Yes, it does!
I'm not sure.
No, it doesn't.
HDMI controversy tested!
In one of my original posts I predicted the difference between an HDV Firewire capture and an HDMI direct capture would only be a few lines of resolution different even if HDMI is higher resolution because each pixel is only 33% longer.
I'm going to post the images from the tests and you can vote on what was or wasn't accomplished.
Realistically you need to look at all the images to judge if that extra 25% of image data is there.
Last edited by Duke; 2008 February 18th at 10:26.
There has been a lot of debate over direct capture by HDMI vs. HDV capture by Firewire. Neither side has had any real proof.
THE CONTROVERSY:
Is HDMI capture really at a higher resolution, or merely up converted data?
CLAIMS:
It has been argued by certain people that the HV20 has 1440 horizontal resolution (for HDV) internally and it is merely up converted to 1920 for HDMI as an economy measure on the part of the engineers.
It has been argued by other people that if the CMOS is truly 1920x1080 it would take one set of circuits to down convert it for HDV and another set of circuits to up convert it again for HDMI. That wouldn’t be economical use by the engineers. Instead it is argued that HDMI data is exported before compression to HDV format.
Last edited by Duke; 2008 February 18th at 09:26.
Known Facts
KNOWN FACTS:
Canon advertises the CMOS chip is a true 1920x1080 chip, which is 2 mega pixels. In fact, it’s a 3 mega pixel chip for capturing photos, but it’s masked off to 1920x1080 for video use.
HDV on tape or through Firewire is 1440x1080 resolution with 1.3 pixel ratio and 4:2:0 color space. It is stretched 33% wider to make a 1920 wide picture and one line of color is blurred into the next line because the next line has no color data. That's what the 0 on the end of the color data numbers means. On the first line 2 out of 4 pixels have color data and on the next line 0 out of 4 pixels have color data.
Direct capture of HDMI is 1920x1080 resolution with 1.0 pixel ratio and 4:2:2 color space. Since the pixels are square they are not stretched and each line has its own color data. The color data numbers mean on every line 2 out of 4 pixels have color data.
Last edited by Duke; 2008 July 21st at 09:32. Reason: Added color data info.
The Test
THE TEST:
If the HDMI is up converted data there will be no new data added, only a blurring of the data that exists.
I took an EIA resolution chart and shot HDMI direct capture with the same lights, same tripod, same chart at the exact same time as HD video, then captured the video via Firewire.
Since the images were shot at the same time they are directly comparable. Since both images are 1080 vertical resolution should be identical, only horizontal resolution should vary. Therefore, I focused on the horizontal resolution part of the chart.
As much as possible I tried to select the same areas of the chart to blow up. In most instances you can see the areas that change darkness in both images.
The enlarged areas are blown up 16 times larger. This is based on the premise that each pixel will be 1/4 inch when blown up to project on a theater screen for a movie. This is the only legitimate use for this amount of resolution. You will never notice the difference on a TV or web based video.
It should also be noted that this is supposed to be a black and white image. However, there is no true white paper, no true black ink and no perfect lights. Since white light is a mixture of wavelengths I used HMI, flourecent and halogen lights mixed together. Still there are color issues though I tried to reduce them.
Last edited by Duke; 2008 February 18th at 09:27.
600
At 600 lines of resolution everything looks pretty good. There are gradients in color in the upper right, and both bottom corners of both of the images. You can find the matching data in both and the steps are pretty equivalent.
Last edited by Duke; 2008 February 18th at 02:04.
700
At 700 lines of resolution, at first glance everything looks bad on both images. At first I thought that area of the chart didn't print correctly.
In fact it's bad enough that I checked another 700 area of the chart. (There is more than one area for horizontal resolution on the chart.) That check is the second image. It also looks poor and the HDMI and Firewire images contain the same steps in the image.
However, after researching resolution charts, it turns out it has to do with the width of the line crossing pixels. It’s a known issue and disappears when you go up or down the resolution chart.
Last edited by Duke; 2008 February 18th at 10:24.
800
At 800 lines of resolution things look better in both images, but the HDV image is a little faded and the black lines have lightened up on the HDV image.
Last edited by Duke; 2008 February 18th at 02:05.
900
At 900 lines of resolution the third line has blurred into the surrounding area in HDV. This is the last time that HDV has four distinguishable lines. HDMI still looks crisp and dark. HDV by Firewire has just lost data.
Last edited by Duke; 2008 February 18th at 02:06.
1000
At 1000 lines of resolution the middle two lines in the HDMI image are visible, but losing strength. However, the middle of the HDV image is blown out and would lose many details.
Last edited by Duke; 2008 February 18th at 02:07.
1100
By 1100 lines of resolution the first and fourth lines in the HDMI image are strong, and the second line is visible. However, the third line is barely visible. It has large amounts of blur and bleed-over.
On the other hand only the first line is strong on the HDV image. The middle two are lost and fourth line is actually overpowered by a new dark area that should be light.
Last edited by Duke; 2008 February 18th at 02:08.
Original chart images
Here are the two images of the horizontal resolution area of the chart side by side. These are not blown up or changed, except for being cropped.
To my eye the Firewire capture still looks a tiny bit softer. What do you think?
Duke
Last edited by Duke; 2008 February 18th at 09:32.
Yeh me too..Originally Posted by Duke
Extra data
I've added another image to the 700 section. It's sort of a retest since that area is bad. Same result for both HDV and HDMI.
