This is terrible -- this can't be right!
 

I just received my HV20 and took it out to shoot some test footage. I tried it in auto mode, program mode and "cine." I played back the footage via HDMI into my Toshiba Regza 42" HD television. Bear in mind that my basis for comparison is my standard-definition, 3-ccd Sony VX2000 camcorder.

I was shooting, primarily, trees in a park near my home.

Basically, anything with high-contrast fine detail, e.g. leaves, strong verticals in bark, etc., looked just awful -- displaying the kind of motion and digital artifacts that I associate with Bayer-filtered sensors on inexpensive single CCD standard definition camcorders. These areas were, literally, vibrating with artifacts.

What's going on here? Is this over-sharpening? Is this the best I can expect from a single-ccd machine?

If so, it's going back and I'll wait until I can afford a Sony FX7 or a Canon AH1. The detail and color are fine, but I can't live with this kind of artifacting.

This can't be right -- everyone is raving about this camera. Am I missing something here?

could you post some stills?

I have a Hv20 and have shot leaves and other nature stuff with verticals and seems fine to me (i have an A1 to compare to) in Cine mode there is no sharpening of any sort so should be alot more smoother.

I captured some video into Premiere and exported it to a WMV9 file, thinking I'd post it as an example of what I was seeing. Now, this is weird: on my computer, it looks perfect! No artifacts. (Incidently, my 3 GHz P4 with 1 gig of RAM is barely up to editing HDV -- playback is jerky).

Now, I'm going to try a different test. I'm going to capture to my laptop and Premiere Pro, export to an AVI file and and play it from my laptop direct to my HD television (the laptop's docking station has a DMI port and I've got it connected to the TV with an HDMI to DMI converter).

I too was shocked at the pq. The color and sharpness of the HV20 is great, but pan on a tripod or track moving objects and I was disapointed with the artifacts.

In good light, I tested with a tripod and an open field of grass at the bottom of the frame and sky at the top. The horizon and sky look fine, but the blades of grass smear when you pan. Stay still and it's full color and sharpness. Move and it is a little jumpy and smeared.

I take this to be a result of 2 things. HD pictures have more resolution and that sharpnes can be more obvious when the camera or subject moves. I think the biggest factor is the amount of compression needed to squeeze the HD picture into the mini DV tape bandwidth. Static pictures come through with a great picture, but a moving images show the limitations of how much data can be stored.

I can be all wrong, but it was the conclussion that made sense to me.

I have the same problem when panning. There is tearing and smearing when panning. It doesn't look way too bad as .m2t at 60fps, but most NLEs will resample at 30fps and then it just looks bad, more than it should.

Well, some more mixed results -- my laptop isn't fast enough to play the rendered mpeg file in real time, so I still can't see the results on my HD television. This weekend I'll do some more experiments with sharpening turned off, 24p and component out from the camcorder to the television. We'll see, but I have a feeling this camera is going back.

Okay, I did one more test. I tried the component outputs and found that crawling and vibration is dramatically reduced, almost to the point of vanishing. I'm wondering if this isn't either an interlace issue or a pull-own issue.

Oh well. More experiments later.

Taking the camera back won't solve the problem. Plenty of people are getting fantastic images out of this camera. You need to learn a little bit about the camera.

In Cinemode, you can't actually control your shutter speed, unless you know a bit more about how to do it. In 24p, you want to be at 1/48 period.


Try shooting in TV mode, and lock the camera at 1/48 shutter speed. You will lose the benefits of Cine mode, but gain more control. You can custom set the camera turning everything down to get close to a Cine mode look..

Second thing is that with any 24p camera, you need to extremely careful about quick pans or jittery camera movement. 24p needs a stable shooting platform.

Okay, you understand that I'm not shooting in 24p, right? This is straight 1080i/60, and the motion artifacts that I am seeing have nothing to do with 24p unless, for some reason, my television thinks it needs to do pull-down. Shutter speed differences don't cause motion artifacts.

I appreciate your suggestions, but I don't think they have anything to do with what I'm seeing. By the way, I have no interest in shooting in 24p -- I'm not trying for a film look. I'm quite happy with good, clean hi-def video.

