picture quality and bitrate

[Cees van Kempen]

recently purchased my nanoflash and i am very happy with. However, I did some test shooting with different bit rates, varying from 18 up to 100 Mb/s (1080p25). Camera EX3. Took shots of exactly the same subject (red car, with yellow shopping bag in front of a green hedge). Played back the footage from the nanoflash on my 60" Kuro flatscreen. I tried very hard to seen any difference betwen the different shots, but even when the picture was changing from 18 Mb/s (4:1:0 is suppose) directly to 100 Mb/s (4:2:2) I could not see any difference in picture quality (it all was stunning anyway). Am I doing something wrong to judge the effect of the bit rate, or should I get my eyes checked?

[John Richard]

In your low motion, low fine detail shot the only thing you would probably notice different is the depth of color. The 4:2:2 footage should appear to have a "richer" or "deeper" color.

With the higher data rates, you will notice a difference in footage with lots of fine detail like rushing water, blowing grasses, leafy trees, closeups of detailed machinery and/or lots of motion. CD has some comparisons of sparks off a grinder that is a combination of high detail and fast motion on their website which show the differences dramatically.

Try shooting some scenes with high detail and fast motion and then run your comparisons.

[Mike Schell]

Hi Cees-
You may need to stop on an individual frame to really see the difference. If you do any color correction or composting, you will notice a dramatic improvement with the 4:2:2 and 100 Mbps data-rate.

We just finished a comprehensive test with a system know as Video Clarity. This product models the human visual system to measure overall picture quality. We did find some surprising results when we tested bit-rates from 18 to 280 Mbps. In general the higher the bit-rate the better the quality, but there was not much difference between Long-GOP and I-Frame once you got above 100 Mbps.

I plan to publish the full report in the coming weeks with detailed charts.

Best-

[Steve Phillipps]

I think you'd also start to see the difference once you went through editing, grading and then broadcast, this is where the codecs tend to get really stressed.
I've heard no end of people say that when they plug their HDV cameras straight into their HD TVs and play back their footage it looks better than anything they've seen broadcast - and yet when it is broadcast it falls to pieces.
Steve

[Ronan Fournier]

Iv'e made a test with a shot captured from an HDV camcorder, Canon XL-H1 (4.2.0 MPEG2 at 25Mb/s) and the the Nano through the HD-SDI output of the camera (4.2.2 MPEG2 at 100Mb/s).
Here is a JPG comparison : http://img11.imageshack.us/img11/1028/25mbvs100mb.jpg

Because of the JPG compression, the differences are less visible than on the .tif version. Anyway, for my eyes it appears that the 100Mb picture is more detailled (dog's nose fur), the hightly lighten areas are more subtles (snow), the outlines are more precises (leaf). It's not obvious onthis test but it is true that colors are more intense thanks to the 4.2.2.
Of course all this is subtle but it gives to the picture a better feeling of relief and clarity.

[Rafael Amador]

Zoom and compare.
rafael

[Dan Keaton]

Dear Cees,

There are cases where HDV can be very good, but it will always be 4:2:0. It will have less color information than 4:2:2.

Wikipedia has a good description of the diferences listed under Chroma Subsampling:

Chroma subsampling - Wikipedia, the free encyclopedia


Then there are cases where there is lots of detail and the differences are obvious.

Our higher bit-rate modes are insurance against having too much detail or too much motion for the lower-bit rate codec's such as HDV or 35 Mbps 4:2:0.

The differences will be obvious wjen you do greenscreen work. The nanoFlash makes this very easy.

As Steve Phillips and Alister Chapman have said, the higher bit-rate 4:2:2 footage produced by the nanoFlash holds up much better in post.

[Piotr Wozniacki]

With lots of detail and random movement, using nanoFlash has enabled higher detail settings on my EX1 without that electronic look, or even some macroblocking.

The latter is common to even the highest quality HDV codecs, but I managed several times to stress the great XDCAM EX 35 Mbps codec so much that it appeared on my SxS cards - but NOT in the nanoFlash clips of the same scenery, recorded at 100 Mbps.

Frankly, should nanoFlash be available when I still used my old, good V1 HDV camera, I guess I'd have gotten an even higher quality boost than I'm getting now using it with my EX1...

[Thane Silliker]

What are the Long GOP and I-Frame modes best suited for? What are their differences?

Edit:
OK, never mind this question. Answered it myself within the pages of my copy of Steven Ascher's excellent book, The Filmmaker's Handbook.

"I-Frames in terms of the Long GOP MPEG-2 compression is the first or reference frame that it fully recorded.

Some forms of compression, such as Sony's 50Mbps MPEG IMX (a variant of MPEG-2) and Panasonic's AVC Intra (a form of MPEG-4) are a one frame GOP, or I-frame only. I-frame only will have fewer artifacts due to the lack of predicted frames, and easier editing due to the fact that the editor's CPU no longer has to recreate the predicted frames.

