PWM for servos, motors, and LEDs plus.

***************************************************************************************
NOW VERSION 3.

See viewtopic.php?p=347830#p347830
***************************************************************************************

I have just released pigpio version 2, a C library.

Available from http://abyz.co.uk/rpi/pigpio/pigpio.tar

Video at http://www.youtube.com/watch?v=-sfzIygVQVM

N.B. The pigpio library uses the DMA (channel 14 by default) and the PWM peripherals.

It is likely to be incompatible with the accelerated X driver being tested as that may saturate the DMA bus and destroy its use for timing purposes.

Consider the pigpio library to be experimental and use at your own risk. That said I have not had a problem.

Its main features are:

• provision of PWM on any number of gpios 0-31 simultaneously.
• provision of servo pulses on any number of gpios 0-31 simultaneously.
• callbacks when any of gpios 0-53 change state.
• callbacks at timed intervals.
• reading/writing all of the gpios in a bank (0-31, 32-53) as a single operation.
• individually setting gpio modes, reading and writing.

NOTE: ALL gpios are identified by their Broadcom number.

There are two significant enhancements over version 1.

• The PWM cycle is now 787Hz (previously 394Hz).
• The levels of all (54) gpios are read and time stamped 200 thousand times per second (previously 1 thousand times per second).

This sounds very useful. I'm interested to see how you implemented it. I'm guessing that part is inside the pigpio.c.cpt file, but I don't know how to make sense of that file. What format is it? What program can work with it?

Thanks, and have fun!

The cpt is the encrypted C source.

I'm happy to release the source to anyone who wants it. My only condition is that you don't publish it.

Hello joan,

This library looks very cool. I would like to write a wrapper unit in pascal.
I have done this before with Gorden's WiringPi library.
In order to do this from pascal, I need an .o file of the library, which I can then link to from pascal code.
As far as I can tell, there is no way to compile the encrypted C file. Please provide the C source.
Once I have the .o file, I will be able to complete the wrapper unit.

Have a great day!

Thanks,
Alex

joan, great library. Thanks!

woodengineer, you can get the .o file by:

Code: Select all

# ar -x libpigpio.a

I've done it to create a shared .so version of the library needed for my project which uses node.js and the FFI module.

On my pi, in the pigpio folder I run tar -xf pigpio-V2.tar and get
tar: This does not look like a tar archive
Even tar -t ditto gets me nothing
any ideas?

On my pi, in the pigpio folder I run tar -xf pigpio-V2.tar and get
tar: This does not look like a tar archive
Even tar -t ditto gets me nothing
any ideas?

Download the tar to a known location, then copy to your home directory. Install as follows.

cp Downloads/pigpio-V2.tar ~

cd ~

tar xvf pigpio-V2.tar

cd PIGPIO/

cp libpigpio.a-hard libpigpio.a

make install

make

sudo ./checklib

@joan (or anyone else)

Is there a working way of using this from Python?

regards

Simon
PS I know all things are possible - I'm looking for "here is one I did earlier"

@joan (or anyone else)

Is there a working way of using this from Python?

regards

Simon
PS I know all things are possible - I'm looking for "here is one I did earlier"

I don't know. Not ALL things are possible.

On balance I think it could be done. Someone with any experience of integrating Python and C could probably give a considered answer after half an hour or so.

I'll start the timer

I'm just wanting to outsource PWM to background processes as doing it "manually" e.g GPIO.11 on wait GPIO.11 off is not too bad but once you get into wanting 2 pins doing it at same time (plus a couple of servos!) then I soon start to have pulses arriving before they've been sent in the first place

Simon

I'll start the timer
...

I'm afraid the timer will start if and when someone experienced in Python/C integration notices the thread and is interested enough to answer!

http://www.youtube.com/watch?v=aomqyL9VVl4 (persevere over the first half minute) shows an early C attempt at generating servo pulses. Glitches are fairly apparent as ill-mannered processes refuse to yield the processor.

[rurwin]

I'm no expert, but I was reading the docs this morning

There's no reason why most of the stuff couldn't be wrapped to work from Python, but callbacks may be a problem.

I'm no expert, but I was reading the docs this morning

There's no reason why most of the stuff couldn't be wrapped to work from Python, but callbacks may be a problem.

Richard (the servoblaster) and myself had a discussion re the relative merits of kernel versus userland daemon versus userland library implementation of gpio control.

I prefer a userland library as that most closely mirrors my usage.

However I think Richard's preference for a userland daemon for PWM and servo might better fit other's usage, I know Richard has a servo daemon, another chap has a PWM daemon. Perhaps they could be merged? That would at least be language neutral.

Love the semaphore servo's
Simon

This is very useful indeed. I only wish it was fully open source, because it would be more in line with the linux and raspberry pi philosophy

For those who prefer open source and/or are familiar with servoblaster, there is another project called pi-blaster specifically for dc motors which also uses DMA direct access to provide pwm.

Hi,
I am using this great library with some quadrature encoders and the Raspi never misses any edges any more like it did when I started playing around with the GPIOs.
For my software it would be great to have an additional void* in the callback-function that one can provide when registring the callback.
The prototype would look like this:

Code: Select all

typedef void (*gpioAlertFuncEx_t)(int gpio, int level, uint32_t tick, void *userdata);

and the function for registering the callback

Code: Select all

gpioSetAlertFuncEx(unsigned gpio, gpioAlertFunc_t f,  void *userdata);

For educational purposes I am rewriting my application using C++ and what I have read so far is that this addition would also simplify interfacing with C++ classes.

...
For my software it would be great to have an additional void* in the callback-function that one can provide when registring the callback.
The prototype would look like this:

Code: Select all

typedef void (*gpioAlertFuncEx_t)(int gpio, int level, uint32_t tick, void *userdata);

and the function for registering the callback

Code: Select all

gpioSetAlertFuncEx(unsigned gpio, gpioAlertFunc_t f,  void *userdata);

...

I've no objection in principle. I just couldn't think of any practical use at the time for a userdata parameter. Do you have a particular use in mind?

Thanks joan for your fast response.
I will try to explain what I try to achieve.
I am interfacing the Raspi with several quadrature encoders each using 2 GPIO pins. In my application the info for each encoder (which pins are used, the counter etc.) is stored in a structure.
I want to provide this pointer when registering the callback and I want to get back this pointer when the callback is actually called. Then I have the relation between the pins and the encoder info and I can proceed with the info. Otherwise I would have to store the info about the relation between the pins and the encoder on my own. I have already done this, but the way using the extra pointer is more elegant.
But why I asked for this addition is because I now want to interface with some C++ classes. I would register a pin for a callback and give the pointer to my encoder object. In the callback I can cast the given pointer to my class type and call the methods I need. As far as I have read this is the way how interfacing between a C library and C++ classes can be done.
I probably could do these additions myself but as far as I can see the sources for this great library are not released yet...

Hi Alex,
I was wondering how you had got on with wrapping joans library ? I have been playing about with Gordon's wiringpi library inside lazarus. Gordon's library only allows pulse with modulation on GPIO port 18, and he doesn't give full instructions on how to set it up. I am trying to make an IO package within lazarus to make it easier for other programmers to quickly implement lcd display drivers, motor controllers, etc within lazarus. I realise that you have been there and done some of these things before me and I don't want to duplicate your efforts unnecessasarily.

Jack

Hi Alex,
I was wondering how you had got on with wrapping joans library ?
...
Jack

I am shortly to release an upgrade which will expose 95% of the functionality to other languages. It might be worth holding off for a day or so.