Noise comparison: 35/4:2:0 vs. 180/4:2:2

[Gints Klimanis]

Besides Mike Schell's informative bitrate vs. quality chart, I'd like to see more information to help me decide when to use 100 MB LongGOP vs. 140+ MBps I-Frame only. 140 MB LongGOP seems to deliver more shimmering but few if any other gains than 100 MB LongGOP. 100 MB LongGOP is perceptibly better than 35 MB SxS, and comparison frame-grabs always show less mosquito (macroblocking) noise in the Nano footage.

Right now, I feel like I have to do these quality tests myself.

[Dave Nystul]

While I have followed this thread with some interest, our facility is Avid based. As such, I shoot I- frame at 220 Mbps in MXF. It does get tiresome moving media to drives but when I see the differences between the SxS footage and the Nano footage it really becomes a no brainer. If Long GoP does introduce noise at certain bit rates I'm grateful for those who are learning the limits of this technology and raising awareness for the benefit of all.

[Andrew Stone]

I don't see it as the limit of the nanoFlash technology or really the high bitrate but an indication of the limitations of the source. The nanoFlash at higher bitrates is capturing in more detail the material coming into it. If the camera is noisy, it will capture it more faithfully than at a lower bitrate where there is smearing going on.

There is a reason why camera ops, DPs and others bathe their scene's in light other than for aesthetic reasons. It's to get the image out of the noisy range of the image sensor's range. You want less noise, use more light.

[Gints Klimanis]

Don't you see additional noise/detail with 220 MBps I-Frame only? I do, and I'm not surprised. Some of the cleanliness of the EX1/EX3 was simply due to Sony's encoder working at a lower bitrate.

The important discussion point is whether LongGOP magnifies MPEG "I-to-B/P frame shimmering" more and more as the bitrate increases. Are 140 and 180 MBps worse than 100 MBps? From my own tests, 50 MBps LongGOP is not enough to eliminate enough macroblocking artifacts. 100 MBps I-Frame-Only is too soft due to detail loss. CD is putting forth their opions, but we need to coax them more. Until this discussion, I did not know that at 140 MBps, I-Frame only is preferred. I upped my bitrate from 100 to 140 MBps LongGOP a few months ago. Now, I have to do some serious testing myself to decide if I-Frame-Only is really vetter for me.

[Gints Klimanis]

Originally Posted by Andrew Stone

I don't see it as the limit of the nanoFlash technology or really the high bitrate but an indication of the limitations of the source.

While we all have seen the noise, the emerging issue is whether there is too much fluctuation in detail/noise from frame to frame in LongGOP MPEG-2. That is, is the faithfully-compressed noise too great or is the interframe detail fluctuation the primary objection?

[Piotr Wozniacki]

Originally Posted by Gints Klimanis

Piotr, is the shimmering more obvious on your 50+ inch TV?

Oh yes it is, Gints. The general rule in this discussion should be that we all examine and compare clips and stills on really big displays - watching them on a 20" (even professional and calibrated, but small) monitor will never show what we're talking about. And after all, we're delivering for big screen (often much bigger than my 50") - aren't we?

Originally Posted by Gints Klimanis

That is, is the faithfully-compressed noise too great or is the interframe detail fluctuation the primary objection?

This is the most rudimentary question, and it took me a long time experimenting before I was able to define it first, and answer it then. The noise is there, because it's in the source camera in the first place (for various reason - irrelevant to this discussion). But it's the excessive fluctuation of detail (and hence the noise) in the high bitrate nanoFlash files what makes it come to live of its own, and seriously jeopardize the obvious advantages of the nano recording!

[Eric Liner]

Originally Posted by Andrew Stone

If the camera is noisy, it will capture it more faithfully than at a lower bitrate where there is smearing going on.

This was also my original impression. But, this thread has got me wondering otherwise. However, I've noticed that a lot of this discussion revolves around use of the Nano with the EX3, a camera that I've always considered to be relatively noisy to begin with.

I'd love to hear if any others are experiencing this issue with alternative cameras. Or aren't having issues, for that matter. I know there are some on this list using some pretty sophisticated cameras in conjunction with the Nano...if you're out there, what settings are you capturing to and are there some that you avoid?

I shoot primarily with an HDW 730 and canon glass. No question, in my opinion, that the highest bit rate I-fame settings look the best and get me closest to an HDCAM recording. But, I typically (for storage reasons) shoot 180 or 140 Long GOP. When I see noise, it's usually in the shadows or underlit portions of the image--which I believe is more a result of my camera settings and chip set. On a 50" consumer set (I only have a 20 inch production monitor) the results look pretty good. But the consumer set has all sorts of scaling and noise reduction, and de-interlacing algorithms going on, so I'd expect it to look good and I'm not sure it's a fair judge.

