Nikon D800 with video features

[Jon Fairhurst]

It might have the same number of pixels, but each photosite would be smaller. Smaller photosites collect less light and therefore have higher noise.

[Josh Dahlberg]

Originally Posted by Sareesh Sudhakaran

The uncompressed 8-bit 422 is the big difference, isn't it? We'll know soon enough whether its worth anything in a few months.

That's the great hope: it will be a huge difference if the image being output via HDMI is largely crisp and artifact free. But simply capturing direct to ProRes via HDMI isn't going to rid footage of moire, aliasing and softness due to poor scaling.

[Jacques Mersereau]

Does anyone seriously think that Canon, after introducing a $16K camera, is now going to shoot itself in the head and put out a 5DMkIII that isn't seriously hobbled? I am happy to be proven wrong, but ... it just wouldn't make any sense for Canon to do that.

Time will tell what the D800 is all about. I am invested heavily in Canon, but my current hopes are centered on Nikon now. If I had $16K it would buy a Scarlet or used RED ONE package.

[John Vincent]

Canon might - the C300 is still a niche camera. I can see only rental houses, highend big city shooters, and perhaps TV productions buying into the camera - in other words, it will hit it's sales ceiling relatively quickly. Don't forget lots of people still LOVE the Alexa and have no plans of diving into the C300. Bigger productions will still either go RED, high end Sony, or (gulp) film.

There's another factor as well, and that's lens sales. If they're selling the Mark III like hotcakes, then they're also selling lenses, batteries, grips, etc in big numbers. While a $16,000 camera won't sell in anywhere enough numbers to generate substantial lens sales, a $3,000 Mark III likely would.

One other thing - it'll still be a DSLR. That means short battery life, no XLR or SDI ports, no internal 4:2:2 50Mbps recording, no built-in ND wheel, no lanc, and of course, the love/hate DSLR form factor.

[David Heath]

Originally Posted by Jacques Mersereau

Does anyone seriously think that Canon, after introducing a $16K camera, is now going to shoot itself in the head and put out a 5DMkIII that isn't seriously hobbled? I am happy to be proven wrong, but ... it just wouldn't make any sense for Canon to do that.

I think it's tending to be overlooked that the 5DMkIII is first and foremost a camera designed for high quality STILL PHOTOGRAPHS. In that case, it is anything but "hobbled"!

I also think the engineering issues of getting video from a sensor designed for still use are not being taken anything like seriously enough. The more you learn about it, the more you appreciate the challenges, and in many respects far from "hobbling" the surprise may be that video from DSLRs is as good as it is, without compromising the still performance.

Price differentials normally have to do far more with economies of scale than absolute tech. From what I'm hearing, C300 sales are already pretty high - but are inevitably going to be small compared to the DSLR market. But that's because so many people buy DSLRs for stills!

Originally Posted by Sareesh Sudhakaran

Is it safe to presume that the D800 will be 'sharper' in DX mode?

No, impossible to say without knowing a lot more. That may not seem logical, but AFAIK at the moment *ALL* such cameras with multi-megapixel sensors have to ignore a percentage of the photosites to get a high enough read/processing speed. It's conceivable that in full frame mode only (say) 25% of the sites are read, in windowed mode it may be (say) 50% of the window. (Which may be more than 25% of the total!)

Originally Posted by Sareesh Sudhakaran

The uncompressed 8-bit 422 is the big difference, isn't it? We'll know soon enough whether its worth anything in a few months.

It may be a big difference in terms of ease of connectivity, but don't read too much into it. Any HDMI signal is going to be uncompressed and 422 by definition (and may be 10bit) but that by no means says that all HDMI signals will be the same! It would be quite possible to have a rubbish front end, which could still give an uncompressed 8-bit 422 output. It's quite conceivable that a good front end feeding an internal H264 codec will outperform a bad front end feeding uncompressed 422.

To continue with the food analogy, if you start with poor really ingredients, the best chef in the world will struggle to cook an exquisite meal. That's why "uncompressed 422 output" means little in itself.

