linux-media.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
To: Jacopo Mondi <jacopo@jmondi.org>
Cc: "Dave Stevenson" <dave.stevenson@raspberrypi.com>,
	"Krzysztof Hałasa" <khalasa@piap.pl>,
	"Mauro Carvalho Chehab" <mchehab@kernel.org>,
	"Sakari Ailus" <sakari.ailus@iki.fi>,
	linux-media@vger.kernel.org
Subject: Re: [PATCH 02/10] media: ar0521: Add V4L2_CID_ANALOG_GAIN
Date: Thu, 6 Oct 2022 18:05:52 +0300	[thread overview]
Message-ID: <Yz7u0LT2kwhyhc8x@pendragon.ideasonboard.com> (raw)
In-Reply-To: <20221006150015.7p7qlab4dbaldnsg@uno.localdomain>

Hello,

On Thu, Oct 06, 2022 at 05:00:15PM +0200, Jacopo Mondi wrote:
> On Thu, Oct 06, 2022 at 03:44:54PM +0100, Dave Stevenson wrote:
> > On Wed, 5 Oct 2022 at 20:07, Jacopo Mondi wrote:
> > >
> > > Add support for V4L2_CID_ANALOG_GAIN. The control programs the global
> > > gain register which applies to all color channels.
> > >
> > > As both the global digital and analog gains are controlled through a
> > > single register, in order not to overwrite the configured digital gain
> > > we need to read the current register value before modifying it.
> > >
> > > Implement a function to read register values and use it before applying
> > > the new analog gain.
> > >
> > > Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
> > > ---
> > >  drivers/media/i2c/ar0521.c | 64 ++++++++++++++++++++++++++++++++++++++
> > >  1 file changed, 64 insertions(+)
> > >
> > > diff --git a/drivers/media/i2c/ar0521.c b/drivers/media/i2c/ar0521.c
> > > index 89f3c01f18ce..581f5e42994d 100644
> > > --- a/drivers/media/i2c/ar0521.c
> > > +++ b/drivers/media/i2c/ar0521.c
> > > @@ -5,6 +5,8 @@
> > >   * Written by Krzysztof Hałasa
> > >   */
> > >
> > > +#include <asm/unaligned.h>
> > > +
> > >  #include <linux/clk.h>
> > >  #include <linux/delay.h>
> > >  #include <linux/pm_runtime.h>
> > > @@ -35,6 +37,11 @@
> > >  #define AR0521_HEIGHT_BLANKING_MIN     38u /* must be even */
> > >  #define AR0521_TOTAL_WIDTH_MIN      2968u
> > >
> > > +#define AR0521_ANA_GAIN_MIN            0x00
> > > +#define AR0521_ANA_GAIN_MAX            0x3f
> > > +#define AR0521_ANA_GAIN_STEP           0x01
> > > +#define AR0521_ANA_GAIN_DEFAULT                0x00
> > > +
> > >  /* AR0521 registers */
> > >  #define AR0521_REG_VT_PIX_CLK_DIV              0x0300
> > >  #define AR0521_REG_FRAME_LENGTH_LINES          0x0340
> > > @@ -55,6 +62,7 @@
> > >  #define AR0521_REG_RED_GAIN                    0x305A
> > >  #define AR0521_REG_GREEN2_GAIN                 0x305C
> > >  #define AR0521_REG_GLOBAL_GAIN                 0x305E
> > > +#define AR0521_REG_GLOBAL_GAIN_ANA_MASK                0x3f
> > >
> > >  #define AR0521_REG_HISPI_TEST_MODE             0x3066
> > >  #define AR0521_REG_HISPI_TEST_MODE_LP11                  0x0004
> > > @@ -77,6 +85,7 @@ static const char * const ar0521_supply_names[] = {
> > >
> > >  struct ar0521_ctrls {
> > >         struct v4l2_ctrl_handler handler;
> > > +       struct v4l2_ctrl *ana_gain;
> > >         struct {
> > >                 struct v4l2_ctrl *gain;
> > >                 struct v4l2_ctrl *red_balance;
