How to use the NCO module- demo code

[viki2000]

OK, then here we are without interrupts:

[viki2000]

And next step was analyzing the long width for “on” and “off” with better precision.
For example when I try 10ms then the jitter error + the error delay generated by code ads itself. But the main error noticed comes from increment. We have a fixed space of 1048575 to fill with a fixed increment. At high delays, the increment value multiplied by needed times does not match exactly the accumulator of the NCO.
Therefore, instead of working with NCO overflow flag, no matter if is NCO interrupt or polling NCO overflow flag in a loop, I found better next solution:
- Increment only with 1 pulse the NCO accumulator.
- Check the value of the NCO accumulator in a loop.
The precision obtained is 20us jitter at 10ms width “on” or “off”.
Here is the XC8 working code:

[viki2000]

Sometimes, when I test I use the real chip, and other times simulators or debuggers.
As simulators I tried mostly MPLABX and Proteus.
PIC10 Simulator does not work with PIC10F322:
http://www.oshonsoft.com/pic10.html
The same with Real PIC Simulator:
http://digitalelectrosoft.com/pic-simulator

For the present tests I tried besides MPLABX, the Proteus 8.2 SP2 which has a small “bug”, or better say a missed update.

PIC10F322 has updated the datasheet regarding the NCOxCLK: NCOx INPUT CLOCK CONTROL REGISTER.
The new devices and the new datasheet 2011-2014, page 125 bottom, have:
https://drive.google.com/file/d/0BwXmKaSw75eKU2hQRGdRTUFBemc/view
bit 1-0 NxCKS<1:0>: NCOx Clock Source Select bits
11 = LC1OUT
10 = HFINTOSC (16 MHz)
01 = FOSC
00 = NCO1CLK pin

The old datasheet 2011, page 135 bottom, indicates a different order:
https://drive.google.com/file/d/0BwXmKaSw75eKaVNySl9FVHdKWVU/view
NxCKS<1:0>: NCOx Clock Source Select bits
11 = NCO1CLK
10 = LC1OUT
01 = FOSC
00 = HFINTOSC (16 MHz)

Proteus 8.2 SP2 works with the info from old datasheet 2011.

Any suggestions how to debug the not functional last CCS code above?

[asmboy]

it looks to me like you did solve it with your post of
Thu Jul 16, 2015 1:12 am

Any info produced by Proteus - is of no help really.

But i would not be surprised to find that the CCS code for the NCO on
your 10F... part is sketchy. This is not so much CCS fault as the
fact that in the 20us/250us mode , jamming the control register pair as you do is not performed in a vacuum.

In the most successful version you coded here is what is happening:

The accumulator is already "counting up" for every i-cycle that occurs in your code stream. Your test for which value to load into the double buffer compare register presently checks for the LONGER time cycle first.

I think You should reverse the order and test for the ELSE case first .

When you zero the accumulator at the end of the test/load code you are "off the timing mark" by the time it has taken to do the tests.

Slightly less jitter will occur if you test for !input(pin) case FIRST ..

But all in all - a bigger problem is still there.

the NCO was designed i think to generate steady frequencies without the jamming you do on every cycle of the counter .

I think you might get better performance yet if you
clear the flag on every detection BUT if the recorded state of the INPUT control pin has not changed - then DO NOT re-jam the counter on every pass.

if the control pin has not changed - just clear the flag and exit.
Only set the alternative register value if the control pin has changed state .

That ought to further cut down on jitter.

Unless i am missing something about your schematic ( which you never posted) i can't suggest anything else.

[viki2000]

Thank you for the suggestions.
I do not think is a hardware jamming of the NCO as long as the XC8 code with 1 increment and the same poling of the accumulator works.
I find rather a problem of CCS when the code is compiled in assembly and when NCO functions are wrong declared in “10F322.h” as 16bits instead of 24bits, knowing that NCO accumulator is 20bits.
The INPUT control pin is in fact the same as OUTPUT control pin. I set it on or off.

I have no schematic. I use only a breadboard with PIC10F322, stabilized 5VDC and one DSO 350MHz.
To program I use ICD-U64 programmer inside the CCS IDE and PicKit 3 programmer with MPLABX.
It is just a test, not a real application.

I tested one more NCO function in XC8, which again is not working in CCS.
It is a microseconds delay function using NCO working in the range 20us-250us with a precision of +/-1us + oscillator jitter.
I increment the NCO slowly and I am looking at the NCO accumulator. It is more precisely than polling NCO overflow flag or using NCO interrupt, because NCO increment does not match always the NCO accumulator exactly.
I tried NCO1INCL also with 1, but the delay function cannot coop with such speeds and there is no signal generated. The NCO1INCL with 2 works, but is still too fast to get a good precision for the function in the range 20us-250us.

[PCM programmer]

[viki2000]

CCS C complier version 5.048
Here is the LST file:
https://drive.google.com/file/d/0BwXmKaSw75eKMjc4SUoxd1prbmM/view?pli=1

[viki2000]

Here is the official answer just received from technical support: