First, Sony has answered this 4:2:0 vrs 4:2:2 question, several times;
http://www.dvinfo.net/conf/851508-post16.html http://www.dvinfo.net/conf/924899-post26.html Second, when monitoring HD-SDI out in "live", the image delay is not long enough to have passed through the codec chip. Third, every time I do an HD-SDI compare to the native codec (as in blue or green screen work), I can clearly SEE the increase in chroma resolution... here's a sample from a clip I had on hand... This is a small crop, enlarge approx. 600%... http://ftp.datausa.com/imageshoppe/o...compressed.png Look at the borders of the red and white, notices how ragged the 4:2:0 images is compared to the 4:2:2 uncompressed sample. Note also the difference test I added in, which shows the areas where there are significant differences in the image... Regards, Jim Arthurs |
Quite honestly, the HD-SDI out was icing on the cake for me. I could care less whether it provides BETTER recording than the SxS cards. I understand where everyone is coming from, but I paid less than $6500 for a Sony Handycam that shoots in the dark better than anything near it's price range, and damn near better than anything that shoots DV. The fact that I can nearly get away with 0db gain and rate this thing nearly what I rate my DVX at, is incredible.
The HD-SDI lets me roll to external storage at full raster with at least the same quality as on board recording. For those who are shooting green screen and doing VFX, what were you shooting on before? And did it give better or worse results? If better, what did the camera cost? If worse, how much more is the EX1/EX3 than what you were shooting? I am looking at the Convergent box as my Firestore replacement. And that was even before Focus filed Chapter 11. If the EX1/EX3 doesn't give the pro results you need, step up to a true pro camera. But it seems a little harsh to beat up Sony on a $6500 camera because it might not offer the same class of performance as the $50k cameras. At least that's my view. |
Thanks Jim...
Well, that pretty much closes this discussion. I remember you mentioned your SDI capture card was not capable of 10 bit a while back. I'm glad you were able to show the direct relation between SDI and SxS. This clearly show how impressive uncompressed (no mpeg macroblocking) with 4:2:2 are over the captured SxS 4:2:0 XDCAM. I'm sure many will feel XDCAM 4:2:2 may be good enough... I imagine a lot of the time it would, but when you're looking for the very best possible quaility that can hold up in post, SDI is your best bet. Especially green screen. |
Quote:
Jim, your picture tells more than 1000 words, indeed - thanks! Edit: As I'm eagerly waiting for my NanoFlash to arrive, I'd still appreciate it very much if Convergent Design commented on this further - with their own comparison, and (hopefully) confirmation that the EX1/3 HD-SDI output is not only 10 bit, but a true 4:2:2, as well. |
The differences are minimal. Is it really worth the dollars to buy the Flash XDR? Many other factors may decrease sharpness and increase noise, so will the SDI out really make a difference?
In normal conditions, the EX1 gives such an extraordinary good picture that I doubt if such a little difference will be noticeably. My major concerns are noise in poor lighting conditions and white-outs in good light. Will a HDSDI 422 10-bit address these concerns? |
Quote:
You're right - the macroblocking is not an issue with the EX codec under normal viewing conditions, but for more elaborate editing (compositing, chroma keying) - the less of it the better, too. But of course, no recording device/codec will make for poor exposure settings ! |
Quote:
It all depends on the application. I know that the BBC will not accept footage from the EX1/EX3 for their high-end shows, due to the 4:2:0 and 35 Mbps. They want 50Mbps 4:2:2 as a minimum. On the other hand, I do understand that they will accept EX1 + nano/XDR footage. Likewise, Discovery HD ranks cameras in bronze, silver and gold classes. I understand that the EX1/EX3 is currently rank in the silver class. I am hopeful that the EX1 can be raised to the gold class with the addition of the nanoFlash running at 50 or 100 Mbps. Also, don't forget the significant difference in media costs. 16GB SxS cards are running around $700, while 32GB CF cards are $115. This difference can really add up if you don't have access (or time) to offload the cards to a laptop. If you shoot any long events, the payback can come fairly quick. |
Quote:
Quote:
Quote:
|
Quote:
Quote:
Thanks for feedback, Piotr and Mike. |
I like Perrone Ford's comments. It's all a matter of perspective. The cost savings for storage media for the Flash XDR nearly pays for the device, as opposed to the SxS cards, especially if you are shooting a long event/documentary or cannot afford to pay to have someone whose only job is transferring files all day. Get the tools that'll do what you want for where you want to be in the marketplace.
|
Quote:
There is of course a limit to what the camera can do, but there is no question that the on-board recording format is the first limitation to the performance of the EX1 or any of the better HDV cameras. Regards, Jim Arthurs |
Quote:
|
Thanks for posting the comparison pics, Jim.
