Resfest - Info on new HD cams - Canon = wow!
- Thread starter mgalvan
- Start date
thisiswells
Retired
What d'ya know? Their behavior sure seems fishy to me...(cue: fictitious print advertisement)Digigenic said:
..yeah i just read that.. IMHO, thats a big DOH! for canon why not pixel shift if your not pulling 4:4:4 for the uncompressed output and thats the reason for cramming in pixels and using interlaced chips resulting in getting worse overall performance than could have been had. if its only 4:2:2 your sort of just spinning your wheels...
I think Barry did a test and found it got 6-7 stops with the dvx. With 8-bit color, I think that's about the limitaiton without some fancy stuff going on in the DSP. The hvx will probably have a little less dynamic range inherent in the ccds, but the DSP has surely improved considerably, so I bet it will end up being better than the dvx. The varicam claims 11 stops for what it's worth. 100 ISO film has about 8-9 stops, but they're all in the highlights, which, IMO, is where it counts.
I have faith in Canon, though, believe it or not. The digic 2 chip is amazing (not sure about the "digic dv" chip...they both lose for stupid names) and definitely compresses dynamic range beyond what should be availible in 8 bits. I wouldn't count Canon out. They have Sony CCDS (come on, you all know it's true) which are very clean and noise-free, and they have the best DSP in the world stolen from their still cameras. If they had DVCPRO-HD, I would consider the HVC trumped (I mean, they even put in an option for mini35 adapters, how incredible is that?! Everyon (me) was begging Panasonic to do that, then Canon comes out with it out of the blue?). Anyhow, all I can say is that I'm impressed. Both Canon and Panasonic are doing interesting things. But, for the love of god, $9,000 for 1/3''CCDs? Seriously, use some 2/3'' chips and save a bunch on the lens or something....it's just not right.
I have faith in Canon, though, believe it or not. The digic 2 chip is amazing (not sure about the "digic dv" chip...they both lose for stupid names) and definitely compresses dynamic range beyond what should be availible in 8 bits. I wouldn't count Canon out. They have Sony CCDS (come on, you all know it's true) which are very clean and noise-free, and they have the best DSP in the world stolen from their still cameras. If they had DVCPRO-HD, I would consider the HVC trumped (I mean, they even put in an option for mini35 adapters, how incredible is that?! Everyon (me) was begging Panasonic to do that, then Canon comes out with it out of the blue?). Anyhow, all I can say is that I'm impressed. Both Canon and Panasonic are doing interesting things. But, for the love of god, $9,000 for 1/3''CCDs? Seriously, use some 2/3'' chips and save a bunch on the lens or something....it's just not right.
Digigenic
Well-known member
:grin: ideally, I'd rather do a video, reminicent of the "flip-flop" political smear campaigns going on during the 04 elections, where they use still images of the culprit and fade it to a red and black two tone image while the voice gets deeper and more sinister accusing them of indecisiveness. That would be good.thisiswells said:
Policar said:
Exactly... about 4 stops of that 11 is in the milky shadows, which everyone crushes to hell anyway. I think comparing "stops of latitude" is very misleading, as it does not give a true picture, it is just a lab test.
I would prefer maybe a "Hudson Mojo Rating" scale to give a true impression of real world performance.
Policar said:
I think the theoretical limit for 8-bit color is 8 f-stops. That's because each bit is twice as high as the next lower bit, and each f-stop lets in twice as much light as the next lower one. You can probably play some games with the levels to get more in there, but probably not much more.
We should also be looking at how we display the image, because most monitors and TV sets are still 8-bits per color. So maybe more than 8 stops of dynamic range is a waste?
Barry_Green
Moderator
Really? Trumped? That discards DVCPRO50, plus 720p at variable frame rates, plus P2 recording, plus manual zoom and precision manual focus, plus uncompressed audio, plus a lot of other stuff...Policar said:
I don't think there's any way you can equate them and say that the HVX is trumped, you can just say that they're different. Now, if Canon had included all that stuff, plus interchangeable lenses, plus a real (and affordable) manual lens, and the footage is comparable, then yes I'd say the HVX got trumped. But right now it seems they have very different feature sets.
