p.2 #1 · potentially interesting 5D2/5D3 video news
super35 wrote:
hey skibum5, that jpg I linked was not my shot and not my processing. I found the link on another forum. The actual raw DNG of the same shot I found here.
Summary: This image was made on a 5D3 by capturing the Live View buffer to a 14-bit DNG using a development version of Magic lantern firmware. Images in this mode are shot without shutter action. This opens up possibilities for timelapse or other uses (burst mode?, video?) of a totally silent raw stills mode at 1933x1288 resolution.
thanks.
yeah, man, assuming that is what sources the video, and you'd guess it certainly would have to be, what the heck are they doing to take the beautiful signal they got off the sensor? It seems the camera hardware can grab a truly top quality 1080p signal. Are they going DNR insane? Applying a gaussian blur to help sell 1DC/1DX/C100 series?? Does something in the digic pipeline suck THAT badly in taking that down into 1920x1080x8bits
p.2 #2 · potentially interesting 5D2/5D3 video news
There is a video on B&H's youtube channel where a Canon USA employee basically admits that they fell backwards into video recording on DSLR's. The way he explained it was that after developing live view they realized the pretty much had video only no way of recording it. So the engineers added video recording almost as an afterthought and the first product with this was the 5D2. At the time Canon had no idea how popular making videos with DSLR's would become.
Now Canon has an entire product line (Cinema EOS) born out of recording live view to video but it certainly had humble beginnings.
BTW, I forgot to say he (Canon USA guy) said video recording on DSLR's was almost a no added cost feature. In the 5D2 all the hardware needed for video was already in the camera to support the live view function.
p.2 #3 · potentially interesting 5D2/5D3 video news
percieved short comings of the 5D3
Perceived? The poor video quality of the 5D2/5D3 (and the 7D) has been demonstrated objectively by several direct comparisons with other cameras in the same or lower price range. There is one particularly eye-opening comparison which shows that the GH2 in 720p mode out-resolves the 5D2 in its (claimed) 1080p mode.
There are people who think Canon can do no wrong and others who think Canon can do no right. The truth is somewhere in the middle. But there is no doubt that Canon is crippling the video capabilities of their still cameras, and its presumably deliberate so the still cameras won't compete with their cinema camera franchise. At least I hope the crippling is deliberate -- if its not, it represents incompetence.
As far as the 1D-C is concerned, the price is simply outrageous for what it is (a 1D-X with a firmware upgrade and a larger heat sink).
p.2 #4 · potentially interesting 5D2/5D3 video news
BluesWest wrote:
Perceived? The poor video quality of the 5D2/5D3 (and the 7D) has been demonstrated objectively by several direct comparisons with other cameras in the same or lower price range. There is one particularly eye-opening comparison which shows that the GH2 in 720p mode out-resolves the 5D2 in its (claimed) 1080p mode.
There are people who think Canon can do no wrong and others who think Canon can do no right. The truth is somewhere in the middle. But there is no doubt that Canon is crippling the video capabilities of their still cameras, and its presumably deliberate so the still cameras won't compete with their cinema camera franchise. At least I hope the crippling is deliberate -- if its not, it represents incompetence.
As far as the 1D-C is concerned, the price is simply outrageous for what it is (a 1D-X with a firmware upgrade and a larger heat sink).
It's a little maddening to see that it's actually getting a high-quality, crisp and reasonably aliasing-free signal out of the sensor (see that DNG) only to have who the heck knows what filter it over at some later stage (for no obviously apparent reason).
If they (ML) can't quite manage to sustain 24fps with the 5MB DNGs I wonder if there is a way to feed those just the minimal size clipping to the compressor and some way to get better quality out or if 10bits could be sent along. They might need access to Digic coding though which has not been hacked. Or maybe some tiny little silly thing in the system was not created properly and it won't be possible to do that even with full programming access and documentation.
p.2 #7 · potentially interesting 5D2/5D3 video news
It depends on if you shoot log or other flat profiles. Normal pro-res is 4:2:2 8bit is a big industry standard acquisition format. Even the very large production houses don't tend to use pro res HQ 10 pit unless they are doing effects work. 125 Mbps pro res 4:2:2 is very nice out of a c-300 to a ki pro. Heck the 50Mbps stuff recorded in camera is great even with rec709 color space. Photos are a whole other conversation.
p.2 #8 · potentially interesting 5D2/5D3 video news
saneproduction wrote:
It depends on if you shoot log or other flat profiles. Normal pro-res is 4:2:2 8bit is a big industry standard acquisition format. Even the very large production houses don't tend to use pro res HQ 10 pit unless they are doing effects work. 125 Mbps pro res 4:2:2 is very nice out of a c-300 to a ki pro. Heck the 50Mbps stuff recorded in camera is great even with rec709 color space. Photos are a whole other conversation.
