HV20 not really 24p??? Is this guy right?

[c.g._eads]

I was set on getting this camera then started reading a lot of bad experiences about the camera. The main thing is the "wobble" problem. Since I planned on shooting a feature with a lot of shots in the car, it's basically eliminated the camera as an option. Then I also found this. Is this true?

http://reviews.cnet.com/digital-camcorders/canon-hv20-mini-dv/4864-6500_7-32172625.html?ctype=msgid&messageSiteID=7&messageID=2463982&cval=2463982&tag=uolst

[vsansal]

I was set on getting this camera then started reading a lot of bad experiences about the camera. The main thing is the "wobble" problem. Since I planned on shooting a feature with a lot of shots in the car, it's basically eliminated the camera as an option. Then I also found this. Is this true?

http://reviews.cnet.com/digital-camcorders/canon-hv20-mini-dv/4864-6500_7-32172625.html?ctype=msgid&messageSiteID=7&messageID=2463982&cval=2463982&tag=uolst

HV20 records 24p with 2:3 pulldown inside a 60i signal but it is captured as 24p and after removing the pulldown you will get your 24p back. It is not an issue of HV20 only. Other camcorders also use the same method. So the statement is not true HV20 has real 24p recording mode. If you look at the PAL model of HV20, it can record 25p and since it needs to be recorded inside an 50i (PAL) signal no pulldown needed. It is true progressive.

[Matt Agnello]

I've seen a couple forum posts on other forums about this. According to those posts, the HV20 does record 24p in a 60i data stream using 2:3 pulldown. This normally wouldn't be a problem except that the HV20 does not flag the extra frames, so you need software to figure out which frames are the extra ones. Using third party software, it is possible to remove the extra frames (or extract the 24p signal from the 60i signal), but because of the HDV compression method, there are slight differences in the interlaced frames that can create image distortion when recombined. However, as far as I know, the image distortion is so slight that the human eye is barely capable of seeing it.

I just purchased my HV20 recently and have yet to establish a workflow for extracting 24p footage, but as far as I know it is possible using FCP and may be possible using other editing programs or 2:3 pulldown programs designed to for this kind of thing.

Here's a post (http://www.dvxuser.com/V6/showthread.php?t=102170&highlight=pulldown) I found about doing extraction without FCP.

Here's one for doing extraction with FCS 2 (http://www.dvxuser.com/V6/showthread.php?t=98411&highlight=pulldown).

Edit: I want to note that this is all second hand knowledge. I haven't done any of this myself yet.

[Barry_Green]

That (original post) guy is almost completely wrong. (there's a reason I say "almost" and it's the same reason Matt Agnello points to).

The HV20 is absolutely, unquestionably, without a doubt true genuine 24p. It operates at 24fps, exactly like it should. It does NOT do a "fake" 24p like the Sony FX1/Z1 did, it's not some hack-job mishmash created from interlaced footage. It's genuine 24p.

It's recorded using 2:3 pulldown, just like almost every other 24p camera on the market uses. That is not an issue.

What *is* an issue, and the reason I say "almost", is that Canon didn't put in frame/field flags to allow for easy extraction in post. That's a disappointment; it means that if you want to strip out the pulldown and access the pure raw 24p data you have to go shot-by-shot (because the cadence can change for every shot) and strip out the pulldown; there certainly should be automated tools that can do this as well. I'm pretty sure I've seen a script for VirtualDub that does this. But with that said, keep in mind that we edited 24p-within-60i for years and years; that's the way it always used to be done. Before about 2004 there were basically no editors that could edit raw 24p anyway; so anything that'd ever been shot on film was telecine'd using 2:3 pulldown to 60i, and edited as 60i. It's really no big deal.

So yes, the guy who posted that is off base. It is a 24p camcorder using the exact same technique that video broadcasting has used since day one (remember in the early days of broadcasting there were no such things as "video cameras"; footage was shot on 24fps film and broadcast as 60i using 2:3 pulldown.) Every Hollywood movie and every big-budget drama you watch on network television is shot at 24fps and broadcast as 60i using 2:3 pulldown. To imply that they're not 24fps is just mistaken and silly.

