This was shot yesterday while testing a lot of settings etc, We shot with native and Nija using PrRes422 in the C100.
And we also shot side by side with the FS700 internally and with the NanoFlash.
I will upload more tests as time allows.
This was unlit with mixed lighting from a North facing window and incandescent light inside. We tested studio lighting as well more on tat later. I used a setting of 6000K set from a slightly warm card.

This would be a typical very quick setup. No grading. Wide dynamic range setting.

C100 native & Nija on Vimeo

Thanks for posting. Will view in detail later, but have a question: What do you set the C100 to (24p, 24pf, etc) and what do you set Ninja to? I am renting both right now and can't get the Ninja to record 24p

Thanks.

Steve I would shoot 24p unless you have clipwrap and want to convert the footage.
Thanks for posting. Will view in detail later, but have a question: What do you set the C100 to (24p, 24pf, etc) and what do you set Ninja to? I am renting both right now and can't get the Ninja to record 24p

Thanks.

Steve I would shoot 24p unless you have clipwrap and want to convert the footage.

I'd like to, but the Ninja won't seem to recognize 24p. Whenever I set camera to 24p the Ninja greys out the 24p option.

So, you guys are really telling me you see a difference there? The shots look good but, I can see no difference at all. I downloaded the raw and still don't see it.

The difference would mostly come out in some added grading flexibility I'd say. Also if shooting with more camera movement and/or moving subjects.
(under exposed and very red material too)

So, you guys are really telling me you see a difference there?

Check out the gold stripe in the interviewee's black skirt. Also her hair in the side shots. I think the transition between 0:14 and 0:16 is the biggest give-away. However, there's an evil part of me that thinks that the captions are transposed at the time of posting. The aliasing reported on 4:2:2 Ninja resembles the aliasing I've seen reported as native AVCHD on a couple of other clips.

We are, basically, debating the number of angels that could dance on the end of a pin, and we're all falling into the trap of gathering around 'fresh meat' and poking it just like we did with the EX1, the AF101, the 5dM2/DSLR and the FS100 to name just the few I've shamelessly jumped in on.

But with that said, I see 'better oranges' on the wall next to the picture of Olaf's partner and cat, I see less aliasing in the skirt and hair in all the shots labelled 'AVCHD' which I think is actually the Ninja.

......However, there's an evil part of me that thinks that the captions are transposed at the time of posting. The aliasing reported on 4:2:2 Ninja resembles the aliasing I've seen reported as native AVCHD on a couple of other clips.


Matt I double checked and I have not mixed them up. Check the screen shot below:

Olof - thank you so much for that! Curioser and curiouser...

I hope to get some quality time with a C100 tomorrow and will just have to shoot for myself - with the internal and my Pix - as I'm concerned and confused. Do you see what I'm seeing on your master files?

What's more, good to see FCPX in action... Is the project set to a PSF/Interlaced setting (e.g. 1080i) so that the interlaced stuff looks fine but the progressive stuff is 'deinterlaced'? I'm clutching at straws to work out the difference between shots.

Hopefully this is just an issue caused by a combination of NLEs' interpretation and Canon's curious choice of PSF over a pure progressive AVCHD format, and that editors must know how to ingest and interpret it correctly. Memories of the XLH2...

BTW (for lurkers) - also good to see some 109% whites in there. In 'proper' edits, Olof, I guess you'd pull your whites down to 100%? On export, FCPX truncates everything to 100% white (a bit like PPro IIRC), so there's detail out of the windows that we're not seeing yet. I whitter on only because it was one of the things that people threw at early FS100 footage - highlights in the >100 range got truncated and people thought that the resultant video had very 'videoish' highlights...

Matt, I was showing you the raw files no grading, as I have on my other demos as well, except where I not if it has been graded.

The FCPX project is a 30P, I have had quite a time interpreting the C100 nonstandard PSF, they call it PF30, and FCPX has a difficult time of it.

I always manually set my project with Custom Settings in this case 30P. Some PSF footage is interpreted correctly and some is not, it is not consistent. Sometimes I have to override the field order settings to get it to come out right.

