Cineframe 30 De-mystefied

[Barry_Green]

Okay, there's been much speculation about CF30, and people have been wondering how it can look so good, with no real discernable loss in resolution... some have suggested that maybe it's using a "smart de-interlacer", which would only de-interlace moving segments, leaving static shots at full resolution (de-interlacing discards one field, leaving a half-resolution image). *"Smart de-interlacing" is the kind of technology Marcus van Bavel's DVFilm Maker software uses to preserve as much resolution as possible.

I've extracted the vertical resolution segments from a CamAlign chart. *This shows the difference between FX1 HDV 60i and FX1 HDV CF30. *There is a substantial loss of resolution. *Furthermore, since this was a static locked-down still shot, this should discard the notion that there's any "smart de-interlacing" going on with CF30, and the same should (note: *should*) be true with CF25.

This is a 300% blow-up in PhotoShop of the vertical resolution extraction from the CamAlign chart, to make the resolution portion more easily discernable.
http://www.icexpo.com/FX1-60i-vs-CF30-res.jpg

In interlaced mode the camera looks like it's delivering about 750, maybe 775 lines of vertical resolution. *In CF30 mode, it's down to about 575.

Now, the interesting thing (to me) about these numbers is that first, for 775 lines of res, that's really very close to Steve Mullen's educated guess that the FX1 should be delivering about 820 lines. *In fact I'd say it's close enough to fall within the margin of error. *The FX1 in 1080 mode delivers a real-world image of about 775 lines of resolution, due to the same line-pair-summation/vertical blur/kell factor issues that cause DV cameras with 480 vertical pixels to only deliver a maximum of about 360 lines.

Second, with a CF30 resolution of an observed 575 lines, that is awfully close to the number you'd expect from a straight de-interlace. *If the FX1 were to skip the low-pass filter and line-pair summation and just do a straight de-interlace on the raw field data, you'd get 540 lines. *Which is, again, close enough to the observed 575 to fall within the margin of error. *Which means to me that yeah, that's what they're doing: a straight field-doubling de-interlace.

Which means that when in CF25 mode you're getting about the same vertical resolution as you'd get from a PAL DVX or XL2 in 25p/thin mode (although still with twice the horizontal resolution).

Furthermore, extrapolating forward from that, if the FX1 can only deliver about 775 lines of observable vertical resolution, that's incredibly similar to the 720 lines you'd get from DVCPRO-HD's 1280x720p. *(with 720p you could skip all that row-pair summation and low-pass filtering and flicker removal and get access to all 720 lines of resolution). *You'd get the same vertical resolution, and 1280 pixels across vs. 1440 pixel-shifted pixels across. *Although this now enters the realm of speculation, I think it's reasonably safe speculation to say that Panasonic's 1280x720P DVCPRO-HD is going to deliver just about the same real-world observable resolution as HDV's 1440x1080i! *(not to mention you also would have true progressive scan, true 24p, true 60p, etc).

You gotta love 'em for trying, but the more I look at it, the more I think the only rational thing to do (regarding the FX1) is just wait to see what the others respond with. *If you absolutely *have* to have HD today, the FX1 is the only choice (and a good choice), but if you don't have to buy today, the FX1 does not look to be the panacea that everyone was hoping for, and there's certainly room for improvement.

[Kidster]

Barry, you're the shit! I didn't understand a single thing you wrote, except for the last part. I believe I will wait. I think we all appreciate you putting your time and effort into your research and answering all the mundane questions that I and others ask you.

[David Jimerson]

Yeah, seems like our own Barry is quite the DVX fanboy.

What's your deal? You may not like his results, but he's constantly said if he's wrong, he wants to know about it. Do your own tests, post them, and show where he's wrong.

Or don't. Either way, lose the personal insults.