I was comparing my HDR-HC3 to my HV20 with moving video in this thread
http://www.hv20.com/showthread.php?t=3341 about post 9

I said I had taken my HC3 hiking and filmed while walking and the video came out fine, on a different trip I did the same thing with the hv20 and the footage was un-watchable. They were saying it is rolling shutter.

I guess the sony cams are better for hand held stuff and the HV20 is a better tripod cam because it is sharper.

Wow, my mistake, and my apologies. Somehow I got the impression you were shooting in 24p. I guess because you indicated you were using Cine mode.

Okay, there is no doubt this camera is touchier than my Sony's with respect to needing stabilization.... I don't know reason why technically. It definitely isn't a VX2000. Hand held, you have to have the stabilization on or you get some nasty footage. And stabilization should be off on the tripod. I leave it on with the monopod. In any HDV you face motion a issues that are accentuated by the codec, and add to that the additional sharpness of the image, and it all becomes more noticeable. I think my FX1, being heavier, is better stabilized. I also believe when stabilization is off on the Canon, something makes it more sensitive to bumps and sudden moves, than my FX1.

As far as rolling shutter, I have occasionally experienced the jello like wobbly images in pans etc, but that isn't the norm.

But lets face it, for now sub $1,000.00 HD camera, you will not beat the image and sharpness of this camera.

Question: Do you have the same issue with component to your HDTV. And is it possible your settings on the TV are wrong.

Other thing I suggest is actually editing capturing and editing in it to get a true feel for what it does.

Don't dismiss 24p. It has a great feel to impart to your films...and handled right, this camera, at this price is a great entree into the 24p world.... and this is only reason I bought it.

I don't think this as a stabilization issue. I have seen similar motion artifacts from single-ccd Bayer-filter SD camcorders, and also in lower-bandwidth, over-compressed mpeg. "Shimmer" is probably a better description of the effect. If you have digital cable or satellite, you'll occasionally see it even on standard definition channels that are over-compressed to save bandwidth.

I also don't think this is a rolling shutter issue, as it only appears in areas that include high-contrast, small detail movement, i.e. high-frequency data. A rolling shutter artifact should effect any object moving in the same direction, regardless of size.

Well, sure, but that's not really an important factor for me -- I'd rather spend more money and get satisfactory quality. Yes, the Canon has a sharp image, good color saturation and reasonable (though not outstanding) low-light performance. I do travel video, which entails shooting lots of high-frequency detail, e.g. buildings, trees, etc. If it can't reproduce what I shoot well without introducing unacceptable amounts of high-frequency motion artifacts, then it is completely useless to me.

That's an interesting question. I tried component last night and the motion artifacts were considerably ameliorated. It may be because component is "softer" in that it passes less high-frequency detail, or it may be something in my television. My TV can do 3:2 pull-down and it is possible that, on the HDMI input, for some reason it thought it was getting 3:2 material and was trying to do that. My TV may also not be doing a good job of de-interlacing 1080i input (though my satellite receiver is 1080i and the TV handles that fine). Alternatively, the Canon may just have lousy HDMI circuitry.

I've already done that, though only on a small scale. Unfortunately, neither my 3.2 GHz P4 editing computer or my Core 2 Duo laptop seem to be up to the task of playing back HD, so it's difficult to evaluate the output. As best as I can tell, the high-frequency artifacts are there, but not as pronounced as direct-from-the-camera. The more I think about this, the more I think that this may be a Bayer-filter sensor artifact that is amplified by either over-sharpening by the camera, a poorly designed HDMI output on the camera or, possibly, a bad filter capacitor in the camera.

People like 24p because they think it imparts a "film" feel to their video (it takes much more than matching the frame rate of film to provide a true film look, but that's a topic for another thread). I'm not trying to produce low-budget films. I shoot travel video for a very limited audience, i.e. my wife and myself, and those friends and family who are interested in our travels. My goal is to recreate, as closely as possible, the immediacy of the experience. I'm not looking to make something that feels like a feature film but, instead, a faithful documentary of the sights and sounds of the various places we visit. My interest in the HV20 was not its 24p capability, but it's supposedly-superior imaging capability at 1080i. I'm still not convinced that the problem isn't confined to the camera's HDMI port or, possibly, that the particular camera I have is defective. This weekend I'll perform some more extensive tests (including turning down the camera's internal sharpening which can definitely negatively impact high-frequency detail).