[Erik Bien]

Barry,

When ya gonna write "The HV20 Book" and set us all straight? :Drogar-BigGrin(DBG)

[Barry_Green]

It's one reason I got an HV20. Problem is, it takes so long to do one of those book/DVD packages properly (usually six months), the HV20 will probably be replaced with an HV30 and HV40 by the time the book/DVD came out! Already there's an AVC-HD version, the HR10, due to come to market.

We may still do something but I wouldn't want to tie it to a particular model when dealing with the consumer market. The DVX has lasted for four years, but in the consumer marketplace there's much more rapid model turnover (even if it's largely cosmetic, they seem to change model numbers much faster on the consumer side).

[Matt Agnello]

I was worried about this demonstration (http://www.box.net/shared/static/15ydiu43co.jpg) that the author of the linked post shared. You can see the distortion pretty clearly in the picture, although the frame is extremely magnified. I don't think this would be a problem for most people; perhaps it's likely to be a problem if on a big enough screen, though.


...keep in mind that we edited 24p-within-60i for years and years; that's the way it always used to be done.I did not know that 24p was editing in 60i for so long, though. Very interesting tidbit.

[logic108]

My question is this : 'If you don't buy the HV20 what other camera will you buy? Also, 'In what way would that camera be better than the HV20?'
For me the camera that is better than the HV20 with a 35mm adapter is the HVX200 - not the A1.

[Barry_Green]

You can see the distortion pretty clearly in the picture, although the frame is extremely magnified.
Well, yeah, if you don't extract the pulldown then you're going to have interlaced lines. This isn't a problem with the technology, it's a problem with the author not understanding the technology.

If this was truly a problem, then no film could ever have been edited on video!

For best results, one would extract the pulldown out and restore the footage to its original 24p state. The HV20 doesn't make this easy, because it doesn't include flags in the data stream. But that doesn't mean it's impossible, it means you have to take another step and run a pulldown-extraction program.

The sky is not falling. Cats and dogs are not living together. There is not complete pandemonium. All you have to do is learn what the product does, and adapt your work to compensate for it. If you want to edit pure 24p, you can do that. You just have to strip out the pulldown from the recorded file first (just like every film transferred to video has had to do, since the dawn of broadcasting).

There are easier ways; there are better ways (such as the 24F mode on the Canon XH/XL series, and the 2:3:3:2/24pA pulldown on the Panasonic & Canon series). But for those circumstances where you don't have those methods, you have to resort to stripping out the pulldown. In FCP I believe you can do it in Cinema Tools in one step. After Effects can do it. And there are utilities for Windows that will do it too.

[Elton]

In FCP I believe you can do it in Cinema Tools in one step.

It's actually the new Compressor 3 that can automatically strip out the pulldown--Cinema Tools doesn't actually analyze the cadence.

[angelo913]

I have used Canopus' ProCoder 3 and applied the Inverse Telecine filter to the HDV24P(60i) footage to create a true 24P file. I used the default setting of the filter and it did an excellent job with no interlacing effects just 24P film motion.

As a test I did the same thing as above but without the filter and 24P file showed deformed objects in motion.

...Angelo

[David Newman]

One of the issues with the CNet review claims, is that it quotes CineForm. CineForm does not back his claims. The camera is a 24p signal on 60i, that is pretty standard. However, there is a minor chroma crosstalk every 4-th frame (not a luma crosstalk with interlaced as this user shows.) This is because the 4:2:0 encoding of HDV, can't have separate chromanance information of each field, so when you get a 24p frame that stradles a 30p frame slot, there is some chroma crosstalk. However if you remove pulldown correctly (which all the CineForm tools do) you will not notice this. Luma is separatable and that is what your eye tracks, so the image looks fine. This is primarily a keying issue, and that I recommend live capture form the HDMI port at 4:2:2, so there are no issues will pulldown chroma crosstalk, or MPEG compression.

[DavidG]

does anyone know of software that will extract the frames that isnt in final cut pro? i have after effects but never heard of this before, i just ordered my HV20 and trying to get to know its many wonders. Dave.

