With stabilization Sony HC1 has 25% lower res INCORRECT

[Radek Svoboda]

The image stabilization is electronic. It says sensors are used.

Here is translated Japanese brochure. See all detailed information. U.S. product pages are joke. 10:1 zoom in U.S. brochure is 41-480 mm.

http://hdvforever.com/hdv/hvra1j/default.htm
http://hdvforever.com/hdv/hdrhc1/default.htm

Animated pictures in brochures mean video. It appears there are two video modes, full scan, uses 1920x1080 pixels and 1440x810 mode. Image stabilization needs extra pixels; Super Steady Shot is probably using those extra pixels. So with SSS resolution drops 25% horizontally and vertically. Without SSS you'll get same resolution as FX/Z1.

Effective pixels in FX/Z1 always 1440x1080, because horizontal pixel shift.

Because Bayer filter, which means 25% resolution decrease, in HC1 you'll get 1440x1080 effective pixels, without SSS and 1080x810 with SSS.

There is something else. Full scan mode is described in prosumer camera; consumer model could have lower resolution all time, but it is unlikely.

More notes: The microphone is very short shotgun with 2 elements for each channel. That is revolutionary. Camera seems have manual exposure. It does have zebra patterns.

Radek

[Chris Hurd]

So with SSS resolution drops 25% horizontally and vertically.

That's a misleading statement. It almost seems as if you see this as a fault or an issue somehow. The CMOS image sensor in the C1 is identical to the A1. The resolution is *not* reduced; even with Super SteadyShot ON, the video image is still 1440x1080, which is equal to this camera's big brother, the FX1. Remember, Super SteadyShot does *not* impact image quality or resolution. I have therefore edited the title of your post to reflect that your claoim is not correct.

Your statement about the color filter is also misleading and fundamentally incorrect. The Bayer pattern incurs some slight loss of resolution, but you are ignoring the complicated reconstruction algorithms employed by the DSP to recover that information.

I really do not appreciate the sort of sweeping claims you're making, not just because they're technically inaccurate, but primarily due to the fact that you have not yet actually seen the image output of this camcorder with your own eyes. This is the DV Information Network, not the DV Supposition Network. Be advised that I have personally escorted a few misguided individuals from our boards for the exact same sort of misinformation that you're attempting to spread here.

My primary issue with this is that you're stating your claim as if it were a fact, even though you really do not know it to be true. If you had phrased it in the form of a question, that would be altogether different -- even encouraged, as I appreciate all questions. Radek, you list your occupation as a student. I'd like to encourage you to step back and assume a role of *learning* around here. This can be an excellent place to learn. Thanks in advance,

[Michal Laskowski]

Super SteadyShot does *not* impact image quality or resolution

I have not heard of any electornical image stabilization (EIS) which is not using additional pixels to do its job. It's really a mistery if HC1 doesn't have optical image stabilization (OIS), because CMOS resolution is 1920x1440 (can we be sure about it?). We've got as much as around 400 pixels to stabilize image vertically (we need "only" 1920x1080), but nothing for horizontal shake. To keep proper aspect ratio while decreasing horizontal resolution we need to decrease vertical resolution - so we've got something what Radek mentioned (1440x810). But the output is always 1440x1080. With interlaced content it's unacceptable to resize image verticaly - the loss of quality is enormous and it's almost not possible to do this without deinterlacing first. I can't belive Sony could do this, it must be incorrect.
At 2000$ tag price it would be really a surprise to not have OIS (is there any other consumer camcorder at this price range without it). But maybe Sony engineers disovered that only vertical shake is important to stabilize image ? We have to wait to see it.
BTW. As some info appears about limited manual control available on HC1, it looks more as HC1 is similar to HC90 with exception to:
- CMOS intead of CCD (is CMOS cheaper?)
- DSP microprocessor ready for HD (costs about 50-80$?)
- worse low-light performance (7 lux versus 5 lux)
So what justify 2000$ price, HC90 with 3Mpx CCD (able to do HDV 1920x1080, but lacking proper DSP) costs 600$.
I hope that most informations available now about HC1 are incorrect.

[EDITED]

Because Bayer filter...

We shouldn't forget that FX1 has 3 CCD with 960x1080 resolution. This means that horizontal resolution is interpolated! As far as I know it's achived by shifting green CCD. With smart algorithm I think you can get more horizontal details from 1 CCD with Bayer filter than from shifting one CCD. The advantage remains in vertical resolution, here FX1 doesn't need interpolation, HC1 needs. But as Chris mentioned the loss is almost unnoticable.

