AG-DVX200 4K Large-Sensor Handheld Camcorder

Just fiddling with noise and sensitivity today. Using the DSC Labs CamAlign - ChromaDuMonde - Billups VF/X28 chart... which has an 11-step gray scale, plus 28 color chips including four skin tones, plus chroma blue and chroma green, plus a gradient sweep of black to white and a gradient sweep of cyan to magenta. It looks like this one, except with 28 color chips instead of just 16.

I set exposure to where the "X" crossed exactly at 50 IRE, and the peaks were at 90 IRE. I kept the light constant and adjusted the ISO up in each step, and adjusted the iris to compensate.

From 500 ISO up to about 4000 ISO, there was practically zero observable chroma noise anywhere in the image. No color blocks, no splotchy noise. None. From 4000 to 8000, if you pixel-peeped like a madman you could tell there was a little noise in some of the color boxes, moreso in the magenta box than anywhere else, but quite mild. 16,000 and 32,000 ISO showed a lot of noise everywhere, as to be expected.

In strictly luma noise (fine grain in the blacks and grays), there's pretty much zero from ISO 500 up to about ISO 1250. And by zero, I mean -- no reasonable person, at a reasonable viewing distance, is going to see any noise whatsoever. If you do like I did and sit three feet away from a 65" screen and stare in the blacks, yes you might discern that there's the tiniest bit, but only an extreme pixel-peeper would do that... At around 1250 there's a little detectable in the black and darkest gray shades, but I would find it completely acceptable; in fact I think it's totally acceptable up to at least ISO 2000, and that's only because I'm hedging my bets -- in reality I think it looks pretty decent up to about ISO 4000, but -- I grew up on film, a little granularity doesn't scare me, and I know that the art of pixel peeping has become extremely demanding, so -- yes, you may see a little bit of noise in the image at 3200 ISO, primarily in the darkest two gray chips. But I would have no hesitation shooting at ISO 2000 or even 2500, and only the slightest blip of concern going to ISO 4000. Above 4000 there does start to be noise in the shadows; it doesn't look "bad" but it's definitely getting grainy down there. Even then, it's strictly luma noise -- there was no detectable chroma noise in the image. And that's a really good thing.

I viewed the footage in the NLE using 8:1 magnification, and on my production monitor in pixel-by-pixel mode, and on a 65" HDTV from about 3 feet away. In 8:1 magnification it's easiest to see the noise (duh); on the production monitor it was not easy, and at a reasonable distance (10 feet) from the 65" screen it was pretty much spotless up to at least 3200 ISO. You had to get close and pixel-peep before you could see the noise in the shadows.

All this testing was done in 1080p, HD NORM gamma, NORM1 matrix, using "low sensitivity mode." Obviously you gain about a stop when using the "high sensitivity mode."

Again, keep in mind this is at the state the camera is in today, on version 0.32 of the firmware. Changes, advances, and improvements are possible as the image is finessed by the engineers over the next couple of months.
 
Last edited:
Sounds pretty darn good. Barry, could you tell us what kind of dynamic range you're getting? With and without log? I know it's a pre-production camera, but if you can test for noise, then maybe DR is not out of the question.
 
Barry_Green said:
Just fiddling with noise and sensitivity today...

...From 500 ISO up to about 4000 ISO, there was practically zero observable chroma noise anywhere in the image. No color blocks, no splotchy noise. None. From 4000 to 8000, if you pixel-peeped like a madman you could tell there was a little noise in some of the color boxes, moreso in the magenta box than anywhere else, but quite mild. 16,000 and 32,000 ISO showed a lot of noise everywhere, as to be expected.

In strictly luma noise (fine grain in the blacks and grays), there's pretty much zero from ISO 500 up to about ISO 1250. And by zero, I mean -- no reasonable person, at a reasonable viewing distance, is going to see any noise whatsoever. If you do like I did and sit three feet away from a 65" screen and stare in the blacks, yes you might discern that there's the tiniest bit, but only an extreme pixel-peeper would do that... At around 1250 there's a little detectable in the black and darkest gray shades, but I would find it completely acceptable; in fact I think it's totally acceptable up to at least ISO 2000, and that's only because I'm hedging my bets -- in reality I think it looks pretty decent up to about ISO 4000, but -- I grew up on film, a little granularity doesn't scare me, and I know that the art of pixel peeping has become extremely demanding, so -- yes, you may see a little bit of noise in the image at 3200 ISO, primarily in the darkest two gray chips. But I would have no hesitation shooting at ISO 2000 or even 2500, and only the slightest blip of concern going to ISO 4000. Above 4000 there does start to be noise in the shadows; it doesn't look "bad" but it's definitely getting grainy down there. Even then, it's strictly luma noise -- there was no detectable chroma noise in the image. And that's a really good thing...
Click to expand...

