I have to admit I did not buy my FX1 for the HDV, I brought it to use as a 16:9 dv camera and because it was a vast upgrade over the PD-170. I should add most of what I shoot is news and current affairs stuff so I have been rather suspicious of the effects of heavy compression. I've been reading this forum regularly and have become aware of many apparent benefits of shooting in HDV then downconverting in-camera for use as SD DV. It is said here often by credible people who do know what they are talking about that this procedure creates an image that has less noise and more clarity. I have now carried out a number of experiments, shooting a shot in DV and then the same shot in HDV, which is then downconverted incamera. I have been comparing the shots to see if I can notice a difference, I've been blowing them up to examine detail and simply putting them side by side. Now the compression is not a huge problem, but I can't see a grain of difference in the actual quality. I don't wish to disparage the opinions of those who know more than me here, but is there some kind of experiment I can do to be able to SEE the difference. I want to be convinced there's a basic quality advantage to shooting SD in HDV?

All I can offer is the quick test that I did which convinced me that the downconverted footage was better. I shot a few minutes of the same scene locked down on a tripod in both DV and HDV modes. Then I captured both clips as DV using FCP.

If you use a PC the details will be a little different, but the technique should be pretty much the same. I put each clip on a separate layer in the timeline. On clip A I used FCP's crop parameter in the motion tab of the viewer to cut the left side of the image out. On clip B I did the same thing. Then I dragged clip B to the left edge of the frame.

The result was a screen split down the center which had the right half of clip A on the right side and the right half of clip B on the left side. So in other words, I could watch half of clip A and the same half of clip B simultaneously. In looking at still frames this way I could see a slight improvement in the HDV clip, but not a lot. However when I viewed the two clips in motion simultaneously the difference was much more apparent.

Try this sort of test and see what you think. Be sure that you shoot in full manual mode and that the shutter speed, iris and gain settings are identical in both clips.

Haven't tried it myself on the Z1, but the BBC did extensive testing and came to the conclusion that there was no quality advantage to be had. With the possibility of dropouts being higher with HDV I resolved to shoot SD unless I actually needed HD.

I'm just *guessing* here that the quality difference when shootiing in PAL (50i) mode may not be so noticeable as it is in NTSC (60i) mode since you're working with 96 more scan lines to start with.

Whenever I have a question about this sort of thing I like to read everything I can find on the topic. However, if I really want to know the answer, and if there's a simple experiment which can be devised, then I do a test myself.

With all due respect Graeme, John asked "is there some kind of experiment I can do to be able to SEE the difference". He may very well conclude that there isn't any advantage to shooting in HDV mode, but that will be based on his own observation and not something he read on the internet...

... ... came to the conclusion that there was no quality advantage to be had. With the possibility of dropouts being higher with HDV I resolved to shoot SD unless I actually needed HD.

I agree Graeme,

I asked a question earlier about switching tapes between DigitalMaster (when shooting HDV) and Sony Premium (when shooting DVCAM) as I have the same thinking on dropouts. Don't want a 15 frame freeze at the HDV supply side causing same freeze on a DV master/edit.

My question never got answered, but what do you think? Do you still use the DigitalMaster tape for SD?? Would switching tapes be a bad thing? Usually taboo, but DigiMaster is rather expensive for olde worlde DV recording. :)

I am about to do similar HDV/DV tests to most likely (it seems) confirm sticking with SD on some projects. It also lowers the load on my Z1 with less need for running tapes just to downconvert.

Best regards,
Richard

Boyd - you may be correct about the PAL vs NTSC comparison, hadn't thought of that.

Richard - I have run about 100 hours of SD through my two Z1's using exclusively Sony Premium mini-dv casettes. So far I've had no issues, but I haven't tried them on HDV. Haven't ever used anything else in the camera, and I am wary of doing so (whether it's justified or not I have no idea)

Yep I am not seeing the difference as yet. It could well be exclusive to NTSC. Then again there is the issue of downconverts via software. Douglas Spotted Eagle swears Vegas makes a difference, but I would not want to buy Vegas just try this out on PAL. It would be interesting to see if Vegas can make a noticable difference to PAL footage, anyone here got a line on this? The BBC say expensive tape decks will make a difference as well, but I have yet to meet anyone here in London who is advertising this service, which is what normally happens with this kind of thing, so maybe there's little benefit to be had from PAL HDV either way.

