Sony announces HDR-CX12 AVCHD Cam

[Steve Mullen]

Originally Posted by Dave Blackhurst

As far as archival storage, downloading the files onto a hard disk is quick and painless (have to do that to edit ANYWAY), and while I might hope that what I shoot is so important that it should be saved for posterity, once I do a mixdown, I can burn the source files to DVD's if I really want to keep them around... very cheap storage.


Don't tar the format with such a broad brush.

Speculation, but based on plenty of practical real world experience with knockoff memory cards... and yes, I blamed the camera at first too, so I'm not pointing fingers, just offering a highly probable explanation.

You're making a huge set of assumptions. I have video from a decade ago that I haven't edited -- some from the 80's that are are now on miniDV. So no -- one doesn't aways HAVE TO EDIT what one has shot. I keep an HDV/DV VTR for this. Keeping original media is the simplest way -- tuck it in a box. No need to BUY anything. No need to DO anything.

All Mac software transcodes -- there is no "dialing-in" in to do.

Pinnacle claims native editing requires a 2.66GHz Quad-core computer. Which itself is a bit of a lie, because Canopus notes that to simply play AVCHD at 1920x1080 at perfect 30fps requires a 2.66GHz Quad-core computer.

Moreover, Canopus outright states native AVCHD editing is "not possible today." That's because they, Apple, and I, define "editing" as a realtime, multi-stream procedure equal to what we can do with MPEG-2. So, I'm not, as you say, "taring the format with such a broad brush." I'm simply REPORTING what the folks who make the software state.

If you really really want to debate your ideas of what's happening with the Z7, there is a forum you can do that on. I only REPORTED what was said there. But, here's a hint -- it has to do with BURST data rate not SUSTAINED data rate. Only the latter is rated. The Z1 requires knowledge of the former. Other cameras do not.

[Ken Ross]

The Sony's picture can be made to look exceptionally good with very little manual intervention. Most cams have a tendency to overexpose and simply reducing exposure in some situations is the best way to maximize quality. However with the Canon I found no way (aside from "Cinema" whose overall picture I did not like) to adequately control exposure. The dynamic range of the chips/processing is simply not as good as the Sony. The strobing effect (in any cam) is usually not visible with high shutter speeds unless you have rapid motion. However, ironically, it's that same rapid motion that causes undesirable effects when shooting in 24p or 30p...at least IMO. In that case there's no way to get rid of it.

However I do agree that if someone needs 24p/30p, their only choice is the Canon. If not there are very viable, and IMO, better options.

[Dave Rosky]

Originally Posted by Steve Mullen

Or, they may say I'll wait a couple of months for the Pana that offers a ring-based control system. (In fact, even if one doesn't care about 24p, some might prefer to wait for the Pana just to avoid the touchscreen (Sony) and to get a VF (Canon).

Steve, which Pana might this be? Are there rumors of such a cam?

[Ken Ross]

Dave, it's not a rumor, but it has 1/6" CMOS chips so I'm not expecting the greatest picture quality.

[Dave Rosky]

Originally Posted by Ken Ross

Dave, it's not a rumor, but it has 1/6" CMOS chips so I'm not expecting the greatest picture quality.

Ken, thanks for the info. A little OT for this particular thread (but maybe not since it's comparative information), I Googled around and had no trouble finding it. On the sensor issue, it appears that with this new cam, Pana has switched to CMOS, so even though the sensors are the same size as in the SD9, the jury might still be out on whether the low light performance is the same as the SD9. In general, CMOS sensors seem better in this regard. There also appear to other changes to the image processing and lens that may improve other aspects of image quality. For someone who does a lot of outdoor shooting and would prefer the combination of light weight, a VF, and no HDD (not to mention the generally good Pana manual controls), the SD100 might be worth waiting to see. Depending on how the image quality turns out, this could the middle ground camera some of us would like to see.

Clearly, the AVCHD camera market is beginning to change really quickly.

[Steve Mullen]

Originally Posted by Dave Rosky

n the generally good Pana manual controls), the SD100 might be worth waiting to see. Depending on how the image quality turns out, this could the middle ground camera some of us would like to see.

Clearly, the AVCHD camera market is beginning to change really quickly.

I have the same hopes for the new Pana.

But, I keep asking, how can three 1/2 MP CMOS chips that are only 1/6-inch be expected to up go up against a 1/3-inch, 3MP, EXMOR CMOS? Frankly, the chips would be fine for an SD camcorder. (Remember, 447K was a popular pixel count for DV.)

I suspect Pana's strategy is to build HD camcorders from SD components thus creating very low R&D costs, part costs -- and thus generate very high margins. Then assume there are enough buyers who like Pana and/or who don't read reviews.

