Allow to couple/uncouple CE pin #23

jjtmp · 2023-11-19T15:00:38Z

No description provided.

Part I/II

Part II/II To use couple function, just use "couple()" in the lcd initilization code. Then clock is always enabled which makes drawing a bit faster on regular pins and helps reducing interference when using two or more displays on the same data pins.

platisd

Hi Jonas, thanks for your contribution.

Can you explain how would the new functions would be used in practice with a bare minimum example?

platisd · 2023-11-20T08:45:49Z

src/Nokia_LCD.h

+    void couple(bool yes = true);
+    void uncouple();


Is couple(false) the same as uncouple?
If so, then let's keep only one.

platisd · 2023-11-20T08:46:50Z

src/Nokia_LCD.cpp

+    if (!coupled)
+        digitalWrite(kCe_pin, LOW);


Consistent styling:

Suggested change

if (!coupled)

digitalWrite(kCe_pin, LOW);

if (!coupled) {

digitalWrite(kCe_pin, LOW);

}

platisd · 2023-11-20T08:49:57Z

src/Nokia_LCD.cpp

+    if (yes && !coupled) {
+        digitalWrite(kCe_pin, LOW);
+    }
+    else if (!yes && coupled) {
+        digitalWrite(kCe_pin, HIGH);
+    }
+    coupled = yes;


Hmm, couldn't this be changed to something simpler, i.e.:

Suggested change

if (yes && !coupled) {

digitalWrite(kCe_pin, LOW);

}

else if (!yes && coupled) {

digitalWrite(kCe_pin, HIGH);

}

coupled = yes;

digitalWrite(kCe_pin, yes ? LOW : HIGH);

coupled = yes;

platisd · 2023-11-20T08:50:38Z

src/Nokia_LCD.cpp

+    if (!coupled)
+        digitalWrite(kCe_pin, HIGH);


Consistent styling:

Suggested change

if (!coupled)

digitalWrite(kCe_pin, HIGH);

if (!coupled) {

digitalWrite(kCe_pin, HIGH);

}

platisd · 2023-11-20T08:54:37Z

src/Nokia_LCD.h

+    /**
+     * couple()/uncouple() is useful for improving speed and 
+     * for accessing two or more displays on the same pins
+     * (but different CE pins)
+     */


The claims are that using these two new functions:

Improve speed

Allow accessing more than 1 display on the same pins

I can understand the 2nd point, but can you elaborate more on the 1st?

Two answer all your remarks and maybe more:

Why this proposal

I was experimenting with two displays. But they constantly resisted to work as expected. I suspect electrical interference between the CE lines. Another advantage is that the contrast setting menu can be shown on all displays the same time so if your application misconfigured one LCD but shows the contrast setting menu on the other LCD too you still can read what's going on.

How-to use

The key of couple() is to stop the send() method from modifying the CE pin. Obviously you have to manage it by yourself then.

The bonus of couple() is that you can switch from one display to another without changing all these lcd.print() invocations in main and subroutines. I use this - but to be frank it would be more clear to wrap a digitalWrite such as activate_lcd1() {digitalWrite(CEofLCD1, LOW);} - I probably will switch to this soon for sake of clarity.

Examples

See
a) Full http://z-info.jbechtel.de/srv/23kopierer_savev179web.ino
b) Reduced http://z-info.jbechtel.de/srv/23kopierer_web_short.ino
c) Mini http://z-info.jbechtel.de/srv/23kopierer_web_veryshort.ino

The a) example is an actually working one. But it contains lots of code which you don't need for LCD (and also lots of dead code).
The b) example is untested. I just took the a) example and removed everything which is not contrast setup menu/counter printer.
The c) example is pretty minimal, just has the initialization phase of b) example. Also untested.

Please don't mind the "lcd.uncouple()" in all a)b)c) examples initialization. It is wrong and most probably should be just removed.
Also please don't mind that it is not neccessary to instantiate lcdL and lcdR objects. lcd object is just enough. All lcdL and lcdR tasks could be surrounded by activate_lcd1() like wrappers (see above 2.).

Improve speed

It was just my feeling that on Arduino the PIN access is pretty slow, compared to directly coding for ATMega. Web search engine results support this. (digitalWrite() seems to switch off PWM first, etc.). So omitting digitalWrite(), even just the surrounding one for CE, helps a bit. Of course one could be pretty much faster when not-at-all using digitalWrite() but reducint calls is not so bad, either.

I have incorporated most of your suggestions and expanded the documentation comment in Nokia_LCD.h. Is it better than before?