Find me a place to upload a 90 meg file, you can clearly, easily see the difference between the two. I'm still shocked over the BBC allegedly not seeing a difference between the two. It's clear as night and day. Anyone who attended the HDV Solutions tour saw this clearly.
It's a file that we downsampled in-camera, and downsampled in Vegas, and there is not just a small difference, it's a very large difference.

I cant do a ninety meg file, but I have done Boyd's experiment and the results are that HDV looks significantly worse in terms of detail, can anyone here host a simple JPEG screen grab for me and you can judge for yourselves..?

DSE, I can host the file for short while...Email me at Sean at SeanLee dot tv and I can give you access.

John - it helps put your questions in perspective if we know where in the world you are; I had no way to know you were in a PAL country but suspected this might be a factor. Maybe you could take a minute to click on the Controls link above, then Edit Profile and add your location for future reference?

The test I did was brief and certainly not very scientific. I'll try to find some time later today to do another NTSC test, and maybe give it a shot in PAL mode as well.

One thing is for sure though. When playing around I switched the component output to 480p with the camera in 1080i mode. Wow, that produced an image on the monitor which was a huge improvement over DV. In fact, on my 1280x768 17" Sony 16:9 LCD monitor I could hardly tell the difference between 1080i and 480p. Unfortunately you can't send 480p over firewire, only out through the component cable.

My FX1 component output only has 1080i and 576i. I take it the Z1 does 480p then?

Ps I mailed my comparison pix to Sean Lee, I hope he's ok with that. Maybe he'll put them up for us to compare.

There is another issue here as well. While downconverting in the NLE may look better, or may not, there is still the advantage of shooting HDV in that you can zoom in on it a lot and still retain great quality since the original frame size is so much larger to begin with. Panning around a large frame within a standard definition project has really allowed me to pull off some tricks that would have ordinarily meant that I needed to just live with a wide shot.

Picture shooting a choir from a wide angle, and then being able to pan around to get some really nice closeups after the fact.

John, this whole downconverting thing is what i've been hooked on for a while, and finally I think i've found A solution, AND YES there is a difference firstly the BBC report DID say there was a quality difference BUT it was only achieved when using a expensive Snell&Wilcox Ukon, and thats what got me motivated to do these tests to see if I too could achive some good downconversion.

Cutting to the chase...

If you downconvert to DV25 , don't bother DV25 is a horrible format (kiddin).

Since the final output is gonna be SD DVD, do this:

(1) Apply sharping filter to HD source + any Colour correction
(2) Render out 720*576 uncompressed (now someone said that this file will have full 4:4:4?) .avi or quicktime (I use MJPEG, its easy on my Harddrive) this quicktime OR avi will be fantastic, if viewed from a distance its almost like the original HD source!

(3) Encode your DVD from this file!

Now I dont have any files that are downconverted via camera to compare, it could be that the camera does a better job then my method.

P.S I've just had another brainwave!!, I've watched footage from my laptop via s-video to SD TV, and my screen was set to 1280*768 and I played the .m2t files at half size (to fit in my screen) (using VLC) and this image has looked stunning, absolulety stunning, COULD IT be that a simple laptop TV output has a very high quality scaler that I could record this via say a DV passthough I could potentionally get results as good as a Snell&Wilcox Ukon???

My FX1 component output only has 1080i and 576i. I take it the Z1 does 480p then?

Yes the Z1 does 1080/60i, 1080/50i, 480i, 576i, 480p and 576p. This makes me wonder if there's something different about how the different models downconvert HDV to DV as well...

Interesting. There might indeed be a difference with the Z1 vs FX1. It would be good to test this, it's been discussed before without conclusion here.




Anhar - I know what you're getting at there. i've seen that method being posted somewhere before. i think there are other steps involved as well. But we're simply interested in camera conversion direct to DV sd.

Yes the Z1 does 1080/60i, 1080/50i, 480i, 576i, 480p and 576p. This makes me wonder if there's something different about how the different models downconvert HDV to DV as well...

