imx8mm lcdif->dsi->adv7535 no video, no errors

[Dave Stevenson]

On Wed, 3 Aug 2022 at 13:31, Adam Ford <aford173 at gmail.com> wrote:
>
> On Wed, Aug 3, 2022 at 7:17 AM Dave Stevenson
> <dave.stevenson at raspberrypi.com> wrote:
> >
> > Hi Adam
> >
> > On Wed, 3 Aug 2022 at 12:03, Adam Ford <aford173 at gmail.com> wrote:
> > >
> > > On Wed, Aug 3, 2022 at 1:20 AM Marco Felsch <m.felsch at pengutronix.de> wrote:
> > > >
> > > > On 22-08-02, Adam Ford wrote:
> > > >
> > > > ...
> > > >
> > > > > > I did some reading about the internal timing generator.  It appears
> > > > > > that it's required when video formats use fractional bytes, and it's
> > > > > > preconfigured to run at 720p by default, but registers 28h through 37h
> > > > > > configure it for other video modes.
> > > > >
> > > > > I think there may still be some issues with the DSIM since some of the
> > > > > clock frequencies are set in the device tree.
> > > > >
> > > > > From what I can tell, the pixel rate is calculated based on the
> > > >
> > > > By pixel rate you mean the HDMI pixel rate from the ADV? If so then yes.
> > > > The ADV has an divider which is already configured by the driver but
> > > > meaningless since the driver is lacking of setting the "manual-divider"
> > > > bit within the same register.
> > >
> > > I was thinking about the pixel clock from the DSI to the ADV.  I did
> > > see the manual-divider bit was missing.  I tried enabling that bit,
> > > but it didn't appear to make much difference.
> > > >
> > > > > burst-clock-frequency and that generates a byte clock.  For 891000000,
> > > > > the byte clock is 111375000.
> > > >
> > > > The burst-clock-frequency is the hs-clk and DDR. So the MIPI-DSI clock
> > > > is burst-clock-frequency/2 which is in your case: 891000000/2 =
> > > > 445500000. This clock is than divided by 3 within the ADV and you get
> > > > your 148500000 pixel clock. This divide by 3 is detected automatically
> > > > by the ADV due to the missing bit (see above).
> > > >
> > > > > Modetest timings for 1080p show:
> > > > >
> > > > > index name refresh (Hz) hdisp hss hse htot vdisp vss vse vtot
> > > > >   #0 1920x1080 60.00 1920 2008 2052 2200 1080 1084 1089 1125 148500
> > > > > flags: nhsync, nvsync; type: driver
> > > > >
> > > > >
> > > > > When looking at modetest, there is a clock for 1080p which appears to be 148500.
> > > > > 111375000/148500 = 750.
> > > >
> > > > Please see above.
> > > >
> > > > > The rest of the entries in my table do not divide evenly.  I don;t
> > > > > know if that explains the lack of display, but it's something to note.
> > > > > It seems to me that instead of fixing the
> > > > > samsung,burst-clock-frequency to 891000000, we should make the desired
> > > > > PLL related to the desired pixel clock so it divides evenly.
> > > >
> > > > Please see above.
> > > >
> > > > > Looking at NXP's kernel, I also noticed that their esc_prescaler is
> > > > > based on the byte clock divided by 20MHz.  With some small code
> > > > > changes to get the PLL based on the desired pixel clock instead of
> > > > > hard-coded,  I was able to set
> > > > >
> > > > > samsung,burst-clock-frequency = <1500000000>;
> > > >
> > > > This is not correct since the burst-clock-freq. specifies the hs-clock
> > > > for the data lanes (see above).
> > >
> > > But I don't think the clock should be fixed. I think it should vary as
> > > the resolution changes.  From what I can tell, NXP's DSI code doesn't
> > > hard code this value, but it does appear to cap it at 1.5G.  I did
> > > soom looking into the NXP frequency calculation and it is capable of
> > > adjusting resolutions to some extent and from what I can see the
> > > 891MHz clock is only set when 1080p.  At 720p, thier kernel shows the
> > > output frequency at  445.5 MHz.  The way the DSIM is currently
> > > configured, it's fixed at 891MHz, so I don't expect the output feeding
> > > the adv7535 to be correct for the different resolutions.
> > >
> > >
> > > >
> > > > > samsung,esc-clock-frequency = <20000000>;
> > > >
> > > > This is correct, we also use a esc-clock of 20MHz.
> > > >
> > > > > With these settings and the above mentioned code changes, 1080p still
> > > > > appears, however when attempting other modes, the display still fails
> > > > > to load.  I also noticed that the phy ref clock is set to 27MHz
> > > > > instead of NXP's 12MHz.
> > > >
> > > > That's interesting, I didn't noticed that NXP uses 12 MHz as refclock
> > > > but I don't think that this is the problem. Since we have other
> > > > converter chips using the bridge driver and they work fine. I still
> > > > think that the main problem is within the ADV driver.
> > >
> > > Do the other converter chips work fine at different resolutions?
> > >
> > > >
> > > > > I attempted to play with that setting, but I couldn't get 1080p to
> > > > > work again, so I backed it out.
> > > > >
> > > > > Maybe I am headed in the wrong direction, but I'm going to examine the
> > > > > P/M/S calculation of the timing on NXP's kernel to see how the DSIM in
> > > > > this code compares.
> > > >
> > > > I think the pms values are fine.
> > >
> > > I compared the P/M/S values between this driver and NXP's and they
> > > calculate different values of PMS when running at 1080P.
> > > NXP @ 1080p:
> > > fout = 891000, fin = 12000, m = 297, p = 2, s = 1, best_delta = 0
> > >
> > > This kernel @ 1080p:
> > >
> > > PLL freq 891000000, (p 3, m 99, s 0)
> > >
> > > at 720P, the NXP Kernel
> > > fout = 445500, fin = 12000, m = 297, p = 2, s = 2, best_delta = 0
> > > (working)
> > >
> > > at 720P, this kernel:
> > > PLL freq 891000000, (p 3, m 99, s 0)
> > > hs_clk = 891000000, byte_clk = 111375000, esc_clk = 18562500
> > > (not working)
> > >
> > >
> > > >
> > > > > If someone who understands the interactions between these different
> > > > > components has suggestions, I'm willing to run some experiments.
> > > >
> > > > Did managed to get access to the ADV7535 programming guide? This is the
> > > > black box here. Let me check if I can provide you a link with our repo
> > > > so you can test our current DSIM state if you want.
> > >
> > > I do have access to the programming guide, but it's under NDA, but
> > > I'll try to answer questions if I can.
> >
> > Not meaning to butt in, but I have datasheets for ADV7533 and 7535
> > from previously looking at these chips.
>
> Thanks for the feedback.
>
> > Mine fairly plainly states:
> > "The DSI receiver input supports DSI video mode operation only, and
> > specifically, only supports nonburst mode with sync pulses".
> > Non-burst mode meaning that the DSI pixel rate MUST be the same as the
> > HDMI pixel rate.
>
> Mine also states the DSI source needs to provide correct video timing
> with start and stop sync packets.
>
> If I remember correctly, it seemed like Marek V wanted the hard coded
> samsung,burst-clock-frequency to go away so the clock frequency could
> be set dynamically.

