F5 and F55 detailed specifications (thanks to Sony Europe)

[Nate Weaver]

And for a completely different approach to the term RAW, the C500:
http://blog.abelcine.com/2012/11/05/what-exactly-is-canon-raw/

Cheers.

You know, also found out the Canon "raw" is 10-bit. Between the baked in gain and WB processing, and now learning it's also 10-bit, I'd say they're definitely pushing the definition of raw. Is the A/D output of the sensor really only 10-bit? My guess is it's more, and it's being decimated for easier transport. Even Canon DSLRs have 12bit A/D from sensor.

I'm quite sure (no sarcasm, it really does) it makes very nice pictures, but based on the 12 stop latitude I saw in the "Man & Beast" short, and now finding that their raw implementation is what I'd call creative...plus the whole rube goldberg external recorder rig...

...I guess I'd say based on available information, the C500 wouldn't be my first choice for 4K. Probably wouldn't even be my second, actually.

 

[avro]

I think that is wrong, F55 is 1250 and F5 is 2000.

I also want to know if these values are SLOG2 or standard gamma modes for XAVC, etc..

If they are indeed SLOG2 based, then the real ISO's are HALF those values. If they are standard gamma values, then SLOG2 are actually DOUBLE.

Guys making movies do not need 10000000000 iso

 

[Admit One]

Is the A/D output of the sensor really only 10-bit?

It has to be more than 10-bit or it would have only 10 stops dynamic range. It seems they are transcoding to log and dropping bits in the process. Most don't seem to be able to just leave the RAW data alone. Hopefully, Sony will.

Cheers.

 

[starcentral]

You know, also found out the Canon "raw" is 10-bit. Between the baked in gain and WB processing, and now learning it's also 10-bit, I'd say they're definitely pushing the definition of raw. Is the A/D output of the sensor really only 10-bit? My guess is it's more, and it's being decimated for easier transport. Even Canon DSLRs have 12bit A/D from sensor.

I'm quite sure (no sarcasm, it really does) it makes very nice pictures, but based on the 12 stop latitude I saw in the "Man & Beast" short, and now finding that their raw implementation is what I'd call creative...plus the whole rube goldberg external recorder rig...

...I guess I'd say based on available information, the C500 wouldn't be my first choice for 4K. Probably wouldn't even be my second, actually.

I read that same Abelcine article yesterday and felt it was very Franken-4k...

I do however like the fact the FS700 since it is bayer sensor is actually capable of sending 12-bit RAW over 3G single link, now if there was only a recorder that supports it. But Jon from AJA recently posted some reassuring comments that makes it very likely we can expect something soon.

 

[starcentral]

It has to be more than 10-bit or it would have only 10 stops dynamic range. It seems they are transcoding to log and dropping bits in the process. Most don't seem to be able to just leave the RAW data alone. Hopefully, Sony will.

Cheers.

The bit depth does not limit the dynamic range of the sensor, ie. the F3 is 12.5 stops recording to 8-bit SXS cards, or 13.5 stops with SLOG enabled recording 10-bit through SDI.

 

[Nate Weaver]

It has to be more than 10-bit or it would have only 10 stops dynamic range. It seems they are transcoding to log and dropping bits in the process. Most don't seem to be able to just leave the RAW data alone. Hopefully, Sony will.

Cheers.

The bit depth does not limit the dynamic range of the sensor, ie. the F3 is 12.5 stops recording to 8-bit SXS cards, or 13.5 stops with SLOG enabled recording 10-bit through SDI.

You're both right.

You can squeeze as many stops as you want into any bit depth you want, but if you ever want to later EXPAND the contrast back in color correction in a mathematically correct sense., and have as little posterization as possible, mathematically, you need 1 bit of depth for every stop of latitude the sensor creates, to not be "throwing any information away". That's how binary bit depths work.