Regards

[Ron Evans]

CHris I agree with some of your comments AND some of Radek's. NEITHER of you know if pixels are reduced when Super Steady Shot is on for this camera at this time. The camera appears to have electronic image stabilization so must use extra pixels from somewhere for this to work. These could be extra pixels that are there all the time or as Radek infers reduce effective source pixels while SSS is on. Please don't tell me that Super Steady Shot is optical and doesn't use extra pixels, I have a Sony TRV50 with Super Steady Shot and it is electronic, so is my Sony PC10. However both these cameras use large pixel count CCD's giving plenty of data for the DSP to work with and retain full pixel count for the DV image. In this case you are correct that the resolution is not reduced with Super Steady Shot but there is clearly less data available to the DSP for image enhancement so this could impact exposure depending on light levels. Your comments about the tone of Radeks post have some merit( have you thought that English may not be Radek's first language?) We will all have to wait just a few weeks to find out. Until then we will all make the assumption that information from any source on the WEB( that includes company sites) is just that. It becomes fact when it becomes absolutely verified personally or by several independed sources that say exactly the same thing( one source doesn't cut it even if its Adam Wilt!).

Ron Evans

[Radek Svoboda]

Originally Posted by Michal Laskowski

CMOS resolution is 1920x1440 (can we be sure about it)

The chip has square pixels and 4:3 aspect ratio; at 16:9 you get 1920x1080 pixels. If you go the the Lorin's Japanese site on the prosumer model, it's well explained. For video, 1920x1440 pixels are used in "full scan mode". There is another video mode with 1440x810 pixels. Does someone see anything different? The translation to English is not very good.

Radek

[Chris Hurd]

Super SteadyShot is Sony's brand for its highest quality image stabilization. It can be either an optical process (OIS) or electronic process (EIS). In either case, Super SteadyShot implies that there is no discernible loss in image quality. When it's electronic, it usually involves a surrounding area of pixels on the the image sensor that have not been previously dedicated to the image area itself.

Besides, normal EIS technology has dramatically improved within the past 24 or 36 months and is nowhere near as bad a hit on image quality as it used to be, thanks to primarily to improvements in the DSP. That's a crucial concept to understand. Image quality trade-offs due to EIS are inherent in older camcorders, not new ones. There is no longer such a dramatic difference between OIS and standard EIS as there used to be.

I firmly agree with Ron Evans that nobody really knows for sure about the C1 and A1 just yet, and that we'll only have to wait just a few more weeks to find out. My primary point is that I would like to strongly caution some of our members here to put forth their ideas in the form of a question -- think of the American game show, "Jeopardy" -- I'll take image stabilisation for $200, Alex. There is no faster route to alienization around here than to preach conjecture as fact; especially when it's inaccurate. Thanks in advance,

[Michal Laskowski]

The chip has square pixels and 4:3 aspect ratio; at 16:9 you get 1920x1080 pixels. If you go the the Lorin's Japanese site on the prosumer model, it's well explained. For video, 1920x1440 pixels are used in "full scan mode". There is another video mode with 1440x810 pixels. Does someone see anything different?

I've already knew that CMOS in HC1 has 4:3 aspect ratio ... I was rather referring that Sony could announce only effective pixel count (1920x1440), but the total nubmer of pixels can be higher and maybe is used for stabilization. I don't think there is any 1440x810 video mode. 1440x810 is a resolution of pictures/stills taken during video recording ("animated pictures"?). The stills are resized from 1920x1080 to 1440x810 to nicly remove interleced mode that occur during video recording. At the same time 1920x1080 is squeezed to 1440x1080 to apply HDV compression.

Regards

[Gerald Lunn]

As I understand from the specs, the CMOS sensor has 2.97M gross pixels - but only uses 1920x1080 - or just over 2M pixels for normal 16x9 video. So it seems to me that there are plenty of pixels available for electronic stabilization without compromising the resolution.

Considering the 1440x810 still resolution, I think that because the H-resolution of the video must be reduced from 1920 to 1440 for recording in the HDV standard, then all video data would go through this data reduction as a first step. Then in order to keep a square pixel shape for the still picture the vertical resolution must be similarly reduced (x0.75) hence 810.

However, there is also the possibility that Sony actually converts the vertical resolution of the 1080 video frame to 810 to reduce the processing power for the MPEG encoding required for recording - then interpolates back after MPEG encoding (much easier because much less info in the MPEG encoded signal) to record the correct 1440x1080 on the tape.

Of course, 1440x810 derived resolution would still be recognised as HD because the unofficial definition of HD is anything greater than 1M pixels.

If you are not recording video then you can get the full pixel count in a still picture - i.e. 1920x1080 in 16x9 mode or 1920x1440 in 4:3 mode.

Gerald

[Gerald Lunn]

As I understand from the specs, the CMOS sensor has 2.97M gross pixels - but only uses 1920x1080 - or just over 2M pixels for normal 16x9 video. So it seems to me that there are plenty of pixels available for electronic stabilization without compromising the resolution.

