AF101E Video Noise on Black parts of the image - looks like snow
- Thread starter jourman
- Start date
Barry_Green
Moderator
It's not intended to be a general discussion of noise. It's intended to be a specific examination of one potential (and I'm pretty sure correct) hypothesis.
oh sorry, i guess i am misreading oldcorpse's quote.
Originally Posted by OldCorpse View Post
Understandably, possible noise is an issue that's very important to the readers here. Instead of having multiple threads all on noise, why not have a big one where we can consolidate all the concerns and experiments as we try to get to the bottom of this? It's very distracting coming across multiple disconnected threads, because then we go over the same ground again and again. Mods?
Ok using the new footage playing on my Win7 computer, nVidia 480GTX video card and Dell 30" monitor, calibrated using the Win7 tool I see a ton of swimming noise on the DSC labs color card shot in the upper left corner and upper right corner. This is playing with the Zoom Player. Which I believe uses Direct Show and windows to decode. I also see this using VLC. The test pattern shot seems ok. The shot of the objects seem ok but there is noise in the box holding up the frog, it's not as bad as the color card but it's still there. Swimming noise. I converted the files with Neoscene and get the same results using the Zoom Player. I also get the same picture with windows Media Player. My quicktime won't open .MTS files (maybe that's a mac thing?) Note I am about 2-3 feet from the very large screen though. If I step back to 7-8 feet (most I have room for) the noise isn't as noticeable but it's still there.
I took the untouched Private folder, put it on a Class 10 Sandisc SD card and stuck it in my PS3 hooked to a Panasonic plasma TH50PZ700U. On the plasma the noise on the DSC labs color card (file 02) is noticeable even from 10 feet away. The shot of the objects you can see the noise on the box holding up the frog but it's not that bad from 10 feet away. Up close it is noticeable. The color pattern file seems fine, at least it doesn't have the swimming noise. The color bars on my plasma do seem a bit fuzzy on their edges (more than on the computer/dell) but it doesn't seem to have swimming noise.
The DSC labs card noise is very bad in the upper left hand corner, just bad in the upper right. The center black seems ok though.
Back to the original question, what playback system under Win7 doesn't show noise? I'll accept it's something about the decoder software I'm using, but if that's true there must be a blessed path, tell me what it is and I'll use it and tell you what I see. But using the same playback tools I've always used I'm seeing a lot of swimming noise.
I took the untouched Private folder, put it on a Class 10 Sandisc SD card and stuck it in my PS3 hooked to a Panasonic plasma TH50PZ700U. On the plasma the noise on the DSC labs color card (file 02) is noticeable even from 10 feet away. The shot of the objects you can see the noise on the box holding up the frog but it's not that bad from 10 feet away. Up close it is noticeable. The color pattern file seems fine, at least it doesn't have the swimming noise. The color bars on my plasma do seem a bit fuzzy on their edges (more than on the computer/dell) but it doesn't seem to have swimming noise.
The DSC labs card noise is very bad in the upper left hand corner, just bad in the upper right. The center black seems ok though.
Back to the original question, what playback system under Win7 doesn't show noise? I'll accept it's something about the decoder software I'm using, but if that's true there must be a blessed path, tell me what it is and I'll use it and tell you what I see. But using the same playback tools I've always used I'm seeing a lot of swimming noise.
Barry_Green
Moderator
Wrong thread to report in. But to answer your question, I'm playing back using Windows Media Player 7, and there's practically zero noise in the chart. I have to really pixel-peep to see *any*. If I stretch out the gamma I can make out a little bit of noise in the blacks in the upper left and upper right, but seriously -- I doubt anyone would be able to see it. I'd have to hand 'em a magnifying glass and describe to them what noise would look like, before they'd see anything. And that's looking from 18" away. In fact, the only place on the entire chart where I see anything where there's noticeable noise at all, is in the green diagonal chip in the lower left. On that, the noise is visible.
So that's very interesting indeed. Especially in that you're seeing it on the plasma through the blu-ray. Did you run the THX calibration on your TV?
So that's very interesting indeed. Especially in that you're seeing it on the plasma through the blu-ray. Did you run the THX calibration on your TV?
Barry_Green
Moderator
Not sure about threads to report in but one thing that might be a factor is what is doing the decoding inside Windows Media Player (or in my case Zoom player). Turns out Zoom player will tell you, and it looks like when playing .MTS files it's using the ffdshow MPEG-4 video codec. Which I don't think is very good(but very compatible which is why I use it). And my Windows Media Player would also use it (since it's setup as the renderer for .MTS files). I'm not exactly sure how to tell what native Media Player is using for the decode, in Zoom Player you can right click and bring up the filter properties graph information. I didn't think VLC used the direct show stuff though but it may also be based on the ffdshow work. In fact I'd bet it is.
Yes I've ran the THX calibration on my TV though its options for adjustments are slightly limited. But I have no idea what the PS3 is using to decode the .MTS files. It may have the same problems ffdshow has.
Yes I've ran the THX calibration on my TV though its options for adjustments are slightly limited. But I have no idea what the PS3 is using to decode the .MTS files. It may have the same problems ffdshow has.
I agree 100%. Seems pointless to do all this testing with footage shot with prototypes and pre-production cameras. How about waiting a week or 2 until the production models are in people's hands and see what the feedback is regarding noise? If there seems to be an issue then we start testing it.
RieniO
Well-known member
Now h.264 might be blamed for this noise, could someone please confirm whether or not the SDI output is h.264 compressed format as well? I ask this because apparently one of those framegrabs was taken from SDI output recorded to nano. And nano records in prores format which isn't h.264, right? So where does h.264 come at play when recording from SDI or HDMI output?