[vsansal]

does anyone know of software that will extract the frames that isnt in final cut pro? i have after effects but never heard of this before, i just ordered my HV20 and trying to get to know its many wonders. Dave.

You can use after effects to remove 2:3 pulldown

[rick deckard]

Well, yeah, if you don't extract the pulldown then you're going to have interlaced lines. This isn't a problem with the technology, it's a problem with the author not understanding the technology.

Hello, I am the original author, I do understand the technology, and I did extract the pulldown using Digital Fusion. And with any 24p footage I record and capture from the HV20, I perform a proper 'pullup' or '24p extraction' or 'reverse pulldown' or whatever you care to call it (people argue about the exact term), yet there is residual interlacing on every fourth frame. For me, it is quite noticeable during playback, depending on how colorful your footage is, and particulalry in the red channel.

Example clip from my HV20 (6.5mb), before 2:3 pullup/24p conversion-
http://www.box.net/shared/static/b7qayhgfqa.mpeg
Here is the same clip after 2:3 pullup in Fusion.
http://www.box.net/shared/static/va357ns51c.mov
Here is the 4th frame, after pullup that still has interlacing in it-
http://www.box.net/shared/static/6anr20frau.jpg

Would love to see if anyone can process the 24p from the first clip and get a different result (what does your 4th frame look like?), and show me where I am confused about the 24p in the HV20.

David Newman, CTO of Cineform gave a precise techical explanation of the problem I see on all my 24p footage. From the Adobe Premiere forum-

"CineForm Aspect HD does support automatic extraction of 24p from HV20's 60i stream. However, Ridlen is correct that the 60i encoding of 24p is not completely reversible, whereas the in 24F in the XH-A1 and XL-H1 is it. The reason the lack of completely reversibility is the 4:2:0 60i encoding which has only chroma value for pixels over two scan lines -- these scan lines may contain data from different frames, cause chroma cross-talk. The luminance is perfectly correct. Luma is encoded at 60i, and chroma at 30p, you can extract 24p from luma but not fully from chroma. If the pulldown is extracted correctly, the chroma cross-talk will only appear on every fourth frame. As our eyes are far more sensitive to luma, it is very hard to see this in motion. The only time this can be an issue is when keying, for that I recommend using HMDI and shooting live via the Black Magic Intensity card (using the new 1.5 drivers.) All CineForm tools support real-time pulldown removal using that card from the HV20. When you do a live capture the 24p in 60i is encoded as 4:2:2, with has separate chroma values for each scanlines, now the pulldown is completely reversable. "

Yet, everytime I bring up the 4th frame interlacing issue with the HV20's 24p footage, users get defensive, support gets evasive, Newman denies there is any problem, everyone reacts so weirdly about this. I was booted off the DVinfo forum for no discernible reason whatsoever after stating my observations very objectively, and even though again David Newman confirmed "...everything David Ridlen is saying is true." But users (including the administrator) attacked me as being an 'idiot' and a 'troll' and refused to help or consider the examples!?! I dont get what the hell is wrong with everyone. Mr Newman was the only one attempting to be helpful and asked me to post clips to try out himself, but then I was not allowed to. So far, no one has demonstrated a different result from what I am getting. I would just like one person to try extracting the above clip and please show me exactly where I am wrong about this.


You just have to strip out the pulldown from the recorded file first (just like every film transferred to video has had to do, since the dawn of broadcasting).

I am stripping out the pulldown, exactly as I have done for the last 10 years, with no prior problem whatsoever, using Digital Fusion, and After FX. The HV20 24p has presented the first ever problem for me with pulldown. Then I read that post by David Newman that explains exactly why I am seeing interlacing on every 4th frame, particularly in the red channel. Playback looks bad to me, a visual FX professional of 12 years, and to my wife, a digital compositing expert. While everyone else thinks it looks great (?). No one has yet shown that I am doing anything incorrectly. But if there is any degree of interlacing lurking in the 24p footage, just as David Newman has said there is, then I dont see how anyone can call it true progressive. Maybe up is down.