[Dylan Morgan]

So many people are pissed off at Canon because they think they have let down their customers. They switch to Nikon and brag about how Canon dropped the ball or that the D800 is a 5DII killer. Why are they comparing the 5DII to the D800 anyway? I laugh at them for getting so mad as they sell off all their gear. It's Canon's 25th anniversary this year and they will deliver as good if not better than what Nikon is offering. I love the specs of the D800 and look forward to people shooting with it. People are so impatient and stupid.

[Josh Dahlberg]

Originally Posted by David Heath

I also think the engineering issues of getting video from a sensor designed for still use are not being taken anything like seriously enough. The more you learn about it, the more you appreciate the challenges, and in many respects far from "hobbling" the surprise may be that video from DSLRs is as good as it is, without compromising the still performance.

David, as someone who's finicky about image quality but has very little understanding about the engineering behind it, I always appreciate your insights. I'm curious to hear your thoughts on what Panasonic's doing that's special with the GH2. I shot extensively with 5DIIs before picking up a GH2 - while it obviously doesn't have the low light performance or DOF control of the Canons, it eats them for breakfast in terms of detail and suppression of artifacts.

Phil Bloom's comparison using a hacked GH2 shows it's very close indeed to C300/F3 resolving power, and in my own observation GH2 footage output to a blackmagic card, footage blown up to 400% is on par with my XF300 (which the BBC tests put at around 900-1000lines). There's no comparison with 1st generation video implementation in Canon DSLRs.

Unfortunately, GH2 HDMI output is only 4:2:0, which is rather a big handicap. And it's a little too small/fiddly to be taken seriously (clients often look with kindly bewilderment at it's toy-like profile and say, "you're shooting with that?" = rarely had that issue with a kitted out 5DII/70-200 2.8 combo).

My hunch is the next series of Canon DSLRs will deliver - they've had more than three years to upgrade video implementation, and with the 1DX they've already promised improved codec and much reduced rolling shutter and unwanted artifacts. It will be very interesting to see how the D800/D4 fair when we get some less compressed, full res samples.

[Jon Fairhurst]

Originally Posted by Dylan Morgan

Why are they comparing the 5DII to the D800 anyway?

It's a valid comparison in that they are the two firms' semi-pro full frame cams. Given that the D800 is three and a half years newer, I would hope that it's the better camera.

But the other shoe has yet to drop. Canon hasn't had any pressure to upgrade the 5D2 - until now. This situation won't last long...

My guess, however, is that the decision between Canon and Nikon for video will no longer be clear. Those who want uncompressed recording and can budget for a Ninja might choose Nikon (assuming Canon's HDMI is still "dirty".) Those with Canon lenses might stick with Canon. We'll need to compare quality and features to know which camera is best for which users and for which shooting situations. But I doubt that there will be a clear, overall winner like there has for the past three and a half years.

[Jacques Mersereau]

<<<So many people are pissed off at Canon because they think they have let down their customers.>>>
So you believe that the $16,000 C300 is the right camera for the right price? For me it is a let down. A HUGE let down. Totally over priced for a 30 fps, 1920x1080, 8 bit machine.
Others find it a great deal. That's cool - so tell me, have you purchased yours yet?

<<They switch to Nikon and brag about how Canon dropped the ball or that the D800 is a 5DII killer>>
Me, myself and I do NOT want to switch. I'm a Canon guy. That said, I need a camera I can afford that produces cinema quality HD video with the ability to record with decent bandwidth. I love what the 5DMkII does for its price point, BUT I need higher bandwidth and real HD signal output, which it does NOT do.

<<Why are they comparing the 5DII to the D800 anyway?>>
I'm comparing the D800 to the C300 and more importantly, $3K to $16K.
You sound like a 5DMkII owner who feels threatened. This is not about that.

<<I laugh at them for getting so mad as they sell off all their gear.>>
I haven't yet, but may have to. Had the C300 been $9K and had 10 bit 4:2:2 out, I would probably find the money to buy that. But $16K? I'd buy a Scarlet if I had that much money.