The internal footage is easy to deal with, just rewrap in ClipWrap and you will have no issues. I understand that the Ninja 2 latest firmware can do the pulldown correctly however I was using the Ninja 1 it sees it as 60i.

It is not unusual for an external recorder to show more noise, internal codecs tend to smooth things out. The EXcams are this way for example, but there is more detail in the Nano/Ninja footage and you can grade it much more w/o it falling apart or showing banding.

I think most of the time there is no reason to record externally. One great thing about the C100 is it can record to both cards at the same time. I have often used external recorders just to get redundancy in single cam shoots.

The bottom line for me is the C100 is a keeper and I will be keeping mine. I am now building my kit for it.

So, you guys are really telling me you see a difference there? The shots look good but, I can see no difference at all. I downloaded the raw and still don't see it.
The issue with codecs is that nowadays you are very unlikely to see much difference between uncompressed and most camera codecs with straightforward viewing of footage - even AVC-HD is pretty good, let alone better ones. (That wasn't true of the first generation of AVC-HD cameras where first generation coding problems were very visible.)

But that's not to say the differences aren't important. But they only really start to show to the eye once footage gets manipulated or decompressed/recompressed. Equally, one codec may show flaws with a lot of motion, a second may handle that well, but be poor with lots of fine detail. When somebody like the EBU do codec trials, they do them with specific types of material to stress the different factors.

And that's why codec comparisons are virtually impossible to do sensibly outside of very controlled conditions. At the very least, you need to do an image subtraction process in something like Photoshop - take framegrabs with the different codecs of the same frame, then subtract one from the other, and add a uniform mid gray. You'd expect the result to be mid gray if the two frames are identical. If you see anything, it must be due to differences between the frames, so you have a way of quantifying the differences. (I'd typically expect to see a faint outline around detail.)

It's further complicated by the point Olaf makes - that there may be differences other than purely down to compression codec, and noise reduction may be a good example of such. Electronic noise reduction can certainly make images look superficially better to the eye - but at the risk of problems further down the chain. That's why simple "to the eye" viewing of raw footage can be as misleading as helpful.

Also worth mentioning that another reason for using differing codecs is less for sheer quality, more to make life easier for the computer. Such as AVC based codecs can have good quality/bitrate efficiencies - but at the cost of needing much more computing power. It's likely to be more satisfactory to record in the codec you intend to edit with from the first, than have to do a transcode.

@Olof,
Are you seeing any difference in dynamic range between the internal AVC-HD and the Ninja?

Spent a day shooting tests with a C100 (internal and Pix220), just going over the rushes in FCPX - all the aliasing I've been seeing is simply 'Malignant PSF' - a Canon speciality it seems (note this is part *ten*):

Pro Video Coalition: TecnoTur by Allan Tépper (http://provideocoalition.com/atepper/story/psfs_missing_workflow_part_10_fcp_x)

Great article.

So this explains a lot of the Ninja vs C100 shoots I've seen. One is seen as progressive, the other as interlaced, and whichever format the timeline's set to... wins.

NB: Should have mentioned - these from INTERNAL codec. I'm still pixel peeping today's footage - I bet there's little or no visible difference between them unless under duress of extreme motion (a bit like FS100).

Excellent stuff Matt, and well researched. A good example of why when you get a result that is unexpected, it's a good idea to really question whether there may be a problem with methodology of testing. This (NLE not dealing with psf footage correctly) is an excellent example.

Secondly, a good demonstration of why a single test with a chart can reveal so much more than hours of "real" footage. The asymmetry shows up the likely cause of the one test being inferior to the other, and is a very strong clue that it's nothing to do with codecs.

And finally, a very nice clean result from the C100 on resolution circles, at least up to 800lpph. From memory, much better than the FS700, which is similar vertically, but worse horizontally. Yes, only one aspect of performance being tested, but encouraging. Since the FS700 and the C100 share the same fundamental dimensions of sensor, it's interesting to speculate why they do differ.

