Canon HF G30 Cinema Mode

[Dave Baker]

That's the one Don!

One display in particular shows bad banding, on the laptop I bought for editing a few months ago. A display I hooked to it shows a touch of it on one clip (raw camera clips), the display on the other laptop, which is not for video although capable, doesn't show it on any of the clips. A transcode of a few clips didn't improve things, neither did any driver settings. So not entirely Cinema Mode then!

One problem solved and another one to consider. The idea of a laptop for editing rather than a new desktop is so I can do some editing while away from home, but it certainly won't be much fun like that. I don't see it as a warranty claim, it hasn't broken, so I'm stuck with it as it is for now.

Thanks Don, I hadn't considered the display. Why would I, it's a new top of the range laptop? I have bookmarked the links for an in-depth read later.

[Bryan Worsley]

Glad you solved it. And there was me seeing if I had any Cinema Mode footage from my old HV30 or HV20 archived somewhere - which I don't, but I seem to recall there was an issue there with posterization on fine gradients, not too mention increased noise in the shadows. There again it could have been my monitor ;-)

[Dave Baker]

Cheers Bryan.

It looks like I shall be trying another Linux distro on that laptop, the other (good display) one has a
different distro. It could be that simple. If not, I shall have to swap them over.

[Ian Thomas]

been out today with the g40 and had no problems with the zoom bit strange to say the least!!!!!

[Dave Baker]

Update for Don and Bryan

It isn't the display, it's the driver. I booted a live distro and the offending clips played beautifully, so I installed it, still good and I'm now putting all my stuff on it. It's a relief it's only one machine!

At least I now know I can use Cinema Mode on my G30 if and when I like. Thanks again gents.

[Don Palomaki]

...not too mention increased noise in the shadows.

More bits to shadows can lift noise out of the black and it becomes visible.

[Bryan Worsley]

Originally Posted by Ian Thomas

been out today with the g40 and had no problems with the zoom bit strange to say the least!!!!!

Bizarre. Makes you wonder about possible environmental interferences - vibration, electromagnetic etc.

Originally Posted by Dave Baker

It isn't the display, it's the driver.

That would make a good bumper sticker..........for the wife that is :‑o

Originally Posted by Don Palomaki

More bits to shadows can lift noise out of the black and it becomes visible.

Right - I noticed it a bit in Cinema Mode on the HF-G10 also, but it's much cleaner on the HF-G30. Didn't check out Cinema Mode on the HF-G40 but increased shadow noise was one of the things I didn't like about Wide DR mode. Seemed to be pushing the gain/noise floor dynamics to the limits.

[Dave Baker]

That would make a good bumper sticker..........for the wife that is :‑o

Not sure I'd want to tell mine that!:-)

I'm not so bothered about noise in the shadows and it's not just shadows, the whole image is noisier than MM, but a little denoiser can work wonders, depending on how bad it is of course.

I proved it was the Nvidia proprietary driver by installing it and getting the banding back, but it won't go away just by uninstalling. I'm sorting that out now. For the record, I will post the details when I've finished in case anyone else has a similar problem.

[Bryan Worsley]

Originally Posted by Dave Baker

Not sure I'd want to tell mine that!:-)

I've tried everything else.

Originally Posted by Dave Baker

.....but a little denoiser can work wonders, depending on how bad it is of course.

Except when there is 'real' banding, in which case denoising may make matters worse, at least in an 8-bit integer processing environment. And what benefits can be gained from processing 8-bit YUV sources in a high bit depth/floating point imaging system (like Resolve) rest largely on dithered noise in the source. Likewise there's no magical transformation that takes place on converting an 8-bit source to a 10-bit intermediate - that would be alchemy; it's still 8-bit data, just padded to 10-bit. It's dithering in the inter-conversion that makes any difference. FFMPEG (and so MLT, in the context of KDenLive) applies dithering (by default) when converting 10-bit to 8-bit, btw.

If conspicuous noise is confined to the shadows, I'd be more inclined to treat it selectively with a mask. If it's in the highlights, I'd look at the trade-off.

[Don Palomaki]

Noise is easier to notice/see in shadows. Two IRE units of noise fuzz on a 10 IRE image is 20%, while the same 2 IRE units of fuzz on an 80 IRE signal is only 2.5% and less obvious to the eye.

[Bryan Worsley]

Thanks for that bit of info Don.

Originally Posted by Don Palomaki

Noise is easier to notice/see in shadows

That's always been apparent to me but I've never had it explained in empiric terms before.

[Don Palomaki]

As a quick test using an XA20 I shot video of the 6 gray blocks on a Macbeth Color Checker Chart in Manual and in Cinema modes. I set the brightest white (90% reflectance) square just into 100% Zebra (set to I00 IRE). Shutter was 1/30, aperture about F/4.0 adjusted to provide zebra, with 12 dB gain, halogen light.

The attached chart below shows the resulting IRE levels of the six gray squares for each mode. It is apparent that Cinema allocates lot more IRE range/space for shadows with under 3% reflectance.

[Bryan Worsley]

Interesting. You can also make that out in those fuzzy waveform plots I posted earlier - the Cinema Mode curve has that upturn at the bottom end - what would you call that in terms of gamma curve 'anatomy' - "toe-stretch" or something ?

[Dave Baker]

That noise comparison is interesting, thanks Don.

For the record:

I do not have the make/model of the laptop display, but the machine itself is a StationX Lancaster, the 15.6" display is matte, 1920 x 1080 IPS.
The graphics card is NVIDIA GP 106M (Geforce GTX 1060 Mobile 6GB)
The driver is the Video-Nvidia proprietary Linux driver.

To reiterate, banding in certain video clips showed up badly when the Nvidia proprietary driver was installed, but not without. Simply uninstalling the driver did not stop the banding. The OS was Manjaro KDE, but I found the same effect with other distros. A re-installed Manjaro without the driver does not have the problem and I have disabled the card in BIOS.

Just to make life more interesting, my backup drive had suddenly become read-only and unmountable, so I have just spent several hours attempting to recover the files. A success thankfully, good old Puppy Linux, I always keep a copy on hand just in case!

[Don Palomaki]

[Quote]...what would you call that in terms of gamma curve 'anatomy' - "toe-stretch"...{/Quote]

Here is how Canon describes it in the XHA1 manual. Black, knee and gamma all interact.

As to why the banding it is more apparent in Laptops, I suspect that in the interest of saving power they try do things in the least processing intensive way, and use displays that draw less power. The additional processing to do a full 8-bit or dithering is not necessary for most purposes, so they save a bit of power by doing 7- or 6-bit and skip on the dithereing. Then banding can become apparent under certain image circumstances. Changing drivers and/or firmware may adjust this, but at a power consumption cost.

I wonder if just uninstalling the drive did not effect it because the drive installation may have included a firmware update. I know that happens with the BlackMagic Design Intensity cards.