Considering the 1440x810 still resolution, I think that because the H-resolution of the video must be reduced from 1920 to 1440 for recording in the HDV standard, then all video data would go through this data reduction as a first step. Then in order to keep a square pixel shape for the still picture the vertical resolution must be similarly reduced (x0.75) hence 810.

However, there is also the possibility that Sony actually converts the vertical resolution of the 1080 video frame to 810 to reduce the processing power for the MPEG encoding required for recording - then interpolates back after MPEG encoding (much easier because much less info in the MPEG encoded signal) to record the correct 1440x1080 on the tape.

Of course, 1440x810 derived resolution would still be recognised as HD because the unofficial definition of HD is anything greater than 1M pixels.

If you are not recording video then you can get the full pixel count in a still picture - i.e. 1920x1080 in 16x9 mode or 1920x1440 in 4:3 mode.

Gerald

[Michal Laskowski]

Considering the 1440x810 still resolution, I think that because the H-resolution of the video must be reduced from 1920 to 1440 for recording in the HDV standard, then all video data would go through this data reduction as a first step. Then in order to keep a square pixel shape for the still picture the vertical resolution must be similarly reduced (x0.75) hence 810.

I don't think the reason was to have square pixels. There is a need to somehow defeat deinterlacing artefacts (stills are progessive, video interlaced), so reducing resolution after deinterlacing to 1440x810 hide them. Otherwise you would get really badly looking shoots (stills).
It doesn't matter if still are independently done from 1920x1080 to 1440x810 or after the first step of reducing video to 1440x1080. The second option is practical - needs less power and gives exactly the same output.

However, there is also the possibility that Sony actually converts the vertical resolution of the 1080 video frame to 810 to reduce the processing power for the MPEG encoding required for recording - then interpolates back after MPEG encoding (much easier because much less info in the MPEG encoded signal) to record the correct 1440x1080 on the tape.

Not possible. HDV standard has fixed resolution 1440x1080. Interpolation can't be done after encoding. It must be encoded to 1440x1080 with MPEG2 and then directly go to the tape. Today it's not a problem for DSP to encode 1440x1080 MPEG2.

Regards

[Tommy Haupfear]

Regardless, I think the HC1 should have optical image stabilization for the price but I really can't recall any single CCD/CMOS Sony cam with anything other than electronic/digital image stabilization.

[Gerald Lunn]

Originally Posted by Michal Laskowski

I don't think the reason was to have square pixels. There is a need to somehow defeat deinterlacing artefacts (stills are progessive, video interlaced), so reducing resolution after deinterlacing to 1440x810 hide them. Otherwise you would get really badly looking shoots (stills).

I certainly agree on the interlacing artifacts, so the resolution must inevitably be reduced. However in addition, for jpeg stills to look correct on a display/printer etc, the pixels must surely be square so there is an additional reason to choose that particular format. For stills from video recorded in DV format on the HC1 they choose 640x480 rather than 720x480 for the same reason.

The thought about power reduction in the mpeg conversion was that there is always an electrical power budget for a portable device so any way to reduce the actual power consumption of the conversion process would be welcomed. Less processing speed = less power consumed! I forgot however that the HDV signal is interlaced so it would actually be a lower data rate than a de-interlaced 810. Anyhow, somehow Sony has managed to get the total power down to under 6W (excluding LCD) which is quite an achievement. FX1 is 7.4W under same conditions, with most of the additional power being for CCD processing probably (much of the CMOS initial processing is done on the chip). I understand from a FX1 writeup that the chip used for conversion to mpeg actually takes only 200mW so that supports your argument that mpeg conversion is no problem!

Gerald

[Andre De Clercq]

The stabilization concept in the HC! is most probably one that combines OIS and motion vector analysis. (cfr Canon super range OIS) The system involves a standard OIS with gyro sensors and VAP or lens shift optical correction elements. In addition those systems perform motion vector analysis for the lower vibration frequency components (gyro's have difficulties to detect low frequency shaking), but instead of changing the read out zone of the sensor the motion vector information is fed back into the OIS actuators. So. OIS is used EIS is partly involved. and there is no need for extra pixels,

[Radek Svoboda]

On resolution loss with single chip camera:

http://www.dvinfo.net/conf/showthrea...page=193&pp=15

Originally Posted by Steve Nordhauser

On Bayer, I've heard the 25% number before, also 30%. It depends on what color you look at. For red or blue primary colored objects, it is much worse. For something with a mix of RGB, it might be nearly zero loss with the proper algorithm. So, I'll give you a very qualitative 'it depends'.

Radek

[Chris Hurd]

Originally Posted by Andre De Clercq

OIS is used EIS is partly involved. and there is no need for extra pixels,

Excellent explanation, Andre, and very much appreciated. Many thanks,