and Weisscam) Cameras designed specifically for high speed photography are really, really good at that application (and really really lousy at 24 fps traditional shooting; I got stuck using a Phantom to do an interview two weeks ago- arrrrggghhh)
Of course I would always rather use my Epic for everything all the time
Cheers,
Harry
On Nov 19, 2011, at 12:04 AM, XXXXXXXXXXXXXXX wrote:
Hi Harry, hope things are well. I have a Red Epic M question and thought you might know. I'm shooting a commercial on the Epic and I did some tests, but they look terrible. I'm shooting at 2k 300fps, T4 on a zeiss ultra prime and Red Prime - both look incredibly soft. I know it's not the lenses, can't seem to figure out what we're doing wrong with the camera. When we shoot at 25fps 5k - it looks fine in an HD sequence, but only at 25% (filling the frame), if i set it to 100% the image looks soft and has a shitty glow around the edges. I tried in after effects and it was the same.
I've searched Reduser but found nothing too helpful. Any ideas?
Cheers,
XXXXXXXXXXXXXXXXX