I guess it also depends what you are shooting too. Carefully planned shots/shots with all sorts of supplemental lighting vs. open-ended natural world wilderness videography. You can certainly produce nice stuff with the 5D3, of either type though. That said, sometimes 10bits or more and more DR would be nice though and for wide angle natural world stuff the low micro-contrast doesn't work out as nicely and it can be tricky to avoid a bit of a plasticky feel to things, for more zoomed in nature stuff the sharpness tends to feel better and the lowered micro-detail doesn't seem as apparent as often.
p.2 #9 · potentially interesting 5D2/5D3 video news
I don't think the 8 bit vs 10 comes into play with DSLRs. They have very limited DR for video vs something like a c300 or Arri Alexa recording at 8 bit.
p.2 #10 · potentially interesting 5D2/5D3 video news
saneproduction wrote:
I don't think the 8 bit vs 10 comes into play with DSLRs. They have very limited DR for video vs something like a c300 or Arri Alexa recording at 8 bit.
Which is odd since they, even the Canons never mind Exmor, get 10bits easily at 8MP normalized. So why does the video from the Canon DSLRs have low DR at 2MP normalized? Even at full res RAW they get more than 8bits DR. So how do they get so little at ultra-downsized 2MP? Some weird things seem to happen at some stage in their pipeline when it comes to video.
It seems hard to believe they wouldn't be able to make use of 10bits for video.
p.2 #11 · potentially interesting 5D2/5D3 video news
Ya it is weird but 5DII has around 5 stops or so of DR. C-300 has much much more and I can tell you that from experience. Alexa is even better even with pro-res 4:2:2. Can't underestimate the scene file impact.
p.2 #12 · potentially interesting 5D2/5D3 video news
saneproduction wrote:
Ya it is weird but 5DII has around 5 stops or so of DR. C-300 has much much more and I can tell you that from experience. Alexa is even better even with pro-res 4:2:2. Can't underestimate the scene file impact.
Well the 5D2 line skips so at least you wouldn't expect it to get all that much of a normalization factor, then again you wouldn't expect it to be worse than 100% view stills mode either.
The 5D3 is odd since it picks up a nice SNR boost for many tones and yet
p.2 #15 · potentially interesting 5D2/5D3 video news
skibum5 wrote:
Could be, but all the same that file you provided looked pretty crisp. How did it go from a crisp frame buffer to a smeary video file? Why the need to smear it down in the video engine? And it does make you wonder about 10bits compressed ability even if 14bits RAW does seem very dubious.
It seems curious they hinted at an equally perfect buffer from the 5D2 since I didn't think the sensor on that could be read so fast. That seems hard to believe. The EOS HD guy may have been blowing smoke in that regard. Maybe it is an aliased messy looking buffer.
Back to the 5D3:
5*24 = 120MB/s rate
36MBish * 6fps = 216MB/s but it does bog down a bit even on the fastest card after a few seconds
I think 1080P24 all-i is something like peak 90MB/s which is certainly less than 120MB/s and it seems like RAW to fast cards tops out at about 80MB/s sustained so it would seem to be a huge 50% too slow (although I thought I saw some rumors about driving it at a constant 100-something MB/s)
Anyway, forget the RAW video, the main things you really wonder about are 10bits and the crispness.... ...Show more →
It's up to 90Mbit/s, or 90 megabits. Roughly 11/12 megabtyes per second. 8 megabits = 1 megabyte.
The lower 30Mbit/s IPB codec can be recorded to a class 6 card (i.e. 6 mEGAbytes per second).
p.2 #16 · potentially interesting 5D2/5D3 video news
StillFingerz wrote:
Skibum...here's a good review of the 5D3's video capabilities by Phillip Bloom, thought it might shed some light on the 'softness' the 5D3's video initially has; it can be sharpened quite well in PP...hope this helps and btw the new firmware is starting to be released; down under and Canon is installing it at their service centers...Jerry
Review of the Canon 5Dmk3
Phillip also did a video comparison of the 5D3, D800 and D4...if you're interested.
