Does the Ninja 2 improve on AVCHD 108060p at 28Mbps?

[Mark Rosenzweig]

I borrowed a colleague's Ninja 2 to see if it produced video that was superior to what my camcorder could produce using its own processor. The camcorder has a clean HDMI signal and uses AVCHD compression, 1080 progressive 60fps, 8 bits per channel with 4:2:0 chroma sampling. I shot with the Ninja and the camcorder simultaneously, the Ninja getting its signal from the uncompressed 108060i put out by the camcorder and using Apple Pro Res 422 (145Mbps), which uses 10 bits per channel with 4:2:2 chroma subsampling.

To maximize quality I edited and converted the Apple Pro Res 422 60i clips to Avid DNxHD 60i (145Mbps), which also uses 10 bits per channel with 4:2:2 chroma subsampling, in Sony Vegas Pro. The AVCHD 108060p clips were edited and combined without recompression. The edited videos were then separately uploaded to Vimeo.


The camcorder-produced AVCHD 60p at 28Mbps version is here:

Atomos Ninja 2 Test Video Comparison 2: AVCHD 28 Mbps 60p from the camcorder on Vimeo

The Ninja2 Apple/Avid high-bitrate/high sampling version of the same scenes from the uncompressed output is here:

Atomos Ninja 2 Test Video Comparison 1: Apple Pro Res 422 (145Mbps) 60i Converted to Avid DNxHD 60i (145Mbps) on Vimeo

You can open two windows and run the two videos simultaneously. Or, if you are a member of Vimeo, you can download the originals. The original AVCHD file is 450MBs. The original AVID file is 2.5GBs. They are the same length - high bitrates have a cost.

The scenes have a lot of red flowers and a lot of detail so a challenge for any codec. Sorry about the shaky cam. I was holding the camcorder and Ninja with one hand, but that also challenges the codecs.

[Chris Duczynski]

You know, I was going to buy one for my Z5, but I'm not going to bother. There's been a few posts with comparisons and frankly I find it hard to justify the cost. From your vimeo test, I just can't pick the Ninja 2 as having any advantage.
What did you think when you saw the native footage side by side?

[Mark Watson]

From what I've read, if you do some color correcting or special effects with the video, then you will plainly see the 4:2:2 looks better.

Mark

[Mark Rosenzweig]

I agree with both comments: there appears to be little advantage straight on. But if you edit - color grade, etc. - and go through multiple compressions, starting with the essentially intermediate intraframe, high bitrate, better-color-sampled codec has an advantage in both ease of use and maintenance of quality. Supposedly AVCHD breaks down, particularly in color, when edited and transcoded.

Also, if there is a 50Mbps-minimum broadcast requirement, you can deliver that.

But AVCHD is an impressive codec, and 108060p has advantages over 108060i too (no recorder, none, can record 108060p from HDMI).

[Jeff Pulera]

I find this "test" to be flawed. Any camera can record a nice-looking compressed image when there is little or no motion, and perfect lighting. Shoot a fountain or waterfall or pan some trees full of leaves once and let's see how the footage looks, or shoot a dark stage event. Push the codec a bit! The camera used is at the higher end of the compressed world, shooting 28Mbps, while most AVCHD cams are maxing out at 24Mbps. And then there is HDV at 25Mbps, but using a much less efficient MPEG-2 scheme that really falls apart with high motion detailed scenes.

I have used the Ninja with Sony HDV cameras for shooting stage events (musicals and dance recitals). At first glance, you won't necessarily see a difference between the native HDV clip and the Ninja's ProRes clip. I edit with Premiere, and if I change the magnification in the Program monitor to blow up the picture, and then compare the two clips, major differences become apparent!! I will put one clip above the other on the timeline, synced up, then toggle the top clip on/off to compare, or add CROP filter and then "wipe" the crop back and forth and the differences then become quite evident. Watching one video, then watching another video, of nice clean footage, you'll never "see" a difference.

On one stage musical I shot with HDV, I compared the HDV and ProRes clips as described above. On a medium shot, the performers faces on HDV were little more than skin-colored blobs. Toggling over to the ProRes clip, eyes and mouths suddenly appeared as well as other facial details!

