Navigator PCA9685 clock choice

[ThierryB]

Hi all,
Just a question about the electronic design of the Navigator board.
Concerning the PCA9685PW chip, why did you choose an external clock of 24.576Mhz instead of the internal one of 25Mhz?
Thanks
Best regards.
Thierry

[EliotBR (Eliot)]

Hi @ThierryB, welcome to the forum

@ThierryB

Concerning the PCA9685PW chip, why did you choose an external clock of 24.576Mhz instead of the internal one of 25Mhz?

As I understand it the internal one wasn’t very accurate, which is frequently the case for internal clocks in chips. I’ll confirm with the design team.

[EliotBR (Eliot)]

@EliotBR

I’ll confirm with the design team.

Following up, the issue was indeed inaccuracies with the internal clock.

In case it’s relevant, I asked for clarification on whether that’s

@eliotbr

… offset accuracy (e.g. it’s precise and consistent but one chip may run faster than another), or … inconsistent (e.g. the clock may run faster than expected at one point in time, and slower at some other point)

and was told

@jwalser

It’s a … device-by-device variance [of] initial offset. Some devices work with the scaling factors calculated using ‘typical’ values, some devices need to be tuned higher or lower to work right with the ESC deadband. We wanted them all to work with the same programming out the door.

The external clock also vastly improves the timing spec. … across the whole temperature range.

[ThierryB]

Thanks a lot for your responses. Clear for me. So not really a issue with the frequency but with stability as well as variance from a chip to another.
Thanks.
Regards

[EliotBR (Eliot)]

@ThierryB

not really a issue with the frequency but with stability as well as variance from a chip to another.

Not an issue with the regularity of the frequency, but yes an issue with the accuracy of the frequency, and inconsistency between chips - e.g. specifying 50 Hz with two chips using their internal clocks may yield one output signal with 51 Hz and one with 49 Hz (indicative values - I don’t know the actual variance), and that scaling also applies to (affects) the duration of each pulse (which is the data component in servo-style signals).

If you’re using the chip as e.g. an LED driver then the main control would generally be with duty cycles rather than pulse durations, and the clock being slightly scaled would not have much of an effect on that, other than potentially causing a slight flickering when taking a video.

If precise timing of the pulse duration is important (like with a servo-style signal) then you need a sufficiently precise and accurate clock signal. The inaccuracy of the internal clock can be compensated for by scaling adjustment on an individual basis (which we didn’t want to do, because measuring, storing, and applying that calibration data is extra work and logistically challenging), but imprecision or drifting caused by electrical noise and/or temperature changes is only really avoidable by maintaining a constant temperature and electrical noise levels (which for the Navigator’s intended usage environments is impractical), or by using an external clock that is less prone to those issues.

[ThierryB]

Clear for me. Thanks for your explanations. I will not make the economy of an oscillator then.
regards