Full frame shoot out: D4 vs D800 vs 5DmkIII
I shoot video all the time with the 5d3. You really, really, really can push even the IPB a long way in post. With some not-as-sharp lenses I have I've used the 60 sharpen amount with no ill effects, and with Adobe Premiere Pro's "Highlight/Shadow Recovery" tool you really can get a lot of information back. After shooting with several GH2's and a Panasonic AF-100, I'd say it's about 90% as sharp as a GH2 after grading, and about 10000000 times better for shallow DOF and stills.
There is no perfect camera. The Blackmagic requires a huge amount of disk space, has a 2.4x crop factor, and a very tedious workflow. For me I'd just like to have a 1dc, but even then, the idea of grabbing stills from video shots is sort of a waste as you have to shoot with a slower shutter in video mode. Of course this isn't going to be as crispy as a nice 1/1000 shutter still, now is it?
p.2 #17 · potentially interesting 5D2/5D3 video news
You know I've been reading a lot of people complaining about the video quality of the 5d3. IMO there's no better hybrid stills/video camera out there.
It's discreet, small form factor, records for 29 minutes straight without overheating, no "must have terabytes of SSD drives lying around" cost, records pretty good audio if I'm honest, and is a stills beast.
I would love to see the following added however:
1. Some sort of manual focus confirmation like the Sony cameras have. Sometimes I'm not 100% sure if I'm in focus while shooting especially if my subject is moving. Like the center rectangle turning green or something.
2. Native focus peaking/zebras/constant histogram without loading ML.
3. Articulated screen a la Sony a99/a77.
But for about $3k what device could one buy that bests it in both stills and video? The Nikons/6D/5DII/(hell even GH2/3) show more artifacts and moire. The GH2/3 is better but still no focus peaking and a far cry on the stills side of the house. The BMC has a very expensive accessory package to purchase to make it useable, has less shallow DOF, 2.4x crop and is a 4 pound brick that doesn't take stills. The Sony cameras take really lousy video especially the a99 (I owned one for a month thinking it'd be the holy grail) - it uses line skipping and is mushy, low detail garbage.
Rambling on, I know. Just shoot some video with the 5d3, turn in-camera sharpening off and turn down the contrast and bring it all back in post, and it's a very pleasing image.
p.2 #19 · potentially interesting 5D2/5D3 video news
Shield wrote:
I shoot video all the time with the 5d3. You really, really, really can push even the IPB a long way in post. With some not-as-sharp lenses I have I've used the 60 sharpen amount with no ill effects, and with Adobe Premiere Pro's "Highlight/Shadow Recovery" tool you really can get a lot of information back. After shooting with several GH2's and a Panasonic AF-100, I'd say it's about 90% as sharp as a GH2 after grading, and about 10000000 times better for shallow DOF and stills.
There is no perfect camera. The Blackmagic requires a huge amount of disk space, has a 2.4x crop factor, and a very tedious workflow. For me I'd just like to have a 1dc, but even then, the idea of grabbing stills from video shots is sort of a waste as you have to shoot with a slower shutter in video mode. Of course this isn't going to be as crispy as a nice 1/1000 shutter still, now is it?...Show more →
Doesn't sharpen 60 or even 30 make it look all digital blocky looking and weird though? Which sharpen tool and what settings are you using though?
p.2 #20 · potentially interesting 5D2/5D3 video news
sqdstf wrote:
He seems to be too optimistic though. Sounds like 12-14fps may be the max. (unless they can find a way to feed a 1920x1080 cropped version of those DNGs into the compressor engine and the engine doesn't clip to 8bits even if it did that at least it might maintain much better details)
But man I just don't get why Canon seemingly left so much on the table. Sure uncompressed is too much for the HW to handle but why in the world could they not have written out a compressed file with original DR left intact and not utterly massacred and with a few more bits? And what on earth are they doing to reduce the beautiful crisp 1920x1080 the fancy sensor the engineers developed for them and then turning it to utter mush! It's not the compression engine because using the HDMI out with ProRes HQ doesn't really seem to bring back any extra detail all (certainly not static detail). It just seems like they left a ton of ability on the table. They would've had an utter revolution if they had not mushed down the resolution that it is actually capturing and written it out with more bits and less DR compression done and put out a really crips compressed file with lots of DR room.
Unless this DNG is not what they feed the compression engine with and they need to use some faster, crummier source. Otherwise I don't get what the heck Canon software coders are doing to ruin the amazing video output their hardware designers gave them. I mean marketing couldn't tell them to trash DR, clip bits and apply a Gaussian blur filter to protect the video division could they?? They spent time improving the compression engine so why would they force them to do something like that?