[rick deckard]

One of the issues with the CNet review claims, is that it quotes CineForm. CineForm does not back his claims. The camera is a 24p signal on 60i, that is pretty standard. However, there is a minor chroma crosstalk every 4-th frame (not a luma crosstalk with interlaced as this user shows.) This is because the 4:2:0 encoding of HDV, can't have separate chromanance information of each field, so when you get a 24p frame that stradles a 30p frame slot, there is some chroma crosstalk. However if you remove pulldown correctly (which all the CineForm tools do) you will not notice this. Luma is separatable and that is what your eye tracks, so the image looks fine. This is primarily a keying issue, and that I recommend live capture form the HDMI port at 4:2:2, so there are no issues will pulldown chroma crosstalk, or MPEG compression.

I know you have to rep a company and step carefully in all this, but someone posted on the Adobe Premiere forum as "David Newman "CTO of Cineform" verifying exactly the problem I am seeing in the HV20's recorded 24p footage- http://www.adobeforums.com/cgi-bin/webx/.3bc3641e/13

"CineForm Aspect HD does support automatic extraction of 24p from HV20's 60i stream. However, Ridlen is correct that the 60i encoding of 24p is not completely reversible, whereas the in 24F in the XH-A1 and XL-H1 is it. The reason the lack of completely reversibility is the 4:2:0 60i encoding which has only chroma value for pixels over two scan lines -- these scan lines may contain data from different frames, cause chroma cross-talk. The luminance is perfectly correct. Luma is encoded at 60i, and chroma at 30p, you can extract 24p from luma but not fully from chroma. If the pulldown is extracted correctly, the chroma cross-talk will only appear on every fourth frame. As our eyes are far more sensitive to luma, it is very hard to see this in motion. The only time this can be an issue is when keying, for that I recommend using HMDI and shooting live via the Black Magic Intensity card (using the new 1.5 drivers.) All CineForm tools support real-time pulldown removal using that card from the HV20. When you do a live capture the 24p in 60i is encoded as 4:2:2, with has separate chroma values for each scanlines, now the pulldown is completely reversable. "


To reiterate, "24p is not completely reversible...you can extract 24p from luma but not fully from chroma." So is the HV20 24p recorded footage interlaced, or is it progressive? Apparently the HV20 can playback as both at once. Can we call it 24Lp/Ci for 'lumi progressive/chroma interlaced?' I still see the HV20 as more accurately termed '24p ready' just like a TV may be advertised as 'HD ready' since it requires you to purchase additional parts to interpret the 'HD' signal, the HV20 requires you to purchase additional parts to interpret the '24p' signal. Make sense?

But by the above explanation, the HV20 is just not a "fully," (true?) progressive camcorder. Although you say it just shouldnt be too noticeable.

[vsansal]

Chroma talk is because of HDV compression. If you capture from HDMI using a capture card it is possible to bypass compression thus chroma talk. So HV20 has true 24p but HDV compression causes problems.

[rick deckard]

Chroma talk is because of HDV compression. If you capture from HDMI using a capture card it is possible to bypass compression thus chroma talk. So HV20 has true 24p but HDV compression causes problems.

Yes, but who purchases a portable camcorder with alledged features of "1920x1080" and "24p" and fully expects this to mean having to purchase additional third party hardware and software, and to always remain in a studio situation to digitize one clip at a time, hooked to a computer thru no more than a six foot HDMI cable in order to get actual 1920x1080, true 24 progressive frames per second out of it?

From Wikipedia- "A camcorder is a portable electronic device for recording video images and audio onto an internal storage device. " Therefore a '24p camcorder' should be portable and portably record 24p to an internal storage device- in this case, mini DV tape. The HV20 camcorder only records 30 fps with interlacing to tape. Does not record 24p. Never has. Never will. Canon will clearly tell you this, the HV20 spec sheet will clearly tell you this, the forums will clearly tell you this. The HV20 does not record 24p. It is at best '24p ready.' Just like an HD ready TV, the HV20 is neither capable of independently recording or playing back at 1920x1080 resolution nor 24 progressive frames, as any buyer would naturally expect.

