PIC24EP512GP806 - I2C not working after sleep/wake-up

[benoitstjean]

Device: PIC24EP512GP806
Compiler: 5.026 (all of a sudden my 5.116 version no longer opens for some reason)

Hi guys,

Its been a while since I posted here and this pertains to a piece of code written a long time ago and I'm just curious about the behavior.

I believe this was actually TTelmah who had given it to me.

Basically, my device goes to sleep and the following code is called:

[gaugeguy]

Before you initialize the I2C module (or any other module you want to use) you need to turn it back on in the appropriate PMD location. The data sheet will detail which bits disable/enable which peripheral.

[benoitstjean]

Actually, I may have just realized that I may be confused:

1) I call getenv(), print-out the value;
2) I set all the values to 0xFF then I print them again without calling getenv() and the values are the same;

Unit goes to sleep.
-wait-
Unit wakes-up.

Then I don't call getenv(), I simply print the values again and they are again the same.

So I guess I'm really confused as to how I should be getting/setting values.

[gaugeguy]

Read section 10 of the data sheet on Peripheral Module Disable. specifically 10.4 and 10.6

Your existing code is also suspect as the PMD registers are 16 bit and your code is treating them as 8-bit.

[benoitstjean]

Oh yeah, good catch on that one but I'm still trying to wrap my head around the usage of getenv and setting these value registers. This code has been there for years and I never looked back at it....


I changed #byte to #word just to see if it would maybe change something but it does not...

[benoitstjean]

UPDATE: So I've commented-out all the code pertaining to PMD registers and after waking-up from sleep, I2C is still off because my sensors aren't working contrary to when I do a real power-up.

In all honesty, I have no idea how the use the getenv() function for what I want to do and CCS docs, well, it's CCS docs. It's good for what it's worth.

The problem here is that I know what my end goal is, I just don't know how to achieve it. I've looked at section 10.4 and 10.6 of the PIC manual and that's all nice and fine but it doesn't explain how to do what I want to do using the CCS compiler.

So:

1) How do I use getenv() to get/read the current status of PMD registers (so that I can set each PMD register back to these values upon waking-up);

2) How do I disable all PMD registers to save on power;

3) When the MCU wakes-up, how do I set-back all PMD registers to the same states as they were before going to sleep;

4) How do I confirm that what I've set the PMD registers to has worked (using just an fprintf is fine);

5) How do I force I2C to be up and running after waking-up from sleep, if that has anything to do with the PMD registers;


Thanks again and sorry for the confusion.

Ben

[gaugeguy]

You have many different issues.

If the I2C is not working after sleep, and the PMD register access code is commented out, then the problem is in code you have kept hidden. There may also be hardware issues. What happens to the I2C pullups during sleep?

genenv("FSR:xxx") returns the address of a FSR. It can be used with any FSR.

Your chip uses PMD to refer to the Peripheral Module Disable as well as the Parallel Master port Data. Check the assembly code to make sure getenv is returning an address for the correct register.

The Peripheral Module Disable registers can be read and written at any time. If a module is disabled all of its setup registers are shut down. When the module is enabled again all of the setup registers must be set correctly again. It does not pick up where it left off.

[benoitstjean]

Thanks for your response.

> You have many different issues.
Nope, only I2C is not working after a wake-up from sleep, everything else works as expected (e.g. both UARTS, SPI1 and SPI2, timers, DMA etc). So to re-iterate, when the device does a normal boot-up or reboots via the reset_cpu() function, the entire circuit works as it should. When the device goes goes to sleep then wakes-up, only I2C is not working... and as I am writing this message, the code for the PMD registers has been commented-out and the issue of I2C not working is still there at wake-up.

This leads me to the following: for testing as I am writing this message, in the wake-up sequence, rather than running the standard wake-up code, I added reset_cpu() and this caused I2C to be properly initialized and the sensors now work.... so considering that my sleep code does not disable PMD's at all because I removed the code, just the action of sleeping and waking-up affects I2C... which brings me back to my initial post's subject: usage/syntax/implementation of getenv()...


> There may also be hardware issues. What happens to the I2C pullups during sleep?
I wouldn't call them hardware issues because the hardware works fine at regular power-up or after a call to reset_cpu(). The wake-up sequence follows pretty much the same path as the standard power-up sequence other than the modem is not initialized - it is simply taken out of airplane mode (circuit has embedded modem, SPI and I2C devices). In that wake-up sequence, the same I2C code from the standard boot sequence is executed because it is in a function and that function is called only on two occasions: at regular power-up and at MCU wake-up.

Therefore, in that 'boot function', to prevent the I2C bus from locking-up in the case where the sensors did not respond for whatever reason (not present or faulty), this is what happens to SCL and SDA:

[gaugeguy]

I think you need to take a step back from getenv and read up on what #byte and #word actually do and how they are used.

[benoitstjean]

getenv("SFR:PMD1") tells the compiler to replace this with the absolute address of the PMD1 special function register. So this returns something like 0x0700 for instance.

#word PMD1 = ... tells the compiler to "create a new 16-bit variable called PMD1 and map it directly to that address". The code will treat PMD1 like a normal C variable but it reads/writes that hardware register.

Basically, it's a direct access to that register via #word PMD1. Fair enough. I understand.


This said, I've re-added the following at the beginning of my power down sequence:

[gaugeguy]

You have not shown your complete code so I am going to make an educated guess with my crystal ball.

You initialize the I2C module but then when you go to sleep you do not de-initialize it so you can start over again fresh on wake up.

[benoitstjean]

Yea, that is in fact what seems to be happening...

So then how would I "de-initialize" I2C at sleep and "re-initialize" at wake-up?

Because clearly my boot/reboot code does something to the I2C module because it works consistently and flawlessly until it goes to sleep and wakes-up....

Thanks again,

Ben

[gaugeguy]

Have you looked up what i2c_init does?

[benoitstjean]

From the manual: