Video Quality Vs Bit-Rate

[Mike Schell]

While we have been quite busy at NAB, I wanted to take a few moments and post this chart which compares the video quality vs compressed bit-rate for the nanoFlash and Flash XDR. We used the Video Clarity System Monitoring, Alarming, Perceptual Quality Testing which models the human visual system using a "Just Noticeable Differences Scale, where "0" indicates absolute best quality (no discernible differences between the original and the compressed frame of video).

You will notice that the video quality drops off quite rapidly below 50Mbps and rises quite slowly above 220 Mbps. As many users have discovered, 100 Mbps Long-GOP is the sweet spot for video quality and storage requirements. However, you do continue to see improved quality up to 280 Mbps, but with dimenishing returns. Above 100 Mbps, we found little difference between Long-GOP and I-Frame, which quite frankly surprised us.

We used a high action football video sequence for all tests. Results may vary from one test sequence to another, but I think the general conclusions will remain the same. You do get a significant improvement in video quality going from 35 or 50 Mbps to 100 Mbps (and beyond).

I am sure this chart will generate a ton of questions. Please recall that the results are from an independent 3rd party test system, developed over a 5 year period to model video quality. The results are completely non-biased and repeatable.

Best-

[Piotr Wozniacki]

Dear Mike,

Thanks for sharing. As you can imagine, there are however other factors equally important as the acquisition quality:

1. Bitrates available when performing any necessary rendering in an NLE

2. Delivery methods.

As far as point 2 goes, the current highest quality consumer format (BD) is much too limiting, unfortunately. But we cannot do much about it, so let's consider point 1.

It would be great if CD worked with the major NLE vendors so that - for mxf/mov formats - their products can preserve at least 100 Mbps.

The previous Vegas Pro version (8.0c), even when used with its 50 Mbps, 422 template, produced full 100 Mbps files when smart-rendering the 100 Mbps mxf timeline. Unfortunately, it doesn't work with 9.0d any more...

Do you think there is a chance of talking SCS and others into it?

[Darek Sepiolo]

Piotr,

As fas as I know Final Cut preserves the full quality of the clips but to take advantage of the extra bits you need to convert XDCAM files to ProRes codec. Then you will have c 180 Mbps using ProRes HQ for 1080p25 clips.

This is what I've red on one of the forums but I should get my own nanoFlash on Friday and I'll be doing excessive testing during the weekend.

Darek

[Piotr Wozniacki]

Hi Darek,

Nice to meet you here :)

I don't know much about Final Cut as I'm on a PC.

However, even the solution you mention requires conversion. I guess also using Vegas, one can preserve the nanoFlash acquired bitrate using intermediates like Cineform.

It would be great to be able to use the native clips...

[Rafael Amador]

Hi Mick,
No much surprise on the chart, although it doesn't reflex well the difference GOP/Intraframe.
I really would like to see the same test comparing the different Prores flavors.
Piotr,
FC will accept all the data rates and will keep them as long as you don't render.
If you render everything will get crunched to 50Mbps that is the standard XDCAM 422 (SONY).
But I don't see any reason to render back to XDCAM.
Most of the people edit native NANO and export to Prores.
rafael

[Darek Sepiolo]

It's quite natural while working with FCP to convert to ProRes. It's the native codec of FCP and it's widely recommended to do this. If you're not in a rush it's not a big deal. But for sure it'll be nice to have native support for 100Mbit.

[Peter Corbett]

I've been working with XDCAMHD 422 clips in Final Cut Pro with a SD timeline and it's been a pain in the butt with slow responsiveness and slow rendering. The view from FCP users on another forum was to convert everything to ProRes, but with 500+ clips this is very tiresome. Not sure what the answer is. Go with the lightweight Nano or go with with a Ki Pro with native ProRes but more bulk and less portability?

Peter

[Andrew Stone]

If you are considering the KiPro (which I was at one time) make sure you check to see where AJA is at in offering HD resolution. Last time I checked they were still not offering all of the standard frame sizes in the standard frame rates when recording. In theory, I think the KiPro has a lot to offer with 10 bit / 4:2:2 / Pro Res files as being the most attractive thing about the product. Cross and up/down conversion is extremely useful and a VTR like interface for studio or field review of clips.

After a lot of thought I decided to get the nanoFlash. In hindsight, it was the right decision, at least for me, and not the least of which is the fact they have delivered a product that does not have glaring omissions of essential functionality (see above) for an HD recorder.

[Eric Liner]

Thanks for posting the test, Mike. I think it confirms some of our early observations, among them that 220mbit I-frame (once the highest setting available) is visually better (color, clarity, detail, etc) than 100mbit Long-Gop.

I'm also surprised that the differences are so unnoticeable between high bit rate Long GOP and I-frame as to not differentiate between them.

Lastly, (and maybe it's obvious or I'm interpreting the results incorrectly) I'm surprised that the luma takes a bigger hit than chroma through the compression process. I would have thought that retaining chrominance info through the heavy mpeg compression would have been more difficult than retaining luminance.

Thanks again for keeping us informed.

[Gints Klimanis]

I'm pleased to see a nearly linear increase in quality up to 140 MBps at nearly the same slope as that between 50 and 100 MBps.

Since I also record on an EX1 at 35 MBps on 16 GB SDHC, I may as well fill my Nanoflash 2x32GB flash memory to capacity. It would be good to record at slightly less than 140 MBps, say at 135 MBps. I don't find that a coarse resolution in bit-rate complicates the operation of the Nanoflash and would appreciate a 5 MBps resolution in the BitRate parameter if that is possible.

[Tim Polster]

Interesting. So when using 100mbps, are long GOP and I-Frame interchangeable? or only above 100mbps?

If so, then it would seem I-frame would be the choice for editing reasons.

This situation reminds me of racing car design. You want them to be as aero efficient as possible, but you can't replace shear horsepower for speed. In this case, a high bitrate cures anything a codec might miss.

[Tim Polster]

Just wanted to ask again, when using 100mbps, are long GOP and I-Frame interchangeable?

Or does this only apply to bitrates above 100mbps?

[Dan Keaton]

Dear Tim,

I just found out the answer yesterday.

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

At 100 Mbps Long-GOP is better.

At 140 Mbps, they could be about equal, but I do not have a definitive answer.

We welcome comments from others.

[Tim Polster]

Hey Dan. Thanks for your reply.

Is there anything inherently better with I-frame outside of NLE's ease of handling?

Like better slow motion etc...

Or to ask, when would one want to use I-frame?

[Dan Keaton]

Dear Tim,

One should use I-Frame whenever there is a substantial time difference between frames.

An example would be Time Lapse. When the image can change dramatically from frame to frame (and it is expected to change dramatically from each frame to the next, then I-Frame is a logical choice.)

In this example, long time-lapse intervals, say one hour or more between frames, or one frame a day, there would be substantial differences between each frame.

Also, please note that the nanoFlash/Flash XDR will force, internally, I-Frame Only, automatically, for these situations where I-Frame Only is required.

(Please do not misunderstand, Long-GOP can handle substantial differences from one frame to the next and can handle 100% difference from one frame to the next).

Over and Under-Cranking is another area where I-Frame Only makes sense.


Now, to get really technical:

For extreme high quality, at very high bit-rates, such as our 280 Mbps, I-Frame Only can have a quality advantage. The advantage is ever so slight, but we are talking "extreme high quality" here, so we are splitting hairs.