Please keep this on topic, as Steve Mullen has (thanks) about resolution. Create a new thread about the card. I'd be happy to do it and drop some of these posts in it! (smile)

heath

Also in interest of keeping this thread on topic, what do we know about the horizontal sampling and encoding of the HVX200 and its resolving power? DVCPROHD encodes at either 960x720 or 1280x1080 - a more horizontally compressed resolution than HDV @ 1440x1080. While DVCPROHD wouldn't play a role directly in viewing the uncompressed camera output via the component interface, would it be possible that the internal sampling/image processing still may be playing a role in setting up the horizontal imaging for DVCPRO encoding? With all these tests and everyone focusing on how many lines this camera can resolve, it seems that nobody has made any mention of horizontal resolution capabilities of the HVX. If there is some horizontal processing and compression going on, could this be having an effect on the perceived vertical resolution?

I imagine within a couple weeks we will know for sure. Someone will rip apart an HVX and run it down to the lab.

My sheer speculation is that Panny has done some serious pixel-shifting and we might find the resolution is quite low. I mean 960 by 780 or therabouts is no reason to not give out the resolution, just say how and why you are doing it.

Is it really just a juiced up dvx? a la reel stream?

Steve, would there be any idea to measure the pixel size with Airy disk phenomena?
Shooting charts and closing the aperture and when the resolving power starts to decrease, calculating the size of Airy disk there?

My sheer speculation is that Panny has done some serious pixel-shifting and we might find the resolution is quite low. I mean 960 by 780 or therabouts is no reason to not give out the resolution, just say how and why you are doing it.

If anyone is still interested in this topic -- I've posted my "latest" understanding of HOW the HVX200 works -- which seeks to explain why Adam's tests measured such low resolution when folks shooting with the camcorder are getting very high quality images.

Please note I'm not saying Adam's tests were wrong or unfair. His procedure was intended to remove the benefits of pixel-shifting -- and they did. And, they are very useful for indicating the native resolution of the CCDs. (Something we shouldn't have needed to do.) But, they underestimate real-world performance -- as they do for the other camcorders that use Green-shift.

http://www.gyhduser.com/showthread.php?t=341

This article here:
http://www.coax.tv/DefaultJAN17.htm
and part 2 here:
http://www.coax.tv/Default.htm

may give some insight into the impact of DVCProHD compression on image resolution.

I imagine within a couple weeks we will know for sure. Someone will rip apart an HVX and run it down to the lab.
...no reason to not give out the resolution, just say how and why you are doing it.
Is it really just a juiced up dvx? a la reel stream?

& Hopefully the person who will rip apart the HVX first, to find out the inner workings, will be Juan from Reel-stream !

If anyone is still interested in this topic -- I've posted my "latest" understanding of HOW the HVX200 works -- which seeks to explain why Adam's tests measured such low resolution when folks shooting with the camcorder are getting very high quality images.

http://www.gyhduser.com/showthread.php?t=341

I was in Kyoto for a shoot in mid-January. After the project ended, I took the opportunity of dinner in Osaka (Kadoma) with a Panasonic engineer who I got to know while he was assigned in the U.S. With an HVX on order, I was eager to privately learn the "inside" story about the HVX's HD methodology. The design makes clever use of inexpensive components for a "prosumer" buyer, but actual performance must not infringe Panasonic's more elegant Vericam products. Within a pixel or two, the description provided at dinner exactly matches Steve Mullen's posted description.

Steve's nailed the HVX's reality.

Using this same logic... the Canon's 540 in 24F mode is likely effected the same way correct?

ash =o)

The design makes clever use of inexpensive components for a "prosumer" buyer, but actual performance must not infringe Panasonic's more elegant Vericam products.

Very interesting because a friend I got to know when I worked in Japan said the exact same thing -- which I brushed off as "too paraniod." But, I can see how Panasonic might be concerned because they watched what the VX1000 did to BetaSP products.

Has anyone got delivery of their PAL unit? A while back I heard that the Euro version is/was proving to be a problem. The rumor in Japan was that the Euro version will be delayed while a "fix" is implemented.

IF that's true, and its always a big IF with rumors, then that MAY give us another insight into how the HVX200 works. Here's why. I assumed that ALL recorded formats are from the 1920x1080 buffer -- including PAL and NTSC. But, maybe PAL and NTSC are taken from the CCDs running in interlace mode. (The dvx100 used switchable CCDs.) Were this to be true -- 540-row CCDs might not yield a high-quality 576-line PAL image for DV, DVCPRO, and DVCPRO50. Hence the "problem" with Euro units.

Frankly, I can't believe Panasonic would not have been smart enough to use 960x576 pixel CCDs. In Region 60 camcorders, 540-rows would be used for HD while only 480-rows would be used for NTSC. Likewise, in Region 50 camcorders, 540-rows would be used for HD while all 576-rows would be used for PAL.

Not only would this be smart -- the use of "widescreen" PAL CCDs would keep CCD cost down.

+++++++

I guess I should have been clear that my model is calculating Effective, not Gross resolution. Thus, the physical CCD would have perhaps 10- to 20-percent more elements.

Using this same logic... the Canon's 540 in 24F mode is likely effected the same way correct?