[scharky]

Hey What happened to Bakkake's post?
Barry I'm sure that we are all very appreciative for the amount of work that you are putting into these test results. This will definatly save many people hours of searching though incompitent reviews from various websites, which shall remain nameless. Seriously, the test results do not lie, if done correctly, and from your background I have the utmost confidence that your test results are accurate. Again, it is funny how many people take offense to this, but you are not even saying that the camera is bad. You constantly say that the camera is great, if you plan to shoot interlaced hi def. But I think people don't understand the concept, the right tool for the right task. It's the whole square pegs in a round hole over and over again.

[SamEdwards]

Hi Barry,
I would love to see what DVFilm Maker would do with the same chart. Great detective work.
thanks,
Sam

[Barry_Green]

Me too -- I keep meaning to run that test! I have a version of Maker that supports HDV, so I just have to actually do it.

I also have a chart that's shot in CF24, so it'd be good to compare those two. The CF24 chart shows a rhythmic pulsing in it (a byproduct of CF24's field blending) and it'd be nice to see how Maker handles the same task.

I'll get it posted, just have to remember to do it (and the Internet station is not the editing station, or I'd do it now...)

[chalbers]

Hi Barry ! You've become my DV Guru ! :-)

It's amazing how much time you take to help us all and I'm really thankfull, as we are all I'm sure, for your amazing information that will help us all to get the camera we REALLY need !

Frank

[Aejaz]

Barry, from the results of your tests and whatever perceptions I have developed about HD (due to lack of availability of HD tvs and technology here in europe), I feel that Sony and other propagators of the HD format for the moment just want to sell a format suitable for larger screens ...i.e. a format giving more horizontal resolution...though not necessarily more vertical resolution...or a real world clarity of the image. And given the lack of the HD display technology in general .... perhaps this serves their purpose for the moment.... But as you said....there is room for improvement...definitely.

and no surprise.... this is typical SONY behaviour....making huge claims; sky rocketing the expectations of the customers; coming out with a 'half-backed' cake finally; and leaving the unsatisfied customers high and dry at the end. Yes because 'It's a Sony' :-X

[xray]

As an expert like Faroudja thinks the route from interlaced to deinterlaced and linedoubling is not working, it says enough. (that whas his opinion at the Wrap Up meeting on the IBC Amsterdam.) Systems like the Faroudja de-interlacers use interpolation which actively renders between a spatial and temporal algorithm depending upon changes in motion and lumi.

But I wonder what Sony is doing today working on the Z1...

[monster]

Ok, so it deinterlaces in CF30 but records to tape as interlaced? So how does a person handle the footage after it has been caputed into their editor? Does it still have to be de-interlaced?

[Barry_Green]

No, you'd use it like any other footage. It effectively works the same as DVX 30P footage, which is scanned progressively but recorded to tape as interlaced.

[SamEdwards]

Hi Barry,
Do you think there is less macroblocking (break-up on fast motion) when using cineframe because the frames are less complex?
Can you think of a way to test this?
thanks,
Sam

[Barry_Green]

Hmmm. Interesting supposition. Offhand, I can't think of a reason why your idea wouldn't be correct. No way to test it right now, I don't have access to an FX1 currently, but if CF30 does work the way I think it does, then I think there's a lot of merit to your idea.

[Zig_Zigman]

So why not shoot in 60i, get max resolution, forget the in-camera stuff, and de-interlace or do what you want in post?

[Shaw]

Because you could buy a DVX? :D ;)

[Barry_Green]

So why not shoot in 60i, get max resolution, forget the in-camera stuff, and de-interlace or do what you want in post?
Because you'll get a lot less resolution doing that. *The FX1 is an interlaced camera, and as an interlaced camera it doesn't deliver nearly the same pixel-count resolution as you would expect from a progressive-scan camera. *According to Steve Mullen, the highest resolution the FX1 is capable of is about 820 lines of vertical resolution. *In my tests I measured 775, so I think between his educated guess and my test, we're probably within the margin of error.

De-interlace 775 lines and you'll end up with only about 390 lines, which is not really any better than a standard-def camera can do (except the FX1 would still have twice the horizontal res). *Using the CineFrame mode lets the camera skip some of the internal filtering/processing/blending that it normally does, so you end up with about 575 lines, much higher than you'd get if you de-interlaced in post.