Makes me think of GE toasters and GM cars. Focus on PROFIT at all costs. Which really has worked for Pana. They are a very profitable company while JVC and Sony have really struggled to get profits out of their sales. (I'm not knocking this -- profit is necessary.)

Interestingly in Japan Panasonic is considered an "appliance" company just like GE was. JVC gets respect for their R&D while Sony for its R&D + marketing. Sony has always been seen like Apple is here.

Hopefully, it will motivate Canon to add a VF to the HF series by CES.

Has any camcorder beyond a $500 Panasonic SD camcorder ever used 1/6-inch? Even the HD7 uses 1/5-inch chips.

[Dave Rosky]

Originally Posted by Steve Mullen

Frankly, the chips would be fine for an SD camcorder. (Remember, 447K was a popular pixel count for DV.)

I suspect Pana's strategy is to build HD camcorders from SD component

The 3-chip HD cameras (Panasonic and others) are actually full HD resolution. These cameras use pixel shifting, which is an interpolation method similar in results to the bayer pattern interpolation used in single-chip sensors. In pixel shifting, the green sensor is ofset by 1/2 of a pixel in both the X and Y dimensionsl, and an interpolation is applied. In a single-chip sensor, 1/4 of the pixels are red, 1/4 are blue, and the remaining 1/2 are green. The pixels are arranged in a grid pattern and interpolation is applied.

In both cases you get a "full" HD image, but the interpolation, and associated anti-aliasing filtering is the reason why the spatial resolution of both types of sensor is less than what you would think from the raw pixel count. This is one reason why, for examples, most consumer HD cameras only have somewhere in the range of 600 (plus or minus) lines per picture height of resolution when there are 1080 raw vertical pixels.

Even most pro and prosumer 3-chip cameras use pixel shifting these days, it's not actually a way of building an HD camera from SD components. Here is a good paper describing pixel shifting and how it results in similar spatial resolution to a bayer-interpolated single chip sensor (thanks to Chris Hurd for the link):

ftp://ftp.panasonic.com/pub/Panasoni...200.CCD-WP.pdf

[Steve Mullen]

Originally Posted by Dave Rosky

The 3-chip HD cameras (Panasonic and others) are actually full HD resolution. These cameras use pixel shifting, which is an interpolation method similar in results to the bayer pattern interpolation used in single-chip sensors.

Not going to get in a flame war about this, but what you say is not true. In fact, it is totally false.

1) There is no definition of "HD" that allows the term be used with anything less than 1MP (1280x720) per chip in a three chip system. So, no, the Panasonic cameras do not "actually [have] full HD resolution [chips]."

2) A typical single chip Bayer camera with 3MP photosites has 1M R, 1M B, and 1M G. The Sony EXMOR, however, has about 1M R, 1M B, and 3M G. From these samples, YUV signals are computed: a transformation from RGB to YUV. There is no pixel shifting to increase luma information because, in the case of the Sony, the chip already has more than 1920x1080 luma (G) information. In the case of the Canon, the chip already has the minimum 1MP of RGB.

3) Starting with 520K R, 520K B, 520K G -- the Pana system is completely different. The number of RGB samples are only 1/4 of the 1920x1080 (2M) green (luma) samples. The luma RGB samples are slightly more than SD resolution.

The myth is that through "pixel shifting" these 520K G become equal to the Sony's 3M G samples or the Canon's 1M G. And, that the 520K R and B become equal to 1M R and 1M G.

You can't seriously believe this is possible! To believe this is to believe in magic. Using the words "pixel shifting" and "interpolation" doesn't explain HOW this is could happen. It's merely invoking technical words from marketing materials.

What actually happens with the Pana is the RGB-based LUMA information is increased by about 115% on each axis. (NOT 150%) Then, this information is spread over 1920x1080 pixels. Folks with a HVX200 already know the camera records soft 1280x1080. Now, do you really believe the same system will record 1920x1080 of REAL information?

4) Also, a myth is that this is HOW pixel shifting is used by professional cameras. Pixel shift in pro SD cameras is used to slightly (15%) increase H. rez. from CCDs that ALREADY have full SD resolution. Pixel-shifting is a way of "over-sampling" without actually using more pixels. It was NOT used to try to obtain SD resolution from "under-sampling" non-SD CCDs.

PS: Note the Pana paper has as a premise that small hi-rez CCDs can't be built that are sensitive. That was a nice argument years ago, but it doesn't apply today. Sony and Canon employ 1/3- or 1/2-inch CMOS chips that VERY easily provide 2MP to 5MP on a chip. And, EXMOR solves the sensitivity issue. The reference to the Panasonic paper has no validity on this thread.

[Ken Ross]

Originally Posted by Steve Mullen

But, I keep asking, how can three 1/2 MP CMOS chips that are only 1/6-inch be expected to up go up against a 1/3-inch, 3MP, EXMOR CMOS? Frankly, the chips would be fine for an SD camcorder. (Remember, 447K was a popular pixel count for DV.)