Brian@202020
Veteran
The Nano does not record ProRes it records a different high end format. It's actually MPEG 2 compression but at high bitrates so it's pretty equivalent to ProRes. But yes if the AF100 outputs uncompressed out the HDSDI and HDMI like Panasonic claims it does, you should see an improvement in the Nano's version. If the noise or artifacts are present in the Nano's footage it is either not uncompressed coming out the HDSDI and HDMI, it's a different problem within the camera itself, or just flat out user error. I'm sure if it's the camera in anyway and people will demand Panasonic take action somehow. If you haven't purchased the AF100, this is where it pays to sit back and wait for the outcome.
Barry_Green
Moderator
h.264 is not to be blamed for the noise issues people are experiencing. INCOMPLETE UNCOMPRESSION of h.264 would be to blame. The footage is nigh unto noiseless when viewed properly and uncompressed properly.
HD-SDI is 100% uncompressed.
That would depend on what format your recorder records in. If you were using an HMR10, it would record in the same h.264 format. If you were to use a Ki Pro Mini, there would be no h.264, there would instead be ProRes compression. And if you were to use the Nano, there would be no h.264 or ProRes compression, but instead there would be MPEG-2 compression.
RieniO
Well-known member
Barry, could you as an expert, address that problem with the elevated blacks in op's frame grabs? Isn't that where this problem with the noise is starting anyway? Whatever you do with your cam, like adjusting gamma or over-exposing, blacks should always start at zero.... see my grabs with photoshop histograms
http://www.squarepixel.com/pics/Nano01_histogram.jpg
http://www.squarepixel.com/pics/01_histogram.jpg
Also the nano bit in the filename suggests that this is SDI output material but since OP isn't answering questions in this discussion this remains a question mark to me. I am most curious about the strange histogram. The waveform should have looked the same, with the blacks starting at 20 or something.
http://www.squarepixel.com/pics/Nano01_histogram.jpg
http://www.squarepixel.com/pics/01_histogram.jpg
Also the nano bit in the filename suggests that this is SDI output material but since OP isn't answering questions in this discussion this remains a question mark to me. I am most curious about the strange histogram. The waveform should have looked the same, with the blacks starting at 20 or something.
Barry_Green
Moderator
In properly-exposed, properly-set-up video, the blacks should start at zero, yes. But the master pedestal setting lets you raise or lower the black level. It's possible to raise it so high that even with the lens cap on, the blacks would be recorded at something like 24 IRE! So proper settings are important to get the most of the dynamic range and also get the best contrast and good deep blacks. I normally set the Master Ped to no higher than -2, which results in a lens-cap black level of zero.
RieniO
Well-known member
Ok so you can set the master ped inside the AF100 already?
I don't know what OP used to export those frames as jpeg stills. I assumed he exported it from the timeline in FCP but he might as well have done it from VLC player or other software. Would be nice if he could finish what he started and enlighten us on that. If he exported this from the timeline he must have done some really strange things to get the blacks starting that high. I export still frames from FCP every day for exporting to photoshop and I never saw a weird histogram like this before.
I don't know what OP used to export those frames as jpeg stills. I assumed he exported it from the timeline in FCP but he might as well have done it from VLC player or other software. Would be nice if he could finish what he started and enlighten us on that. If he exported this from the timeline he must have done some really strange things to get the blacks starting that high. I export still frames from FCP every day for exporting to photoshop and I never saw a weird histogram like this before.
Ok, I did a side by side today with an HPX370. The 370 was much, much cleaner than the AF100. I had:
Subject, Diva 400, 3200K Preset. The areas I saw noise were all over, even in skin tones.
F5
Cine D
Det: -6
V-Det: -6
Coring: +2
I had them both on bars set-up BT-LH1700W's
The footage, when imported via L&T to FCP then viewed via AJA IO HD still had the snowy noise.
Currently I have the camera running into the 1700:
ISO 200
Cine D
Det: -6
V-Det: -6
Coring +2
Chroma: -2
Master Ped: -2
Shooting an X-Rite Color Checker, Diva 400, 5600K Tubes, Preset 5600 WB
I am seeing snow in all areas of the image. I have black balanced.
When side by side with the 370 earlier there was no snow in the areas the AF100 had snow. The AF100 image actually got cleaner when I went back to the default F5 and F4 settings. I chose the ones I did based off of some other threads.
Please help.
Subject, Diva 400, 3200K Preset. The areas I saw noise were all over, even in skin tones.
F5
Cine D
Det: -6
V-Det: -6
Coring: +2
I had them both on bars set-up BT-LH1700W's
The footage, when imported via L&T to FCP then viewed via AJA IO HD still had the snowy noise.
Currently I have the camera running into the 1700:
ISO 200
Cine D
Det: -6
V-Det: -6
Coring +2
Chroma: -2
Master Ped: -2
Shooting an X-Rite Color Checker, Diva 400, 5600K Tubes, Preset 5600 WB
I am seeing snow in all areas of the image. I have black balanced.
When side by side with the 370 earlier there was no snow in the areas the AF100 had snow. The AF100 image actually got cleaner when I went back to the default F5 and F4 settings. I chose the ones I did based off of some other threads.
Please help.
RieniO
Well-known member
Lightmast, could you please export some still frames from your FCP sequence (from AF100 and 370 footage) in png format and upload them to a site so we can view them? Or e-mail to rieni@squarepixel.com I'll put them on my site then.
File / export / Quicktime conversion / still image
File / export / Quicktime conversion / still image
