(SOLVED) ECCP Full-Bridge PWM x 2 w/ 18F45K22

[Sergeant82d]

I am trying to get both full-bridge PWM outputs working on an 18F45K22. Using all the usual suspects (setting TRISD directly, writing to CCP1CON, PWM1CON, ECCP1AS, etc.) - in addition to just "setup_ccp1" - I have not been able to get both outputs working. PWM 2 works fine, but on PWM1, only the P1A & P1C pins are toggling with direction changes - no actual PWM output on P1B & P1D.

I believe, based on the ECCP1AS value (see below), that my problem is related to that register, but even though I have tried writing it directly I have not had any luck.

I am afraid I am at a complete loss while looking at the .lst file. I copied the assembly code and tried to use it directly, changing what I *thought* was wrong, but I do not know assembly at all, and was unsuccessful.

I am using Ver 4.110 under MPLAB 8.56 on Windows XP Pro, in hardware. I am using a scope & LEDs to monitor pin status.

I have searched the forums and the web:

http://www.ccsinfo.com/forum/viewtopic.php?t=43129&highlight=eccp

and although I tried everything in the above post - and many more - I could never get Timer 2 to work at all. Both Timers 4 & 6 work fine for me, except as stated above.

I stripped out the printf commands for posting, but here are the register values I have been printing from this code during my testing:

Forward:
PWM1CON = 183
PWM2CON = 101
CCP1CON = 189
CCP2CON = 102
ECCP1AS = 182
ECCP2AS = 100

Reverse:
PWM1CON = 183
PWM2CON = 101
CCP1CON = 189
CCP2CON = 102
ECCP1AS = 182
ECCP2AS = 100

Here is a test program:

[Sergeant82d]

What are the practical differences between using #byte and #define?

I have tried both, and other than #byte taking up more memory space, I get different values when I print the registers (in the post above).

When I use #define, I get values in the 180 (decimal) range and anything I try to write to the registers does not "take"... using #byte, I get whatever value I set.

In this printout, I have set:

PWM1CON = 0X80;
ECCP1AS = 0x0F;

and I get the below values. Compare them, please, to the values in the original post:

Forward: PWM1CON = 128
PWM2CON = 0
CCP1CON = 76
CCP2CON = 76
ECCP1AS = 15
ECCP2AS = 0

Reverse: PWM1CON = 128
PWM2CON = 0
CCP1CON = 204
CCP2CON = 204
ECCP1AS = 15
ECCP2AS = 0

The CCPxCON register values are correct for the mode and direction; PWM2CON & ECCP2AS I did not set; as you can see, when using #byte instead of #define, I see the expected values.

Unfortunately, the PWM outputs on P1B & P1D are still not working.

Don't know if or how they are related to my not getting P1B & P1D working, but I have seen this type of behavior in other code/chips here, and wanted to ask...

Thanks for looking,
Brad

[Ttelmah]

They are _totally_ different......
A bit like saying what is the difference between a screwdriver, and a hammer....

#define, is a simple MACRO. It is part of the basic macro language of C, saying 'when you see this word, substitute this'. You can do some really nifty things with this, since it does support some rarely used features for testing, but it really has nothing to do with the core compiler, just being a 'pre compiling' text substitution.

#byte, is a compiler command, saying 'create a single character variable, and place this variable _at_ the defined location' (it has a second more rarely used meaning, if you define a larger variable first - int16 say, then use _this_ variable in a byte declaration, it relocates the existing variable to the new location). The first meaning is what is involved here though.
In either case, it is putting an actual variable in memory _at_ the defined location.

Now, if you use (say):

#byte FRED=0x80

You now have a variable called FRED, located _at_ memory location '0x80' in the RAM. If you choose a location that corresponds to a physical processor register, then reading and writing this variable, will access the actual register.

If instead you use:

#define FRED 0x80

Then all you have is a text replacement, that if you type 'FRED', the number 0x80 will be substituted.
So if you then use:

val=FRED;

All that will happen is that 0x80, will be put into val....

You can actually access the contents of the address 0x80, by using the syntax:

val=*((int8 *)FRED);

Which says "treat the number 0x80 (FRED) as the _address_ of an int8 variable, and read it's contents". A lot more work.

Obviously, accessing a physical variable takes more code than just substituting a number, which is why #byte takes more space than #define. It is actually doing something!......

When you use #define, you are doing nothing whatsoever with the real registers.

Best Wishes

[Sergeant82d]

Thank you. I had read the descriptions in help files and other texts, but thought for some reason that using a pointer to the #define'd register would still access it.

Thanks for clearing that up for me!

[Ttelmah]

Potentially using a pointer 'to' a value will access it's contents, _but_, doing this, you are not saying how large the thing 'pointed to' is. This then means that you can end up talking to other addresses nearby that you don't expect - hence the need to explicitly 'cast' the pointer to be to a byte or int8, if disastrous results are not to happen....

Remember also that the pointer needs to be used for both reading and writing. I suspect you were missing this at some point, and hence getting the defined value, not the contents of the value.

You also should get into the 'habit' of bracketing defines. So:
#define CCP1CON (0xFBD)

This is 'often forgotten', and can lead to disasters, with the result when the expansion takes place, not being syntactically what is expected.....

Best Wishes

[Sergeant82d]

I noticed how you cast the pointer as an (int8 *), and figured it was something like that...

Looking at the values I was receiving, and doing a decimal/hex/binary conversion and comparing that to the register addresses, shows me that what you said is exactly what happened.

Thanks again!

****

You don't have any suggestions on my PWM problem, do you?
<grin>

Brad

[PCM programmer]

I'm the only one working on the pwm as far as I know. But anyway, here's what I came up with:

I think I made enough changes so vs. 4.110 will fit the Full Bridge section
of this diagram in the 18F45K22 data sheet:

[Sergeant82d]

Thanks! I'll try it when I get home tomorrow and can get hands on my board again.

*********************

EDIT:

IT WORKS!!!

THANK YOU THANK YOU THANK YOU!!! Now I can finally get on to the meat and potatoes of my project!

You are SO the man!

[mcr1981]