I have to agree with Steve on this one. I saw the direct evidence of the hit you take on picture quality with the reduction of chip size in the TG1. Here's a cam that has the same Exmor & Bionz processing as its bigger brothers, the SR11/12. The TG1 couldn't hold a candle to either one in any kind of light. When it came to low light..well, let's just say it wasn't too pretty. To add to this, the Panny uses low rez, pixel shift technology, which I've yet to see produce results similar to what a genuine "HD pixel count" sensor would.

Hopefully Panny has come up with a way around this, but I doubt it. But at least they were smart and added the VF and more manual controls. But as I've said before, all the manual controls in the world won't help if the camera's guts aren't capable of matching the best of the breed.

[Steve Mullen]

Originally Posted by Ken Ross

But as I've said before, all the manual controls in the world won't help if the camera's guts aren't capable of matching the best of the breed.

Ken and really I do agree on this. Unless a $1,000 camcorder is able to record video good enough to pass as that from a more expensive camcorder -- one will never be able to feel one can do paying jobs with it.

IMHO, Sony's SR clearly has quality as high as the V1. There really is no longer the need for three chips in the lower price category. I'd love a 1/2-inch 6MP CMOS single chip camera!

EXMOR not only provides a Bayer filter with enough samples to support Full HD recording, screen grabs on another site show how much lower video noise is at low light levels.

Sony shares EXMOR only with Nikon and it may be one of those FUNDAMENTAL technologies that will give it a very long-term lead. Of course, Canon is also a world leader in DSLRs so you can bet they aren't going to give Sony and Nikon their market without a fight. CES should be interesting.

Ken, Sony may be running close-out sales on the FX as they switch to the newer Intel in July. I'm still considering one given that it is IMPOSSIBLE to play a BD on a Mac!

[Ken Ross]

Originally Posted by Steve Mullen

Ken, Sony may be running close-out sales on the FX as they switch to the newer Intel in July. I'm still considering one given that it is IMPOSSIBLE to play a BD on a Mac!

Steve, do you mean the "FZ" notebooks? They just released the new lines of both notebooks and desktops. Not a ton of change, depending on the lines you're looking at. However there are some nice price cuts in some models.

I've decided on an 'all in one' Sony for the 'general purpose' computer, since both my wife and I fell in love with the simple (yes, Mac-Like) design. The Vaio laptop I bought for my son, with the T8300 processor, seems to play AVCHD as well as anything I've tried. Most of the times it will play it full-screen with few, if any dropped frames. But then at times you'll play the same clip and it will drop some frames.

I think the bottom line is, as you suggested correctly in another post, unless you've got a quad core, you won't get consistent playback of 1920X1080...forget about editing. But the T8300 is 'good enough' for my purposes. Remember I don't plan on doing much editing with AVCHD, so this isn't a deal break. At some point in the future, once the hardware and software catches up (Canopus, are you listening?), then I'll get the more souped up computer for that purpose. But the all in one Sony will serve as a general purpose, family computer. For that it's really ideal.

[Dave Rosky]

Steve, I'll be the first to admit here that I'm not an expert on the subject, and I'm often wrong on things, so I will pose a few more questions/comments (I think one of the nice things about this particular forum is that we can have a technical discussion without a flame war):

1. The Pana paper surely has been infused with some marketing hype in it, and it states a 1.5x increase in luminance resolution. This is almost surely an over-estimate. The 15% number, however, is likely an under-estimate. Based on simple calculations from the geometry, I would actually expect something more like 1.414x (square root of 2), which would still be quite a bit more than 15%. Is there a definitive paper somewhere that mathematically justifies the 15% number? If the 15% number comes primarily from competitors or other biased detractors, then it probably has as much marketing hype in it as Panasonic's 1.5x number, and the reality is likely to be somewhere in between (perhaps the 1.414x dictated by the geometry).

2. Even if the actual number is 1.414x, it would be true that there would still be less sensor resolution than a Bayer sensor with 3MP raw pixel count. However, the end results are what really matters, and they seem to show that the system provides enough sensor resolution to closely match most single-chip systems with 3MP raw pixels in overall resolution (at least for consumer level cameras). The final spatial resolution of the entire system is determined by more than just the sensor pixel density -- AAF, lens MTF and distortion, and analog processing prior to the ADC all contribute to the end result. The reviews I've seen where resolution was measured from a test chart show that the Canon single chip cameras do in fact beat out the Panasonic 3-chip cameras, but only by about 5-10% in spatial resolution. Yes, this is a measurable difference and might even be detectable to the eye under ideal circumstances, but it is small enough that other camera and PQ characteristics probably matter more. To paraphrase what someone here said earlier, if the controls aren't there to achieve what you want to achieve artistically, then 5 or 10% more lines of resolution aren't going to save you.

