A7S + 7Q+: Chromatic noise - where to check?

A

A7s_user

Active member
Yesterday I had another filming with my A7S (Cine4 PP, FF mode, Sigma 11-24mm Nikon mount, connected via HDMI to Odyssey 7Q+, recording in 4:2:2 LT UHD

On set I did not recognize, but at home I saw the fierce chromatic noise of fine red and green dots in the blue sky. The Odyssey shows it (on the 1280x720, that's awesome), but it was recorded! It becomes less or stops in darker areas of the sky and water, but its indcredible intense in the blue just over the skyline on the right side. It did not depend on focal length (used 15-24) nor on the focus (changed a little after changing focal length). The skyline is not in focus, I filmed a barge with fireworks 200m in front of the skyline.

http://84.200.249.7/tmp/15mm.jpg
http://84.200.249.7/tmp/24mm.jpg


I urgently need help please, what may have caused this, since I'm doing an important trip with exactly that equipment soon. I try to reproduce this evening (at the moment its to bright). It did film with that lense just one time before, same settings, without any chromatic noise, but then it was darker, like on the left side of the images. Please watch them unscaled in a dark room.

Where would you assume the error? In the lense (when electronic of A7s/7Q+ did not break after the shot before, I would expect that, since I filmed often before other blue skies with same equipment and settings but other lense)? On the sensor/A7S? HDMI (connectors, cable)? 7Q+?

Thanks for any hint in advance.
 
A7s_user said:
Where would you assume the error? In the lense (when electronic of A7s/7Q+ did not break after the shot before, I would expect that, since I filmed often before other blue skies with same equipment and settings but other lense)? On the sensor/A7S? HDMI (connectors, cable)? 7Q+?

Thanks for any hint in advance.
Click to expand...

This looks like very high ISO noise, is that possible? Was it also hot?

When shooting stills the A7s applies quite aggressive noise reduction at high ISOs but in video it cannot do that.

Having said that i would be fairly certain that a good noise reduction pass would remove that (NeatVideo for example) as the nature of the noise would change frame by frame and temporal noise reduction would clear that up.

What was the ISO? You can test this in the dark and just dial up the ISO and see if you get the same effect.

cheers
Paul
 
Hi Paul,

thanks for hooking in.
I always film with 3200 ISO, never more. This kind of fierce noise (it is clearly noise changing frame by frame, although I never saw something like this, with a cloud of red and a cloud of darker green pixels) never occured before with exactly this camera and this settings. Not internal at 1080p, not external in UHD on Odyssey. It was something "special" (since?) yesterday. I do always use noise reduction in post, but did not try on this footage yet. This may be a solution (will increase temporal if needed, thanks), but I would like to avoid it or at least know, when its going to happen - or if something is broken.

No, was not hot, 22-23°C after a sunny day, no wind. Recorded externally without noise reduction via HDMI.

Will check to repoduce this at tonights blue sky, increasing ISO for a test. Its not going to happen in really dark (lense cap or very dark sky, even yesterday), that did I check already.
 
Last edited:
A7s_user said:
Hi Paul,

thanks for hooking in.
I always film with 3200 ISO, never more. This kind of fierce noise (it is clearly noise changing frame by frame, although I never saw something like this, with a cloud of red and a cloud of darker green pixels) never occured before with exactly this camera and this settings. Not internal at 1080p, not external in UHD on Odyssey. It was something "special" (since?) yesterday. I do always use noise reduction in post, but did not try on this footage yet. This may be a solution (will increase temporal if needed, thanks), but I would like to avoid it or at least know, when its going to happen - or if something is broken (eg. sensor lasered?).

No, was not hot, 22-23°C after a sunny day, no wind. Recorded externally without noise reduction via HDMI.

Will check to repoduce this at tonights blue sky, increasing ISO for a test. Its not going to happen in really dark (lense cap or very dark sky, even yesterday), that did I check already.
Click to expand...

at 3200 that is weird. When you bump up the ISO, like 50k+ you start to see this kind of patterning.

It's definitely camera side, not lens. And more testing to narrow it down or repeat it will be needed. Once you've managed to duplicate it then we can look in more detail.

There's also a small chance it could be cable related, or picking up interference from something either on your rig or at the location.

cheers
Paul
 
Thank you both!
So I will try to repoduce. I would not think its the HDMI, since it was not all over the image, but in some special brightness/color only and I use that (good, strong) cable since weeks.
There was a DJI Phantom III sender 15m away and the Phantom in 150m height. I would not assume so ;)

Will keep you informed.
 
So, its still difficult.

Checking the shot before yesterday at 100% I recognized now clouds of clearly blue dots and few reds. I did not see that before, just oversaw them. Checked all shots before that one and there never was any disturbing dot. So what did I change? Just updating 7Q+ to the May firmware from 5.10 and changing the lense. And had some large scale displays with lots of light and beams.

All colored dots are extremly sharp. They are all one pixel and not comparable to the normal noise within the image. Normal image noise, even when colored, is softer, blurred, more then one pixel in size. These new dots look like some kind of electronic errors/artefacts, but not like regular noise.

I did a lot to reproduce it today in the dark blue nightsky. I additionally used a flash light to hit the lense from different directions, I raised ISO to 25600, I shaked cables/connectors. I was only able to produce very few dots, not massive clouds as yesterday and the shot before.
It seems, but this is not 100% verified, that the one cable suddenly produces these one pixel sized dots, depending of the brightness and light sources hitting the image.
An older, very cheap cable does not produce these well defined dots. It has colored noise, even red dots at 25600 - but they are bigger and blurred.