For me, they actually reinforced the point that the differences are minuscule and won't affect anything in practical terms, including green screen work. 600x magnification, and images look just the same to the eye, unless you digitally extract the difference, which seems to be not much either. And to capture this, you'll have to invest 75% more than the cam's cost? As for using CF cards for media, there's a couple of threads here where people cleverly use EX1 with CF card adapters to record on cheap CF cards. Certain adapter/card combinations work with CF cards, seemingly, without a hitch now. See work in progress: http://www.dvinfo.net/conf/sony-xdca...ex1-ex3-6.html http://www.dvinfo.net/conf/sony-xdca...ory-ex1-2.html Back to this thread though... isn't 422 suppose to have 2x more chroma information than 420? If so, how does that difference show, in layman terms? My greenscreen tests showed zero visible quality difference between the keys made with either footage, SxS vs HD-SDI. |
Quote:
|
If you can't see a difference and feel it's so small, stick with the XDCAM codec. Like Jim pointed out, 4:2:2 is your best bet for clean edge green or blue screen work.
But, I certainly can see the difference, especially as Thomas pointed out in more complex scenes. Working with XDCAM footage over the last year, although it's no doubt one of the best mpeg compression verses data rate, more detailed scenes with motion can show mpeg artifacts. SDI will allow to capture your footage at a higher data rate to minimize these issues and also allow better chroma subsampling (4:2:2). It's all about if you want the best possible. Certainly XDCAM codec does a great job. |
1 Attachment(s)
Here is Jim's image showing just the chroma channels.
The way color compression works is that the lower sampling of color was designed so people wouldn't really notice it with the naked eye. The fact that some people just cannot tell isn't a shock to me at all. When you isolate the chroma channels the way a keyer does you really start to notice the difference. Here you can clearly see there is a difference between 4:2:0 and 4:2:2. At first I thought maybe the camera was up converting or filtering the chroma channels before it sends it out SDI but I don't think so anymore. The image Jim posted helped confirm it to me. Look at the thin horizontal line in the upper left of the images. In the 4:2:0 version the line is clearly much thicker. This is because 4:2:0 can never have objects a pixel thin since it really works in blocks that are 2x2 pixels in size. So a thin line will will be twice as thick in the chroma channels in a 4:2:0 image. If the video was chroma smoother or up sampled you would still end up having blocks of color that are bigger then the pixels on the luma channel. A chroma up sampler doesn't remove extra pixels, it just blurs the chroma channels only or fills in the missing chunks. On the 4:2:2 version of the image the chroma line is clearly much thinner or what we would expect how it should look compared to the luma channel. Therefore I don't think it was up sampled at all but a true 4:2:2 output from the camera. If it was up sampled that thin line would be much thicker. With all that garbage said however people have made a very valid point that 4:2:2 or 4:2:0 doesn't have to be a deal breaker. In many case even for keying 4:2:0 will give some pretty darn impressive results. In my experience as a compositor after I up sample the chroma about the only thing I get is a bit of color bleed on edges due to the larger cover area of the chroma pixels. |
Thomas, Jim's EX1 sample is an excellent example of 4:2:0 vs 4:2:2 from the EX1.
Thanks for digging this up. |
It's worth mentioning again that the advantages of 4:2:2 over 4:2:0 are far greater with an interlace system than a progressive one - I'm assuming the images we've seen posted are from a progressive original?
Remember that when digital component video first came about, video was interlace, period. Hence the reasoning behind halving chroma resolution horizontally, but not vertically. |
Funny that no one brings up motion artifacts, except for the allusions made by Thomas Smet. As far as I'm concerned, the 35Mbps bitrate from the EX1 is more than satisfactory UNTIL the camera starts panning, vertically or horizontally. As Thomas points out, it's in full frame motion(like a pan) where 35 mbps gets overwhelmed by the amount of compression it needs to do. Image blurring, aka motion artifacts become painfully apparent, especially on fine detail. Compression rates on the order of 50-100 mbps effectively reduce the amount of compression by 1/3-2/3rds, and hopefully, the motion artifacts. This has nothing to do, really, with chroma sampling rate or chroma keying. At least, this is my biggest hope for the nanoFlash I've ordered.
Please don't tell me to reduce my pan rate. I know the arguement, it's valid, up to a point. Motion artifacts can occur in other instances where in frame motion is a big part of the image frame. |
A VERY vaild point David!
|
I hear you Bill.