And what about the capabilities and settings of the display device that the stuff is being viewed on?

We all assume it's being viewed on correctly set-up and appropriate display devices.

Now I'm sure I'll get flamed badly for mentioning it; by those who are offended that I have, but for those who are all excited about what HDV can do for them - whether it be downconverting to SD for superior results from the higher rez image, or getting 16:9 or whatever... The results aren't going to magically appear on the display stuff you got for SD material.

What I am wondering is that in this day and age who wants to downconvert HDV video ? When I watch the news I want to see the news in high definition not crummy standard definition. If you downconvert HDV you might as well shoot in black and white. When color cameras where first introduced did people ask how good where these color cameras for producing black and white footage ? When motion picture cameras were first invented did people wonder how good a still picture these cameras could make? What amazes me is that since the 1950's we are still stuck with the same 480 lines of resolution. Had there not been a great depression television would have boomed in the 1930s and we would have had only 300 lines of resolution as standard definition for the next 100 years. During these last 50 years there has been many oppurtunities for picture quality improvement but the increases in bandwidth went to more programming choices not to any improvement in picture quality. Most people vehemenently resist change so high definition no matter how good it is will always face fierce opposition by people who are unwilling to change.

I shoot and edit HDV. I show HDV to the majority of the people who see my videos. But now and then, I need to export to a standard DVD. It happens. It is annoying, but some people just don't have HDTV yet, so recommending a AVeL Linkplayer2 just isn't an option.

I shot a couple tests and here are my first impressions. These were exported from the FCP Timeline with Quicktime conversion as JPEG's at highest quality. The only alteration to the images was stretching the PAL to 1024x576 and the NTSC to 854x480.

NTSC Test: http://www.greenmist.com/hdv/NTSC-lake.jpg

If I enlarge to 200% in Photoshop I can see a definite improvement in the downsampled HDV compared to straight DVCAM clip. Take a look at the treeline against the sky and the dead twigs, they are sharper. Not a big difference, but it's there. Also notice the color is a bit different, but I didn't change any settings. This might just be due to rapidly changing light around sunset, but they were shot within one minute of each other. Playing the clip in motion makes the difference in sharpness pretty noticeable to me.

PAL Test: http://www.greenmist.com/hdv/PAL-lake.jpg

These two are much closer together, as I suspected. I still think the HDV downconverted image is slightly better, but the difference is subtle. Playing the clip in motion seems to favor the downconverted version to my eyes. But I can understand why the BBC didn't think the difference was significant in PAL mode.

I don't know why you are seeing inferior results with the downsampled HDV John, I certainly don't see it here. Perhaps the FX1 and Z1 really do use a different approach to downconversion?

As far as how I viewed these (aside from Photoshop), I have a Samsung 22" HD LCD monitor (1280x720) which I connected to the camera's component output. Looking at the native HDV clip compared to any of the DV clips shows a huge difference in resolution. After watching in HDV the DV clip looks out of focus. This was a good test for me since I enjoy shooting scenery like this. It helps convince me that I need to gear up to start working directly with HDV and forget about all this downconversion stuff :-)

I also watched the clips in the FCP canvas at 100% on my Apple 23" Cinema Display. With all the switching between 50i, 60i, DV and HDV I was having a hard time keeping the camera and FCP settings correct, and I started to have problems monitoring over firewire. It's getting late, so I'll take a closer look at all this tomorrow, and will try to put some other examples online.

Steven has it like most folks do. There are very limited choices as of right now, for HD to the masses. Avelink, Buffalo, Windows Media HD, Nero Digital/AVC. So, a lot of folks have no choice but to downsample.

Oh, one other curious thing. Look closely at the top of the downconverted images. The NTSC one has a black line at the top of the frame. The PAL one has a black line at the top, a normal line, then a white line below that.

Interesting...

Boyd, can you host a couple of JPEGS for me from my test? You'll see my results, I was looking for sharpness of detail? Like what you've shown here there is a subtle difference in the two, I am interested to see what others think?

... ... Like what you've shown here there is a subtle difference in the two, I am interested to see what others think?

John, Boyd,

I am confused! All images (the four halves) look the same to me! I see the same ripples on the water. Exact same tree/twig formations. So how can they be shot at different times?