> > > @@ -167,6 +176,36 @@ static int ar0521_write_reg(struct ar0521_dev *sensor, u16 reg, u16 val)
> > >         return ar0521_write_regs(sensor, buf, 2);
> > >  }
> > >
> > > +static int ar0521_read_reg(struct ar0521_dev *sensor, u16 reg, u16 *val)
> > > +{
> > > +       struct i2c_client *client = sensor->i2c_client;
> > > +       unsigned char buf[2];
> > > +       struct i2c_msg msg;
> > > +       int ret;
> > > +
> > > +       msg.addr = client->addr;
> > > +       msg.flags = client->flags;
> > > +       msg.len = sizeof(u16);
> > > +       msg.buf = buf;
> > > +       put_unaligned_be16(reg, buf);
> > > +
> > > +       ret = i2c_transfer(client->adapter, &msg, 1);
> > > +       if (ret < 0)
> > > +               return ret;
> > > +
> > > +       msg.len = sizeof(u16);
> > > +       msg.flags = client->flags | I2C_M_RD;
> > > +       msg.buf = buf;
> > > +
> > > +       ret = i2c_transfer(client->adapter, &msg, 1);
> > > +       if (ret < 0)
> > > +               return ret;
> > > +
> > > +       *val = get_unaligned_be16(buf);
> > > +
> > > +       return 0;
> > > +}
> > > +
> > >  static int ar0521_set_geometry(struct ar0521_dev *sensor)
> > >  {
> > >         /* All dimensions are unsigned 12-bit integers */
> > > @@ -187,6 +226,21 @@ static int ar0521_set_geometry(struct ar0521_dev *sensor)
> > >         return ar0521_write_regs(sensor, regs, ARRAY_SIZE(regs));
> > >  }
> > >
> > > +static int ar0521_set_analog_gain(struct ar0521_dev *sensor)
> > > +{
> > > +       u16 global_gain;
> > > +       int ret;
> > > +
> > > +       ret = ar0521_read_reg(sensor, AR0521_REG_GLOBAL_GAIN, &global_gain);
> > > +       if (ret)
> > > +               return ret;
> > > +
> > > +       global_gain &= ~AR0521_REG_GLOBAL_GAIN_ANA_MASK;
> > > +       global_gain |= sensor->ctrls.ana_gain->val & AR0521_REG_GLOBAL_GAIN_ANA_MASK;
> > > +
> > > +       return ar0521_write_reg(sensor, AR0521_REG_GLOBAL_GAIN, global_gain);
> >
> > Does this work without nasty interactions?
> 
> It seems so :)
> 
> > The register reference I have says that the bits 6:0 of 0x3056,
> > 0x3058, 0x305a, 0x305c, and 0x305e are all aliased to register 0x3056.
> > That means that the writes from ar0521_set_gains for GAIN,
> > RED_BALANCE, and BLUE_BALANCE will stomp over your ANALOGUE_GAIN.
> 
> but you're right the interactions between the registers are not 100%
> clear to me yet.
> 
> The fact is that libcamera only manipulates ANALOGUE_GAIN while the
> other gains are not changed. I guess if one wants to manipulate the
> single gains individually this is possible, but when setting the
> global gain they will be overwritten ? This seems to be confirmed from
> my experiments where changing the BLUE/RED gains has no visible
> effects on the image as libcamera constantly adjusts the global gain

I'm tempted to drop support for the colour gains really, and turn the
V4L2_CID_GAIN into V4L2_CID_DIGITAL_GAIN. Digital colour gains can still
be useful on platforms that have no ISP, but I think we need an array of
gains in that case, not abusing V4L2_CID_RED_BALANCE and
V4L2_CID_BLUE_BALANCE. Any objection ?

