AF100+Samurai improved results (w dnxhd codec)

[Preston Davis]

I wonder if anyone else has found these results using the AF100+Atomos Samurai+dnxhd codec:

Much improved highlight detail

More low light detail with smoother noise

My comparison between the AVCHD and Samurai footage was made with ungraded footage dropped on Pr Cs5.5 timeline.

Oddly, the proresHQ codec did not yield as obvious results.

Could this mean the sensor is less the issue with the AF100 than the AVCHD processing?

I wonder if the 100a 10-bit will get even better results with the Samurai....

[David Heath]

Originally Posted by Preston Davis

Could this mean the sensor is less the issue with the AF100 than the AVCHD processing?

I've seen test results via the live output with issues very similar to results when recorded internally.

Bypassing the AVC-HD codec for something much less compressed is only likely to bring benefits - but don't expect miracles. I'm fully certain the sensor is a far bigger issue than the coding.

And with a Samurai about a third the cost of an AF101 body, you're paying quite a lot to get a relatively small improvement.

[Preston Davis]

My tests do show a marked improvement with the dnxhd codec, highlight detail and noise smoothing.
Were the tests you saw using dnxhd or proresHQ.? I tried them all side by side and the dnxHD was markedly better! proresHQ yielded little / no improvement, as you noted.
My question is directed at someone who has compared the dnxHD codec (specifically) to the avchd and proresHQ.

[David Heath]

Originally Posted by Preston Davis

My tests do show a marked improvement with the dnxhd codec, highlight detail and noise smoothing.
Were the tests you saw using dnxhd or proresHQ.? I tried them all side by side and the dnxHD was markedly better!

What I saw were direct captures from the uncompressed output, so no codec involved. (So subsequent prores or any other codec coding could only be worse, even if slight, no codec can improve on the quality of an uncompressed output.) Yes, uncompressed or a less compressed codec will be better than the AVC-HD internal recording, but I didn't see any night and day difference - and I wouldn't expect to.

At the highest bitrates AVC-HD is pretty good, and it can be hard to spot flaws due to it in the first generation - issues due to it's use are more down to it not withstanding future post work or recompression as well as more robust codecs.

And any AVC-HD issues in were swamped by factors which could only be put down to the sensor. Codec choice is not going to affect such factors as basic sensitivity or resolution - both matters where the AF100 could be much better relative to comparable cameras.

Both DNxHD and ProRes should be virtually transparent - certainly to first generation - with sensible bitrates. That makes me wonder if the differences you saw may be down to some other factor, an NLE setting (or mis-setting) between the comparisons, maybe? Likewise factors such as highlight handling should be completely independent of codec.

[Preston Davis]

So you saw raw footage that had no codec? My settings are fine in cs5.5.

Maybe I have misunderstood; your opinion is that that is no benefit to using the Samurai?

I think I will trust my eyes on this one. Thanks for the input.

I am interested in hearing from someone who has done this test themselves, not just reviewed footage on line,

which can be skewed by youtube and vimeo encoding.

[David Heath]

Originally Posted by Preston Davis

So you saw raw footage that had no codec? My settings are fine in cs5.5.

Please excuse my suspicions, but have you seen this post about codec differences on the C100 forum? http://www.dvinfo.net/forum/digital-...ml#post1764688 Point being that very negative conclusions were being drawn about the C100 internal codec - but it turned out to be a field order issue in Premiere Pro. Problem wasn't with the C100, but the users testing methodology.

Maybe I have misunderstood; your opinion is that that is no benefit to using the Samurai?

No, that's not my opinion - just look back at the previous post: " Yes, uncompressed or a less compressed codec will be better than the AVC-HD internal recording, but I didn't see any night and day difference - and I wouldn't expect to." And the Samurai will clearly give a less compressed output than AVC-HD, so yes, I'm sure it will give *SOME* benefit.

But I'd expect to see that difference when trying to do post work - far less on first generation playback. And I wouldn't expect much difference between two high quality external codecs such as ProRes and DnxHD.

What I do disagree with was your original thought - "Could this mean the sensor is less the issue with the AF100 than the AVCHD processing?". What I saw (uncompressed) took all AVCHD or any other codec thoughts out of the equation. Yet the other issues remained - hence must be due to sensor and front end processing.

I am interested in hearing from someone who has done this test themselves, not just reviewed footage on line, which can be skewed by youtube and vimeo encoding.

All I can say is that what I saw was straight from the camera, uncompressed. So certainly no youtube, vimeo or any other encoding or alteration. So together with the subject, purely testing front end performance as scientifically as possible.

[Preston Davis]

Thanks for the clarification.