AVCHD vs. 3-chip

[Serge Victorovich]

Now cheap (>$200) Hybrid AVC cameras alike Aiptek HD Action has 5MP 1/2.5 inch image sensor and able to capture 1080p30 at 8Mbps or 720p60 at 6Mbps - encoder is Ambarella.

If you can change plastic lenses of Aiptek to good one...

http://www.youtube.com/watch?v=wcVo0U_xb0Q

Time to hack firmware to push bitrate up to 15-18Mbps in 720p60 :)

[Steve Mullen]

Originally Posted by Ken Ross

I'll take a small hit in sharpness vs the Canon for a more professional looking image any day of the week.

The only way to compare any piece of equipment with another is to put them both on "the bench." It is well known humans cannot be trusted to make objective judgements which is why ALL scientific experiments are done double-blind.

DATA doesn't lie. People do. And, people mostly lie to themselves. Every study of human perceptual judgment shows people SEE what they BELIEVE. Perception FOLLOWS cognition.

I don't know why lay people -- non psychologists -- have so much trouble accepting that what they SEE isn't of very much value, unless they are experienced in making judgements -- like a coffee taster. Having drunk a lot of coffee is not enough anymore than having shot a lot of video.

The "other" site -- using test equipment for an objective comparisons: the Canon was far more sensitive (lux for 50IRE) was 10 vs 14 which is a huge difference. Noise was almost identical -- again using test equipment for an objective comparison: 1.0125 vs 1.13.

The best you can get from people is a "preference." So you "prefer" the look of a camcorder that offers significantly less objectively measured resolution with slightly less noise (but with less chroma information and very possibly less resolution, which, of course, is HOW the noise is made less) and with a 1/2-stop less sensitivity. Fine -- you like the SR. So do I! You like it better than the Canon. I can't say that yet.

So, despite all the exotic Sony tek -- the SR, at best, matches Canon's old fashioned tek. But, that is the "Sony way." Use R&D to find a different way to do something. Even if it doesn't really work better -- marketing uses all the neat sounding stuff to DIFFERENTIATE and justify a higher price.

PS1: the 2x4-cell filter used by Sony on it's ClearVid chip MUST return less luminance rez than the 2x2 used by Canon and JVC. Which is why UNDERSTANDING tek allows one to reasonably well predict test measures which in turn reasonably well predict performance. It's all physics in the end. :)

PS2: I just ordered a VAIO since Sony had a great price.

[Dave Blackhurst]

And we all know of course that a machine can "accurately" determine the aesthetic value of an image...

"Tests" are still run by PEOPLE, and you'll never get me to believe that the bias of a tester never enters the equation... figures don't lie, but...

In the end I'll trust MY evaluations, which of course have every right to be "different" from anyone else's. I personally just see things with Canon's engineering choices and color rendition that I dislike, other's don't like the "Sony look" or engineerign choices. Tah-may-toe, to-mah-to...


You can have a lab or a warehouse full of test gear, but it's the ol' "Mk 1 eyeball" with all it's flaws and potential for errors that counts in the end.

[Robert Young]

I think that the test data attempts to give us an objective anchor that helps to keep our various opinions somewhat moored to a common reality.

However, the eye of the beholder is indeed powerful and frequently has veto power over the physical "facts".

That's what is so beautiful about a camera like the XDCam EX1. It's like a blank canvas that comes equipped with a full palette- you can paint whatever "look" appeals to you.

Consumer Cams like the little Sony SR12 and it's Canon equivalent arrive more or less prepainted. If you like the provided combination of color, gamma, detail, noise, etc.; that's the one you will probably take to the dance, even tho the other one maybe has more control, technically higher rez, higher sensitivity, or whatever. There is no right answer. Like Mac vs. P.C., like what really constitutes the "film look", like what's the best gamma- these discussions go on forever, because they can't be settled definitively- it's personal.

You should have what you like. And in 2 years when you like something different, you should go out and get that.

What we try to do is at the interface between technology and art. Without the art, the technology would have little value by itself. Sloppy, shakey camera work can devalue the most fabulous shot by the best camera in the world. And conversly the exactly right shot, with the right light, well executed, can be captured by some sort of key ring camera and be a masterpiece. I think it is supposed to be like this- room for all to go with the look that they each like. It's fabulous!!!

