Weird PCA9685 restrictions.

[AnonymousMonkeyNo12]

I am currently working with the PCA9685 and I find the restrictions that the current implementation offers a bit limiting.

1. private setPWM function
   Currently, you can basically pass an "on/off ratio" and diozero will turn the signal on at the start of the duty cycle and turn it off after the ratio. This makes it so that you can not offset the start of the signal from the start of the duty cycle. This is because the setPWM function is private and is only called through helpers that limit it.

diozero/diozero-core/src/main/java/com/diozero/devices/PCA9685.java

Lines 275 to 291 in e03ad50

	private void setPwm(int channel, int on, int off) throws RuntimeIOException {
	validateChannel(channel);
	validateOnOff(on, off);
	// Logger.debug("channel: {}, on: {}, off: {}", Integer.valueOf(channel),
	// Integer.valueOf(on),
	// Integer.valueOf(off));
	// TODO Replace with writeWordData()?
	// device.writeWordData(LED0_ON_L + 4 * channel, (short) on);
	// device.writeWordData(LED0_OFF_L + 4 * channel, (short) off);
	device.writeByteData(LED0_ON_L + 4 * channel, on & 0xFF);
	device.writeByteData(LED0_ON_H + 4 * channel, on >> 8);
	device.writeByteData(LED0_OFF_L + 4 * channel, off & 0xFF);
	device.writeByteData(LED0_OFF_H + 4 * channel, off >> 8);
	// SleepUtil.sleepMillis(50);
	}

2. opinionated validateOnOff function
   The validate function enforces that "off" must be smaller than "on", which is not at all necessary for the PCA9685 (See section 7.3.3 on page 16 of the doc https://cdn-shop.adafruit.com/datasheets/PCA9685.pdf ). This makes it so that you can not implement "Example 2" from the same document, page 17. The semantics of the function are also not quite correct, because according to the documentation diagram, "off" is always after "on" despite "off" having a smaller value than "on".

diozero/diozero-core/src/main/java/com/diozero/devices/PCA9685.java

Lines 293 to 309 in e03ad50

	private static void validateOnOff(int on, int off) {
	if (on < 0 || on > MAX_VALUE) {
	throw new IllegalArgumentException(String.format("Error: on (" + on + ") must be 0.." + MAX_VALUE));
	}
	if (off < 0 || off > MAX_VALUE) {
	throw new IllegalArgumentException(String.format("Error: off (" + off + ") must be 0.." + MAX_VALUE));
	}
	// Off must be after on
	if (off < on) {
	throw new IllegalArgumentException("Off value (" + off + ") must be > on value (" + on + ")");
	}
	// Total must be < 4096
	if (on + off > MAX_VALUE) {
	throw new IllegalArgumentException(String.format("Error: on (%d) + off (%d) must be < %d",
	Integer.valueOf(on), Integer.valueOf(off), Integer.valueOf(MAX_VALUE)));
	}
	}

(Examples 3 and 4 on page 18. might also not be possible in the current implementation, but I didn't think about them yet)

---

Why it matters:

Currently if I set 16 LED strips to half brightness, they all turn on and off in sync, causing a voltage peak and inducing a loud pieping/buzzing in a power supply.

To mitigate the noise, it is recommended to stagger the signals. For optimal staggering, some signals need their on cycle to cross a duty cycle, thus having an off before an on.

---

My proposal would be to:

• make setPWM public
• remove the "off must be after on" validation

I also don't quite understand what the InternalPwmOutputDeviceInterface is there for and whether I'd have to modify that. It seems to be shared by a lot of similar devices, but when I use the PCA9685, the functions calls do not go through that interface (and the name suggests it is for internal use?)

Should I do a PR or is there a reason that I am not seeing for why those two aspects are this way?

[mattjlewis]

Have committed change to make setPwm() public and update validateOnOff().

Regarding InternalPwmOutputDeviceInterface - this is the diozero internal mechanism for providing a consistent lowest common denominator interface for controlling any PWM device using the exposed PwmOutputDevice class and higher-level PwmLed device class.

[AnonymousMonkeyNo12]

Thank you, I just did it on my fork as well 😅

The maven-ci deployment seems to be broken because it uses a deprecated action version https://github.com/mattjlewis/diozero/actions/runs/15565223471/job/43827528142

I think you can just bump it without breaking anything for it to work again. According to actions/cache#1510

[AnonymousMonkeyNo12]

hi @mattjlewis

sorry for the late reply. I just got back to the project. The change works for me (compiled from source).

Maybe it also makes sense to set getPwm to public in the same vein?

---

It seems the changes have not yet been released to Maven.

As far as I know Maven does not allow overwriting an existing version and 1.4.1 is from Oct 10, 2024

https://mvnrepository.com/artifact/com.diozero/diozero-core/1.4.1

https://central.sonatype.com/artifact/com.diozero/diozero-core/1.4.1

But the runner ran with "1.4.1" https://github.com/mattjlewis/diozero/actions/runs/15565482191/job/43828391571

https://repo1.maven.org/maven2/com/diozero/diozero-core/1.4.1/diozero-core-1.4.1.jar

Still has the function as private. And the checks below that contain the previous ones.