Shooting seaside footage of water movement can show mpeg macroblocking with XDCAM codec. I'm looking forward to Portable SDI capture devices for the best quality possible. Even capturing at 50mb/s will offer cleaner mpeg artifacts. |
Quote:
|
Mark,
I'm sure you will find Bill is a fan of the EX1. He owns one and has offered a lot to these forums What I'm finding is it's hard to write in a few sentences without someone misinterpreting someones thoughts. Mpeg artifacts are not real noticeable with the XDCAM. There's no doubt that XDCAM 35mb/s is probably "good enough". The real discussion is when your looking to get the best possible from the EX1. We can't help this stuff man.... We all have the strive for the best possible disease. No doubt this is why we own the EX1. I'm still floored on the quality coming off of this camera. |
Mark, I agree that more practical tests are needed.
The more data, the clearer the picture will be. We all agree that 422 HD-SDI was *supposed* to be so much better than the internal SxS MPEG-compressed recording, in theory. Let's see clear evidence of that with EX1, on practical footage. A bit of history: I actually was one of the first people to positively confirm that Sony FX1 did output uncompressed video out of it's Component connector. I was then the first one (correct me if wrong) who actually built a device to capture it that way. I used it to shoot a HD short back in 2006. It was covered by the press, and Sony engineers called me numerous times to see how I did it. So I'd like to think that I know what I'm talking about, usually. My own point has always been that we should always try to capture at the highest quality that can be afforded. Now, with EX1, I thought the same: surely HD-SDI out should be significantly better in both compression artifacts and color fidelity. And then, my tests did not show any difference, to my surprise. The image, especially in terms of no motion artifacts, was so much better with both FX1 (live out of Component) and V1U (live out of HDMI). But not with EX1. Thus it'd be useful to see more tests by others, if possible. Maybe I'm doing something wrong. Or maybe there really isn't a discernible difference w EX1 (again, we are talking practical applications, not theory, please... I'd like to see something that can actually be appreciated by viewing on-screen as part of the moving images.) |
I've tested a PDW-700 which is 4:2:2 at 50Mbps and an EX3 side by side under controlled conditions and I can tell you there is very little difference in the pictures that these cameras record. Is the EX1/3's HDSDi 4:2:2 or 4:2:0? I don't know. What I do know is that the images coming out of this camera are simply amazing. There are many cameras out there with full 4:2:2 HDSDi that cost an awful lot more than the EX1, yet the pictures are not as good. Don't read too much into the numbers, look at the pictures and make your quality judgement that way.
I've been using the XDCAM codec for 3 years and it is very robust. The difference between the uncompressed and compressed images has always been very very small. The XDCAM version of MPEG 2 is remarkably robust, much more so than HDV. |
Quote:
|
I'm surprised David from Cineform hasn't post yet, so I'll fill in a bit for him.
I sent him an uncompressed capture from HD-SDI through a Kona card. 10 frames of 1920 x 1080 60i. He wanted 60i to make it easier to distinguish the chroma sampling. We settled on a red Sharpie cap moving in front of a blue background to give the clearest chroma reference. "I think it is 4:2:2. I can't see how a 4:2:0 to 4:2:2 will generate to amount of chroma information. I realize now that a more detail object would have made it easier to tell, the noise characteristic is not one of 4:2:0 interpolated to 4:2:2." David Newman, Cineform So the conclusion is - probably 4:2:2 native. Or some really clever 4:2:0 conversion. Also, partly because it's informative and partly because everyone is going to ask anyway, I've included a frame from the clip I sent to David and a clip of the same setup captured in HQ to SxS card. I say partly informative because this is a shallow depth of field shot of a pen waving around and it's interlace, so focus/resolution is a bit mushed. That said, there is a clear advantage to the HD-SDI shot. Quieter, cleaner, with a finer tooth to the interlace yuck. This advantage comes from avoiding MPEG macro blocking and the 30 something to 1 compression on the SxS card as well as 4.2.0 vs 4.2.2. Watching the clips in motion, the twice per second softening usually visible in high motion shots on SxS captures is gone in the uncompressed clip. On the secondary question of whether the nanoFlash will be useful, I think it should improve high motion shots and be much easier to haul around than a Mac Pro with RAID. |
I always find it interesting how much people differ in their approach to things. I have always been a pretty big techie, sometimes that's a plus, sometimes not so much so. All I can ever do is offer my opinion...yes O-P-I-N-I-O-N. I will leave proof to the reader. If you are happy with the results you're seeing, I envy you because, in the end, this profession/hobby will cost you much less than it costs me...LOL