Single cmos doesn't have anything to do what your describing. It sounds like your describing mpeg artifacts, but without any screencaps its hard to say. My hv20 produces beautiful images.

It's not the fact that it's a single sensor, but the fact that it uses a Bayer filter that can result in these kinds of artifacts. I've seen in standard definition miniDV camcorders which, of course, don't use mpeg or, for that matter, any form of temporal compression.

Okay, I've posted a frame that exhibits digital artifacts due to what I believe to be oversharpening. The artifacts are more visible if you zoom in a bit on the insets. It's particularly obvious on the cross-hatched palm tree trunks -- light, high-contrast areas have significant "zaggies" on the diagonal lines that are not present on the lower-contrast areas. The sharpening is most visible on the large palm trunk with the vertical striations. This appears to validate my initial impression that the artifact problem was confined to high-frequency detail. It also explains why the artifacts are more visible via HDMI than over component.

http://travelersvideo.com/hv20.jpg

This weekend, I'll experiment with turning down the camera's internal sharpening. However, if it can't be eliminated or, at least, significantly mitigated, I'll probably return the camera.

Looks like good 4:2:0 HDV 1440x1080 HDV footage to me. What do I mean? I see MPEG compression artifacts, and interlace artifacts, and that be normal. This IS an HDV consumer camera, using the commonly accepted GOP technique to capture pseudo-HD. If you notice that on HDMI input, that's good - you're seeing more detail. From what I've read, it'll all get worse with a Sony HC7 as it sharpens the image even more.

I see nothing wrong.

However, if you're not happy with a purchase, return it if that will let you feel better. Be prepared to fork out more dough for a higher end camera, or wait a year or more for a camera to offer more TVL/ph at a similar price point to still get good value on your dollar.

Paul:

Is this a snap taken in the still camera mode, or a frame grab, or what ? It doesn't look like it is in the right aspect ratio to be 16:9 from video.

Sorry, but we're talking apples and oranges. I was interested in this camera because I've read that pros are using it for b-roll and 2nd unit. I didn't expect artifacts as extreme as I've found. I certainly wouldn't consider a Sony HC7 as an acceptable alternative. I'll probably wind up with an FX7 unless turning down the sharpening works.

It's a frame grab pulled into Photoshop and saved as a jpeg.

Ooops -- just looked at it again. I may have pulled it into a Premiere Pro project with the wrong settings. I'll have to look again.

Okay, I looked --

The frame export was correct, but the pixel aspect ratio in Photoshop was wrong.

I was considering the V1U, but the HV20's price and feature set were enough value for me. If you do get an FX7, please share your findings on this thread, and if you can visit the same location as you did with the HV20, that'd be great too.

I've shot extensively with the V1U's older brother/cousin the Z1U and, in my opinion, the Z1U can't hold it's own against the HV20 in terms of image quality/resolution/lack of noise. My personal thoughts were (before the HV20 was announced) if I were going for a larger cam, I'd have gone for the Canon XH-G1 or Sony V1U (which has a much better image processor than the Z1U). Best of luck.

Eh? They have inferior image to the hv20 so what exactly are you looking for? There's nothing wrong with the image you posted. Just basic mpeg-compression. And if you think a 3-ccd camcorder will take that away then think again. The V1 actually has a somewhat worse encoder than the canon cameras in terms of macroblocking. DV-cameras also encode with an mjpeg type codec so they will also feature some crawling effects on borders though not as bad as hdv. Has nothing to do with 1-ccd vs 3 or a bayer-filter.

I'm looking at that screencapture with my 24" Eizo and it looks great. Ofcourse there are compression artifacts if you zoom in 300%, it is hdv anyway.

Absolutely astonishing that an image as beautiful as this doesn't cut it for you in a consumer camera. There has to be something wrong with your Toshiba. I know that some settings in lcd-monitors can increase grain and encoding effects so atleast check your brightness levels etc.

I don't believe the XH-A1 and the FX7 exhibit the same artifact problem.

Then your standards are significantly lower than mine. As I said when I started this thread, high-frequency detail, literally, crawls. The artifacts exhibited in the still that I posted are not static.