In practice though, everybody gets awesome images from a 14 stop Alexa recording to 12 or 10 bit ProRes files. I find that I can only start to see things when I expand high DR images from 8bit recording, things get a little crunchy in places (posterization). Has a lot to do with how noisy the image was to begin with, and how much contrast you're adding.

My only point is that if you're going to call it raw, to date, that has always meant at least 12 bits to work with. Until the C500 came along.

 

[Admit One]

The bit depth does not limit the dynamic range of the sensor

The bit depth of the Analog to Digital converter absolutely limits the dynamic range of the sensor. After A/D you can maintain dynamic range with fewer bits, but you need them during A/D.

Cheers.

 

[Nate Weaver]

...but you need them during A/D.

Ah, right. Makes total sense. Could be why there's no Canon camera rated above 12 stops dynamic range (at least none that I've seen), they're limited at the sensor A/D.

 

[starcentral]

The bit depth of the Analog to Digital converter absolutely limits the dynamic range of the sensor. After A/D you can maintain dynamic range with fewer bits, but you need them during A/D.

Cheers.

sorry I missed that you were talking about sensor A/D, my bad.

:embarasse

 

[FilmDude]

RAW is ONLY raw when all of the information is just from the sensor and thats it, nothing is baked in. That IS raw and thats it. So if Canon is baking in things, and floating around only 10 bits... its impossible for it to be "raw" then its just a marketing ploy and nothing more.

 

[Grug]

RAW is ONLY raw when all of the information is just from the sensor and thats it, nothing is baked in. That IS raw and thats it. So if Canon is baking in things, and floating around only 10 bits... its impossible for it to be "raw" then its just a marketing ploy and nothing more.

Most still cameras do the same thing as the Canons when it comes to sensitivity and raw files - the 'sensitivity' of the sensor is set in camera and the file is exposed with whatever signal amplification the sensor has going on to bring its ISO up. There's only a few raw cameras that use gamma curves to adjust 'ISO' from a base level.

 

[Admit One]

The C300 differs from most other cameras because even though it uses a Bayer sensor pattern, no de-Bayering is required:

"The image sensor is, however, specifically designed to produce a standardized 1920 (H) x 1080 (V) high definition 4:4:4 RGB video component output set... The new Canon image sensor was designed to avoid the demosaicking process entirely. It instead relies on innovations in pixel addressing and associated readout mechanisms to separately extract the three RGB video components" - Canon White Paper

I believe the C500 uses the same sensor. Not sure, but perhaps that is why the C500 RAW output doesn't really seem to qualify as RAW.

Cheers.

 

[Guest]

RAW, Raw, raw.

I used to use the term to mean uncut and ungraded footage straight from the camera. Nothing to do with data, just uncut footage. Until this idea of RAW became the buzz word with dslr's, when I felt it meant all the 1's and 0's from the sensor. But now if it is the minimally processed non RGB data, then supposedly it is raw. Whether it is 8-bit, or 10-bit, or highly compressed seems irrelevant.

 

[starcentral]

The C300 differs from most other cameras because even though it uses a Bayer sensor pattern, no de-Bayering is required:

"The image sensor is, however, specifically designed to produce a standardized 1920 (H) x 1080 (V) high definition 4:4:4 RGB video component output set... The new Canon image sensor was designed to avoid the demosaicking process entirely. It instead relies on innovations in pixel addressing and associated readout mechanisms to separately extract the three RGB video components" - Canon White Paper

I believe the C500 uses the same sensor. Not sure, but perhaps that is why the C500 RAW output doesn't really seem to qualify as RAW.

Cheers.

Canon up to their good old tricks - anyone remember "pixel shift" technology? :happy:

 

[penryn]

Canon up to their good old tricks - anyone remember "pixel shift" technology? :happy:

what is "pixel shift" tech?

 

[cowpunk52]

RAW is ONLY raw when all of the information is just from the sensor and thats it, nothing is baked in. That IS raw and thats it. So if Canon is baking in things, and floating around only 10 bits... its impossible for it to be "raw" then its just a marketing ploy and nothing more.