Thanks Barry. That is really helpful information. It sounds like it's going to perform WAY better in low light than the AF100.
 
Thank you Barry for the helpful info. Are you able to compare your findings on noise with the GH4 sensor? Thanks.
 
do you suppose you could post sole footage that demonstrates once and for all how sharp that lens is? some uncompressed stuff? since there was so much grumbling over that? thanks.
 
philiplipetz said:
Barry you shot in HD and I would expect the downsampled image to reduce 4k noise by about a stop. Is this your experience?
Click to expand...
I think that's what he wrote a couple of posts before. Now I wonder which image has better low light performance: Donwnsampled 4K to HD in post or the native HD?

Or was this just your question, Philip? ...
 
joe 1008 said:
I think that's what he wrote a couple of posts before. Now I wonder which image has better low light performance: Donwnsampled 4K to HD in post or the native HD?

Or was this just your question, Philip? ...
Click to expand...

yes. And I was enquiring about the 4k noise distribution between chroma luma. They might be potentially different as downsampling increaese color depth.
 
:grin:
That was funny. I love that bit where the fat guy is behind the pole :thumbsup:
Sound was a bit out of sink though (IMO), I had to turn the sound off to watch it without laughing.

On a pre realese camera I'm not going to make huge judgement. The only big thing I noticed was a few times where the light was a little to different from one shot to the next (and not the shots with the guns :)). But it's nothing a little tweaking in an editing program couldn't fix.
 
Last edited:
Sumfun said:
Sounds pretty darn good. Barry, could you tell us what kind of dynamic range you're getting? With and without log? I know it's a pre-production camera, but if you can test for noise, then maybe DR is not out of the question.
Click to expand...
VLOG-L is a full 12 stops.

The other gammas range from 10 stops (for HD NORM) up to 11+ stops for FILMLIKE-3.
 
Mike Harvey said:
Thanks Barry. That is really helpful information. It sounds like it's going to perform WAY better in low light than the AF100.
Click to expand...
In general, yes. Especially when using the 14-140 on the AF100. But, keep in mind, you could have put an f/0.95 lens on the AF100 too. I did a test shot once where I did a three-point lighting setup using a single candle and a few mirrors as my light source with the AF100; with that f/0.95 lens, it pulled it off...
 
Josh Bass said:
do you suppose you could post sole footage that demonstrates once and for all how sharp that lens is? some uncompressed stuff? since there was so much grumbling over that? thanks.
Click to expand...
I don't have any way to capture uncompressed at this time, and I am not comfortable posting footage from an engineering sample. Especially because I know for a fact of some things that are going to be addressed. I did post a resolution chart extract which shows pretty definitively that the lens & chip are sharp as hell.
 
philiplipetz said:
Barry you shot in HD and I would expect the downsampled image to reduce 4k noise by about a stop. Is this your experience?
Click to expand...
I shot HD because that's what my bigscreen monitors were, and I wanted to compare against an HPX250 which is only HD. I found that the noise in the shadows at ISO 800 was about the same as what the HPX250 does at ISO 320.

I haven't done testing specifically for noise values between 1080 and 4K, but on the HC-X1000 I definitely observed that noise was lower on 1080-recorded footage, and downsampling 4K footage to 1080 in post definitely reduces the appearance of noise, so I have been expecting that yes, 1080 on the DVX200 should have about 6dB less observable noise than 4K footage. But -- I haven't verified that yet.
 
joe 1008 said:
I think that's what he wrote a couple of posts before. Now I wonder which image has better low light performance: Donwnsampled 4K to HD in post or the native HD?
Click to expand...
Perceived sensitivity is the same. Sensitivity to noise ratio, I would expect HD to be a little cleaner and thus able to withstand more gain than 4K/UHD.
 
Barry_Green said:
In general, yes. Especially when using the 14-140 on the AF100. But, keep in mind, you could have put an f/0.95 lens on the AF100 too. I did a test shot once where I did a three-point lighting setup using a single candle and a few mirrors as my light source with the AF100; with that f/0.95 lens, it pulled it off...
Click to expand...

I think I remember that test. However, right now 99% of what I shoot on my AF100 is with the 12-35mm and 35-100mm. At least for me (well, more like the producer I shoot all my medical stuff for), once I start getting into the 1,600-2,000 ISO range, the noise starts to make the footage unusable. If I'm understanding you correctly, that probably won't happen until 3,200-4,000 ISO, and higher in high sensitivity mode. Would that be a fair assessment-- noise on the DVX200 at 3,200-4,000 ISO is comparable to the AF100 at 1,600-2,000 ISO (as of right now, on a pre-production model, with v0.32 firmware which may or may not improve)?
 
You must log in or register to reply here.
Back
Top