Have you stitched same frames together in error? Or am I doing something incredibly stupid. :)

Just back from my own testing of HDV vs DV. My method was to shoot several scenes, with and without motion, using HDV record for a minute, then switching to DV SP (to keep tape speed same) for another minute. Pans and close ups. Buildings, buses, and park lake/trees.

To view results, I simply attached an s-video lead from Z1 to my Sony SD widescreen monitor and rolled the tape. With my commentary and the Z1 switching display, I focussed on looking for artifacts and fine detail in the resultant SD signal that switched between HDV and DV recorded footage.

Result: The HDV recorded images via s-video are definitely clearer than the DV recorded images. So much so, I have to agree with Spot's comments. Fine detail like Chinese Banyan tree hanging roots from afar are clear in HDV->DV and woolly in DV. Fine shadows of the roots on a bridge are visible in HDV and absent in DV. And so on. Motion artifacts appear the same. Nothing abnormal, and nothing to worry about.

So, I will be sticking with HDV recording (DigitalMaster tape) and getting myself an HDV player as soon as I can afford one. :)

I've done a few experiments on the subject.

DV vs HDV -> DV (in camera or otherwise)

Shooting DV in camera results in the uncompressed original data being downsampled and compressed via the DV codec. For producing "DV footage", this result should be the best, because the MPEG-2 compression is avoided entirely. The key here is that this statement applies to "DV footage" only... not "SD" footage. This is difficult to test perfectly without two cameras in perfect alignment and a perfect beam splitter. Since "perfect" is a bit of a stretch, then any error in the experiment would likely over-ride the conclusions.

In my opinion, if you want DV footage out - you should shoot HDV and down-convert in camera to DV. While there might not be a significant gain in image quality, you at least have a higher quality master for later.

HDV -> 4:4:4 uncompressed SD vs HDV -> DV

Converting HDV via software tools to 4:4:4 yeilds the best result, short of recording an uncompressed HD signal directly. The difference between the 4:4:4 colourspace of a software downsampled image and the 4:1:1 DV is quite noticable. While it's difficult to detect without knowing what you're looking for, watching the 4:4:4 SD play it's immediately obvious which is better. Furthermore, the added DV compression, appearing in the form of mosquito noise and other artifacts is apparant with inspection.

An demonstration of software-vs-in-camera down conversion is given below. Note that differences in colour are due to the handling of the different YUV colourspaces of DV and HDV (YUV.6xx vs. YUV.7xx):
http://s94963366.onlinehome.us/HDRFX1/softvincam.bmp (2 MB)

-Steve

Not sure what you mean Richard? I shot each clip for 60 seconds, one after the other. I kept the exposure and everything else the same. There was no wind so the water was still. I don't think you're seeing any ripples in the water, it was calm. Those are lily pads in the distance and some sort of other "stuff" in the foreground. Nope, these are definitely the correct frames stitched together. These are static shots - your test using motion is a good idea though.

I enlarged a section of each using photoshop to help compare:

http://www.greenmist.com/hdv/NTSC-lake-200.jpg
http://www.greenmist.com/hdv/PAL-lake-200.jpg

There really isn't a huge difference in the NTSC clips, but I think the downconverted one is a little better. When I play them in motion this is noticeable. The differences in the PAL clips are pretty subtle. However if you look at both of those links side by side it definitely shows the advantage of PAL vs NTSC in terms of resolution. You can see more twigs in the PAL version.

John - sure, send me your clips and I'll put them online.

That's interesting Steven. I don't see much difference in your two examples but as you say, it probably shows up better in motion. There's noticeable interlace in these - were they handheld maybe? It would be interesting to compare these with a frame grab from the component output in 480p mode. I wish I had a way to capture from component because I see a big quality difference shooting HDV and downconverting the component to either 480p or 576p.

I'm also unclear as to whether they was any final determination about the component output. I was assuming that it's HDV compressed when set to 1080i, but DSE and others have implied that you get superior results when capturing uncompressed HD from component.