Technically, no - a raw file is an undemosaiced image. But by your rationale, there are no raw video cameras on the market. Every raw camera bakes information in.

 

[cowpunk52]

The C300 differs from most other cameras because even though it uses a Bayer sensor pattern, no de-Bayering is required:

"The image sensor is, however, specifically designed to produce a standardized 1920 (H) x 1080 (V) high definition 4:4:4 RGB video component output set... The new Canon image sensor was designed to avoid the demosaicking process entirely. It instead relies on innovations in pixel addressing and associated readout mechanisms to separately extract the three RGB video components" - Canon White Paper

I believe the C500 uses the same sensor. Not sure, but perhaps that is why the C500 RAW output doesn't really seem to qualify as RAW.

Cheers.

The C500 output does quality as raw because it is undemosaiced. Baking in ISO & WB adds "gain adjustments at the sensor level [to produce] a consistent stop range above and below the middle grey level, even at high ISOs, and reduce the overall noise in the image." (source: http://blog.abelcine.com/2012/11/05/what-exactly-is-canon-raw/).

In short, it's what allows the C500's raw output to have the same high quality low-light characteristics of the C300, notably including a constant middle gray and stop range over and under throughout it's ISO's.

 

[cowpunk52]

Canon up to their good old tricks - anyone remember "pixel shift" technology? :happy:

I think you're mistaking Canon for Panasonic, who most famously used pixel shift tech in the HVX200 and HPX170 (and we all know what disasters those cameras turned out to be). Canon's cameras only applied pixel-shift technology with its native 1440x1080 HDV cams when recording the live signal precompression through HDMI or HDV. And it actually worked surprisingly well (for an "old trick"): http://downloads.bbc.co.uk/rd/pubs/whp/whp-pdf-files/WHP034_ADD25-Canon-XHG1.pdf

Otherwise, the tape recorded the native resolution of the sensor.

edit: although, now I'm remembering (and researching) that canon did employ pixel shifting in the XL1 and GL2 cameras. And that, indeed, was recorded in camera, so that's probably what you're referencing. But whether you think of it as a "trick" or not, you can't deny that it did increase recorded resolution, albeit not to the extent that a natively higher pixel-count sensor would have.

 

[starcentral]

I think you're mistaking Canon for Panasonic, who most famously used pixel shift tech in the HVX200 and HPX170 (and we all know what disasters those cameras turned out to be). Canon's cameras only applied pixel-shift technology with its native 1440x1080 HDV cams when recording the live signal precompression through HDMI or HDV. And it actually worked surprisingly well (for an "old trick"): http://downloads.bbc.co.uk/rd/pubs/whp/whp-pdf-files/WHP034_ADD25-Canon-XHG1.pdf

Otherwise, the tape recorded the native resolution of the sensor.

Pixel-shift is a panasonic technology, but I was referring to Canon who marketed it on their brochures for their XL1 16:9 native CCD camera.

"With pixel shift the green CCD was physically shifted 1/2 pixel horizontally from the red and blue CCDs and the green signal is electronically shifted 1 ⁄2 pixel vertically. That increased the sampling points and Canon said that the result is comparable to 410,000 pixel resolution even though it was 270,000 pixel CCD."

"The resulting image is softer than the PD150’s and defaults to more saturated tones. The desirability of this look is endlessly debated by users of the competing systems; fans of the look consider it filmlike, non-fans call it “smeary”."

 

[cowpunk52]

You know, also found out the Canon "raw" is 10-bit. Between the baked in gain and WB processing, and now learning it's also 10-bit, I'd say they're definitely pushing the definition of raw. Is the A/D output of the sensor really only 10-bit? My guess is it's more, and it's being decimated for easier transport. Even Canon DSLRs have 12bit A/D from sensor.

The C500 can output 2K RGB or YCC at 12-bit 4:4:4, so yes, it is more.