A large mural painted on the back wall of the stage looked like large blocks of just a few different shades in HDV, while in ProRes, every individual pixel had different shades of color and of course DETAIL. And so forth, zooming in on and comparing different areas of the scene.

If I had shot close-ups of a few flowers on a sunny day with that same HDV camera, then no, I likely would not see such major differences with Ninja, although doing a color pass filter on the flowers would most definitely yield better results with 4:2:2 intra-frame versus HDV. Guys, this is like comparing a Chevette to a Corvette when driving around the grocery store parking lot and then saying they both handle and accelerate the same. Get both cars on out on the open road and it's an entirely different story.

Keep in mind that H.264 is about twice as efficient as MPEG-2, so at the same given data rate, an AVCHD camera will produce a cleaner image than HDV (optics aside), and the benefits of recording ProRes with the Ninja may not be as apparent. But if you have an HDV camera, Ninja can breathe new life into your recorded images for sure.

Oh, one more note - why was the ProRes converted to the Avid codec??? The Avid codec is very good and similar to ProRes, but any conversion adds some quality loss, so how is that fair or wise when comparing HDV to Ninja, if a third codec is introduced?

[Mark Rosenzweig]

High bit rates and 4:2:2 sampling should provide sharper details and more accurate color from scenes with a lot of detail - grass blades, fence lines, flower details too. I agree that low light and fast action also creates stress. But, 60 progressive is likely going to beat 60i for action due to interlacing.

There are constant claims of high bitrate hacks for the GH2 really upping quality, with videos of similar scenes with lots of detail (but no benchmark comparisons!).

With respect to:

"why was the ProRes converted to the Avid codec??? The Avid codec is very good and similar to ProRes, but any conversion adds some quality loss, so how is that fair or wise when comparing HDV to Ninja, if a third codec is introduced?"

It's not completely fair, but it is wise: I know of no way to trim and combine Pro Res clips without recompression. I do not have Pro Res encoding capability so I merged to Avid, which has the same properties. Supposedly an advantage of Pro Res is there is very little loss from this process. I would have shot Avid if that was available on the Ninja (it will be). I would have preferred to not have to recompress, I compared the Avid versions to the original Pro Res and I do not see much difference.

I guess I could have recompressed the AVCHD files, and that may have worsened the quality for those clips. But I wanted to compare the best video, not just clips, from both that I could get.

I can do a low light test easily enough. But I do not see why the higher bitrate or greater color sampling helps in low light anymore than bright light. Sensor is going to matter much more. But I am willing to try.

I agree that the Ninja should easily beat HDV, but for the price of a Ninja and ssd disks plus a little more you can just get an AVCHD camcorder. But, the Ninja does produce clips that can withstand editing better than AVCHD so maybe a fine HDV camera + Ninja is a better way.

[Chris Duczynski]

Jeff, can you clarify please. I'm shooting on a Z5 HDV camera to a mrc1 external recorder which records AVCHD files. Do you see any significant difference there?

[Jeff Pulera]

Hi Chris,

The Sony HVR-MRC1 Memory Recording Unit does NOT record AVCHD, important to note. Rather, it records exactly the SAME data that your HDV camera is recording on tape. More convenient workflow perhaps, but exact same quality = exact same file as on tape. No benefit in that area. For HDV, this means recording 1440x1080 Long-GOP MPEG-2 with 4:2:0 color at 25Mbps, versus Ninja recording ProRes at 1920x1080 with 4:2:2 color and data rates of 100, 150, or 220Mbps. No comparison. Ninja also has nice color screen that is much larger than most camera LCDs, so useful as a portable monitor/playback device as well (Ninja 2 has HDMI out !!)

For Mark,

I guess I get spoiled working with Adobe, it would edit the ProRes files natively, so no transcoding step before export to Vimeo, so I understand now why you converted to Avid.

Regarding "Ninja and ssd disks plus a little more you can just get an AVCHD camcorder". Well, yes and no. You can get a "HandyCam" lacking much manual control for that kind of money. But, if someone already owns a Z5 or Z7 for instance that they paid $4-5k for, and that is a higher end camera with a nice lens and XLR and all that goodness, Ninja will allow the user to extend the life of that existing investment, and record a better image than ever before, bypassing the HDV codec altogether.

