HD-SDI 8-bit to 10-bit Ki Pro
Just curious technically what happens when you feed the 8-bit signal out of a camera capturing to 10-bit ProRes on a Ki Pro. You obviously have a 8-bit original source and you can't increase the bit size of the original. With that said, now that it is recorded at 10-bits do you have a little extra headroom or stability in post for grading? Is it noticeable?
Does the camera matter? Not trying to sound flip at all but isn't it about the math regardless of camera make or model?
lets say any camera that outputs 8-bit 4:2:2.
All these cameras output 8-bit 4:2:2:
Sony PDW800 with XDCAM-MPEG422 at 50 megabits
Canon XF300/XF305 with MPEG2 at 50 megabits
Yeah, Flippy McFlipperton.
Just because the footage is compressed to 8 bit compression in the onboard recording device doesn't mean the camera outputs 8bit out of the HD-SDI. You didn't specify what camera or how you're hooking it up, or if you're capturing "live" or recording from in camera playback. For example if you were coming in HDMI from an HDMI 1.1 device you won't get 10bit.
So yeah, Flip, it all matters. Sometimes it's math, sometimes it's art, most of the time it's math art. It can be expensive electronics, cheap electronics, cheap food, or gourmet.
Have your camera generate a ramp, check for banding on a 10bit monitor. You can then see for yourself, no math involved in that case.
If you don't care, then you'll be fine with the component signal that can be digital or analog coming out of your camera in to the Ki. It will record nice pictures if you feed it nice pictures and ProRes will lessen any generation loss in post.
[Craig Meadows] "All these cameras output 8-bit 4:2:2:
Sony PDW800 with XDCAM-MPEG422 at 50 megabits
Canon XF300/XF305 with MPEG2 at 50 megabits
Actually, that is not correct Craig, yet this is a reasonable question.
however only the HDCam and DVCPROHD tape formats are 8 bit on this list.
F900R models (with hdsdi) has always output 10bit
As do all Varicam models after the Varicam f model
The Pdw 800 outputs 10bit via HDSDI when the 24p module is included in the camera.
The Canon XF cameras HDSDI is 10bit too.
What happens is the 8bits native of info is output evenly across the 10bit pipe, since this is being done hardware it will offer the recorder you the highest quality conversion of the bit depth allowing for the best quality thru post and with VFX handling.
Post and Production Workflow Consultant
Production and Post Stereographer
Thanks Gary. I got that camera list from another source online, my apologies for any misinformation.
The reason I ask is that the folks over at RED say with 8-bit the damage is already done (in regards to grading), it does not matter if you you are recording to a 10-bit external device like a Ki Pro if the HD-SDI on your camera is only outputting 8-bit.
Meanwhile, over at Cineform I found a white paper stating that a 8-bit to 10-bit conversion using their codec (keeping in mind it' wavelet based) does help in grading. They demonstrate with a gradient blue sky shot. All 8-bit shows banding, while their 8-bit to 10-bit codec does not.
I am curious to know if the 10-bit ProRes file from a Ki Pro allows for better grading in post (like the Cineform example above) if the camera is only outputting 8-bit. Or, should you just stick with a 8-bit work flow if there is no real advantage for recording to 10-bit.
Any real world examples or Ki Pro user comments greatly appreciated.
I think you're still missing it , Craig. If capturing live (meaning not from tape) you will most likely be capturing a 10bit signal. Yes 10bits are better than 8 and ProRes is better than most in-camera compression codecs from the cameras you mentioned. The KiPro also shines in green screen situations. You will have a bit more flexibility, yes, but the Ki also gets you to the post process in an extremely fast manner, especially compared to the tape camera(s) you listed. And there's no transcoding needed as with some of the tapeless cameras you have mentioned.
So, if you bypass the onboard recorder and capture
Live to KiPro, will will also bypass 8bit.
Cineform and red workflows are completely different animals and usually don't follow more traditional hdsdi workflows.
Jeremy, I do understand what you are saying but not sure you completely understand my question.
If it's a 8-bit camera. We are bypassing recording to the cameras recorder be it tape or card, I understand that. We are using the cameras HD-SDI BNC to send a "live" uncompressed 8-bit signal to a 10-bit device, the Ki Pro. The signal coming out of the camera at that BNC is still 8-bit, correct?
If the camera is only capable of outputting 8-bit via it's HD-SDI BNC connector, you cannot magically increase the color information (i.e. the bits that make up the original color) just by recording to a 10-bit device. What you have is a 10-bit file with color information from a 8-bit source. There's no way you can add extra color information to what's not there.
Yes, the resulting file will be 10-bit but the original color information that the 8-bits used to create that color has not changed.
Now we move to post. If this had been a 10-bit camera outputting 10-bit from it's HD-SDI port we would be fine. Instead we have a 10-bit file but not more original color information than was available if it was 8-bit, but, more bits to push around.