[Kevin Shaw]

Measuring video noise is like trying to define how pretty a sunset looks: there may be objective factors but a human interpretation is required to give any real meaning. The double-blind test here would be to have several people view clips from different cameras without knowing which was which, and comment on both the quantity and quality of image artifacts. And as we've discussed here, the quality of the image is also dependent on the skill of the camera operator to get the most out of the equipment.

[David Andrews]

While on the subject of tests, has anyone examined the difference between xyColor off and xyColor on which is available with the newer Sony AVCHD cameras? If so are they able to explain just what that difference is?

[Steve Mullen]

First, folks are confusing a TOOL with its use. The former is the realm of science and the latter of art. In the world of art it's all subjective.

The goal of science is to make a TOOL that imposes nothing of itself. An amplifier should be as transparent as a piece of wire. A camcorder should record exactly what's in its view. Nothing more and nothing else.

Of course, mics and lenses are often bought and used based upon HOW they sound/look. Most of can't buy that many camcorders and lenses.

Measuring noise is nothing new. Each has a signature and so can be measured. On analog tape there were two types: AM and PM. PM was subjectively the worst. Both were measured. In fact, it's safe to say no review of an analog VTR was without both measurements.

Same for Wow and Flutter. Same for sensors. The idea that because something seems complex it can't be measured is simplistic. For example, if you have the money you can measure compression codecs. How do you think MPEG-2 and MPEG-4 were developed? There are a standard set of test images and sequences everyone uses. Then codecs are measured.

Without standardized Test and Measurement no product would come to market. No could QC be run on the production line. Networks would not buy 100 camcorders because a few shooters played with it for a few days.

For example, person A can claim that the SR12 has good AWB. Person B can claim it has "blue push." You as a reader can't tell a thing by these opinions. If person A has an uncalibrated monitor it most certainly has red push. Red helps warm the blue and his eyes tell him all is well. Person B has a calibrated monitor -- now the image is seen as it IS.

But there's no need to ever SEE a frame of video. Look at the 2 vectorscopes below of AWB and MWB. You can SEE immediately that the uncentered AWB is not correct. So one can conclude that the idea of running the SR12 in AWB is dependent on the viewing monitor and the BELIEF about skin looks like plus what an individual believes is ACCEPTABLE. (When I claim skin is white, look at the photo below.)

Plots and graphs are as SEEable as video itself. Look at the static rez shot of the SR12 vs HF10 on the "other" site. Not only can you SEE the Sony has lower resolution one can SEE by looking at the concentric circles it has diagonal aliasing. Look at the plot of frequency response on a German site. You can SEE what is happening to the signal.

Some of the best testing is done by the BBC. Want to SEE how the HVX200 REALLY performs: http://www.bbc.co.uk/rd/pubs/whp/whp034-add18.shtml

How about a Canon: http://www.bbc.co.uk/rd/pubs/whp/whp034-add25.shtml

How about a Sony: http://www.bbc.co.uk/rd/pubs/whp/whp034-add20.shtml

The idea that camcorders can best be judged by shooters is dear to the heart of shooters. It's just a conceit.

What humans can do is judge manual operation. Touchscreen vs joystick is subjective. Or, is it. Human factors researchers can easily measure the time and accuracy of making adjustments under a wide range of illumination.

PS: Has anyone seen the type of field problem in the uploaded pix?

[Duane Prince]

"PS: Has anyone seen the type of field problem in the uploaded pix?"

Can't say I've seen this before, but It's obvious the red channel is out of "sync" with the blue & green channels. Is this a progressive frame, & which camera shot this, if you know.

[Steve Mullen]

Originally Posted by Duane Prince

"PS: Has anyone seen the type of field problem in the uploaded pix?"

Can't say I've seen this before, but It's obvious the red channel is out of "sync" with the blue & green channels. Is this a progressive frame, & which camera shot this, if you know.

The Sony SR12. It's about 10 frames from the end of a clip converted to AIC so I doubt it's in the actual AVCHD file. Amazing -- AVCHD is the first codec that one couldn't natively play in a player without buying something.

[Giroud Francois]

yes this seems to be a known problem on avchd camera.I have seen a french video forum where some guys where asking same question.
Seems it happens when doing a fast pano.

[Bob Diaz]

Originally Posted by Steve Mullen

First, folks are confusing a TOOL with its use. The former is the realm of science and the latter of art. In the world of art it's all subjective.