Managed to have a pixel-peep at tests on C100 internal and external recorders this afternoon, and relieved to say that on first pass the situation looks very similar to FS100: there is little or no visible difference on rushes from AVCHD or ProRes 4:2:2 HQ, though there are major differences seen in the vectorscope and waveform monitor which will only bite you on the bum after 4-5 generations.

In other words, Broadcasters are absolutely right to demand 4:2:2 high bitrate origination, but corporates and other material destined to be 1-2 generations away from the edit master will be fine with internal recording. But if you are shooting for something that may be sold for broadcast, use an external recorder.

Drum Roll Please...

The first image is a frame from the internal recorder in FCPX, zoomed in 1:1 with the vectorscope beside it. I am panning over a busy scene with texture and a wide dynamic range. I'm looking for signs of quantisation, reduced detail. It's all okay for now BUT LOOK AT THE VECTORSCOPE! It's like looking through a sieve! That, dudes, is the curse of 4:2:0. You won't see this now, but 6-9 generations later, it will get ugly, and every generation after that makes it fuglier.

The second image is matched from the PIX220. Not a lot of difference to the naked eye at this stage. But the vectorscope shows that the Pix version is a cornucopia full of data - the WFM is similarly endowed.

So, just like the FS100, for event, R&G, low/mid corporate and vimeo/festival indy film, you're fine with internal recording. For anything that's destined to be sold to broadcasters, signed up for distribution, syndicated or otherwise buggered around with, use an external recorder to PRESERVE what you see from your original rushes, not necessarily to improve them.

Very interesting and useful. Thanks for the heads up Matt!

Update: just tried the 'trees in wind' torture test on C100.

For an inter-frame (e.g. Long-GoP) codec, this is horrible: unpredictable motion, every part of the screen contains motion, lots of edges. We'd expect any Long-GoP codec to fall apart here.

However, my first pass at this test is that the Canon's AVCHD isn't quite as robust as the FS100.

Doing simple pan tests is one thing - and in most cases this is the sort of pixel torture we experience with cameras - pixels become blurred, motion is constant and going in a predictable vector. So, the sort of run-of-the-mill motion we get is handled well by the internal codec of both the FS100 and the C100.

However, the 'trees' test is awful. Unpredictable vectors, lots of edges, etc - nightmare.

The FS100 does do a fair job of this - if you pixel peep, you're going to see some mess, but it's fairly innocuous - the eye is very forgiving of it when you play back in real time.

However - and these are preliminary results and just FYI - I'm seeing big huge gobs of nasty artifacts which include partially interlaced mouse-teeth on the C100 footage. Okay, so an ordinary, sane, well balanced human would have a hard time seeing it, but I can - and obsess about such things.

Seriously, the difference between this and the FS100 is enough for me to use an external recorder for anything that features dust, smoke, water, extreme action, big crowds and detailed external wides. Will so a side by side when I'm done with the next assignment.

So, you guys are really telling me you see a difference there? The shots look good but, I can see no difference at all. I downloaded the raw and still don't see it.

No different at all.

I rented the Canon XF300, which has the broadcast 50Mbps and 4:2:2, and the C100 with internal AVCHD and compared them. So pixelwise things may not be 100%, but boy is the image way better. So what's the argument all about? Image character or quality?

Paul, just finding this thread. It wasn't clear from your post that you thought the xf300 was better or not from the C100 with the external. I would assume the xf300 was better. Thoughts?

The ladies' sweater is more red with the Ninja. This is also compressed. I'm drinking a good Merlot but I'm sure the sweater is more red. With the Ninja she looks more intelligent, less threatening, more of the marrying kind but with the internal... she may be a husband killer. I would watch out. I wouldn't marry this women quite yet. Perhaps another camera might help explain this mysterious aura better-

I have done some tests with the C100 and the Ninja 2 myself. In most shots I see a small improvement on the Ninja footage, even before grading.

As an example see the attached crops (straight from camera). Somewhat underexposed because the rest of the frame was a bright window.
The C100 seems to do some noise reduction here on the AVCHD, probably because 24 mbit/s can't handle the noise.