The only one who got trumped would be the Sony (1080/24F beats 1080/CF24 six ways from Sunday, and an interchangeable auto-only lens beats a non-interchangeable auto-only lens all day long too) and the JVC is looking less unique as well (1080/24f with interchangeable auto lens, vs. 720/24p with interchangeable but questionable manual lens). And even in the Sony's defense, its street price is half (shall I say that again -- HALF) of the Canon's MSRP (don't know what the Canon's street price will be).
Well, yes 1/3" is rather limited for the price tag of $9,000 -- but when you say "save a bunch on the lens" -- no way... if the Canon lens seriously trumps the JVC's, I'd say they probably put their money in the right place. The JVC lens bothers me. Of course, the auto nature of the Canon lens has its own level of repugnance (depending on how closely it parallels the design of the XL2's auto lens!) but if the glass is good, and there's no chromatic aberrations, I'd take the auto/good-glass version over the manual/aberrational version any day.
The Canon is definitely interesting. I'm looking forward to seeing it.
thisiswells
Retired
Lol. Well done, as usual.
Actually, the specs for XL H1 say they use horizontal pixel shift. (see http://consumer.usa.canon.com/ir/controller?act=ModelTechSpecsAct&fcategoryid=165&modelid=12152 in the image sensor section). They haven't flopped on the issue - they're still using it. Furthermore, I'd wager a guess that the pixel shift is the reason the HD-SDI output is 4:2:2 and not 4:4:4. They are stating "horizontal" only because we all know that you can't get real 4:2:2 if you use vertical pixel shift to gain resolution.
-Spiff
-Spiff
I think that the pixel shift is only used for stills, as it claims they can be recorded in 1920x1080.
With 1.56M pixels effective pixels there is no need for pixel shift to achieve 1440x1080, unless they are using it in some clever improved sensitivity/latitude/noise way........????
With 1.56M pixels effective pixels there is no need for pixel shift to achieve 1440x1080, unless they are using it in some clever improved sensitivity/latitude/noise way........????
if the ccds are already 1440 x1080? (what are they again?) then are they pixel shifting up to a full 1920 x 1080 or something just for the uncompressed out? so if you by some big dollar studio chance happen to be running uncompressed youll get some benefit from full res output and better S/N ratio and latitude than a full res ccd, but for the 99.9% of real world likely useage of the camera you lose out.. thats doesnt make sense. .. someone point me in the right direction here.. what are the CCDs in pixel dimensions.. are they 1440 x 1080 ? ... i cant seem to find the specs in pixel dimensions.
So I went back today to talk to both Panasonic and Canon. The Panasonic booth had a sales rep, but interesting enough, the Canon had some Independent DP there. Once again, the Panasonic rep seemingly didn't know (or thats what he was trying to play off) too much info on the camera. This time around we swithced it to camera mode to see what the image looked like on a high def monitor. And since the camera was only 85% done, all that was coming out was the raw image ... no WB, gamma, nothing ... just the raw image and resolution. So I do have to say this ... the footage is indeed pretty sharp. We zoomed in to some of the areas around the room and we even saw little scratches on the glass of the doors, and we weren't even zoomed in that close. And mind you, this is the interlace and not progressive 1080, so one can assume that the quality will be better in that mode. Otherwise from that ... nothing new to report that you don't know about already ...
And then to the Canon booth and was this interesting once again ...
This time, there was an independent DP named Matt who was there. He was asked by Canon to do a review of the camera, so I had a nice chat with him. He was overall very impressed with the cameras abilities. Though he too wasn't to keen on HDV, he did note that Canon did something to their HDV implementation to minimize any artifacting, including anything that appears due to fast motion. When I asked his opinion on 24F, he was very certain that the 24F modeis IDENTICAL in motion to the 24P mode in the XL2 and noted that Canon will be presenting footage sometime soon in regards to this. He said that the motion from 24F is so much like 24p that in SD mode, the camera implements 24F EXACLTY the same way that the XL2 does with 24P, with a 2:3 or 2:3:3:2 pulldown (as well as the ability to reverse the "telecine" to get back the true 24 frames). According to the Canon rep and this DP, 24F is 24P ... with just different names.
