[Recovered Thread: 116056]

[rgvcam]

gdo wrote:

Looks good! Way better than any "best" shot from a PVR. One other thing I recall, before giving this a rest, I recall fishycomic was able to load the PVR firmware on the MPVR on one of his post. According the his report, he indicated he encounter "grainy video like the PVR" after doing the upgrade. Of course running a firmware that was not design for the specific hybrid will probably not stand up as a valid argument but it is interesting that he notice the "grainy video" after loading the PVR firmware to the MPVR. This to me is another indication that the firmware can/may improve the PVR video.

I have seen the results of the PVR firmware on an MPVR and have to say that the effect wasn't qute the same as the PVR. It was more of a psychodelic effect. Remember the Sensor on the MPVR is 3MP as against 5MP on the PVR so it was expecting 5MP worth of pixels input which is probably why he got those funky effects. Therefore, in my opinion, it is not a valid test and one they would quite rightly dismiss anyway.

[sgspirit]

How many ways can the manufacturer admit the camera is a pos, while trying to make it look good?

The video subsystem is very different from the still picture mode. To start with, video uses only a tiny 640x480 pixel portion of the sensor. It takes 30 stills per second on that area, and uses software to compare the frames to each other, then throws out most of the data on the basis of that comparison. The still mode does nothing like this. Both do use the sensor information to automatically adjust for light type and brightness. Not to mention image stabilization. (Does shutting off the IS have any effect?)

I have a Digilife DDV-920, with a ccd sensor. When I take video with the sun or bright reflections in it, the video frames show light rays extending vertically above and below the highlight, to the top and bottom of the image. My Samsung i6, also with a ccd sensor, does the same thing to a lesser degree. Neither of these cameras do this when I take stills. Nor have the cmos hybrids I've had, do this in video or stills, and this is a known issue with ccd sensors. So it's entirely possible to have problems with video or stills that don't show up in both.

[rgvcam]

sgspirit wrote:

How many ways can the manufacturer admit the camera is a pos, while trying to make it look good?

The video subsystem is very different from the still picture mode. To start with, video uses only a tiny 640x480 pixel portion of the sensor. It takes 30 stills per second on that area, and uses software to compare the frames to each other, then throws out most of the data on the basis of that comparison. The still mode does nothing like this. Both do use the sensor information to automatically adjust for light type and brightness. Not to mention image stabilization. (Does shutting off the IS have any effect?)

I have a Digilife DDV-920, with a ccd sensor. When I take video with the sun or bright reflections in it, the video frames show light rays extending vertically above and below the highlight, to the top and bottom of the image. My Samsung i6, also with a ccd sensor, does the same thing to a lesser degree. Neither of these cameras do this when I take stills. Nor have the cmos hybrids I've had, do this in video or stills, and this is a known issue with ccd sensors. So it's entirely possible to have problems with video or stills that don't show up in both.

I'm not so sure they do just use a small 640x480 portion of the sensor. I am pretty sure they use all the 5MP although I would be happy to proven wrong. If they did, then that would be an even tinier part of an already tiny ccd/cmos sensor which would make the device terrible in low light (not that it's great to start with) especially at the exposure rate of 30fps. Also I can't imagine the digital zoom would work that well with the pixel area of just 640x480. Having to scale different sized sensors down to video size would provide a plausible reason why these two cameras act differently whereas if they had the same 640x480 area, then I can't see the lens assembly creating the distortion you see in the video.

I remember a conversation with my Aiptek and being told that the MPEG encoding was all in software on these devices. I think the Sanyo's have a dedicated chip which no doubt is one of the reasons they have the edge in quality as well.

Anyway it's all speculation unless we are given a schematic and details of how they put these things together.

[sgspirit]

We had a discussion here one time about that. The conclusion was that if they are using an area on the sensor larger than 640x480 to get a 640x480 image, there would be no reason to be unable to do digital zooms with no loss of quality until you got down to a 640x480 area. The fact there is such quality loss suggests they are using a 640x480 area for video. But it's also possible they use a larger area, but haven't implemented lossless digital zooming.

Most of the Samsung Minkets use only a 640x480 or 800x600 sensor. So they take only low resolution stills, but still take what I understand is very good quality mpeg4 video. Their file sizes are twice as large as our hybrids typically do, so they're using less compression and have higher capacity computer systems.

[rgvcam]

I got a response from my contact at Aiptek USA and he said that as far as he was aware, the devices use the entire sensor area in video mode and they don't tend to get cropping in VGA mode which might tend to occur if just a portion was used.