I've never worked with Exynos or imx8, but my view would be that
samsung,burst-clock-frequency should only be used if
MIPI_DSI_MODE_VIDEO_BURST is set in the mode_flags (it isn't for
adv7533/5).
Without that flag the DSI link frequency should be running at the rate
defined by the mode clock, number of lanes, bpp, etc.

>From the DSI spec (v 1.1 section 8.11.1):
"Non-Burst Mode with Sync Pulses – enables the peripheral to
accurately reconstruct original video timing, including sync pulse
widths."
"RGB pixel packets are time-compressed, leaving more time during a
scan line for LP mode (saving power) or for multiplexing other
transmissions onto the DSI link."
How can the peripheral reconstruct the video timing off a quirky link frequency?

Unless the Exynos DSIM_CONFIG_REG register bit DSIM_BURST_MODE [1]
reconfigures the clock setup of the DSI block, then I don't see how
the Exynos driver can follow the DSI spec in that regard.

Hope that helps.

  Dave

[1] https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/exynos/exynos_drm_dsi.c#L803

> I have attempted to do some of this work based on
> what I am seeing in the NXP kernel, and I get get my monitor to sync
> at some resolutions, but the screen is usually all green or all blue,
> so it's not really a success.  The clock part appears to be good
> enough to make the monitor see some sort of signal, so I am going to
> investigate the calculation of the rest of the video timings to see if
> I can fix the color issue.
>
> > Section 6.1.1 "DSI Input Modes" of adv7533_hardware_user_s_guide is
> > even more explicit about the requirement of DSI timing matching
> >
> > The NXP kernel switching down to an hs_clk of 445.5MHz would therefore
> > be correct for 720p operation.
> >
> > If you do program the manual DSI divider register to allow a DSI pixel
> > rate of 148.5MHz vs HDMI pixel rate of 74.25MHz, you'd be relying on
> > the ADV753x having at least a half-line FIFO between DSI rx and HDMI
> > tx to compensate for the differing data rates. I see no reference to
> > such, and I'd be surprised if it was more than a half dozen pixels to
> > compensate for the jitter in the cases where the internal timing
> > generator is mandatory due to fractional bytes.
>
> Thanks Dave!
>
> adam
>
> >
> >   Dave
> >
> > > adam
> > > >
> > > > Regards,
> > > >   Marco