Skip to contentSkip to conversation

favr32: improve touchscreen response

PATCH In review Ben Nizette 4 participants 6 messages Apr 20, 2009 – Jul 29, 2009

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 favr32: improve touchscreen response

The ezLCD+101 board (to which an favr-32 is fitted) has a long, unshielded, nasty lead between the touch panel and the ads7843 touch controller. In order to get satisfactory response then, we need to employ every noise-reduction trick in the driver's arsenal. After extensive fiddling I've found some good settings:

1) We keep vref on all the time to dramatically reduce settling times (at the cost of a tiny increase in power consumption).

2) Despite 1 the settling time is still non-zero. 500uS is plenty of time for the signals to settle

3) Despite 1 and 2 there's still a little bit of noise around. By setting a pen recheck delay we make the panel feel less touchy and twitchy.

Someone with more time and patience myself might be able to tune this numbers further but these settings are now perfectly acceptable for normal use.

Tested on ezLCD+101 though should only improve response on other ezLCD+/ favr-32 boards too.

---
 arch/avr32/boards/favr-32/setup.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)
arch/avr32/boards/favr-32/setup.c +4 -0
index 58203d1..f1152fa 100644
@@ -72,6 +72,10 @@ static struct ads7846_platform_data ads7843_data = {
 	.debounce_max		= 20,
 	.debounce_rep		= 4,
 	.debounce_tol		= 5,
+	
+	.keep_vref_on		= true,
+	.settle_delay_usecs	= 500,
+	.penirq_recheck_delay_usecs = 100,
 
 	/* The ads7843 pendown irq is just connected to regular
 	 * gpio and therefore can only trigger interrupts on
Signature · 1 line
1.6.0.2
#1 lore raw

arch/avr32/boards/favr-32/setup.c static struct ads7846_platform_data ads7843_data = { · line 81

Quoted 6 lines · Ben Nizette
arch/avr32/boards/favr-32/setup.c +3 -0
+	.keep_vref_on		= true,
+	.settle_delay_usecs	= 500,
+	.penirq_recheck_delay_usecs = 100,
 
 	/* The ads7843 pendown irq is just connected to regular
 	 * gpio and therefore can only trigger interrupts on

Whoops, looks like this applies on top of

[PATCH] ads7846: Allow both-edge interrupt mode

Sorry. Let us know if you want me to rebase :-)

	--Ben.
Acked-by on v1 1/1
#2 lore raw
In reply to Ben Nizette
Quoted 16 lines · On Mon, 20 Apr 2009 12:36:53 +1000 Ben Nizette <[email protected]> wrote:

The ezLCD+101 board (to which an favr-32 is fitted) has a long, unshielded, nasty lead between the touch panel and the ads7843 touch controller. In order to get satisfactory response then, we need to employ every noise-reduction trick in the driver's arsenal. After extensive fiddling I've found some good settings:

1) We keep vref on all the time to dramatically reduce settling times (at the cost of a tiny increase in power consumption).

2) Despite 1 the settling time is still non-zero. 500uS is plenty of time for the signals to settle

3) Despite 1 and 2 there's still a little bit of noise around. By setting a pen recheck delay we make the panel feel less touchy and twitchy.

Quoted 9 lines

Someone with more time and patience myself might be able to tune this numbers further but these settings are now perfectly acceptable for normal use.

Tested on ezLCD+101 though should only improve response on other ezLCD+/ favr-32 boards too.

Quoted 19 lines
---
 arch/avr32/boards/favr-32/setup.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)
(quoted diff) +0 -0

b/arch/avr32/boards/favr-32/setup.c index 58203d1..f1152fa 100644

arch/avr32/boards/favr-32/setup.c +4 -0
@@ -72,6 +72,10 @@ static struct ads7846_platform_data ads7843_data =
 { .debounce_max		= 20,
 	.debounce_rep		= 4,
 	.debounce_tol		= 5,
+	
+	.keep_vref_on		= true,
+	.settle_delay_usecs	= 500,
+	.penirq_recheck_delay_usecs = 100,
 
 	/* The ads7843 pendown irq is just connected to regular
 	 * gpio and therefore can only trigger interrupts on
Signature · 2 lines
Best regards,
Hans-Christian Egtvedt
#3 lore raw
In reply to Ben Nizette

Hi Haavard,

This has hce's ack, any chance of getting it in .31?

Thx,

	--Ben.
Quoted 42 lines · Ben Nizette

The ezLCD+101 board (to which an favr-32 is fitted) has a long, unshielded, nasty lead between the touch panel and the ads7843 touch controller. In order to get satisfactory response then, we need to employ every noise-reduction trick in the driver's arsenal. After extensive fiddling I've found some good settings:

1) We keep vref on all the time to dramatically reduce settling times (at the cost of a tiny increase in power consumption).

2) Despite 1 the settling time is still non-zero. 500uS is plenty of time for the signals to settle

3) Despite 1 and 2 there's still a little bit of noise around. By setting a pen recheck delay we make the panel feel less touchy and twitchy.

Someone with more time and patience myself might be able to tune this numbers further but these settings are now perfectly acceptable for normal use.

Tested on ezLCD+101 though should only improve response on other ezLCD+/ favr-32 boards too.

---
 arch/avr32/boards/favr-32/setup.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)
arch/avr32/boards/favr-32/setup.c +4 -0
index 58203d1..f1152fa 100644
@@ -72,6 +72,10 @@ static struct ads7846_platform_data ads7843_data = {
 	.debounce_max		= 20,
 	.debounce_rep		= 4,
 	.debounce_tol		= 5,
+	
+	.keep_vref_on		= true,
+	.settle_delay_usecs	= 500,
+	.penirq_recheck_delay_usecs = 100,
 };
#4 lore raw
Quoted 3 lines · Ben Nizette

Hi Haavard,

This has hce's ack, any chance of getting it in .31?

Yeah, should be a good chance :-) Sorry for taking so long.

It doesn't apply, however. I ended up doing this:

arch/avr32/boards/favr-32/setup.c +4 -0
index 46c9b0a..75f19f4 100644
@@ -72,6 +72,10 @@ static struct ads7846_platform_data ads7843_data = {
 	.debounce_max		= 20,
 	.debounce_rep		= 4,
 	.debounce_tol		= 5,
+
+	.keep_vref_on		= true,
+	.settle_delay_usecs	= 500,
+	.penirq_recheck_delay_usecs = 100,
 };
 
 static struct spi_board_info __initdata spi1_board_info[] = {

It appears to compile, but please verify that it's still correct.

Haavard

author Tested-by
#5 lore raw
Quoted 25 lines · Haavard Skinnemoen
Quoted 3 lines · Ben Nizette

Hi Haavard,

This has hce's ack, any chance of getting it in .31?

Yeah, should be a good chance :-) Sorry for taking so long.

It doesn't apply, however. I ended up doing this:

arch/avr32/boards/favr-32/setup.c +4 -0
index 46c9b0a..75f19f4 100644
@@ -72,6 +72,10 @@ static struct ads7846_platform_data ads7843_data = {
 	.debounce_max		= 20,
 	.debounce_rep		= 4,
 	.debounce_tol		= 5,
+
+	.keep_vref_on		= true,
+	.settle_delay_usecs	= 500,
+	.penirq_recheck_delay_usecs = 100,
 };
 
 static struct spi_board_info __initdata spi1_board_info[] = {
Quoted 1 line

It appears to compile, but please verify that it's still correct.

Looks good, thanks!

Quoted 2 lines

Haavard