Re: [PATCH] arm64: dts: rockchip: decrease rising edge time of UART2

From: Katsuhiro Suzuki
Date: Mon Mar 04 2019 - 08:59:53 EST

Hello Tony, Robin,

On 2019/03/04 5:41, Tony McKahan wrote:
On Sun, Mar 3, 2019 at 2:51 PM Robin Murphy <robin.murphy@xxxxxxx> wrote:

On 2019-03-03 6:45 pm, Tony McKahan wrote:
Hello Katsushiro,

On Sun, Mar 3, 2019 at 12:31 PM Katsuhiro Suzuki
<katsuhiro@xxxxxxxxxxxxx> wrote:

Hello Tony,

On 2019/03/04 0:13, Tony McKahan wrote:
On Sun, Mar 3, 2019 at 9:04 AM Katsuhiro Suzuki <katsuhiro@xxxxxxxxxxxxx> wrote:

Hello Heiko,

Thank you for comments.

On 2019/03/03 22:19, Heiko Stuebner wrote:

Am Sonntag, 3. MÃrz 2019, 13:27:05 CET schrieb Katsuhiro Suzuki:
This patch increases drive strength of UART2 from 3mA to 12mA for
getting more faster rising edge.

RockPro64 is using a very high speed rate (1.5Mbps) for UART2. In
this setting, a bit width of UART is about 667ns.

In my environment (RockPro64 UART2 with FTDI FT232RL UART-USB
converter), falling time of RockPro64 UART2 is 40ns, but riging time
is over 650ns. So UART receiver will get wrong data, because receiver
read intermediate data of rising edge.

Rising time becomes 300ns from 650ns if apply this patch. This is not
perfect solution but better than now.

Signed-off-by: Katsuhiro Suzuki <katsuhiro@xxxxxxxxxxxxx>
arch/arm64/boot/dts/rockchip/rk3399.dtsi | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)

your changing a core rk3399 property here, so I'd really like to get
input from other board stakeholders on this before applying a core

Could you either include the submitters of other rk3399-boards in the
recipient list so that they're aware or limit the change to rockpro64 for
the time being (aka overriding the property in the board-dts) please?

OK, I'm adding other boards members.
by ./scripts/ arch/arm64/boot/dts/rockchip/rk3399-*.dts

RockPro64 directly connect UART2 pins of RK3399 to external connector.
I think maybe other RK3399 boards are facing same problem, but I cannot
check it because I have RockPro64 only...

I'm happy if someone tell me other boards situation.

I'm pulling out other rockchip boards momentarily to see what kind of
population we have.

Note these are not all running 5.x kernels, however none of them have
the UART2 drive levels modified to my knowledge, and regardless, none
show over 100 ns.

board: rise/fall

rk3399-roc-pc: 90ns/90ns
rk3399-rockpro64 V2.0: 90ns/45ns
rk3399-rockpro64 V2.1: 40ns/41ns

I'm having to eyeball off a 20MHz analog scope (thank goodness for "yes"
being able to generate a nice periodic signal), but for my NanoPC-T4
with a cheap eBay FT232R adapter both rise and fall times are certainly
<100ns. FWIW I've not noticed any issues with letting the Bluetooth
interface on UART0 run flat-out at 4Mbaud either.

Robin, Thanks a lot. Your results show that cold solder (or some
electric parts on board) of my RockPro64 is broken or something wrong.

Please make sure there's not a large amount of flux or something
around the terminals on your board, that seems excessively high.

Thank you for valuable information. For more deeply discussion,
I tried other conditions and watch the rise/fall times.

1) Not connect
The rise/fall times are 40ns/5ns when nothing connect (impedance is
very high) to external pin of RockPro64.

What UART device are you using with RockPro64? If you use some device
with RockPro64 and board shows rise/fall times = 90ns/45ns, my device
is not suitable for RockPro64 by some reason. So it's better to drop
my patch.

The adapter is an FTDI FT232RL breakout board, attached with some
generic Dupont connector jumpers.
Interesting your RockPro is showing this symptom, perhaps there is a
cold solder joint somewhere introducing resistance?

2) Other SoC
I have other SoC board rk3328-rock64, Rock64 shows rise/fall times =
90ns/80ns when same UART-USB device is connected to UART pin.

I measured similar on my Rock64 as well.

Tony, thanks for your info about environment.

It seems my RockPro64 problem. I don't have enough electronic knowledge,
but try to check my RockPro64 as much as I can.

I think it shows rk3399's (or RockPro64's?) drive strength is a little
weak. So it's better to increase the drive strength of UART of rk3399.

I do not think this is a bad idea generally, it simply allows for more
available current from the interface. I'll let others be the judge of
that, however.

There may be RK3399 users who would care about the possible EMI impact,
so it's probably still best to limit such a change to boards which
definitely need it.

Ah, good point.

As to speeds achievable, I've only run into drive level issues with
SD/SDIO, but that's all the way up at 25-50 MHz.

My patch has bad effects for EMI issues of all RK3399 boards.

So conclusion is, my patch should be dropped. Sorry for confusing.

Best Regards,
Katsuhiro Suzuki



Best Regards,
Katsuhiro Suzuki

Best Regards,
Katsuhiro Suzuki


diff --git a/arch/arm64/boot/dts/rockchip/rk3399.dtsi b/arch/arm64/boot/dts/rockchip/rk3399.dtsi
index beaa92744a64..e3c8f91ead50 100644
--- a/arch/arm64/boot/dts/rockchip/rk3399.dtsi
+++ b/arch/arm64/boot/dts/rockchip/rk3399.dtsi
@@ -2000,6 +2000,11 @@
drive-strength = <8>;

+ pcfg_pull_up_12ma: pcfg-pull-up-12ma {
+ bias-pull-up;
+ drive-strength = <12>;
+ };
pcfg_pull_up_18ma: pcfg-pull-up-18ma {
drive-strength = <18>;
@@ -2521,8 +2526,8 @@
uart2c {
uart2c_xfer: uart2c-xfer {
rockchip,pins =
- <4 RK_PC3 RK_FUNC_1 &pcfg_pull_up>,
- <4 RK_PC4 RK_FUNC_1 &pcfg_pull_none>;
+ <4 RK_PC3 RK_FUNC_1 &pcfg_pull_up_12ma>,
+ <4 RK_PC4 RK_FUNC_1 &pcfg_pull_none_12ma>;

Linux-rockchip mailing list

Linux-rockchip mailing list