The goal of science is to make a TOOL that imposes nothing of itself. An amplifier should be as transparent as a piece of wire. A camcorder should record exactly what's in its view. Nothing more and nothing else.
...

Steve, you've hit upon an interesting point there, the "Art" vs. the "Science" of the technology.

The science can give us the parameters of a given item like, lines of resolution or noise level. However, we lack a good working model as to how the impacts the viewer when all the parameters are combined.

The art tells us how it impacts the viewer, but can't be measured like voltage. A hero rushing to defuse a bomb with only 60 seconds to go may take 80 seconds of screen time, but few will sit there with a stop watch and catch the error. The excitement of the action, overrides the time distortion.

The only way to compare any piece of equipment with another is to put them both on "the bench." It is well known humans cannot be trusted to make objective judgements which is why ALL scientific experiments are done double-blind.

DATA doesn't lie. People do. And, people mostly lie to themselves. Every study of human perceptual judgment shows people SEE what they BELIEVE. Perception FOLLOWS cognition. ...

So true, people color their decision with what they think is true. The only real comparison test is where the viewers don't know which video goes to which camera.


Still, I wonder if we obsess over which camera is "the best" and fail to focus on generating a video that grabs the viewer so strongly, that they never notice any difference there might have been...


Bob Diaz

[Ken Ross]

Originally Posted by Steve Mullen

The Sony SR12. It's about 10 frames from the end of a clip converted to AIC so I doubt it's in the actual AVCHD file. Amazing -- AVCHD is the first codec that one couldn't natively play in a player without buying something.

Steve, why do you persist with AVCHD. The format seems to annoy you endlessly and this comes across clearly in virtually every thread you post. If I felt as you do, there is no way I'd stick with anything AVCHD.

P.S. Never saw anything like that posted shot. The AWB on my SR12 looks on-target the majority of times on my calibrated Pioneer Kuro...surely more so than the HF10 did on that same display.

By the way, just another data point, I've found over the years that two cams of the same company, same model, can have somewhat different AWB setups. My video buddy and I have tended to get the same cams over the years and rarely did the colors look exactly the same on the same display pointing at the same scene. The same was true of sharpness. So I even take some of those 'exact' charts with a grain of salt. They may be accurate for THAT particular unit, but should not be taken as gospel for any other unit.

[Steve Mullen]

Originally Posted by Bob Diaz

Still, I wonder if we obsess over which camera is "the best" and fail to focus on generating a video that grabs the viewer so strongly, that they never notice any difference there might have been...


Bob Diaz

I think we talk about cameras because in reality -- we can't talk about art. Right now even the best HD website, ExposureRoom uses an H.264 upload and a Flash encoded download that can't play smoothly. We can't yet share our art. Plus, when we watch deinterlaced footage on our computers, it's not really the same as watching on an HDTV.

And, I'm not trying to find a BEST camcorder. Having reviewed camcorders since 1992, I never pronounce a WINNER. What I try to do is to characterize -- in words -- how a camera performs. The reason I prefer measurements is because it eliminates my current feelings and what marketing materials I've read. And, if one is familiar with measures they really do paint a picture of what one sees. Or better put, they flag things to look for.

When the BBC can't come-up with resolution numbers for the HVX200, but can for every other HD camcorder -- it screams there is something weird about this camera. That tells one to look very very carefully for aliasing when you visually test.

And, I really overstated my case! :)

Of course, one doesn't use ONLY numbers. Sometimes measures point-out problem areas to confirm by looking. Other times, one sees something and the measures confirm it -- or don't. That's why reviews take a month and not a day or two!

I'm still deep in testing the Sony SR. The real key is not to decide if it is the BEST, but can one find ways to get it to perform like it were the best. It fact, for the 24p shooter no matter how good it the SR is -- it isn't for you.

[Ken Ross]

Originally Posted by Steve Mullen

The best you can get from people is a "preference." So you "prefer" the look of a camcorder that offers significantly less objectively measured resolution with slightly less noise (but with less chroma information and very possibly less resolution, which, of course, is HOW the noise is made less) and with a 1/2-stop less sensitivity. Fine -- you like the SR. So do I! You like it better than the Canon. I can't say that yet.

So, despite all the exotic Sony tek -- the SR, at best, matches Canon's old fashioned tek. But, that is the "Sony way." Use R&D to find a different way to do something. Even if it doesn't really work better -- marketing uses all the neat sounding stuff to DIFFERENTIATE and justify a higher price.