<<It's Canon's 25th anniversary this year and they will deliver as good if not better than what Nikon is offering.>>
Well, either you're guessing or you broke your NDA. Either way, your statement remains to be delivered. From my experience, Canon is always FAR behind everyone else. The 5DMkII was really a still cam that Canon added video so PHOTO journalists could shoot a bit of video and save the BBC some money. Canon leadership in Japan was probably shocked to find that, although requested for years and years previously - here on the former XL1 Watchdog website, there is actually a market for film makers who want to use 35mm glass for HD video acquisition.

<<< I love the specs of the D800 and look forward to people shooting with it. People are so impatient and stupid.>>
Here we agree.

[Dylan Morgan]

This isn't a comparison to the C300 (Nikon D800 with video features thread). I was speaking generally about how people were reacting to Nikon's recent announcements. We're comparing apples to oranges with the C300 and D800. The C300 was designed for a specific market and type of shooting that will sell regardless of it's cost. Yes, I'm disappointed with the price and specs, but I'm not in the market for one so I don't care that much. BTW I don't own a 5DII and I don't feel threatened. You say that Canon is always far behind, but the 1DX looks to be par with the D4 as a flagship camera. I agree with you about HDMI, but can we be real and state what percentage of people with actually drop the money on an external recorder?

[Dylan Morgan]

Another thing I should point out about the C300 in comparison to DSLR cameras. Have you ever downloaded a full 1080p video shot on one? It has true 1080p clarity that the D4 or 5D don't come close to IMHO. I downloaded the "Why" video and viewed it on my monitor and it looked no different (DR set aside) than any DSLR video. Anyone who has owned a Sony EX1 will know what I'm talking about. You can tell
true 1080p on a 24+'' when you got it. DSLR video looks softer and not as sharp, period.

[Sareesh Sudhakaran]

Originally Posted by Jon Fairhurst

It might have the same number of pixels, but each photosite would be smaller. Smaller photosites collect less light and therefore have higher noise.

Actually won't it be the same? It's still an APS-C size with 16 MP, as opposed to the 7D with 18MP. Same pixel pitch, same sensor size. If the sensor readout is only within this area, it will be (theoretically) sharper and about the same quality as a 7D in DX mode.

Don't know if this is true, but I've read somewhere that 36MP is the same as APS-C pixel density, only its on a fuller frame.

[Sareesh Sudhakaran]

Originally Posted by Jon Fairhurst

Any HDMI signal is going to be uncompressed and 422 by definition (and may be 10bit) but that by no means says that all HDMI signals will be the same!

The HDMI protocol does not distinguish between compressed and uncompressed data. You can send any kind of video via HDMI. But I get what you're trying to say, and I agree with you.

I have faith in Nikon to deliver uncompressed. After all, they've been shouting about this from every rooftop since the first press release. And, they have nothing to lose, unlike Canon.

The cool thing about Nikon is its glass. It can be used on EF mounts (and pretty much any other modern mount) without loss in quality. Can't say the same about Canon's glass. In this 'war', the bodies are going to come and go at a frightening pace, and the better mount to invest in is Nikon's. Just my feeling.

[Jon Fairhurst]

Hi Sareesh,

On the photosite size question, I thought we were comparing the D4 in full frame mode vs. the D800 in crop mode. I wasn't referring to the 7D.

Regarding the HDMI item, it was David Heath who wrote that rather than I. That said, HDMI is always uncompressed. HDMI receivers aren't required to have MPEG-2, h.264 or any other video decoder.

I think what David was referring to is that even though HDMI is uncompressed 4:2:2 at 8 or 10 bits, that doesn't mean that the source signal was ever that good to begin with.

As an example, there are DVD players with 1080p upscalers and HDMI outputs that output uncompressed 4:2:2, but the signal quality for that product would still only be as good as MPEG-2 encoded standard definition, even if the signal capacity is much, much better than that. I think that David was making a joke that if the video from the D800 was ugly, the HDMI output wouldn't help.;)

Regarding Canon, their DSLRs put out uncompressed 4:2:2 from the HDMI output today (at least when not recording.) The two problems are that 1) Canon makes the signal "dirty" with the focus box and other menu items, and 2) Canon scales the 16x9 signal to fit within a smaller window - and the scaling is done with yet more pixel skipping. Canon's HDMI output is uncompressed but it's not clean. :(

[David Heath]

Originally Posted by Josh Dahlberg

David, as someone who's finicky about image quality but has very little understanding about the engineering behind it, I always appreciate your insights. I'm curious to hear your thoughts on what Panasonic's doing that's special with the GH2. I shot extensively with 5DIIs before picking up a GH2 - while it obviously doesn't have the low light performance or DOF control of the Canons, it eats them for breakfast in terms of detail and suppression of artifacts.