No, it is not just basic mpeg-compression. It appears to be a combination of over-sharpening with poor mpeg-compression. I've been transcoding my standard definition video to mpeg for years to burn my own DVDs. Not one of them exhibits this kind of artifact.

Sorry, but you're completely wrong. If, by "DV-cameras" you mean miniDV cameras, they use the DV-25 standard which is a non-temporal compression codec. DV-25 is not mpeg-like, because mpeg is a temporal compression codec and the typical artifacts which affect mpeg are the result of that temporal compression.

Next, Bayer-filters are used only on 1-CCD machines -- A Bayer-filter is a grid of primary color filters that allow a single sensor to reproduce colors. Obviously, there is no need for a Bayer-filter on a 3-ccd machine. Because the Bayer-filter is a grid, it can introduce high-frequency artifacts. I have an old TRV-20 that does this (if you like, I can post frame grabs that demonstrate this) and this was precisely the reason I moved to my VX2000, which is a 3-ccd machine and has virtually no perceptible artifacts.

These are motion artifacts -- that's the problem. They are obvious and distracting without any magnification.

There may, indeed, be a problem with my Toshiba. The fact that there are visible and obvious artifacts in a frame capture suggests otherwise. The question isn't whether the HV20 is a good for a consumer camera. It's whether it is good enough for my applications. The very limited amount of video I shot in my initial test indicates that it is not. It would be nice if it was because I wouldn't have to consider the added cost, weight and size of a prosumer machine. That is, however, exactly what I had to do to get an adequate for my purposes standard definition machine.

I wonder if CMOS versus CCD has something to do with your complaint? I'm much less knowlegeable than you are about video, but I have read many posts about artifacts that seem to come into play on CMOS cameras that do not on CCD's. Especially when it comes to horizontal motion. It's a given to me that you absolutely must pan as slow as possible with CMOS.

The only thing different is the rolling shutter. It seems people don't really now how to handle 24p material or have most of them even seen interlaced material.

... Do you know what mjpeg is? It's basically a jpeg codec such as dv is. It also exhibits jpeg artifacts in images and could potentially be seen as "crawling" effect on certain types of contrasty edges. DV-codec also has artifacts that's why I mentioned it. Offcourse the artifacts will be lower, but still. They are there.

What do you mean even by "you're completely wrong". Completely wrong in what? DV-codec has basic jpeg-artifacts and they could be seen as high frequency crawling on certain edges. What is wrong in that statement may I ask?

I don't buy it. If they are motion artifacts then it can't have anything to do with a bayer-filter. The RED camera also utilises a bayer filter and the results are astonishing. I'm still a believer in one ccd-machines provided the resolution and sensor size is there.

You have to be talking about mpeg-artifacts and if so an XH-A1 or the V1 will DEFINITELY not save you. They have your basic mpeg-artifacts just like in those screencaps.

I'm not shooting 24p. I thought I had made that clear.

Wasn't directed at you. There so many threads where people don't know how interlacing and pulldown work even in the xl h1-forum.

Yes, I do. Mpeg is not, "basically a jpeg codec such as dv." Mpeg compresses by identifying a reference frame and then compressing subsequent frames based on the delta between the reference frames and the frames that follow. DV-25 (and DV-50) only compress within in each frame. The temporal compression of mpeg is what introduces the motion artifacts. Because DV-25 does not use temporal compression, it is not susceptible to this particular form of digital artifacts.

Wrong in suggesting that mpeg and DV-25/50 are the same type of compression that result in the same kind of artifacts. The lack of temporal compression in the latter is an enormous difference between the codecs.

Nothing. However, that's not the statement that resulted in my saying you're completely wrong.

Of course it can. Bayer filters can result in "jaggies" because the grid creates a screen-door effect when resolving sharp, high contrast lines. If I get some time, I'll post some examples from my TRV20.

And there are ways to address the artifacts caused by Bayer filters, both in sensor design and post-sensor processing. The RED camera isn't a consumer machine; the latter is not likely include more expensive design features that can address this.

I'm not talking about mpeg-artifacts, per se, nor has anything you've said persuaded me otherwise. However, as with the difference between my VX2000 and TRV20, I'd expect that a prosumer machine wouldn't exhibit such blatant and distracting artifacts. I haven't seen the output of an XH-A1 yet, so I can't say for certain.