So, I would not expect that a "cheap / bad" cable can blur noise/dots that way, that they seem to be gone (the greens and blues) or blurred (the reds). If not, then its probably the one cable producing the colored noise. I am really not sure, since I could not reproduce the amount of them seen before. Need different lights. I am even not sure, if its maybe the joining of /Q+ and A7S, since some time the one cable seem to be more fine, next switch to on its worse...

The good thing is: temporal noise reduction fully removes the 1 pixel dots. But I need to investigate more with the cables and light sources hitting the image in the dark...
 
Last edited:
A7s_user said:
So, its still difficult.

Checking the shot before yesterday at 100% I recognized now clouds of clearly blue dots and few reds. I did not see that before, just oversaw them. Checked all shots before that one and there never was any disturbing dot. So what did I change? Just updating 7Q+ to the May firmware from 5.10 and changing the lense. And had some large scale displays with lots of light and beams.

All colored dots are extremly sharp. They are all one pixel and not comparable to the normal noise within the image. Normal image noise, even when colored, is softer, blurred, more then one pixel in size. These new dots look like some kind of electronic errors/artefacts, but not like regular noise.

I did a lot to reproduce it today in the dark blue nightsky. I additionally used a flash light to hit the lense from different directions, I raised ISO to 25600, I shaked cables/connectors. I was only able to produce very few dots, not massive clouds as yesterday and the shot before.
It seems, but this is not 100% verified, that the one cable suddenly produces these one pixel sized dots, depending of the brightness and light sources hitting the image.
An older, very cheap cable does not produce these well defined dots. It has colored noise, even red dots at 25600 - but they are bigger and blurred.

So, I would not expect that a "cheap / bad" cable can blur noise/dots that way, that they seem to be gone (the greens and blues) or blurred (the reds). If not, then its probably the one cable producing the colored noise. I am really not sure, since I could not reproduce the amount of them seen before. Need different lights. I am even not sure, if its maybe the joining of /Q+ and A7S, since some time the one cable seem to be more fine, next switch to on its worse...

The good thing is: temporal noise reduction fully removes the 1 pixel dots. But I need to investigate more with the cables and light sources hitting the image in the dark...
Click to expand...

Which firmware are you on for the O7Q+ and i'll match that.

I agree, it's electronic noise, not sensor noise. The fact that they are red or blue (digital extremes) seem to suggest electronic drop outs of signal. All roads are pointing to cabling or connectors.

UHD over HDMI is very high data rates, HDMI is a consumer connector too, so it was never really designed for field use. Are you using HDMI cable locks?

I would doubt that the image being recorded is affecting the dots, flashlights and so on because if it was then that would be a sensor issue. I really suspect that the cable, connectors or interference from, say a DJI wireless remote, is being picked up. Did you have a mobile phone in your pocket? They burst microwave signals all the time.

I've never seen dots like this so far and i've been using a 2m long thin HDMI cable from a gimbal to an O7Q off board. But it's an area i would be concerned with.

If you can let me know which firmware version you're on i'll see if i can match that. If i get the noise then we know it's firmware related.

cheers
Paul
 
Thanks again Paul for your helpful thoughts. I am not blaming the firmware on 7Q+ for the noise, but when suddenly something extremly changes, it must be caused by an event. Firmware is the last May update 2015.5

The dots are

  • green (a deemed, not full green), which can by clearly seen all over the two screens above, especially in the dark parts of the sky. They are all over the image, but can be seen only in totally dark room with calibrated and sharp monitor
  • red (its a full 255,0,0 red), which show as massive cloud in the brighter parts if the sky in above screens. In my tests and the shot before the screens they are much more seldom, just 2-3 per frame and not in all frames.
  • blue (a dark blue, but brighter than the greens): only 2-3 dots in above screens can be found above the water, but they were all over in the shot taken before. And: they were suddenly there, within one recording, when the show became very bright with large comets and stars. There was a laser (which could not hit the sensor because of my position angle) but when laser and/or bright comets where visible, the blue dots are thousands (especially when lase was blue), but almost no with less light.

After these studies and the tests last night the reason seem somehow to be a mixture of HDMI and light brightness / color. Without light there was nothing (at the show before above screens and the tests yesterday; but the screens above show lots of green and red dots without real light influence). Yesterdays tests with the probable "defect" HDMI cable just showed few red and green dots when the flashlight hit the image. Within the shadows there was (almost) nothing without light. The same test with an old cable did not bring up any of the electronic 1 pixels, but at ISO 25600 there were 1-2 blurred red dots in single frames when light came in.

I use some kind of locks on both sides (fixed HDMI-micro-adapter with Sony default plastic lock on A7S side + HDMI-mini-adapter fixed with screews and straps on Odyssey side) and connect with HDMI-A cable only. The adapters stay fixed all the time. I tried to move these locks yesterday, but that had absolute no effect.
 
A7s_user said:
After these studies and the tests last night the reason seem somehow to be a mixture of HDMI and light brightness / color. Without light there was nothing (at the show before above screens and the tests yesterday; but the screens above show lots of green and red dots without real light influence). Yesterdays tests with the probable "defect" HDMI cable just showed few red and green dots when the flashlight hit the image. Within the shadows there was (almost) nothing without light. The same test with an old cable did not bring up any of the electronic 1 pixels, but at ISO 25600 there were 1-2 blurred red dots in single frames when light came in.
Click to expand...

The signal in the HDMI AFAIK is YUV not RGB, so that might explain why the red green and blue aren't all at 100%.

I don't really know how the HDMI signal is physically transported in the cable, but i wonder whether the drop outs are based on frequencies in the cable which is why sending a different luma value shows more or less drop out.

I think the key point is the old cable didn't show anything, except sensor red noise.

Time for some new cables i think!

I'm on the same firmware at the moment and i have some lower light stuff to do later so i'll keep any eye out for it - but i've never seen this in the past

cheers
Paul
 
You must log in or register to reply here.
Back
Top