Canon C100 vs. Sony FS700

[Alister Chapman]

Pixel size is the primary factor that determines sensor sensitivity. If you take a sample of pixels sizes from the cameras currently on sale today you will see a direct correlation between pixel size and sensitivity across the board. It's no surprise that most 2/3" HD cameras are more sensitive than a 1/2" EX1 which is more sensitive than a 1/3" XF305, or that an old 2/3" SD camera will be more sensitive than any of those. Nor is it a surprise that the Sony F3, FS100 and Arri Alexa are almost identical and very slightly more sensitive than the C300 and FS700.

Canon has not taken a march over anyone else in terms of sensor technology. There is absolutely nothing special about the tech in Canons newer sensors. Yes they are reading out the sensor a little differently, but as far as I know the sensor tech itself is really no different to anyone else. Camera sensors (other than back illuminated sensors which only have an advantage in very small sizes) have not changed in terms of sensitivity for quite some time. What has happened in the video world is that sensors and thus pixels have become larger and as a result low light performance has improved. In both the video and stills world noise reduction has also improved. Until new materials that are more efficient at converting photons to electrons are discovered, photo site size is still the key thing that determines sensitivity. Modern pro level video sensors use micro lenses to overcome gaps between photo sites and achieve extremely high efficiencies as a result, typical QE's these days are around 70. If you look back 10 years you'll see that QE has not really changed much, just the use of high QE sensors has become more common place.

Gain is gain, there are differences between applying it in camera and in post, mainly to do with whether you add the gain before compressing and any compression artefacts or post compression and then use the significantly greater processing capabilities of a workstation to reduce artefacts and add noise reduction etc. in my opinion the best way to add lots of gain is to first use the very best codec you can, or uncompressed recording and add a small amount of gain in camera, then add the bulk in post where you have much better control. Sony, Canon, however could bring out cameras with even higher ISO's, it's a great marketing tactic, "buy our new camera because it's 100,000 ISO" and many people start getting excited by the idea of a mega sensitive camera, but the reality is that super high ISO settings are more noise than image. If you really could get sensible images simply by adding more and more gain then the military wouldn't have to use image intensifiers.

[Jim Martin]

The C300/500/100 chip, using the Bayer pattern, is the 1st chip not being de-bayered. Me thinks that is contributing the the overall performance of the chip.....and Alister, we don't get weather here too much so visiting weather is great!

Jim Martin
Filmtools.com

[Paul Cronin]

Ah yes Jim lives in the land of sunshine Alister. We live with the changing seasons.

As for the question at hand. I tired the FS700 twice and liked the camera but for my shooting the C300 is a better camera and better picture right from the cards. Hard to beat 422 50Mb/s with the latitude of this chip. And the glass options are very nice. Yes you can use an adapter on the FS700 and use the same glass but I think the C300 wins. Now if you really need over cranking well the FS700 is worth a go.

[Brett Sherman]

It wll be interesting to see if the AVCHD codec of the C100 negatively affects the low light performance - making the "grain" blocky and unusable. Until someone actually gets their hands on one and tests high ISOs, I'm not sure how they will compare. I'm hoping AVCHD retains the detail. If it works well, I might be looking at getting a C100 as a B-camera.

[Matt Sturns]

Thanks for all the info. Much appreciated. There are so many new cameras coming out...exciting times.. Thanks everyone for the info especially Alister, your wealth of information is invaluable. The fact that you share your wealth of knowledge, and without being arrogant or snooty, is frankly amazing. Sooooo much appreciated. I cannot thank you enough for your expertise!!!

[Alister Chapman]

Modern high end AVCHD encoders are very good. I would fully expect the AVCHD from the C100 to be very good. As the camera won't be noisy at lower ISO's noise should not be a big issue unless your really pushing the footage very hard in post. Same with the FS100/FS700 etc.

There is a push from some broadcasters now for 35Mb/s AVCHD for broadcast use as this would be comparable or better than 50Mb/s mpeg2. 24Mb/s AVCHD is now permitted in Europe for some broadcast news gathering applications.