Okey dokey. Sorry.

MJPEG (Motion JPEG), note the J in mJpeg. That is basically a dv codec. Was never comparing it straight to mpeg.

That's it, I'm gonna take some screencaptures from the HV20 and the Sony hc1 (that has the same image quality as the sony a1u) and set you straight. These cameras give an image quality that for example put the dvx100 to shame and its 3 sensors. I find it very hard to believe your vx2000 could keep up. Having seen video from vx2000 a year ago I still remember it being severely inferior to the hc1 (except in low lighting).

Folks, if you mention Motion JPEG in a thread where MPEG is discussed, please point it out by saying "Motion JPEG" at least once... it's easy to confuse mpeg and mjpeg. And by all means, let's keep it friendly at all times, please.

By the way, single-chip RGB can be nearly as good, equal to, or even sometimes better than three-chip color accuracy depending on certain particulars... in short, single-chip RGB is in the same ball-park as three-chip.

Okey, here we go:

http://hmcindie.pp.fi/hc1/

I have uploaded some hc1, hv20 and dvx100 screengrabs.

All of the hv20 screengrabs are straight from the camera (no aspect ratio change) without any modifications. Shot in progressive & cinemode (25p)

HC1 pictures have been deinterlaced and aspect ratio corrected. DVX100 shot progressive and hc1 and dvx100 have been scaled to 1280x720 (hc1) & 1280x960 (dvx100) to see upfront the resolution difference.

Whitebalance is really off on the dvx100 but try not to be bothered by that.

Anyhoo, I just can't see any sd-camera coming close to the hv20. Yes, there are mpeg artifacts when you look closely but in motion they aren't that noticeable. I consider the rolling shutter artifacts to be worse (for example picture hv20 5).

Paul,


Sounds like some of the problems watching on your TV may be the TV itself. If it looked better with component input rather HDMI, that sounds like the TV's pulldown etc.. Also, you said it looked better on your computer, but the computer can't handle the compression, so hard to really evaluate.

But, since you already have it in your head that the camera is not living up to your expectations, it would probably be better to send it back and wait for another. You will always be looking for problems, whether they are there or not.

Best of luck----Mike

I'm with Mike on this one. You've not found the HV20 up to your expectations so it's out the game. I agree with others here.. you're looking at HDV on an HDMI cable - you're going to see the compression of a high-res image - every little bit of it. An HVX200 wouldn't be safe when being viewed over HDMI! Or.. perhaps it would.. because it's sensor has less resolution than the HV20 so there'd be less to look at over HDMI since the image would be softer than an HV20's... ;)

Suggestion: rent/borrow another HDV camera with HDMI and test using the same circumstances. Again, for this test, have HDMI and HDV to be fair to compare agianst the HV20 that hasn't made the cut. Viewing recorded HDV over Component will not suffice.

Seriously though, an XH-A1 or FX7 have the potential to have even more resolution than an HV20.. and they're still going to be compressed with HDV - which will lead to similar results. I don't see how they can withstand this 'terrible looking image' test. Apples and oranges indeed.

Please share your findings as others have.

I still say Paul should look at this stuff on a different monitor before he throws in the towel, but it is up to him of course. Just in discussions about how the image looks off the computer, and component cables, vs. HDMI directly fed to the Toshiba, he seems to experience different results. This makes me wonder if something is up in the HDMI import process. I do use DVI converted to HDMI as a second monitor cable to my cheapy 32" inch Magnovox HDTV, and I note differences in the out put from camera component, versus output via the HDMI.

The only thing I have to compare with is the FX1, the VX2000, and the HV20. The HV20 runs rings around the VX2000 when you output it to SD. The FX1, with the right adjustments, and the HV20 will run neck and neck generally, with the FX1 3 chips giving better performance in lower light situations. That is my experience.

It may be a problem with your TV indeed. After all, Toshiba's flagship HD-DVD player does not use in-house hardware/software, it uses Silicon Optix chip for deinterlacing and noise reduction.
There are no frames in the interlaced mode, there are just fields. The frame grab you have is made by your NLE. The jaggies on the left bottom image are interlace jaggies that your NLE could not remove because it is not smart enough. Apparently, your TV is not smart enough either. Try connecting the camera to the Pioneer Elite or to a Hitachi Director series TV and see whether you notice the difference.