As suggested by pull request review

platisd

I think we are getting somewhere :)

platisd · 2023-11-20T23:48:20Z

src/Nokia_LCD.h

+     *     ac) reducing electrical interference in case it happens that one CE line
+     *         overspeaks to the other CE line


Sounds a bit farfetched but I haven't tried it so I guess you have a point.
However, it feels like not only the electrical engineering aspects are outside the scope of this library, but -more importantly- this isn't some information we should include in the API documentation.

platisd · 2023-11-20T23:48:52Z

src/Nokia_LCD.h

+     *     ad) manage CE pin by different code
+     *            lcd.couple(); // once and forever, now someone else has to manage
+     *                          // CE pin


I think I understand this one, but does it need to be in the API documentation?

platisd · 2023-11-20T23:51:43Z

src/Nokia_LCD.h

+     *     ad) manage CE pin by different code
+     *            lcd.couple(); // once and forever, now someone else has to manage
+     *                          // CE pin
+     * b) improving speed by ~10% (16 instead of 18 digitalWrite() per data/command byte)


While there are 10% less instructions that doesn't necessarily mean it's 10% faster, especially since branches have been introduced to the code.
Regardless, any performance claim should be backed up by data and I'd argue that even if there is a 10% performance boost, it isn't worth mentioning in the API. :)

src/Nokia_LCD.h

src/Nokia_LCD.cpp

src/Nokia_LCD.h

src/Nokia_LCD.cpp

platisds' points are reasonable; you can manage to use couple()/uncouple() only when needed; the API description can be a straightforward functional description with little example. Co-authored-by: Dimitris Platis <platisd@users.noreply.github.com>

jjtmp · 2023-11-21T22:26:05Z

It is right that the performance claim isn't from measurements and that the interface description is better brief than possibly wrong. One could argue how much access to inner variables should be possible, but of course it's not unacceptable to make access private.
I committed your suggestions.

platisd

👍 🎉

jjtmp · 2023-11-22T20:20:23Z

👍 Thank you! Makes exchange of the .ino file easier between my build pipeline (Makefile) and the project initiators build pipeline (Arduino IDE) without any explanation.
As soon as I have access to the hardware again I will check about the electrical interference topic, just to be sure.

jjtmp · 2024-01-27T15:54:04Z

As signalized I have checked the "electrical interference" issue. Maybe it isn't related to electrical interference, at least I couldn't reproduce this. It's probably just a question of reset pins. The begin() method does set reset pin to low (and high) and then writes makes necessary initialization bytes. So there are following variants to share pins of two LCDs:

[hardware] either use distinct RST pins
[software] or set RST pin to the same value as CE pin in instantiation of LCD object (for those later LCD for which you call begin() after you called it for the first LCD)

Nokia_LCD lcd10(13, 12, 11, 10, 9);
Nokia_LCD lcd7(13, 12, 11, 7, 7); // reset pin "7" = trick the constructor
lcd10.begin()
lcd7.begin()

[software] or use couple(). the couple() must be called for all LCDs. It is not necessary for the LCD you call begin() on. Example:

Nokia_LCD lcd10(13, 12, 11, 10, 9);
Nokia_LCD lcd7(13, 12, 11, 7, 9); // reset pin irrelevant, never touched
lcd7.couple(); // new since version NokiaLCD 2.7.0
lcd10.begin();
lcd7.uncouple();

So now I'm done with that story.

platisd · 2024-01-27T15:55:50Z

Thanks for the follow-up @jjtmp 💯

jjtmp added 2 commits November 19, 2023 15:36

couple/uncouple (Nokia_LCD.h)

8d95a04

Part I/II

couple/uncouple (Nokia_LCD.cpp)

4dd9f46

Part II/II To use couple function, just use "couple()" in the lcd initilization code. Then clock is always enabled which makes drawing a bit faster on regular pins and helps reducing interference when using two or more displays on the same data pins.

platisd requested changes Nov 20, 2023

View reviewed changes

platisd reviewed Nov 20, 2023

View reviewed changes

jjtmp added 3 commits November 21, 2023 00:07

Change couple() description and arguments (Nokia_LCD.h)

69cc0d4

As suggested by pull request review

Consistent code styling; reduce couple() arguments (Nokia_LCD.cpp)

9991fc4

As suggested by pull request review

comment

db61ae6

platisd requested changes Nov 21, 2023

View reviewed changes

platisd approved these changes Nov 22, 2023

View reviewed changes

platisd merged commit 3e9241f into platisd:master Nov 22, 2023
0 of 2 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Allow to couple/uncouple CE pin #23

Allow to couple/uncouple CE pin #23

jjtmp commented Nov 19, 2023

platisd left a comment •

edited

Loading

platisd Nov 20, 2023

platisd Nov 20, 2023

platisd Nov 20, 2023

platisd Nov 20, 2023

platisd Nov 20, 2023

jjtmp Nov 20, 2023

platisd left a comment

platisd Nov 20, 2023 •

edited

Loading

platisd Nov 20, 2023

platisd Nov 20, 2023

jjtmp commented Nov 21, 2023

platisd left a comment

jjtmp commented Nov 22, 2023

jjtmp commented Jan 27, 2024

platisd commented Jan 27, 2024

		* ac) reducing electrical interference in case it happens that one CE line
		* overspeaks to the other CE line

Allow to couple/uncouple CE pin #23

Allow to couple/uncouple CE pin #23

Conversation

jjtmp commented Nov 19, 2023

platisd left a comment • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

platisd left a comment

Choose a reason for hiding this comment

platisd Nov 20, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jjtmp commented Nov 21, 2023

platisd left a comment

Choose a reason for hiding this comment

jjtmp commented Nov 22, 2023

jjtmp commented Jan 27, 2024

platisd commented Jan 27, 2024

platisd left a comment •

edited

Loading

platisd Nov 20, 2023 •

edited

Loading