[Al Bergstein]

I don't know. I tend to agree with A.C. about the two likely looking very similar.But having 'left' AVCHD behind in my HMC-150 for Canon's MXF 4:2:2 world i can say that i do not relish returning to it. Latitude in the 'real world' of mxf seems much better than AVCHD. Green screen seems easier. Grain feels more 'organic' and useable at higher ISOs.. OTOH my buddy's GH2 takes fabulous video, and i can match it pretty well in post. If the C100 was mxf, I'd already have one on order. I can't afford $15k for a camera, at this point, and frankly for web video don't think it's worth it. (i just watched PBloom's 'confluence' which is a great ad for staying with a 7D & 5d!). Then add the change of cards to SD, another cost I'd rather avoid, given my 7D, ( and the fact that i have found CF cards more reliable) it's just very disappointing. I am what seems a perfect candidate for the target market one would assume Canon is aiming for with the c100, i own a 7d, xf 105 and 305, love em, but if I am going to be forced into SD cards and AVCHD, I will take a very hard look at returning to Panny for either their GH3 or a Af100, used, and/or selling off my Canon gear. The stories i've read of Sony build quality on the 700 and 100 make me reticent to seriously consider them, as i shoot in bad weather & harsh conditions frequently. I also assume it will take more post work to match a C100 or Sony AVCHD footage withmy Mxf footage. So pure tech arguements are interesting but ultimately secondary. I will be very curious to see the real shipping C100 footage. I feel that canon is making a big marketing bet here, and could lose.

[David Heath]

Originally Posted by Alister Chapman

There is a push from some broadcasters now for 35Mb/s AVCHD for broadcast use as this would be comparable or better than 50Mb/s mpeg2. 24Mb/s AVCHD is now permitted in Europe for some broadcast news gathering applications.

To compare with 50Mbs MPEG2 you'd probably need more than 35Mbs - the higher the bitrate, the less point there is to H264/AVCHD. The compression advantages are less pronounced at higher bitrates.

And it begs the question "WHY"? Granted the file sizes may be a bit smaller, but we're probably only talking 20% lower or less for comparable quality. Is it really worth it - given that AVCHD needs more processer power for comparable performance.

If the lower bitrate meant much cheaper memory it may be worth it, but even that's not true now - the Canon records 50Mbs to Compact Flash cards.

[Alister Chapman]

Some of the desire for 35Mb/s AVCHD comes down to support from 3rd parties. AVCHD is almost universally supported these days, mpeg 2 in it's various mts ,m2t, mxc, mp4 wrappers is not as widely support in lower cost systems. I've seen some demos of 35Mb/s AVCHD and it's extremely good. Would also be possible to make it 10bit and 422 with only some small encoder changes.

When it comes to making cameras match, the codec used is the very last of your worries. Camera matching tends to be a function of the manufactures colorimetry, for example Panasonic look different to Canon who are different to Sony. The codec itself does not affect color, dynamic range or latitude, again this is a function of the camera front end. This is easy enough to demonstrate. Take the HDMI out from a Sony FS100 (AVCHD) and record it using a NanoFlash as an Mpeg2 MXF at 100Mb/s and you will be very hard pushed to see any difference. Obviously there is a difference, but this difference will only really become apparent when you start to push and grade the material in post where the 100Mb/s material will not degrade as quickly as the 24Mb/s AVCHD.

[Al Bergstein]

Ah thanks. That makes sense. That bit had slipped my mind. I do remember that I seemed to have a lot less latitude to play with in AVCHD, but that, as you say, is likely to be more an issue of the back end than the front. I suppose if I went directly to HDMI out capture, I would see the latitude expand, even in the lower end products. I certainly do when I go out of my little Panny TM900 and directly into a high end monitor.

A key issue will be what the Canon Log looks like and how much it helps. Given the sensor and the log, it's likely to be quite good. That's my bet for now.

[Glen Vandermolen]

35mbps AVCHD? I thought AVCHD's maximum bit rate was 28mbps.

[David Heath]

Originally Posted by Glen Vandermolen

35mbps AVCHD? I thought AVCHD's maximum bit rate was 28mbps.

Quite true - what Alister is referring to is a suggested extension to the specification. It doesn't exist as a standard yet.

[Alister Chapman]

Originally Posted by Al Bergstein

I do remember that I seemed to have a lot less latitude to play with in AVCHD, but that, as you say, is likely to be more an issue of the back end than the front. I suppose if I went directly to HDMI out capture, I would see the latitude expand, even in the lower end products. I certainly do when I go out of my little Panny TM900 and directly into a high end monitor.

The codec will not change the latitude of a camera, that's down to the front end. A poor codec will introduce image artefacts and perhaps reduce the resolution and soften the image, but latitude should not change. 24Mb/s AVCHD is normally pretty good these days.

[Tom Roper]

Originally Posted by David Heath

To compare with 50Mbs MPEG2 you'd probably need more than 35Mbs - the higher the bitrate, the less point there is to H264/AVCHD. The compression advantages are less pronounced at higher bitrates.