You were advised to use 24p. The point is not just film look, you also will get proper progressive frames if you use right software to extract them. Or, you can watch this telecined video directly on the TV. If the TV is smart enough it will perform IVTC an will display clear frames without any jaggies whatsoever. You can go to Silicon Optix website and order their testing DVD for $20 to verify how good (or bad) is your TV.

The compression artifacts don't seem ostensible to my unprofessional eye even on the magnified parts of your frame.
Well, if you don't like motion artifacts and don't want to shoot in 24p then buy a better TV. Or buy a camcorder that shoots in 60p like a JVC, but to watch this stuff you will need a TV with 1080p input and AFAIK Regza cannot take that, it takes 1080i only. Umm, maybe it can take 720p, I am not sure.

Sorry if this has already been mentioned, but do you see any improvement when dialing down the Sharpness setting in the camera?

You hit the nail on the head. I did some more tests earlier this morning. I turned down the camera's sharpness and shot some high-frequency subjects (leaves, fences, etc.). I tried both 24p and 1080i.

With the sharpness turned down, the problem was ameliorated dramatically, but still present. After pondering the visiible difference in the HV20 output between component input and HDMI input on my TV, I turned off all sharpening in the television. The problem is gone!

There are still some minor mpeg artifacts, but only if I look really hard to see them. The "crawling," and "shimmering" is gone. Interestingly, it was still present, though lessened, when I viewed the footage I shot earlier this week with the camera's sharpening set to "medium."

When I get a new television, I always adjust it from the out-of-the-box settings by, among other things, turning off sharpening. I was using a spare input on my Toshiba, so it was still set to the out-of-the-box default, which included too much sharpening.

I want to run some more tests later. It's cloudy and gray this morning, so I want to confirm that camera does okay in bright sunlight -- I suspect, though, that the problem was the sharpening and the camera will do fine and I'll be keeping it, and selling my VX2000.

Now that the motion artifact problem is resolved I do like this camera. I have some minor quibbles with the physical design, e.g. the lack of good attach points for a neck strap and an awkwardly-placed start/stop switch, and I'd prefer better low-light performance. These, however, are not enough to override the appeal of a high-quality high-def image.

Incidentally, I can see the appeal of 24p. I have no use for it for the kind of shooting that I do, but I can see why those who want to try their hand at film production would want to try this camera.

"Incidentally, I can see the appeal of 24p. I have no use for it for the kind of shooting that I do, but I can see why those who want to try their hand at film production would want to try this camera."

One thing you might try is to tape, edit, and burn to DVD in 24p mode. results are great. I'm thinking of getting one of these canon HV30 cameras as a knock about danger cam to go with my JVC HD110. 24p 1/48th of a second to 24p DVD looks great.. and yes... as long as you turn the Detail (sharpness) to off... if you want any added sharpness (edge enhancement/sharpening) add it as a filter in your NLE. I even do sports in 24p now. Looks killer on LCD and plasmas on anamorphic DVD. just something to to try if you haven't.

I'm astonished at this conversation as an onlooker. This is an argument over a CONSUMER CAMERA! This is embarrasingly too small and best buy looking to use in any kind of professional environment as a primary cam. This is an HDV camera that I bought for $699 and am planning on using as a deck. It is a great little cam but no one should mistake for professional especially if they have a lot of experience and knowledge. It's amazing that it does as well as it does. Its competition in the SD land is not a DVX or a VX2000 it is a Elura 100 or a Panasonic GS120. Jeez.
Bill

Bill,

You're lucky to get the cam for only $699. Most paid $900-1100! But more importantly, is the fact that shooters are using this cam because of what is packed into the small form factor. Students are using it because it lowers the price of entry.

It's an incredible vacation cam. Last spring, I asked a friend what cam I should take on a trip to Europe. He asked me if I was going with my wife or the cam... I took the HV20 on a monopod and had a fabulous trip with my wife:-)

Even if it is consumer looking, the accessory product market for this cam shows it's versatility. The HV30 will only continue the legacy of a great little cam. It will only be supplanted by an HD version of the GL2 with full manual and a larger lens. It may even sport SD card recording...