At the risk of repeating myself, my opinion of capturing 10-bit 4:2:2 at 100mbps vs. native capture of 8-bit 4:2:0 at 35mbps, has absolutely nothing to do with 4:2:2 vs 4:2:0. It's all about the mathematics of squeezing quality out of 35mbps worth of data vs. 100mbps worth of data. As a clue to what this might be like, I invite anyone with Avid or DVCPROHD to compare the same footage from DVCPRO50 to DVCPROHD(100mbps). If you see a difference, then you will see an even bigger difference with (nano)Flash. Let me tell you, after a summer of doing 16mm film transfers(which was done at 2k 4:4:4), there is a HUGE difference. However, for anyone who never strays far from their 1600x1200 computer monitor, you'll never SEE the difference. Hell, I can see the difference between DNx115 and DNX175 on a production monitor. It stands out like a slap in the face. Not to mention DNx220 or even Red's 4k images. Vive la differance! |
Quote:
However EX1's video quality is so much better than all the previous cameras in this price range. I even sold my V1U (a great camera on its own) because I wasn't happy how the video was intercutting between it and EX1. |
Quote:
|
Quote:
|
Quote:
|
Gints, gald to be of help. Z1/FX1's video is really easy to break. (Unless you are capturing live from Component out, in which case it does not have much motion artifacts. I even posted comparison video caps Tape vs Live Component out here a couple years ago.)
V1 also not the best - I did a test at NY intersection, zoomed in so 2/3 of the car's length would take the whole screen. With cars (relatively slowly) moving in different directions in the background, and people doing the same in the foreground, the image fell apart a couple of times real bad. I assume this was one of the situations when codec could not predict the motion, and thus could not cope. When Adam Wilt said that EX1 is much better and it's virtually impossible to break its image, I could not believe it. Would 35Mbs vs 25Mbs really make much difference? Apparently it did :) |
Perhaps a bit late in this discussion to be pointing this out however I do recall Sony saying the SDI signal is derived from the camera's component outputs. If correct then the signal has been though a chroma smoothing process in the D->A converters.
Avid has long maintained that the best way to capture DV is via component from a VCR so equiped. Clearly the VCR cannot put back that which the DV compression lost so the same result can be achieved in post. Even if what is coming out the SDI port is from 4:2:0 processed video that's been resampled into 4:2:2 the critical question for keying is how much chroma resolution is available, not the sampling scheme. |
Quote:
You're joking right? |
Quote:
|
Quote:
I can't see a difference between Jim Arthurs pics? I have been told however that in panning there would be a difference and all I ask is evidence to show this |
Quote:
Ever since I equipped my editing work-post with the 50" HDTV plasma, hanging above my computer monitor - I can watch virtually every pixel of the gorgeous EX1's output. But since the better is the good's enemy, I am looking forward to use the NanoFlash for an even better HD experience. This is why the 4:2:0 vs 4:2:2 debate is so important to me. However, even with the difference in Jim's comparison being evident, you're right we still haven't arrived at a 100% certain conclusion. Mike of Converget Design has promised his engineer would provide us with some, but we're still waiting. Let's assume now that - as Bob suggests - the SDI signal goes through some processing, and "what is coming out the SDI port is from 4:2:0 processed video that's been resampled into 4:2:2". But even then the processing is doing its thing very well! Add to it no macroblocking or mosquito noise (due to much more relaxed compression), and the nanoFlash is still a worthy upgrade. Now, if it turns out that the SDI output is a "true" 4:2:2 - the better for us! Especially for those, whose ego feels better when all specs are the best possible, also on paper - especially when it comes to spending more money... Oh, and one more thing: we have witnessed many instances of pure incompetence from Sony "reps"; on the other hand what higher-rank Sony's officers like Mr. Martinez were saying, always proved to be right so far; why not believe in what he said this time? |
Hi Piotr
Thanks But Im still puzzled? Ive downloaded the image and zoomed in? By zooming in surely I am compensating although maybe I'm missing something here again? Im starting to wonder if there really isnt a true 422 and even whether a faster bit rate is making much discernable difference? At least this is the picture Im seeing at the moment.. I fully expect though to be able to see a difference being demonstrated and am eager for that. But certainly at this moment in time I wouldnt take a gamble there even was one certainly not on what Ive seen so far. |
Quote:
|
All times are GMT -6. The time now is 07:36 PM. |
DV Info Net -- Real Names, Real People, Real Info!
1998-2024 The Digital Video Information Network