And it begs the question "WHY"? Granted the file sizes may be a bit smaller, but we're probably only talking 20% lower or less for comparable quality. Is it really worth it - given that AVCHD needs more processer power for comparable performance.

If the lower bitrate meant much cheaper memory it may be worth it, but even that's not true now - the Canon records 50Mbs to Compact Flash cards.

It's just my subjective opinion that for equal bit rates (25+ Mbps), AVC mpeg4 h.264 encoding still looks better than mpeg2 even at the higher bit rates where AVC yields diminishing returns. AVC by default applies some deblocking.

AVC can employ one of two forms of entropy encoding, CABAC (context adaptive binary arithmetic coding) or CAVLC (context adaptive variable length coding). The former is responsible for making h.264 coding small and efficient, but is processor intensive, for both editing and playback. It is the default AVC coding entropy for most NLEs. The latter is much less processor intensive for editing or playback, but also loses 15-20% encoding efficiency, so the quality advantage over mpeg2 for near equal file sizes is less.

Although we know that most NLEs either offer a choice of CABAC or CAVLC, or if only one will have CABAC as the default, we don't know, or at least I don't know what coding entropy AVCHD cameras use internally. The point about that, is when not knowing the distinction, it's harder to try and equate performance at certain bit rates, for example 28 Mbps or 35 Mbps AVCHD versus 50 Mbps MXF, with the exception that the latter AVCHD is new altogether if 10 bit and 422 are confirmed.

My editing experience and preference would overall be about neutral, except a lean toward Sony-Canon MXF chiefly because the camera itself is usually better than most AVCHD cams. That and the fact that certain NLEs like Vegas edit MXF natively is nice. But otherwise, I think the weakness of editing AVCHD has more to do with mediocre quality or lack of control over the specific implementation of final AVCHD rendering within a given NLE than any flaw or weakness within the h.264 codec itself.

While there is no doubt that h.264 is the more advanced codec, it's superiority comes from its ability to maintain higher quality within a smaller file size. Sounds great? Yes..but it's more compressed, thus harder to encode/decode and maintain its advantage.

And although I prefer MXF generally, I still think a final render to AVC h.264 28 Mbps CAVLC looks smoother and better (very slightly) than mpeg2 at higher bit rates, which maintains its blocky artifacting.

[David Heath]

Originally Posted by Tom Roper

It's just my subjective opinion that for equal bit rates (25+ Mbps), AVC mpeg4 h.264 encoding still looks better than mpeg2 even at the higher bit rates where AVC yields diminishing returns. AVC by default applies some deblocking.

AVC can employ one of two forms of entropy encoding, CABAC (context adaptive binary arithmetic coding) or CAVLC (context adaptive variable length coding).................

I don't think there is any question that * for equal bit rates* H264 will outperform MPEG2, the question is whether it's worth it. That's why it may be better to look at the issue with comparable equal quality, and for any level see the equivalent bitrates. It follows from what you say that for H264 (and AVC-HD) it will be less than for MPEG2 - the question is how much less, and is it worth it? Lower bitrate is the upside - higher complexity is the downside.

In general, the higher the bitrates involved, the less advantage H264 has over MPEG2.

And as you rightly say, all H264 and AVC coders are certainly not equal. CABAC and CAVLC are certainly possible differences between encoders, but there's plenty more features that may or not be employed to improve compression efficiency. You very rightly ask about ".....what coding entropy AVCHD cameras use internally. The point about that, is when not knowing the distinction, it's harder to try and equate performance at certain bit rates." Very true. It's why I've tried not to get too much into numbers, other than to say I'd expect AVC-HD to need "more" than 35Mbs to rival XDCAM422 50Mbs. How much more will depend heavily on the individual encoder.

And a key point is the difference between real time and non-real time encoders. If you're compressing within an NLE in a powerful computer it's obviously more feasible to use all the possible "tricks" - it'll just take longer to encode!

But a camera has to do it in real time, and is very unlikely to have the processing power of even a moderate computer. Hence the encoding tends to be much simpler.

I'm quite prepared to accept that at lower bitrates software encoders (non-real time) for H264 may well require less than half the datarate of MPEG2 for equivalent quality. At higher bitrates (>25Mbs), and for realtime camera encoders the difference will not be anything like as great as 2x.

Which brings it round in a circle. A new AVC-HD spec of around 40Mbs may indeed rival XDCAM422 for quality - but is it worth it? Is the datarate drop of about 20% worth the complexity?

I take Alisters point about MPEG2 existing within different wrappers, but isn't such as the Panasonic implementation of AVC-HD different to such as the Sony one?