Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

EPIC HDMI output less noisy than recorded footage?

Rob Ruffo said:
Very interesting. Does 3:1 help? Maybe I'll pick up one of those cheapie BM recorders and do some tests. Here is yet another area where Red One never caused big problems. I have mine for sale here. Maybe I shouldn't sell her after all.
Click to expand...

Ok defo last post...lol

Lowest I could go was 4:1. I tried 3:1 and I got a warning :-/
 
Karl H said:
please do as I cant presently record the HDMI signal from mine. I don't think you'd need to bypass the compression, just have an option to process the ISO/gain before the signal is recorded (I'm sure the wizards at Red can add this)- of course I think that means you're out of RAW space, but in high ISO situations RGB may actually be a better option. I'm no tech expert tho :/
Click to expand...

The compression happens when the R3D files are created.

If I capture directly from the HDMI port, I'm bypassing the compression.

I've done this with DV cameras to get a higher quality image (no mpeg compression and 1920 x 1080 instead of 1440 x 1080), but ... in this case, it will be 2K from HDMI vs 4K from R3D, so ... hopefully in most cases except very low light, or maybe a very busy scene that compression doesn't deal with well, the image will be of lesser quality at 2K from HDMI than from 4K R3D's.
 
Is there a discernible difference between HDMI and SDI recordings for noise? I have ISO800 footage from both SDI and r3d, but that's not "high ISO" so the comparison would be null.

Very intrigued by this; is ISO5000 clean/useable? Would cameras that require off board 4k raw recorders (like the FS700) have similar low light benefits from such an approach (e.g. capture the output at the ISO selected in cam). Apparently other cameras (like the Blackmagic DCC) don't allow ISO adjustment in post even though they're RAW, so maybe they'd benefit as well. Okay, okay, what about if you don't adjust ISO in RCX, but instead you make it brighter via a curve or FLUT? Is it any cleaner or dirtier using one method over the other (using FLUT instead of ISO)?
 
Les C. said:
The compression happens when the R3D files are created.

If I capture directly from the HDMI port, I'm bypassing the compression.
Click to expand...

Yup, yes you can bypass the compression and go to prores or something, what I was trying to say was I see no reason why Epic couldn't do this itself. It can clearly create the live output and with some tweaking I'm sure Red can give us the option to record this in camera. Also going out HDMI will restrict us to 2K obviously so a Red solution would be preferable.

Mike P. said:
Is there a discernible difference between HDMI and SDI recordings for noise? I have ISO800 footage from both SDI and r3d, but that's not "high ISO" so the comparison would be null.

Very intrigued by this; is ISO5000 clean/useable? Would cameras that require off board 4k raw recorders (like the FS700) have similar low light benefits from such an approach (e.g. capture the output at the ISO selected in cam). Apparently other cameras (like the Blackmagic DCC) don't allow ISO adjustment in post even though they're RAW, so maybe they'd benefit as well. Okay, okay, what about if you don't adjust ISO in RCX, but instead you make it brighter via a curve or FLUT? Is it any cleaner or dirtier using one method over the other (using FLUT instead of ISO)?
Click to expand...

Hi Mike - obviously this is just my findings and I've only looked at HDMI (the SDI should be the same in theory), but at 5000ISO - at 2K the noise IMO is usable, I guess it depends on your shot. You can even see this quite clearly on the 5" touchscreen. Crank up the ISO and look at the image live - record then playback - you can even see there the compression on playback is so much worse than live. Try it at anything like 9:1. (I know this isn't a full debayer but it illustrates the point on a 5" screen).

I can't speak for how black magic do it as Im no tech, but the curve wouldn't work either in post. Whether you use ISO, FLUT or curves you're still boosting artifacts in the shadows - imagine filming on a Canon 5D at ISO 100 and then boosting the dark mp4 in post, you know you're going to see some kind of blocking in the shadows, however if you cranked up to ISO 800 first and recorded it, that blocking wouldn't be there. It's a similar but not as extreme principle on the EPIC as in effect RAW is always recording at a fixed ISO.

As I say I could be wrong but if you play any clip back you've recorded on your touchscreen and touch the histogram - you'll get an option to view the histogram in RAW. Click it, and suddenly the histogram (depending on what you've shot) will bunch up to the left. AFAIK this is the info that's actually recorded. Viewing at 800ISO is simply metadata, or in other words a digital gamma curve/correction.

All IMO, until Red chime in, do you own tests and draw your own conclusions.
 
Hm... Interesting thread. If the image through SDI is indeed before compression, a RLF output in low light shoukd indeed be a lot less noisy than a compressed one...

BTW
You do NOT ETTR if you shoot at 5000 ISO. That is the most complete misunderstanding you can imagine... :)
 
lol. That's as bright as it will go in this brothel.

I guess what I should have said was, when I crank up to ISO 5000 the histogram looks normalish, but yes, down at 800 its a dogs dinner (with no lights on it).
 
Karl H said:
lol. That's as bright as it will go in this brothel.