Another interesting thing is regarding latitude. I asked him how the latitude was on the XLH1 in comparison to those reported from the HVX. He didn't know anything about the HVX's perceived latitude, but he said that the Canon was about 7-8 stops!!! He was actually really impressed by the fact that it was able to output such. he did say that he didn't have his light meter with him when doing his review, so he can't say for certain that this is the number, but to his eye, it was around 3.5 stops on both ends. He also did hint that Canon will probably be releasing footage to prove its greater latitude pretty soon (something about clocks in front of bright white walls or something along those lines).
The one thing this guy had a gripe about was regarding the peaking/focus zoom assist. He said that if you useds this feature while recording, it would record it to tape and through HD-SDI as well. Not exactly good if your trying to use it to focus during a live record. It is something that you have to use before you start your shot, although he said he brought this up to Canon and they may fix this before release.
There's more too, but I can't think right now as I am at work busy with a show. Maybe when things calm down ...
And then to the Canon booth and was this interesting once again ...
This time, there was an independent DP named Matt who was there. He was asked by Canon to do a review of the camera, so I had a nice chat with him. He was overall very impressed with the cameras abilities. Though he too wasn't to keen on HDV, he did note that Canon did something to their HDV implementation to minimize any artifacting, including anything that appears due to fast motion. When I asked his opinion on 24F, he was very certain that the 24F modeis IDENTICAL in motion to the 24P mode in the XL2 and noted that Canon will be presenting footage sometime soon in regards to this. He said that the motion from 24F is so much like 24p that in SD mode, the camera implements 24F EXACLTY the same way that the XL2 does with 24P, with a 2:3 or 2:3:3:2 pulldown (as well as the ability to reverse the "telecine" to get back the true 24 frames). According to the Canon rep and this DP, 24F is 24P ... with just different names.
Another interesting thing is regarding latitude. I asked him how the latitude was on the XLH1 in comparison to those reported from the HVX. He didn't know anything about the HVX's perceived latitude, but he said that the Canon was about 7-8 stops!!! He was actually really impressed by the fact that it was able to output such. he did say that he didn't have his light meter with him when doing his review, so he can't say for certain that this is the number, but to his eye, it was around 3.5 stops on both ends. He also did hint that Canon will probably be releasing footage to prove its greater latitude pretty soon (something about clocks in front of bright white walls or something along those lines).
The one thing this guy had a gripe about was regarding the peaking/focus zoom assist. He said that if you useds this feature while recording, it would record it to tape and through HD-SDI as well. Not exactly good if your trying to use it to focus during a live record. It is something that you have to use before you start your shot, although he said he brought this up to Canon and they may fix this before release.
There's more too, but I can't think right now as I am at work busy with a show. Maybe when things calm down ...
Graeme_Nattress
Red Team
The reason the SDI output is 4:2:2 is that you NEED dual link HD SDI for 4:4:4!
Graeme
Graeme
mmm said:
thanks mmm.
i dont know though. so they must be using the pixel shift simply to output at 1920x 1080 full res in 4:2:2 on the HD-SDI out, and the signal would either get downsmapled to 1440 x1080 for regular usage or not implement the pixel shifting(if its possible to switch it off) either way that just seems odd to me.. why not base the best performance of the camera around the most likely usage instead of thehalf a percent or so usage that the HD-SDI will see? but anyway if its really getting 7-8 stops than they are doing alright for themsleves anyway .. but even still it could have been even better is the point. alas not that I should care as im not likely personally buying one, but it sure is shaping up to be an interesting and very pro-like peice of kit. certainly has some nice features. perhaps ill rent one for a day when they are out and see what she can do.
Barry_Green
Moderator
Graeme_Nattress
Red Team
They could be upsampling from 1440x1080 to 1920x1080 for the SDI output. If you looked at the output on a CRT you'd never be able to tell they'd done that 
Graeme
Graeme