Here's my experiment with a PAL FX1 using in-camera convert. In both cases the bottom half is HDV, and the top is DV. What do you guys think?


http://www.greenmist.com/hdv/poore/hdv-dv%20testing.jpg
http://www.greenmist.com/hdv/poore/hdv-dv.jpg



And my thanks to Boyd for kindly hosting this.

Well the differences are subtle, and maybe a little inconsistent between different parts of the images. With this sort of complex texture I suspect that the MPEG compression might be taking its toll on the HDV image. In the second link I can see a lot of "blockiness" in the lower (HDV) part.

The only way you really can compare footage and analize the quality of a video image is using a professional industrial critical monitor. Using web images or consumer even plasma monitors to do so does not make much sense. People are getting sloppy when analizing things these days. About the downconverted footage being superior to SD DV footage, it is cut and clear. Just get an 8045Q monitor or larger (rent one if you dont have one) and observe the difference.

My two cents.

The only way you really can compare footage and analize the quality of a video image is using a professional industrial critical monitor. Using web images or consumer even plasma monitors to do so does not make much sense. People are getting sloppy when analizing things these days. About the downconverted footage being superior to SD DV footage, it is cut and clear. Just get an 8045Q monitor or larger (rent one if you dont have one) and observe the difference.

My two cents.

I so wish I could afford an 8045....what a great monitor.
We've got two Premiere 234's on the LCD side, and a BVM9 with HD/SDI. It's not optimal, but it is a nice little HD monitor, and it's a loaner so we can get by until we can afford a better CRT. We also have a WUXGA projector and an 11" screen, which is where we do most of our test viewing.

The only way you really can compare footage and analize the quality of a video image is using a professional industrial critical monitor.

Well of course it's hard to dispute this, but for my purposes comparing two images on the same monitor teaches me something, which is why I did it. Draw your own conclusions, or ignore it completely, I'm just sharing what I did.

Most of my work has been done with the goal of projecting it on 40' or wider screens with 10,000 lumen projectors. Subjectively, I think my LCD panels have served me very well for this. If you want to broadcast your footage then it probably calls for something else; that's not something I have any experience with.

For the last project I did in Buenos Aires I wasn't happy with the way the crew setup the projectors initially, but when we had a few minutes I played the Quicktime movie fullscreen on my Powerbook while we adjusted the projectors and we got it to match pretty well. Now the Powerbook screen is not a production monitor by any stretch of the imagination, but it was what I had to work with there, so it made my job much easier to adjust the projectors to it instead of the other way around.

As much as I love quality you kind of have to ask yourself if it would really make a difference testing this stuff on that high end of a monitor. Very few if any people will ever actually watch your video on that type of monitor and while it would be great to get a great looking image you also need to find out what works on a consumer monitor since consumers will be watching it.

Sometimes video can look great in my studio but if the client watches the video on a not so good TV then the video may actually look bad. It is always good to make sure your video will look good on a lower end TV as well.

Using a decent level TV can give you good test results if all of the other variables are the same. That TV may not tell you exactly how the image will look for everyone but it will tell you if there is a difference between A and B. If that TV happens to be a consumer grade TV well then chances are that is how it will look for your clients as well.

As much as I love quality you kind of have to ask yourself if it would really make a difference testing this stuff on that high end of a monitor. Very few if any people will ever actually watch your video on that type of monitor and while it would be great to get a great looking image you also need to find out what works on a consumer monitor since consumers will be watching it.
.
To a limited extent, I agree w/you, because a lot of consumers don't have comparable sets.
That said, even on my 99.00 Sanyo monitor, I can clearly see a diff between HDV downsampled and/or zoomed in SD, and video acquired in SD/DV.

One trap that I see/hear a lot of editors fall into is "The consumer doesn't have a great monitoring setup whether it's display or audio, so I don't need one either." That's a scary place to be. We're visual and aural surgeons, and need those microscopes to determine exactly what is in our content, or else someone will trip us up. These days, the range of what consumers have will go from extreme top end to extreme bottom end. Aim for the top, and feel good if you fall only a little short of the goal, rather than aiming lower, and feeling good because you fell into the 5 ring instead of the 1 ring on the target.
Did I mix metaphors or what?!

That's interesting Steven. I don't see much difference in your two examples but as you say, it probably shows up better in motion. There's noticeable interlace in these - were they handheld maybe?