> > I also don't see a call to __v4l2_ctrl_handler_setup from
> > ar0521_set_stream, so whilst there is an explicit call to
> > ar0521_set_gains, ANALOGUE_GAIN won't be set.
>
> See [PATCH 08/10] media: ar0521: Setup controls at s_stream time
> later in the series :)
> 
> > [1] https://github.com/torvalds/linux/blob/master/drivers/media/i2c/ar0521.c#L190
> >
> > > +}
> > > +
> > >  static int ar0521_set_gains(struct ar0521_dev *sensor)
> > >  {
> > >         int green = sensor->ctrls.gain->val;
> > > @@ -456,6 +510,9 @@ static int ar0521_s_ctrl(struct v4l2_ctrl *ctrl)
> > >         case V4L2_CID_VBLANK:
> > >                 ret = ar0521_set_geometry(sensor);
> > >                 break;
> > > +       case V4L2_CID_ANALOGUE_GAIN:
> > > +               ret = ar0521_set_analog_gain(sensor);
> > > +               break;
> > >         case V4L2_CID_GAIN:
> > >         case V4L2_CID_RED_BALANCE:
> > >         case V4L2_CID_BLUE_BALANCE:
> > > @@ -499,6 +556,13 @@ static int ar0521_init_controls(struct ar0521_dev *sensor)
> > >         /* We can use our own mutex for the ctrl lock */
> > >         hdl->lock = &sensor->lock;
> > >
> > > +       /* Analog gain */
> > > +       ctrls->ana_gain = v4l2_ctrl_new_std(hdl, ops, V4L2_CID_ANALOGUE_GAIN,
> > > +                                           AR0521_ANA_GAIN_MIN,
> > > +                                           AR0521_ANA_GAIN_MAX,
> > > +                                           AR0521_ANA_GAIN_STEP,
> > > +                                           AR0521_ANA_GAIN_DEFAULT);
> > > +
> > >         /* Manual gain */
> > >         ctrls->gain = v4l2_ctrl_new_std(hdl, ops, V4L2_CID_GAIN, 0, 511, 1, 0);
> > >         ctrls->red_balance = v4l2_ctrl_new_std(hdl, ops, V4L2_CID_RED_BALANCE,

-- 
Regards,

Laurent Pinchart

  reply	other threads:[~2022-10-06 15:06 UTC|newest]

