I had recommended closing this Karmic bug because I cannot prove it's a regression. However, I have now duplicated the dark webcam behavior on Karmic with guvcview only, and it's very interesting. The problem can be demonstrated very simply:
- Start guvcview
- Select the video & files setting tab
- Change the frame rate setting, trying each of the choices
With the Chicony CHF7129 camera, in low light (think a single 60-90 watt bulb):
- 30 fps and 5 fps settings give a reasonably well exposed image
- 20, 15, and 10 fps settings fade immediately to almost completely black
(just a few reflected highlights from glasses and my balding forehead)
Doing the same test on Karmic with my Logitech Notebook Deluxe Webcam, you get what you'd expect -- the best lighted image is at 5fps and it gets progressively darker as you move toward 30fps. The changes as you increase the fps setting are gradual, and even the darkest one isn't awful or as bad as the CNF7129 at the non-working fps settings.
But doing this test with the CNF7129, you get well exposed and surprising close images at both 30fps and 5fps, and an almost completely black image at all the other fps settings. The 30fps image is the best exposed as if "autoexposure" is on, while the 5fps image is just a little bit dark, but usable. All the other fps settings are unusably dark, but there's still an image. My conjecture is that skype is failing for the same reason -- effectively picking an fps setting that goes extremely dark -- but I can't prove that.
The testing I performed using v4l2ucp gives similar results.
Also, the room light has to be below some threshold to test, but a threshold which really should work ok. It's easy to test and find everything is working, when it's only working because there's a lot more light during the testing. It fails under typical home (incandescent) lighting (60-90 watts), and in coffee shops at night if you don't have a spotlight pointed right at you.
This is a very popular bug (well, unpopular, but often reported), on a very large number of Asus and other laptops. If a fix can be found for this, it makes *a lot* of people happy... (!)
I also tested Windows XP SP3 on the same 1000HE netbook/CNF7219 combo (yeah, I know... sorry, but I needed another comparison), and the Windows driver gets the best image in low light. It's just my impression, but it looks like Windows selects a frame rate in low light that's even lower than 5fps. (Move in front of the camera, and the image gets badly blurred/smeared.)
I've submitted this new information to Hans de Goede (video4linux) as well, and he recommended expanding the discussion to include Laurent Pinchart, which I have done.
I had recommended closing this Karmic bug because I cannot prove it's a regression. However, I have now duplicated the dark webcam behavior on Karmic with guvcview only, and it's very interesting. The problem can be demonstrated very simply:
- Start guvcview
- Select the video & files setting tab
- Change the frame rate setting, trying each of the choices
With the Chicony CHF7129 camera, in low light (think a single 60-90 watt bulb):
- 30 fps and 5 fps settings give a reasonably well exposed image
- 20, 15, and 10 fps settings fade immediately to almost completely black
(just a few reflected highlights from glasses and my balding forehead)
Doing the same test on Karmic with my Logitech Notebook Deluxe Webcam, you get what you'd expect -- the best lighted image is at 5fps and it gets progressively darker as you move toward 30fps. The changes as you increase the fps setting are gradual, and even the darkest one isn't awful or as bad as the CNF7129 at the non-working fps settings.
But doing this test with the CNF7129, you get well exposed and surprising close images at both 30fps and 5fps, and an almost completely black image at all the other fps settings. The 30fps image is the best exposed as if "autoexposure" is on, while the 5fps image is just a little bit dark, but usable. All the other fps settings are unusably dark, but there's still an image. My conjecture is that skype is failing for the same reason -- effectively picking an fps setting that goes extremely dark -- but I can't prove that.
The testing I performed using v4l2ucp gives similar results.
Also, the room light has to be below some threshold to test, but a threshold which really should work ok. It's easy to test and find everything is working, when it's only working because there's a lot more light during the testing. It fails under typical home (incandescent) lighting (60-90 watts), and in coffee shops at night if you don't have a spotlight pointed right at you.
This is a very popular bug (well, unpopular, but often reported), on a very large number of Asus and other laptops. If a fix can be found for this, it makes *a lot* of people happy... (!)
I also tested Windows XP SP3 on the same 1000HE netbook/CNF7219 combo (yeah, I know... sorry, but I needed another comparison), and the Windows driver gets the best image in low light. It's just my impression, but it looks like Windows selects a frame rate in low light that's even lower than 5fps. (Move in front of the camera, and the image gets badly blurred/smeared.)
I've submitted this new information to Hans de Goede (video4linux) as well, and he recommended expanding the discussion to include Laurent Pinchart, which I have done.