Two clock source switching - RESOLUTION

[SuperDave]

Thanks to all for their help, here's what I found including some things that are not obvious but very important.

1. Control the PLL either in #fuses and/or setup oscillator especially if the external clock is not between 4-8 Mhz per 16LF1828 spec. If the PLL is running and the external clock is outside the spec range it's crazy clock time.

2. I ended up using external clock all the time with the external clock switching from 135kHz to 26MHz. #flags was set and left at ECH. This increases current draw during the brief time that 135kHz is active but works.

3. Think through #use delay(). If you set it then call an include file, the include file will be compiled with that delay. If the include file changes it, then the include file will from the change point be compiled with the change. BUT, when you come back after the include, the OLD delay will be active, you need to change it again. At least that's what appears to happen. Extra #use dalay() lines are cheap, easy and clarify debugging.

4. Side note. Using the pullup on MCLR as an input apparently does not work unless you manually clear WPUEN. Activating pullups on other Portx pins do not require manually clearing WPUEN. Reset defaults the pullups to On but WPUEN to OFF(1)!

Thanks again.
------------------------------------------------------

At startup I want a pic18lf1828 to use its internal RC clock. It then uses SPI to change the startup on a "wifi" chip so that the "wifi" chip outputs its 26MHz clock to the ClkIn pin on the Pic. Then the pic switches over to the "wifi" clock. (I need its accuracy.)

Here are the simplified relevant code pieces.

[asmboy]

check the 16F1829 sheet:

OSCON register

[Ttelmah]

You have to rely on using the FCMEM fall back to do this.
The clocks you have are:
1) External clock on the main oscillator input.
2) Internal RC.
Now, the alternative of the timer1 input for the external clock, can't handle this frequency (16MHz max).

So, with ECH selected as the 'master clock', if the 26MHz is not present when the chip starts, FCMEM will allow the chip to boot. Without this the chip won't boot without the clock (comment below).

Now assuming you are not going to do anything 'timed' when running off the internal clock, do as you currently show, and set the clock statement to reflect the high speed clock.

Delays are set at _compile time_, so if you switch clocks, it becomes your responsibility to handle timing things differently...

Then at the start of your program, set the internal clock to the rate you want, as you currently show. So:

setup_oscillator(OSC_16MHZ | OSC_INTRC);

Normally this is done for you, if you specify the frequency in the clock setting, but since you are not using a setting the internal RC can generate, you need to do this. With the clock rate selected to give the one you want.
It'll get to this point using the 'fallback', if the external clock is not present.

With the oscillator set like this, and the clock setup given, time delays in the code will run at 0.615 of the speed specified. So delay_msec(10) will give 1.6mSec etc...

Now you can code round this using your own delays. So:

[SuperDave]

I'm going to do some other tests today but to be clear on what I think the order of events is at startup now (see the code in my first post)

1. The pic is told to default to external clock high (ECH) but use internal RC now.
2. It's not in the code but the pic is getting (but not using) 192kHz on ClkIn from the external chip. I believe this would keep FCMEN (used to be FCMEM) from tripping on a failed clock.
3. The program starts with 16MHz from the internal RC as requested.
4. It writes thru SPI to the external chip to change the 192kHz to 26MHz.
5. The external source and ClkIn change to 26MHz, no question.
6. The pic is told to go back to the default ECH clock (ie. 26MHz).
7. It doesn't.

My first test thought is to move the #use delay to after the SPI/ECH change, though I would think if that's the problem, the observed pulses would be shorter not longer.

My second test thought is to try the external clock alone just to prove the pin and then try switching from internal to external and back without trying to change the external frequency. I have a push button that would let me do this repeatedly.

Somewhere there's a reason (I think).

[Ttelmah]