Thread overview: 71+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-10-05 19:06 [PATCH 00/10] media: ar0521: Add analog gain, rework clock tree Jacopo Mondi
2022-10-05 19:06 ` [PATCH 01/10] media: ar0521: Implement enum_frame_sizes Jacopo Mondi
2022-10-06 14:37   ` Dave Stevenson
2022-10-06 16:33   ` Laurent Pinchart
2022-10-07  4:57     ` Krzysztof Hałasa
2022-10-07  8:08       ` Laurent Pinchart
2022-10-07  7:29     ` Jacopo Mondi
2022-10-07  8:11       ` Laurent Pinchart
2022-10-07 10:32         ` Dave Stevenson
2022-10-07 12:05           ` Sakari Ailus
2022-10-07 12:12             ` Laurent Pinchart
2022-10-05 19:06 ` [PATCH 02/10] media: ar0521: Add V4L2_CID_ANALOG_GAIN Jacopo Mondi
2022-10-06 14:44   ` Dave Stevenson
2022-10-06 15:00     ` Jacopo Mondi
2022-10-06 15:05       ` Laurent Pinchart [this message]
2022-10-07  5:28         ` Krzysztof Hałasa
2022-10-07  8:20           ` Laurent Pinchart
2022-10-12 18:54             ` Sakari Ailus
2022-10-13  9:30               ` Laurent Pinchart
2022-10-07  5:20   ` Krzysztof Hałasa
2022-10-07  7:17     ` Jacopo Mondi
2022-10-07  8:30       ` Laurent Pinchart
2022-10-07 12:01         ` Krzysztof Hałasa
2022-10-07 12:07           ` Laurent Pinchart
2022-10-07 14:02             ` Krzysztof Hałasa
2022-10-17 15:10         ` Jacopo Mondi
2022-10-17 15:57           ` Sakari Ailus
2022-10-17 16:31             ` Jacopo Mondi
2022-10-17 16:37               ` Sakari Ailus
2022-10-17 16:42               ` Dave Stevenson
2022-10-07 11:56       ` Krzysztof Hałasa
2022-10-07 12:11         ` Laurent Pinchart
2022-10-07 14:00           ` Krzysztof Hałasa
2022-10-05 19:06 ` [PATCH 03/10] media: ar0521: Set maximum resolution to 2592x1944 Jacopo Mondi
2022-10-06 14:57   ` Dave Stevenson
2022-10-07 13:06     ` Laurent Pinchart
2022-10-20 11:23       ` Jacopo Mondi
2022-10-07  5:33   ` Krzysztof Hałasa
2022-10-07 12:42     ` Jacopo Mondi
2022-10-07 14:07       ` Krzysztof Hałasa
2022-10-05 19:06 ` [PATCH 04/10] media: ar0521: Rework PLL computation Jacopo Mondi
2022-10-07 13:56   ` Laurent Pinchart
2022-10-12 19:02     ` Sakari Ailus
2022-10-13  9:31       ` Laurent Pinchart
2022-10-05 19:06 ` [PATCH 05/10] media: ar0521: Add LINK_FREQ control Jacopo Mondi
2022-10-06 15:10   ` Dave Stevenson
2022-10-07 14:01     ` Laurent Pinchart
2022-10-07 14:26       ` Dave Stevenson
2022-10-16  1:53         ` Laurent Pinchart
2022-10-17 11:21           ` Dave Stevenson
2022-10-17  9:24         ` Jacopo Mondi
2022-10-17 11:00           ` Dave Stevenson
2022-10-17 12:00             ` Jacopo Mondi
2022-10-05 19:06 ` [PATCH 06/10] media: ar0521: Configure pixel rate using LINK_FREQ Jacopo Mondi
2022-10-06  5:51   ` kernel test robot
2022-10-06 15:42   ` Sakari Ailus
2022-10-07  7:25     ` Jacopo Mondi
2022-10-07  5:52   ` Krzysztof Hałasa
2022-10-05 19:06 ` [PATCH 07/10] media: ar0521: Adjust exposure and blankings limits Jacopo Mondi
2022-10-06  2:08   ` kernel test robot
2022-10-06  4:17   ` kernel test robot
2022-10-06 15:41   ` Dave Stevenson
2022-10-05 19:06 ` [PATCH 08/10] media: ar0521: Setup controls at s_stream time Jacopo Mondi
2022-10-06 15:43   ` Dave Stevenson
2022-10-07  7:23     ` Jacopo Mondi
2022-10-05 19:06 ` [PATCH 09/10] media: ar0521: Rework startup sequence Jacopo Mondi
2022-10-06 15:45   ` Dave Stevenson
2022-10-05 19:06 ` [PATCH 10/10] media: ar0521: Tab-align definitions Jacopo Mondi
2022-10-06 15:48   ` Dave Stevenson
2022-10-06 16:11   ` Laurent Pinchart
2022-10-07  5:42   ` Krzysztof Hałasa

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=Yz7u0LT2kwhyhc8x@pendragon.ideasonboard.com \
    --to=laurent.pinchart@ideasonboard.com \
    --cc=dave.stevenson@raspberrypi.com \
    --cc=jacopo@jmondi.org \
    --cc=khalasa@piap.pl \
    --cc=linux-media@vger.kernel.org \
    --cc=mchehab@kernel.org \
    --cc=sakari.ailus@iki.fi \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).