Frankly Steve, I find some of this just plain silly, not to mention insulting...and I'm being quite restrained here my friend. You have some very very obvious issues with Sony and that shows in most of your Sony related posts. I have no issues either with Sony or Canon or Panasonic or ANY manufacturer and so therefore have no bias. I choose the camera that performs the best to my eyes. I would never go to you for an objective assessment on Sony products given your obvious issues with them. I've been doing video for many years and have ZERO allegience to Canon, Sony, Panasonic, Toshiba or whatever brand you "think" I have a preferece for. As I've said before, my last two HD cams have been Canons. Why? Because they looked better to me than the comparable Sonys of that time. No Steve, NOT because I have an allegience, bias, financial stake or what have you with Canon, but rather because they seemed to PERFORM BETTER than the comparable Sonys of that time.

So now we have the new Sonys and to my eyes they produce a more professional looking picture. Do I care if it's the result of BIONZ or EXMOR processing? Of course not! I could care less about marketing names or schemes, it's the performance that counts. ALL camcorder manufacturers are guilty of shamelessly promoting equipment with catchy names or slogans, Sony is not alone in this. Do you recall Canon was guilty of calling their cams 'full rez' when they were not? They all do it Steve. Steve, I never ever buy ANY equipment based on frequency response graphs (audio OR video) or graphs depicting any other performance parameter, nor do I buy because someone on the net, in a somewhat pompous tone, TOLD ME it was better based on those charts...and that same person told me my eyes were deceiving me for liking or 'thinking' the other cam could produce a better picture. Your words clearly insinuate than anyone that prefers the Sony picture must be doing so despite the 'significantly less resolution', despite 'less chroma resolution' etc. etc. By the way, your comment regarding the Sony having 'significantly less resolution' is also inaccurate since the measured values on the site that most people seem to use for this nonsense, showed a difference of less than 10%...hardly 'significantly less'.

No my friend, I've learned there is far more to audio or video than a bunch of charts and graphs. People who buy A/V equipment this way are bound to make mistakes. Yes, they do have a role when used intelligently, but to use them to make you 'think' you should be seeing something you're not...well, that's the height of absurdity!!!

I did more A/B tests to determine what ACTUAL detail can be discerned between the HF10 and the SR12 than you could shake a stick at, and no matter how many tests I conducted, there was NO CONCLUSIVE PROOF on my calibrated 60", 1080p, Pioneer Kuro that either cam had an advantage. I shot street signs, liscense plates from varying distances in real world conditions and neither cam showed any more or less detail on any consistent basis. Guess what set of parameters I'll use Steve in determining in REAL USE what cam shows more detail, your charts or my eyes? My eyes viewing material on a 60" 1080p plasma win every single time. I showed these same tests to two other people and both could see no consistent difference in detail.

The fact is that when many of the picture performance parameters are combined in the final picture (color, detail, gamma, dynamic range etc.) there is no graph, no pie chart, no nothing anywhere in the world that can tell you which picture is better. You may wish that were so, but the bottom line is the human eye is the final determinant as to which picture is better. Yes, obviously personal preferences come in to play, but those exposed to professional video on a calibrated display generally know which picture looks more 'professional'.

So please Steve, don't tell me what you think I 'should' be seeing and how my eyes are either biased or lying.

[Dave Rosky]

I think there's a middle ground.

Lab measurements are a good thing. Although it's possible for the interpretation of lab tests to be biased, the raw data itself is usually not, and the lab measurements give a needed description of how the hardware behaves along a number of different performance metrics.

The problem is that the subjective reaction to a video image is a combination of many different aspects - sharpness, color, tonality (gamma, etc.), noise, exposure, saturation, etc. Different people react differently to these things. One person might be fairly tolerant of color error while another will be bugged by the slightest shift, likewise for other parameters. Person A might be annoyed by anything but the most tack-sharp image, while person B might actually prefer a slightly soft image.

Test measurements are a good guide, and may help you narrow your search based on the specs that are the most important to you, but in the end the equipment with the best specs might not be the equipment that pleases everyone. Tube preamplifiers, for example, have far worse performance than most high quality modern amplifiers from the point of view of lab measurements, yet there are still people who like the specific response even though it is technically less accurate and insist on using them.