Also note that while the more expensive SSD disks might be needed for recording on vehicles or other situations with lots of bumps and vibrations, very inexpensive laptop drives can be used reliably with Ninja/Ninja 2 for tripod or general hand-held use. I'm talking a 500GB drive for under $100 that provides 5, 7 or 11 hours of uninterrupted recording depending on ProRes quality setting. That's yet another benefit, being able to record long events without the interruption of a tape change. I have shot lengthy stage events close to 3 hours, and while I did run HDV tape for backup, during tape changes, the LIVE HDMI out of camera was continuously recorded by Ninja, no glitch or pause.

So the bottom line is, if you have big investment in an HDV camera and you like the camera, but would like to upgrade the image quality and move to a tapeless workflow without buying another expensive new camera, there ya go.

Thanks

Jeff Pulera

[Mark Rosenzweig]

I agree about the advantages you state.

And I have used a regular spinning disk with the Ninja 2 in the field (Hitachi Travelstar) with no problems, though I worry a little.

While I don't think that low light shots, as suggested above, would show anything vis a vis the comparison (sharper noise?), I may try color grading the two videos already shot and show how they do as second-generations. That is supposed to be another advantage of the 4:2:2. 10-bit codecs (for good reason). In this test I would color grade the original Pro Res files and save to both AVCHD and AVID high-bitrate and color-grade the original AVCHD files and save to AVCHD. Then all are second-generation.

[Mark Rosenzweig]

I took the original AVCHD 108060p camcorder video clips and the original Apple Pro Res 422 Ninja 2 video clips, color graded them using the "Warm" preset from Magic Bullet Looks and converted both to AVCHD 108060i at 22Mbps.

The question is whether the recompression and color grading affects the quality differentially across the two differently compressed videos.

Here is the color-graded video made from the original Apple Res 422 files:

Atomos Ninja 2 Test Video Comparison 3: Apple Pro Res 422 *Color Graded* and Converted to AVCHD 60i on Vimeo

Here is the color-graded video made from the original 108060p files:

Atomos Ninja 2 Test Video Comparison 4: Camcorder AVCHD 108060p *Color Graded* and Converted to AVCHD 60i on Vimeo

The uploaded originals (before Vimeo conversion) of both files can be downloaded.

[Chris Duczynski]

Thanks Jeff,
I just went back over an old Z5 thread and noticed you told me the same thing about 6 months ago for the atomos 1. After viewing marks graded footage (thanks Mark) I'm going for the Ninja 2. It seems a good and economical way to upgrade to broadcast quality full HD.
I'm still using CS5 which hopefully will handle pro-res easily.

[Eric Olson]

Originally Posted by Mark Rosenzweig

I took the original AVCHD 108060p camcorder video clips and the original Apple Pro Res 422 Ninja 2 video clips, color graded them using the "Warm" preset from Magic Bullet Looks and converted both to AVCHD 108060i at 22Mbps.

Your results seem completely reasonable. Both interlaced 4:2:2 and progressive 4:2:0 have chroma planes with the same resolution of 960x540. What camera are you shooting with?

[Jeff Pulera]

Hi Chris,

CS5 on the PC will play the ProRes files, with one catch - there is an Adobe issue that causes it to reject ProRes clips with 4-channel audio, says "Unsupported Format" when you try to import. Just make sure you are recording only 2-channel stereo with Ninja and the ProRes clips play fine in that case.

Note this was later corrected for CS5.5 with a patch update, so the 4-channel clips work with CS5.5.2 and CS6 now.

Thanks

Jeff Pulera
Safe Harbor Computers

Originally Posted by Chris Duczynski

Thanks Jeff,
I just went back over an old Z5 thread and noticed you told me the same thing about 6 months ago for the atomos 1. After viewing marks graded footage (thanks Mark) I'm going for the Ninja 2. It seems a good and economical way to upgrade to broadcast quality full HD.
I'm still using CS5 which hopefully will handle pro-res easily.

[Chris Duczynski]

Thx mate, will keep an eye on it.
I'm using a program called nanocosmos which seems to sort out pro-res really nicely in CS5. But as soon as I right click on one of the 4 audio channels to try and raise the level or eq it - CS5 just stops. Very annoying.
Cheers

[Chris Duczynski]

Sory, don't want to hijack the thread but Jeff, why do my file properties from the MRC1 say AVCHD (m2t) ?? Thats where I was confused.