It's actually from a pan. There is indeed noticable interlace, as handling the fields properly in down-conversion was (at the time) as much of an issue as getting anything else right. I had posted this image a while back, as well as a description. The 4:1:1 is very visible when looking at the green blades of grass against the low-chroma background. First generation DV mosquito noise isn't terrible, but it is noticable on the edges of the goose's neck.

I see a big quality difference shooting HDV and downconverting the component to either 480p or 576p.

I don't buy it. In theory it is always better to work directly from the original digital source... the only "benefits" I can imagine from capturing analog components after HDV compression are the analog smoothing and noise giving the impression of improved image quality.

DSE and others have implied that you get superior results when capturing uncompressed HD from component.

I still haven't really seen this tested. It is widely expected to be true, but I don't know that anyone has the hardware, and has actually done the capture this way.

-Steve

It's actually from a pan. There is indeed noticable interlace, as handling the fields properly in down-conversion was (at the time) as much of an issue as getting anything else right. I had posted this image a while back, as well as a description. The 4:1:1 is very visible when looking at the green blades of grass against the low-chroma background. First generation DV mosquito noise isn't terrible, but it is noticable on the edges of the goose's neck.



I don't buy it. In theory it is always better to work directly from the original digital source... the only "benefits" I can imagine from capturing analog components after HDV compression are the analog smoothing and noise giving the impression of improved image quality.



I still haven't really seen this tested. It is widely expected to be true, but I don't know that anyone has the hardware, and has actually done the capture this way.

-Steve

It's very true, all you need is a Decklink HD, Kona, or Bluefish card and you can test this for yourself. Of course, your app needs to be HD-capable.
For those that will be at WEVA, I'll likely have my personal machine with a BlackMagic card in it, and you can see it yourself. (I'm expecting that the machine provided won't be up to par, and so anticipate carrying my own machine)

I don't buy it. In theory it is always better to work directly from the original digital source...

I certainly agree with the theory. Do you have a Z1? If so, hook it up to a monitor, set the camera to HDV with component downconvert to 480p. Then watch as you switch component back to 480i. The difference is striking, it looks almost the same as 1080i on my 17" 1280x768 Sony widescreen LCD.

I think what's happening is that that 480i downconversion is doing some vertical blurring to prevent thin horizontal lines from flickering (which is common for most interlaced cameras). But in 480p mode it probably doesn't do that.

The 480p probably looks significantly better because it's 480p60 - and hence twice the resolution of a 480i down-convert. I only have an FX1, so that option isn't available to me. However, it's easy to convert 1080i60 to 480p60 in post... and I expect this will yield better results than capturing along components as well. You can expect however to have to do some optimization to first match, and then exceed the results the Z1 acheives in hardware.

-Steve

And why, oh why is there this insistance on using JPEG to show off still images of how HDV to DV/SD compares?

I know you all are going to rail-on about bandwidth and the size of the image if you use un-compressed, but there are compressed alternatives that are superior for displaying this still image stuff from a compressed video source.

Don't forget that you are using a compressed source when using HDV in any form. So regardless of whether you transcode HDV video to an uncompressed format, it remains video from a compressed source... Now, when you go ahead and JPEG any still - let alone the fact that it's also come from an interlaced stream so if you don't do the de-interlace properly... well - you're just re-compressing the compression all over again. And I'd bet that when the JPEG is made that the quality setting aimed for is medium; so even more destruction occurs.

There are formats such as TIFF LZW and PNG that are preferable (not perfect) for providing compression algorithms that are less destructive for your purposes than JPEG. Better still, if you want the best possible quality for comparing still frames - don't use any compression at all on your comparison images.

...There are formats such as TIFF LZW and PNG that are preferable (not perfect) for providing compression algorithms that are less destructive for your purposes than JPEG. Better still, if you want the best possible quality for comparing still frames - don't use any compression at all on your comparison images.

Correction: TIFF LZW and PNG are mathematically lossless compressions these should be used (particular PNG -- as more tools decode it) in favor of raw format like BMP (which is a waste of bandwidth.) Use PNG as is indentical to the raw source.

David Newman
CTO, CineForm