[vsansal]

If 2:3 pulldown is too much of a problem for you, you can always buy the PAL version for a couple hundred more. No pulldown just 25p.

[David Newman]

I now believe vsansal is correct, and the chroma issue I was seeing are MPEG articfacts not a pulldown issues. The pulldown is completely reversable. Sorry, I was in error -- so quote me on that, not on the older stuff. :) Rick, it is time to work out where you went wrong, as no one is backing you up. I need the source MPEG data for your test sequences, but the links you provided didn't work.

[Barry_Green]

Rick, you may have a point -- the interlaced 4:2:0 could perhaps be complicating issues. I hadn't factored that in when I wrote the original response. You might need to use software that intelligently understands how to reverse out the pulldown and at the same time compensate for the 4:2:0 color recording; if the software simply drops fields then yes it isn't going to work 100% right. But if it understands how the chroma is split across fields, it should be completely reversible.

But -- this is a $1,000 camcorder too. It's not really designed to have pulldown removed. Canon didn't even put the flags into the data stream to support pulldown removal, and when users complained enough, Canon said (basically) "sorry, that's not what this product is made for." (their response was posted on DVInfo somewhere).

I'd actually wonder about the V1U as well; it uses the same interlaced 4:2:0 recording system, it should potentially have the same issue on every 4th frame, if interlaced chroma is not accounted for.

As far as the mantra about "not recording 24p", buckle your seatbelt for a surprise -- practically no other camera records "24p" either. Not a $100,000 HDCAM F900 (it records 24PsF with segmented fields instead of frames). Not a $45,000 VariCam (it records 24p embedded within a 60p data stream). Not a DVX or XL2 or DSR450; they all record 24p-within-60i (same as HV20) or 24pA-within-60i (which still uses a form of pulldown but totally bypasses any interlaced-chroma issues). Not a Sony V1U, it records 2:3 pulldown just like the HV20. Not a JVC HD100/HD200/HD250; they record 24p-within-60p using 2:3 frame-repeat flags. The *only* cameras out there that can RECORD a straight 24P are the HVX200/HPX500 using 720/24pN mode. In 1080 mode they still use the same 2:3 or 2:3:3:2 pulldown systems.

Curiously, Canon does offer a straight 24p recording mode on their 24F cameras, but their footage is sourced from interlaced chips so while it's a progressive recording format, the source frames aren't sourced from progressive chips, so I'm sure there are plenty of purists who would complain about that too.

So what you want doesn't exist. And it most certainly doesn't exist in a $1000 palmcorder. Trying to start a campaign to accuse Canon of deception is either naive, misinformed, or ill-guided. They are employing the exact same recording mechanism everybody else does. Your problem (insomuch as a problem exists) may have more to do with the HDV format's 4:2:0 color sampling than it does with anything else. But lots of people seem to be happily extracting pulldown on the Windows platform with no residual interlacing artifacts, so I'd expect you've just got some sort of error in your process somewhere.

[Barry_Green]

... and David posted while I was writing that, so it may turn out that there's no issue with the HV20 and removing pulldown at all.

[manbart]

People have found work arounds for this 24P problem with the HV 20 using freeware. It's somewhat complicated and labor intensive, or you can spend $250 on Cineform's Neo HDV which will deal with this and works with Vegas 7

[sean90291]

For best results, one would extract the pulldown out and restore the footage to its original 24p state. The HV20 doesn't make this easy, because it doesn't include flags in the data stream. But that doesn't mean it's impossible, it means you have to take another step and run a pulldown-extraction program.


To your knowledge, Does Cineform's NeoHDV do what's needed as far as pulldown-extraction? I mean, I know it does pulldown, but given that the HV20 doesn't contain flags, does Cineform's software automatically detect the HV20's cadence and duplicated frames (because as was pointed out, the HV20's cadence changes from scene to scene).

[vsansal]

To your knowledge, Does Cineform's NeoHDV do what's needed as far as pulldown-extraction? I mean, I know it does pulldown, but given that the HV20 doesn't contain flags, does Cineform's software automatically detect the HV20's cadence and duplicated frames (because as was pointed out, the HV20's cadence changes from scene to scene).