My question is do users see a improvement pushing 8-bit original color source around a 10-bit environment? Can you push grading more than if it was 8-bit? If a colored sky or gradient bands in a 8-bit work flow does it also band if you have done this 8-bit to 10-bit capture process?
[Craig Meadows] "We are using the cameras HD-SDI BNC to send a "live" uncompressed 8-bit signal to a 10-bit device, the Ki Pro. The signal coming out of the camera at that BNC is still 8-bit, correct?"
No, I completely get it, Craig. Believe me, I do what you're doing a lot (although I mostly use 10bit native recording cameras).
I still think you are missing it. So, again:
If you are capturing live (that is NOT recording the signal that is being played back from already recorded/compressed material, or put another way, turn on the camera and hit record on the KiPro) the signal that you are capturing is most likely 10bit (depends on the camera). It is 10bit out of the HD-SDI when capturing live. If you are playing back already compressed material (meaning you have captured something on the cameras onboard recorder, and playing it back) that will essentially already be an 8 bit image and yes, as you mentioned, the 'damage will be done'.
[Craig Meadows] "If the camera is only capable of outputting 8-bit via it's HD-SDI BNC connector, you cannot magically increase the color information (i.e. the bits that make up the original color) just by recording to a 10-bit device. "
But if capturing live, it's not necessarily 8bit as the HD-SDI is usually sent pre tape/tapeless compression. I think that's where you are getting tripped up. As I and Gary mentioned earlier, just because the camera has an 8bit on-board recording device, this doesn't mean the live picture it's sending "off of the chip" is 8bit. Bit if you are playing back a already recorded material, the 8bit is there.
So again, it all matters, your workflow matters, you still haven't described exactly what you are trying to do or how and when you plan to get material in to the Ki, but maybe you are just working on theoretics at this point.
[Craig Meadows] " What you have is a 10-bit file with color information from a 8-bit source. There's no way you can add extra color information to what's not there."
I understand, but see above.
[Craig Meadows] "My question is do users see a improvement pushing 8-bit original color source around a 10-bit environment? "
Like you said, you can't magically add information and it really depends on your codec/source footage and how you captured it. The resulting rendered file will more likely resemble your original image, but it won't make it better. With ProRes, you will basically reduce the amount of artifacting/banding you add back to the image on subsequent renders/generation loss, especially when processing in 16bit or higher in your applications.
[Craig Meadows] "If a colored sky or gradient bands in a 8-bit work flow does it also band if you have done this 8-bit to 10-bit capture process?"
The original banding , if present, will be there. Yes.
What camera do you have? Do you have all of those that you listed?
Thanks Jeremy, that helps a lot. The questions come from a interest in the new Panasonic AF100 that's buzzing all the forums lately.
Our camera collection over the years has included BVP3a (yes a tube camera back in the day) DXC537, SDX900, Z1U, HD100, RED, EX3, HMC150, HMC150, and sure I left something out.
We've never had the need for a external recorder but the idea of a AF100 with a Ki Pro Mini is very interesting to me as we are a all AJA Kona/FCP company. I actually don't mind the avccam as a capture codec but would love to know if stepping up to a Ki Pro Mini with the AF100 would give us a very noticeable improvement.
Thanks for all your help.
[Craig Meadows] " actually don't mind the avccam as a capture codec but would love to know if stepping up to a Ki Pro Mini with the AF100 would give us a very noticeable improvement."
Absolutely. Not only in 8 vs 10bit but in 24 or less mbps AVCHD vs 145 or 220 Mbps (ProRes or HQ) compression.
I can't wait to see this camera either. A KiPro mini is going to be an awesome compliment, I would imagine.
Just to expound upon what Jeremy's said:
I think I'm right in saying HD-SDI is necessarily 10-bit. The only question is whether it is a genuine 1024 gradations or a "padded" 256 (8-bit).
The three pivotal points in image capture are:
Camera sensor's A/D
The Sony F900R has a 12-bit sensor A/D, 10-bit processing and onboard 8-bit HDCam recording. Record the live HD-SDI feed and you'll get genuine 10-bit.
Not sure of the Canon XL-H1's A/D, but its 8-bit processing, which is (necessarily) done before HD-SDI output, means you are getting 8 bits padded into a 10-bit signal. So in this case, yes, the damage has been done.
Something can be gained by interpolating the 8 bits into 10 — and some hardware, as well as the likes of Cineform's software, might do this too — but it's not going to be as good as recording genuine 10-bit in the first place.
[Craig Meadows] "Just curious technically what happens when you feed the 8-bit signal out of a camera capturing to 10-bit ProR"
You just add to "0" to the end of each Y', C'b and C'r value.
You increase the data rate in a 20% and don't get nothing in exchange.
Is an 8b picture enveloped in an 10b pack.