I guess what I should have said was, when I crank up to ISO 5000 the histogram looks normalish, but yes, down at 800 its a dogs dinner (with no lights on it).
Click to expand...

THEN try 320. If you're at tungsten you're another stop off... There will be noise with serious technical underexposure. But it can still be nice and the right/available exposure...
 
Karl H said:
To see the noise levels at different ISO's I crank it up to 5000 ISO and it looks surprisingly clean over HDMI.
Click to expand...

I have also noticed that SDI and hdmi looks cleaner at lowlight. The red codec seems to have a threshold that works really good, then suddenly when too much detail (noise) is in frame, it completely breaks down with large comression patterns. To have the option to record the 1080p RGB signal that normally goed to the SDI/hdmi on to the red mag would be great.
 
JanneJansson said:
I have also noticed that SDI and hdmi looks cleaner at lowlight. The red codec seems to have a threshold that works really good, then suddenly when too much detail (noise) is in frame, it completely breaks down with large comression patterns. To have the option to record the 1080p RGB signal that normally goed to the SDI/hdmi on to the red mag would be great.
Click to expand...

LOL
Despite Jims mission on this, I think there are many who can see the use of an uncompressed 444 RLF recording out of the Epic. The bandwith on the SSDs are sufficient for a 1080 recording like that, but we won't get it anytime soon...
 
Gunleik Groven said:
THEN try 320. If you're at tungsten you're another stop off... There will be noise with serious technical underexposure. But it can still be nice and the right/available exposure...
Click to expand...

This was purely a test, intentionally done to see how far I could go into the noise and get away with it.

If I rate at 320 ISO or 5000 ISO it will make no difference to the light falling onto my sensor or what is recorded to the r3d file. That is determined by the shutter/fstop. ISO AFAIK makes no difference whether you boost/drop it during recording, or afterwards in RCX, its simply a tool to help you determine exposure (you're determining an amount of electronic gain). For example, if I shoot at 320 ISO and I don't clip on the RAW histogram, I could crank it up to ISO 5000 and push record too - it still won't clip. Although the image will look awful in your viewfinder, you can simply choose to view it at 320 in RCX and it'll look normal again.

This is kind of the crux of the whole debate, R3D pays no attention to user assigned ISO when recording, it just records RAW. In order to stop compression artifacts in low light we'd need a codec that does take ISO into account when recording.

JanneJansson said:
I have also noticed that SDI and hdmi looks cleaner at lowlight. The red codec seems to have a threshold that works really good, then suddenly when too much detail (noise) is in frame, it completely breaks down with large comression patterns. To have the option to record the 1080p RGB signal that normally goed to the SDI/hdmi on to the red mag would be great.
Click to expand...

would be even better have the 4K RGB signal :-)
 
Well, there are allways tradeoffs, and shooting without enough light for what you want to depict has many (more IMHO) sarious tradeoffs than purely noise in the darks.

You ARE right however that the codec but even more: whiteballancing, can eat your image and make it into a heap of noise under low exposures...

If that is a camera or operator deficiency, is another question and the answer depends on a lot of uncontrollable parameters...
 
This may have to do with how the camera adds gain to the HDMI signal. IIRC, there are different ways that cameras add gain. It's not always just a simple amplification of the signal. I believe they sometimes remap bits to allocate more to the shadows during quantization. The net effect of this is that during playback the shadows are lifted.
 
yup, I'd never actually do a proper shoot in these conditions but at times it would be unavoidable. I have learned through this test to always check the RAW histogram. But there is a way to get much cleaner shots in low light, the sensor is capable of it. If Red could find a way to record this information, for a lot of people it'd be like having a new camera again. 5000ISO, 6400ISO may seem extreme but the principle being you should see improvements even at lower ISO's, so it'd be nice to have. Until then you could always compromise with a 2K live output if you need low light shooting.

I did my tests in tungsten, not the sensors strong point and perhaps thats why the two signals seems so different. Perhaps shooting with daylight sources it wouldn't be as noticeable. I haven't tested that.
 
Gunleik Groven said:
LOL
Despite Jims mission on this, I think there are many who can see the use of an uncompressed 444 RLF recording out of the Epic. The bandwith on the SSDs are sufficient for a 1080 recording like that, but we won't get it anytime soon...
Click to expand...

Hehe, yes it's a bit in the wrong direction for a 5k raw camera :) but who knows, a 1080p dnxhd function on a red may be like the 24p h264 was for the 5D. Many would discard the function as usless, but many would love it and get a camera just for that.
 
What about 2k/60fps or 2k/300fps are those cleaner when recorded off board? I guess not, because they're time-remapped after recording, aren't they?
 
Mike P. said:
What about 2k/60fps or 2k/300fps are those cleaner when recorded off board? I guess not, because they're time-remapped after recording, aren't they?
Click to expand...

I'm not aware of any monitors or recording devices that go above 1080p60.
 
You must log in or register to reply here.
Back
Top