Yes, it does. All you need to do is select inverse pulldown. It is that easy and it does all of this during capture.

[sean90291]

Is there any loss in resolution after Cineform's pulldown? I know deinterlacing causes a loss of resolution (in fact you're not supposed to deinterlace AND do pulldown-extraction with Cineform--you're just supposed to do the pulldown witout deinterlacing apparently).

[Cranky]

The reason the lack of completely reversibility is the 4:2:0 60i encoding which has only chroma value for pixels over two scan lines -- these scan lines may contain data from different frames, cause chroma cross-talk. The luminance is perfectly correct. Luma is encoded at 60i, and chroma at 30p, you can extract 24p from luma but not fully from chroma. If the pulldown is extracted correctly, the chroma cross-talk will only appear on every fourth frame. Whoa, I have never thought about it. Anyway, what looks awful for you as a professional, will (and does) look just amazing for a regular consumer. Even crippled as it is, HDV-2 provides so much more detail compared to DV that consumers are in awe. Heck, talk with professionals about 12-, 10- or even 6Mbps 1080i feed over cable or about HD-Lite and all you'll hear is cursing. But regular consumers think that this castrated HDTV is the best thing they ever saw. They love it. The HV20 is a consumer camcorder after all. To those who disregard Rick's accusations of HV20 as a non-true 24p camcorder, please check out this article: http://www.dvdfile.com/news/special_report/production_a_z/3_2_pulldown.htm It explains how A,B,C,D film frames are pulled down into A1A2, A1B2, B1C2, C1C2, D1D2 field sequence.

http://www.dvdfile.com/news/special_report/production_a_z/images/3_2_pulldown/3_2_a.gif

Judging by this sequence it seems clear that every fourth 24p frame (B) in is split between two 60i frames and its croma is "borrowed" from the frame A. I believe that one frame out of four having croma crosstalk is not noticeable to most consumers, so while HV20's 24p is not truly reversible, it can be considered "true" 24p for a consumer appliance. This is just another small lie in a series of lies, like that HV20 is a true 1920x1080 HD camcorder, while in reality it is a standard 1440x1080 HDV-compliant device, with only 1920x1080 imager. Heck, the whole idea of having true 1080i HD at 25Mbps is laughable, because the compression kills the fine details and macroblocking is inevitable on fast-moving scenes. But HDV produces visually better footage than DV and this is all that matters for consumer market. No one cares about formalities.
What *is* an issue, and the reason I say "almost", is that Canon didn't put in frame/field flags to allow for easy extraction in post. That's a disappointment; it means that if you want to strip out the pulldown and access the pure raw 24p data you have to go shot-by-shot (because the cadence can change for every shot) and strip out the pulldown By looking at the picture above I am curious, does HV20 writes video in chunks of five 60i frames?

I understand the remark about flagging, but considering that even budget DVD players finally became fluent in candence recognition (OPPO with Mediatek comes to mind first, and of course Faroudja and Silicon Optix chips), I suppose that this should not be a big deal for computer software.


---

Michael, Canon Elura User Pages (http://www.elurauser.com)

[Barry_Green]

By looking at the picture above I am curious, does HV20 writes video in chunks of five 60i frames?
Once recording starts, it's done in continuous sequences of 10 fields, with the pattern repeating after 10 fields. But every time someone stops and then starts recording again, the pattern is potentially interrupted. Which means the cadence is "off", and that complicates the whole system.

Some software can "guess" the pulldown sequence (like After Effects). Most NLEs don't. 24P video cameras are designed to put flags in the timecode stream, and NLEs are designed to recognize those flags. Canon didn't design the HV20 for pulldown removal, so they didn't put the flags in. If you're using software that can appropriately guess the pulldown removal, and you break your footage into individual shots, then there should be some automation possible to strip out pulldown.

A custom utility could strip out pulldown and compensate for the missing chroma information; I haven't heard of anyone producing such a product though, unless CineForm's products do it?

[SomewhereinLA]

I don't know much about the HV20, but I wonder if the sequence starts the same way each time you start recording? If so then to cine-compress the clip should be easy no guesswork needed.