H.264, part of the MPEG-4 standard is considered to be twice as efficient as DV, taking up half the storage for the same quality of image, even in I-frame only mode"

Really good book so far.

[Mark Job]

Originally Posted by Dan Keaton

Dear Cees,

There are cases where HDV can be very good, but it will always be 4:2:0. It will have less color information than 4:2:2.

Wikipedia has a good description of the diferences listed under Chroma Subsampling:

Chroma subsampling - Wikipedia, the free encyclopedia

...Hi Dan: I see a real *Great improvement when you use an HDV camcorder like the XL H1 via its HD-SDI output into a nano (XDR in my case). Not only do you get a 4:2:0 to 4:2:2 color improvement, but you also win big time by being able to record from a 1920 x 1080 Full raster HD signal, which is also in fact, uncompressed @ 1.43 Gbps ! This is where otherwise HDV challenged camcorders get a *Huge quality boost. I have conducted tests with my H1 at sample rates as low as 50 Mbps Long GOP, and I can plainly see a significant difference in both fine detail as well as color depth, which holds up remarkably well in multi-generation editing.

EDIT: In theory, the XL H1 camera should be a much harder camera to test the image quality on, since it already produces remarkable quality HDV images via its own internal encoder and tape transport to begin with.

[Cris Daniels]

I've always tested the Nano recordings directly against my SXS 35mbps. Since I record to both the Nano and the SXS at the same time this is simple enough to do.

Even if there is not a huge amount of motion in the shot, you can usually see a difference if you check the individual frames. I just sync up the shots in the FCP timeline and click the eyeball on the top video track to toggle it on or off. Even on shots that dont really let the Nano shine, I do see a difference in things like fine detail (hairs, grass, etc...). Granted the difference might not always be huge, but I like to think that 5-10% here and there can really add up in the end when you extrapolate that out through the entire shooting/editing/rendering process. Sometimes the Nano is clearly superior, sometimes less obvious, that is what I am trying to say I guess.

If I have a shoot that I really want nailed, I shoot at least 140mbps LongGOP. I've actually never tried the I Frame recording with the Nano so I cant really comment on that.

If it is more run and gun or something like wildlife where I am not certain exactly when something great will happen, I will shoot at 100mbps LongGOP to get a longer recording time.

[Paul Cronin]

Cris I have shot 6TB now with both 4.2.0 35Mbps on the camera and everything from 50Mbps to 280Mbps I-frame at 422 and I think the color spacing is a huge advantage with the Nano over the 4.2.0. This is most of what I see. As for 50Mbps to 100Mbps I do not see the difference even after grading. So I would save your hard drive space and use 50Mbps. I know I will be jumped all over since everyone says that 100Mbps is the SWEET spot and that is fine. But with 6TB of footage I just don't see it.

[Piotr Wozniacki]

I agree with Paul - 50 Mbps 4:2:2 gives me not only better picture quality (which keeps well enough with the kind of post I usually do), but also makes my material acceptable by any HDTV broadcasters here.

[Cris Daniels]

Well dont get me wrong I would never jump on anyone if that is what you find works best. I'm just another end user. Anyway, I don't always see a big difference in the color sampling, some scenes are remarkably well captured by the Sony EX Codec. Overall though, who could argue with the advantages of shooting 4:2:2.

My shooting bitrate is almost always determined by a variety of factors, but I'm with you on the huge storage needs. Thankfully those G-Tech drives are not very expensive....

I am not even talking about after the fact, when I transcode all my files to Cineform, and the file sizes increase again!, it all gets pretty outrageous.

Unfortunately the cameras that could most use the Nano, wont yet work correctly with it. This isn't Convergent Designs fault, it is about the DSLR's like the Canon 5Dmk2 that dont output a recordable raw HDMI feed. Talk about a camera that could use the help of the Nano. I find the DSLR codecs easy to break, and many times can corrupt the footage in ways you cannot correct.

As far as meeting broadcast standards, I think it is fabulous that this device allows cameras that have more limited on-board recording capabilites the option to be used in higher quality productions.

[Paul Cronin]

Cris you are right some times it is hard to see the 420 vs 422. But most of the time I see the difference and it makes the Nano worth it just for 422. I did do a job last week with no Nano. EX1 and PMW-350 both at 35Mbps. The client looked at the footage yesterday on a 40" Sony Bravia and loved it. They had me play it over and over and loved it. So there are times for me when native codec it fine. But this was a very controlled set with nice lights and minimum motion.

DSLR would improve with Long GOP 50Mbps 422. But I don't think the camera companies will let CD make the dollars when they can do it in house. It will come but will it be for everyone, I don't think so. Choices are great and give us our creative freedom.

I have never used Cineform since I shoot the same coded on all of my cameras. But it looks like a great product.