Firstly, I don't have a GH2, haven't used one, and haven't seen any proper test results from one, so I can only talk in generalities. Yes, I've seen Phil Blooms test results and (as stated on the thread) I would be wary of concluding too much from them. (Too many variables.)

From Phil Blooms postings, it did strike me that the F3 and the C300 were way out in front of the rest, (maybe shouldn't be surprised), the FS100 and GH2 somewhat behind, and the AF101 lagging at the rear - at least in terms of resolution.

In general, (and forgive me if you already know) if you want to use a Bayer sensor for video, the approach most likely to give good results is to read the whole sensor, do a deBayer as for a still photograph, then downconvert in high quality to 1080 output resolution. And do that at frame rate. That's what the F3 does, and even with a relatively low pixel count sensor (just over 3 megapixel – optimum for 1080) the power consumption is quite high. I leave you to speculate what the heat and power requirement would be if you wanted to do the same with a 12-16 megapixel sensor, even if you could get the silicon to manage it. It does not happen in cameras in this price range.

So, again in general, if you want to get video off a chip with a high pixel count, what do you do? Basic principles are:

1/ Don't read all the pixels on the sensor.
2/ Don't deBayer - do something like the C300, (Just look at the low power consumption of the C300 compared to the F3)
3/ After 1 and 2 above, make the raw output lower than 1080 resolution, then upconvert - don't downconvert a high res signal.

You can get a clue as to what is happening by power consumption, sensitivity, and most especially zone plates. If the camera isn't reading all the photosites, that's going to affect the sensitivity. Make allowances for sensor size, one chip v three chip etc and it's possible to deduce quite a lot from relative sensitivities.

Have you seen Alan Roberts zone plate test of the 5D? If not, http://downloads.bbc.co.uk/rd/pubs/w...on_5D_DSLR.pdf . Really nasty isn't it? (Sorry, don't know how later models compare.) Things to look for are symmetry, true resolution (circles centred to bottom left), aliases or "false resolution" (circles centred elsewhere), coloured circles (hint at pattern used for sensor/readout) etc.

And to really answer your question about the GH2, I’d like to see a properly done zone plate from it. I’ll leave one question in your mind. It’s been commented on that it has been found to be sharper than the AF101. On the face of it that sounds strange – how is a £3,000 (plus lens) dedicated video camera from a single manufacturer beaten by a far cheaper model from the same manufacturer, primarily designed for stills. Ridulous, doesn’t it seem?

The only answer that makes sense to me is that fundamentally they are reading the chip in a very similar way – but the AF101 has an optical low pass filter designed for video, the GH2 doesn’t. That may make the GH2 look better on side by side viewing of raw camera footage – but likely with much more aliasing. I can hear people now thinking “so what? If it looks better, who cares about the aliases?” And the answer is in what was said before – the real nasty about aliases is that they can screw up codecs. It’s quite possible that two cameramen may shoot the same scene with a GH2 and an AF101 and the GH2 look better at first sight. Put them both through a broadcast transmission chain and the GH2 may fall apart far faster than the AF101. I’ll leave you with what Alan Roberts says at the end of his look at the 5D.

Aliases in moving pictures are much more of a problem because, when the image moves, the aliased frequency content moves in the opposite direction to the image motion, causing a rippling effect on edges. Since motion-sensitive compressors such as MPEG2 and MPEG4 depend on the cleanliness of edges to measure motion, these aliases can cause the compressor to allot undue bit-rate to motion and/or result in excessive compression artefacts. Either way, pictures with aliasing at the levels seen here are not acceptable as HDTV.