[Steve Mullen]

Originally Posted by Dave Rosky

Steve, I'll be the first to admit here that I'm not an expert on the subject, and I'm often wrong on things, so I will pose a few more questions/comments (I think one of the nice things about this particular forum is that we can have a technical discussion without a flame war):

1. The Pana paper surely has been infused with some marketing hype in it, and it states a 1.5x increase in luminance resolution. This is almost surely an over-estimate. The 15% number, however, is likely an under-estimate. Based on simple calculations from the geometry, I would actually expect something more like 1.414x (square root of 2), which would still be quite a bit more than 15%. Is there a definitive paper somewhere that mathematically justifies the 15% number?

2. The reviews I've seen where resolution was measured from a test chart show that the Canon single chip cameras do in fact beat out the Panasonic 3-chip cameras, but only by about 5-10% in spatial resolution.

Remember when Pana would state the resolution of the CCDs used in the HVX200. I have a math model that converts Pixel Count to TVL. It covers a dozen camcorders from SD to CineAlta. Once the measures of rez. were made I ran it backwards and got 960x540. At that point, Pana finally gave us the number. The model uses 115%. And, works for all pixel shift camcorders.

But here's an important point. I do not use STATIC REZ measures. My model assumes the camera is under a slight amount of movement. So, I use numbers like those Adam Wilt gets. He and I both agree static tests over-estimate rez. Video almost always involves motion.

By doing this discounts the potential contribution of pixel-shift. So my 115% is a parameter based upon "real world" conditions. If you see the world the way Adam and I do -- then this is fair. It rewards cameras that use more pixels.

One could also conclude that it's unfair. For example, the vast majority of rez. measures are static. Who are Adam and Steve to decide to test cameras another way?!? However, I do read that now that folks can SEE the difference between 1080 from the EX1 and the HVX200 1080 -- that folks are really SEEING how much more detail the EX1 captures. My feeling is the HVX200 is far better at obtaining a DVCPROHD 720p video.

If one uses static tests, the number is much close to your estimate. Which is why the Pana can measure much closer to the Bayer camcorder numbers.

Except the SR is not a typical Bayer camera. The ratio of green to red and blue pixels is 6:1:1 in the Sony. In the Bayer pattern, it’s 2:1:1. The 3X greater green information obviously increases luma resolution. A 3-chip FullHD camera would have 2MP of RGB.

The SR has 3,810,000 pixels. Divide by 8. That's 476,250 Red and 476,250 Blue. And, 2,857,500 Green. I'm not sure how HOW that gets to YUV, but 3MP of G has to provide a whole lot more luma than 1/2MP of G.

And, that was my real-point. Is it reasonable to think that ANY process can fill 1920x1080 recording with REAL luma detail from three 1/2MP chips?

PS: It is a perfectly reasonable business model to save tons of money by not doing R&D to create exotic tek if your market doesn't care about NEW tek. News is driven far more by workflow than CCD specs. And, the BBC and other such networks have specs that demand 2/3-inch chips, but say nothing about pixel count. In fact, some feel lower pixel counts look more like film. And, both markets don't like long GOP MPEG-2. DVCPRO HD and AVC-Intra are perfect for both.

The danger is that with the number of review sites and the web, customers may look past your marketing claims and compare based on how well camera test. (This is what happened with the JVC HD7.) So far, Panasonic hasn't tested well -- which is why my hopes are tempered.

[Ron Evans]

Steve, this may explain why my SR11 looks sharper than my FX1 at times. I was trying to identify with some shots the other day and could not decide whether it was the lower noise in the SR11, the colour depth or whatever but the SR11 picture really pops out on my Panasonic plasma when the FX1 is just a really nice picture. Can't really describe the effect but it could just be a little more resolution, shot was in 1920x1080? IF only the SR series had a few more manual controls, especially shockless exposure change etc.

Ron Evans

[Kevin Shaw]

Regarding tape vs tapeless costs, for the latter one should include the time and materials necessary to make at least two archive copies of your source footage. For an hour (8GB) of AVCHD that's going to be ~$2.50 for two dual-layer DVDs or ~$3 for corresponding hard drive space, plus the value of the time required to make the backups. That's not a cost advantage compared to recording HDV on $3-5 miniDV tapes with no archive time required.

Regarding editing, AVCHD is inherently difficult to work on directly and hence best converted to an I-frame intermediate. Current reports suggest this takes longer than capturing HDV to such an intermediate in real time, so no advantage there.

In the price range of good AVCHD cameras I'd suggest considering the Canon HV30 or Sony HC9 as more practical alternatives.