HPX371 Noise Issue?
 

Playing around with an HPX381 and getting something odd.
Every time I do a move there is massive noise in the background. The first frame grab here is when the shot is static and the second one 2 frames later as it starts to move.
This doesn't look right to me, any ideas?
Steve

Which scene file and which gamma are you using? What is your pedestal set at?

Dan

I stuck in the BBC settings, so CineLike D gamma, Cinelike matrix, Master Ped +3.
Steve

Just looking at stills is always awkward, but it looks like the camera is using quite a bit of temporal electronic noise reduction to try to help mask the noise off the chips. It seems to be doing a reasonable job on static subjects, but is confused by movement (not surprisingly) - I assume it is relying on comparing static subjects on a frame-by-frame basis. What gain setting were you on?

I should have said David, it was -3db gain.
Same codec as the Varicam and never saw anything even remotely like that.
Steve

Here's a quick clip, when the gull's head clears frame look at the background foliage. This test clip is pretty much how it looks natively too, so it's not extra .mov or web compression.
Looks the same played straight from camera to TV via HDSDI.
Steve

Hey Steve,
What frame rate are you shooting at? Shutter Speed? Looks like lack of dynamic resolution is blurring the background so much that the codec is a little overwhelmed but not sure the noise increases based on my computer screen.
I do remember people complaining about noise in the dark areas with the HPX-300 and I have seen noise on my LCD Laptop screen from my HPX-500 so I always think Panasonic has more noise in the blacks than Sony at least on the lower end units.

Steve:

You are aware that Cinelike D is the noisiest gamma setting, right? I would shoot the same scene with the same settings, but shoot one pass with Cinelike D and one with BPress. Compare and let us know what you think.

The 371 should look pretty clean for a 1/3" chip. Barry Green did some evaluation when the 370 was released and if I recall, the 370 was a hair cleaner than the EX1 in some settings and just a shade less clean in others, but that is very clean, I describe the EX1 as "sterile" it is so clean. I don't think you have a defective camera, I think you just may be using the wrong settings to obtain the cleanest image.

Dan

It's 1080/25P AVC Intra 100, 1/50th sec shutter speed.
I don't think it's noise as such as you'd see it the same on the still as well as moving image, it definitely looks like compression but I would have thought that the codec was the strongest thing on the 371. I've certainly never seen this on any other camera I've used, and I've used a lot.
Steve

This seems like the most logical answer. The movement is uncovering what the chips are actually putting out while the static frame gets all of the attention from the noise reduction. Which if true does not paint the camera in too favorable of a light. Nor does it look like CMOS has been much of a step forward if you ever need to move the camera.

Here's another cropped area.
The first shot is on static, the second is about 3 frames later as I start to pan.
Steve

I don't think that's really relevant here - the issue is not how noisy the camera is in itself, but how the noise increases on areas of movement. A different gamma setting may make a difference to the overall noise level, but is unlikely to affect what Steve is seeing.
The question Steve's findings raise is how is it achieving that cleanness? The 300 was slated for poor noise figures, the 370 was brought out as the answer. I still can't think of a more likely answer than what I previously suggested - that electronic noise reduction was incorporated. And Steves examples seem to indicate it falls apart under certain circumstances.

Steve, have you still got the camera? If so, are you able to compare the AVC-Intra recorded output with the live output? (Or record the latter on something like a nonoFlash?) I find it hard to believe it's a codec issue, and if the live/recorded feeds show the same effect it would rule it out - would back up my theory about it being due to the front end.

The other thing I'd try is seeing how it behaves in low light with gain. Any noise reduction circuitry may not work as well then as at low gain settings.

Agree with all that David.

You can actually see it happening in the viewfinder, so I guess, yes that eliminates the codec maybe.

I questioned after Andy Shipsides' test exactly what was happening. Panny talked about this brand new sensor, but how much better could it be than the one they'd put in the 300 six months earlier? His tests showed less noise but also less sharpness as if it was largely software at work.

There is a very strange phenomena under discussion elsewhere where part of the image shifts while the rest stays put, very weird. Again it seems as if it's software working its magic but causing something odd to happen that it didn't do in the lab.

Steve

It would be interesting to compare live SDI output to a high quality, shaded monitor to AVCINTRA recording of the same. I agree with David, it would be good to try to troubleshoot the issue by isolating whether or not this is coming from the imagers/processing or the recorder end of it. Do you have a good relationship with Panasonic Europe/U.K.? What do they say about this? Your post is the first I have read of this particular issue with the 371.

Dan

It's not a pretty picture!
Here are a couple more grabs, there is some weird streaking going on here.
These are while moving the camera, and something is not keeping up. Look at the second shot and you've got multiple images of the plug holes.
These are crops and with levels boosted a bit just to make the problem a bit more visible, but it is very noticeable on the original moving footage and through the viewfinder while shooting.
Steve