Me too. HDV looks softer in the last shot.
The HDMI capture looks stronger all the way through. Thanks Duke for this test.
Would it be fair to say that capturing live through HDMI gives your footage more dynamic range? If so then using cinemode would be even more of a plus.
Last edited by Ian-T; 2008 February 18th at 12:02.
No one can read the Gospels without feeling the actual presence of Jesus. His personality pulsates in every word. No myth is filled with such life...Albert Einstein
If your not even sure when comparing pictures side by side than whats the point? Nobody would ever tell the difference.
....but if you want to blow up the picture...then it just slaps you right in the face....
No one can read the Gospels without feeling the actual presence of Jesus. His personality pulsates in every word. No myth is filled with such life...Albert Einstein
I agree. The firewire gamma looks different too. I think there are enough differences in all the comparisons to shoot via the HDMI if you have the kit and reasonable suspicion to do so. OTOH the possible losses via the firewire hookup may not be sufficient to warrant shooting 100% that way, just in case.
Spoiler Alert!
SPOILER ALERT! DON'T READ HERE UNTIL YOU VOTE:
I never said I wasn't sure. I just wanted to give everyone a chance to see for themselves.
If the image has 25% more data and around 1100 HDMI is degrading, you would expect HDV at 1440 to degrade between 800 and 900. Guess what!?
As I said many times it makes no difference if you're only going for video, DVD or web release. However, there are a certain segment here with aspirations as Indie film makers.
We need the cheapest camera that gets the job done. This lets us put money into sound, lights, etc.
The fun thing is most prosumer cameras costing thousands more still can't do this because they don't have HDMI out. My Sony FX1 didn't come close to this picture quality.
Duke
Last edited by Duke; 2008 February 18th at 16:07.
Not directly, but as an indirect consequence that's true. Literally the dynamic range is the same, but 4:2:0 gives color data on two out of four pixels on one line and then sort of averages the data on the next line that has no color data at all.
That blurring of color data reduces the dynamic range of what's there. If you look at two images (I did an image comparison some time back and if I can find it I'll put the link here) the direct capture has brighter colors because the color data hasn't been mixed.
On the other hand 4:2:2 gives twice the color data, period. HDMI is 4:2:2.
Duke
i still think you're totally confusing the numbers with the fact that hdv is compressed. try downsampling the hdmi signal to 1440x1080 4:2:0 and chances are it will look just as good. hdv compression is pretty lousy, and it's for that reason direct hmdi capture is a good idea, not because it may or may not win in the numbers. this is a static shot, but take a look at something in motion and you'll see an even bigger difference.
/matt
mattias, i'm not sure who it is you are responding to but i agree with what you are saying. These static shots only tell us part of the story...but the real evidence comes out when there is motion. I hate that sometimes staticky blurry look you get when panning with HDV....you will not have that problem with HDMI capture.
Hey Duke...this might be a bit of a stretch for you...but if you have time...could you film something with lots of motion via HDMI and captured to tape at the same time so that we could compare and see what the HDV compression is doing in motion shots?
No one can read the Gospels without feeling the actual presence of Jesus. His personality pulsates in every word. No myth is filled with such life...Albert Einstein
Duke - I think you're onto proving something quite controverial and interesting. I was just wondering if you have been able to produce a 1920 video on a BluRay or standard DVD in HD-DVD format yet. I imagine the size increase would allow you to only demo a few minutes at best.
If you are able to take Ian up on his request, could you keep the time of the shot down to where the file size is under a couple hundred meg please.
Maybe you already have somewhere, but I'd like to see a list of hardware and software you have that allows you to shoot & capture HDMI. I think most of us would rule this format out as nothing more than a curiousity unless you're really into experimenting. I could be wrong, but in my opinion, at this point in time, there isn't a good reason, other than passion, to attempt to acquire the equipment needed to shoot this format.
Of couse, as you know, a lot of us 20 owners are passionate.
my point is that hdmi is so much better than hdv on moving shots, i.e. everything but test charts, that the numbers are completely irrelevant. even on static shots it's clear to my eye that hdv compression is the main reason the quality isn't perfect, any references to difference in color sampling and resolution is jumping to conclusions and only acts as a smokescreen, imo.
/matt
Compression
I think it's a combination of everything. 1440 instead of 1920 is a form of compression. 4:2:0 instead of 4:2:2 or 4:4:4 is a form of compression. Long GOP is a form of compression. Those engineers had to use every trick in the book to get it to fit on mini-DV at that data rate.
Even on still images uncompressed is twice as large as lossless compression.
When you look at supposedly solid color pictures in close up on the HDV you really notice the colors. HDV produces banding and sudden color jumps because of compression and color space issues.
However, that being said you only notice these in extreme situations, but that wasn't really the point.
The point is that HDMI makes a difference in resolution and color when you are doing something extreme like blowing it up 16x bigger to project on the big screen. It really doesn't make a difference if it's compression, colorspace, resolution.
Then progressive scan to get rid of the comb effect of interlacing, and other improvements. The POINT is that HDMI capture CAN make a difference if you need the absolute best image that can be seen.
[The flip side is a movie is all about the story and the feelings it evokes, not the pretty pictures.] Of course I'd like to produce both. 28 days later was partially shot on an XL1, and Open Water was partially shot on a digital camera.
Time to use an HV20 on a real movie.
Duke
Posting Permissions
Reply With Quote