I use the Nano because it allows me to work in difficult conditions where tape storage and maintenance is a real burden, and it delivers high quality results. As soon as I can afford to work with all 64GB CF cards and bus powered FW drives double in capacity (but not physical size) I'll shoot everything at 220 or 280. Until then, I'll hop back to HDCAM tape when the situation allows and shoot high bit rate Long GOP when it doesn't.

Cheers,
Eric

[Dave Nystul]

Gints,

I see much more color information comparing the two. Most of the time I'm shooting with the PMW-350, a quieter camera than the EX-3. Inherent noise is reduced and so is probably not an apples to apples comparison. Even when shooting the Nano with our EX-3 the benefits far outweighed any noise issues. I'll admit it would be nice to A/B each source to see a scope comparison. Additionally, most of my viewing is done on Sony 24" LCD monitors. I don't have access to anything larger.

Best,

Dave

[Piotr Wozniacki]

Originally Posted by Mike Schell

I understood that this thread was intended to discuss video quality.

I'm still counting this will happen, Mike - but we need a more technical input from you and Dan...

[Dan Keaton]

Dear Piotr,

Our best technical advice is based on the very thorough and objective results that we obtained by testing all of our modes on the Video Clarity System.

Summary:

At 100 Mbps and Below, please use Long-GOP for the best images.

Above 100 Mbps, please use I-Frame Only.


(In a post above, I explained why you may see the difference in "frame by frame" quality, when you have an image with lots of detail, or a very noisy image, and use Long-GOP above 100 Mbps.)


With that said, how may I help you?

What technical detail would you like us to provide?

[Piotr Wozniacki]

Originally Posted by Piotr Wozniacki

MPEG-2 usually does differ in the number of frames between consecutive I-frames in the GoP, depending on whether it's for the 50 Hz area (in which case, it usually has 12 frames between I-frames), or the the 60 Hz area (when it usually has 15 frames). Needless to say, [setting GoP length to 12] may be in favor of the 50 Hz area encoding quality.

You also say you're using Closed GoPs; this s good as with closed GoPs, the "All short" GoP length scheme can be taken advantage of, whereby all GoPs start with an I-frame and are the same length. Also (unlike with the "First short" or "None short" options), all GoPs are created as short GoPs - all this should potentially increase the encoding quality. But, are you using the "All short" GoP length for sure?

Dear Dan,

Let me answer with the above excerpt from one of my previous posts (I bolded out the most relevant parts).

Now, I'm not an expert in advanced L-GoP settings, but (even before acquiring my nanoFlash) I have played with various settings when rendering out both HDV and the XDCAM EX into MPEG-2. Below are the two relevant tabs of the MPEG-2 Vegas template.

I can tell you that even though the impact of tweaking these parameters is quite subtle, I have found that for 1080/25p, I achieve best results (including the most consistent frame quality within a GoP) when using the exact settings as shown in the attached template pages.

May we know whether or not the nanoFlash L-GoP encoder uses similar settings?

Thanks

Piotr

PS In the below sheets, please disregard the Template, Notes, and Format fields' contents - these haven't been updated to reflect the real settings (in this case, for nanoFlash L-GoP 50 Mbps/422 t/l).

[Dan Keaton]

Dear Piotr,

We use the commonly accepted standard of 12 frames per Long-GOP for 720p and 15 frames for 1080.

Sorry, but I do not currently have the level of detail that you are requesting.

I hope you understand that we have to create files that are usable with a wide range of Non-Linear Editors. Tweaking codec parameters, and then testing them against every possible editor, with all of the frame rates and options can be a very lengthy process.

[Piotr Wozniacki]

Dear Dan,

The settings I'm talking are by no means tied to a single encode (MainConcept), or NLE (Vegas).

They're general MPEG-2 settings that all encoders (software and hardware) use, and their purpose is getting the best compromise of quality vs. the required encoding speed, storage etc.

If you *know* otherwise, I'm all ears.

[Eric Liner]

"Our best technical advice is based on the very thorough and objective results that we obtained by testing all of our modes on the Video Clarity System.

Summary:
At 100 Mbps and Below, please use Long-GOP for the best images.
Above 100 Mbps, please use I-Frame Only."


Hi Dan,

Are you telling us to avoid using Long-GOP above 100mbps? Are we talking about a point of diminishing returns or "deteriorating" returns? That's not how I interpreted the chart that was distributed with the VCS test results. I'll go back and check that thread.

However, if this is the case (that our video will actually suffer as a result of using 140 or 180 Long GOP), why would you even include these as optional settings?

Eric

[Eric Liner]

Hi Dan,

Just FYI, here was a quote from the thread following the VCS test:

"Our current feeling is that at 180 Mbps I-Frame Only and Long Gop are about equal in quality."

Have you changed your opinion on this?

Just to be clear, my understanding was that 180mbps was a little bit starved for creating high quality I-frames... Whereas180 would be more than adequate when taking advantage of the more efficient Long-GOP structure.

Thanks,
Eric