Hi,
On 09/24/2012 03:36 PM, Frank Schäfer wrote:
Hi Hans,
Am 24.09.2012 12:47, schrieb Hans de Goede:
Hi,
Sorry for the slow response ...
On 09/20/2012 01:54 PM, Frank Schäfer wrote:
Hi,
Am 20.09.2012 11:08, schrieb Hans de Goede:
<snip>
Many webcams have RGB gains, but we don't have standard CID-s for
these,
instead we've Blue and Red Balance. This has grown historically
because of
the bttv cards which actually have Blue and Red balance controls in
hardware,
rather then the usual RGB gain triplet. Various gspca drivers cope
with this
in different ways.
If you look at the pac7302 driver before your latest 4 patches it has
a Red and Blue balance control controlling the Red and Blue gain, and a
Whitebalance control, which is not White balance at all, but simply
controls the green gain...
Ok, so if I understand you right, red+green+blue balance = white
balance.
And because we already have defined red, blue and whitebalance controls
for historical reasons, we don't need green balance ?
Maybe that matches physics, but I don't think it's a sane approach for a
user interface...
No what I was trying to say is that the balance controls are for hardware
which actually has balance controls and not per color gains (such as the
bt87x chips), but they are being abused by many drivers to add support
for
per color gains. And then you miss one control. And in the case of the
pac7302
driver the "original" route was taken of using whitebalance to control
the green gain. Which is wrong IMHO, but it is what the driver does know.
A proper fix would be to introduce new controls for all 3 gains, and use
those instead of using the balance controls + adding a 3th balance
control
being discussed in the thread titled:
"Gain controls in v4l2-ctrl framework".
Ok, it seems I'm misunderstanding the meaning of color "gain" and
"balance"...
Anyway, I would say that it makes sense to have per color AND global
controls, so a V4L2_CID_GREEN_BALANCE would be missing.
Weather it makes sense to use them at the same time is a different question.
And why do you think the controls in question are gain controls instead
of balance controls ?
"Balance" suggest balancing it against some fixed value, where as once
there are 3 for all of r, g and b, there is no fixed value, then the registers
are just controlling an amplification factor (which could be less then 1.0)
and that is normally called a gain.
If you look in most sensor datasheets they talk about color gains not
balance controls ...
Anyways this is mostly just semantics. We are working on defining standard
controls for per color gains, and once we have those we can map them to
reg 0x01 - 0x03.
The Windows driver calls them balance controls, too (which could of
course also be a Windows driver or API issue...).
And as said other drivers have similar (albeit usually different)
hacks.
At a minimum I would like you to rework your patches to:
1) Not add the new Green balance, and instead modify the existing
whitebalance
to control the new green gain you've found. Keeping things as broken as
they are, but not worse; and
I prefer waiting for the results of the discussion you are proposing
further down.
I see in your next mail that you've changed your mind. So I would like to
move forward with this by adding your 2 patches + 1 more patch to also
make the whitebalance control (which really is the green gain control)
use 0x02 rather then 0xc6. To do this we must make sure that 0xc6 has
a proper fixed / default setting. So what does the windows driver use
for this? 1 like with 0xc5 and 0xc7 ?
And can you do a 3th patch to make the whitebalance control control
0x02 rather then 0xc6 like you did for red and blue balance?
No, we shouldn't do that.
Reg 0xc6 (currently called "white balance temperature") definitely works
different compared to register 0x02.
Whatever it does exactly, it's not green gain or balance adjustment.
Will try to figure out next time.
The Windows driver doesn't use this register for an (user settable)
image control. It just sets its value to 55, which I fixed with one of
the patches from my previous patch series, so we have the correct
default value now.
Ah, interesting!
0xc6 is also different from regs 0xc5 and 0xc7: settable values are
0-255 compared to 0-3.
True, I should have noticed that.
So let's not touch 0xc6 / "white balance" until we know its real meaning
and just apply the first 2 patches.
OK, will do.
Regards,
Hans
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html