That's why Adam wobulated all cameras. Any camera that uses green-shift would measure worst-case resolution in the test. My my model estimates both the Sony and Canon to have far more than the measured resolution in real-world situation.

The 24F mode should not be different horizontally than the 60i mode.

Jarred posted this tonight:
http://www.dvxuser.com/V6/showpost.php?p=448028&postcount=1

Saw it. Steve was money.

Now let's get over it and enjoy the creativity the camera will facilitate. No more pixel wars.

Canon should come clean about 24F IMHO. It's probably a very good 48i real-time deinterlace. Virtual 24p that works great too.

It's all good in the hood.

YOU CAN SKIP THIS POSTING AS I'M IN ERROR. BUT I'M LEAVING IT SO BARRY'S RESPONSE MAKES SENSE.

Questions:

How does the HVX200 get 1080i from the 1080p60 DSP buffer without line flicker or line twitter in the video. (Row-Pair summation seems NOT to be used because vertical resolution doesn't drop!) I've posted how I think it's done, but it was just a wild guess. And, if even my guess is correct -- see my earlier post -- the 1080i video has NO interlace artifacts. Will it look like HD "interlaced video?"


Also -- the question remains as to how NTSC and PAL are generated for DV, DVCPRO, and DVCPRO50.

1) Are both derived from the 1080p60 buffer in the DSP?

If so, then we have the same question about line flicker and twiter. Of course, Row-Pair Summation could be used during the downconversion to SD. However, effective vertical will then be only 435 TVL. Is that enough for PAL?

2) To obtain "real" SD interlaced video, like the DVX100 CCDs, the HVX200 CCDs could be switched to interlace mode. However, my computer model indicates that if interlaced" NTSC and PAL were pulled directly from the CCDs, effective vertical resolution would be only about 435 TVL. And, again I'm not sure this will be OK for PAL.

1) Are both derived from the 1080p60 buffer in the DSP?
According to the article, yes. Everything starts at the 1920x1080 DSP buffer, and all other modes are downrezzed/created from that.

Of course, Row-Pair Summation could be used during the downconversion to SD. However, effective vertical will then be only 435 TVL. Is that enough for PAL?
Well, that's how many discernible lines PAL currently offers, so it should be. Applying the .75 interlace factor to 576, we get 432 TVL of resolution for interlaced PAL.

In any case, the "interlaced" NTSC and PAL video have NO interlace artifacts. Again, will it look like SD "interlaced video?"
Of course it does. In DV 60i mode it looks exactly like any other interlaced DV camera. And in 1080i mode it looks exactly like any other interlaced 1080 camera. I've got an FX1 and an HVX here right now, and in 1080i mode playing back on my interlaced 1080i set they look exactly the same as regards motion and interlacing).

2) To obtain "real" SD interlaced video, like the DVX100 CCDs, the HVX200 CCDs could be switched to interlace mode.
Not according to the article; there basically is no interlace mode on the CCD, it's all progressive all the time. Which is why interlaced mode doesn't get a boost in sensitivity, like the DVX does. But it looks exactly like interlaced mode. The DVX and the HVX, when both set to 4:3 60i, look extremely similar.

Oh, and by the way -- I sent you an e-mail, but I'll say it here too -- congrats on figuring out the pixel count down to the exact number. You did a great job on that!

Barry it was you insisting you had "non-wobulation" numbers that were significantly higher than Adam's numbers that made me realise that BOTH measurements had to be "valid."

Plus, you insisted the HVX200 LOOKED so much better than the worst-case tests. So Panasonic had to have designed a unique CCD + DSP "system." Which is what Jan was saying in her interview.

I enjoyed exchanging e-mails with you, it was interesting to see the process you went through to achieve your conclusions. But I'm really amazed at how well you nailed it. Your descriptions seemed to match all the evidence, but I still wanted to see official factory confirmation -- and that's exactly what we got.

Well done!

Well, that's how many discernible lines PAL currently offers, so it should be. Applying the .75 interlace factor to 576, we get 432 TVL of resolution for interlaced PAL.

I've always used 75% but my model achieves minimum error with a value of 66%. That's one of many problems with any math model -- it can come up with a low overall error over a dozen camcorders -- yet have individual assumptions that don't match reality.

The model predicts NTSC to have 317 -- which actually comes close to the textbook answer of about 330. It estimates 380 for PAL.

So whichever number is "really" right -- 435 is greater than needed -- so the can DSP supply all. Very clever and a lot of number crunching!

You are also correct about interlace. For anyone who I totally confused:

With "real" interlace video -- when there is motion -- each field in a frame is different. We call them interlace artifacts -- and if you are going to film they are bad. But, when we view them on an interlaced display -- these slight differences between fields provide the eye with motion information at 1/60th second -- not 1/30th second intervals. So the artifacts actually help smooth motion.

When the HVX200 captures frames every 1/60th second, it obtains samples at a 60Hz rate. Each 1/60th second half the lines in a frame are discarded and the other half put into a field. So we generate 60i video exactly as we do when we have a camera that uses "interlace scanning."

So what does it all mean?

It means we can quit worrying about it and go shoot now. :)

Green's layman translation is all I need!

heath

It means we can quit worrying about it and go shoot now. :)

Good enough for me! I just need the eagle to land...