NEED HELP with multiple servo motor

BLDD · Joined: 24 May 2015 Posts: 8

Hi, i'm new here and new in the ccs world!
I'm doing to my mechatronic's school project an Animatronic Hand using 5 servo motors. To the position of each servo i use a glove with a flex sensor for each finger and then read the sensor signal with my adc.
But it's not my problem at the moment, i can control a servo with my finger perfectly, but i just can control one at once ... when i try to add some reads of the other adc channels in my program everything goes wrong :/ I know its a coordinate / delays program because i should respect the 20ms low output to control.
I just don't have idea how to do it. I read a lot about this, should i do the servo control with timer interrupts???

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

You're missing the setup_adc() line.

Also, you need to post your PIC and the PIC's oscillator frequency.

Also, the line below implies that you have external Vref voltages
connected to your PIC's Vref pins. Do you ? What voltages are they ?

temtronic · Posted: Sat May 30, 2015 5:43 pm

and... this line...

servodelay = servodelay*.8+(value*2.8+250)*.2;

won't execute as you expect.

you declared servodelay as a 16 bit integer yet use floating point math....

Jay

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

Actually the math will work OK. CCS will promote the int16 to a float
before doing the math. Because he puts the result into an int16, the
fractional part of the floating point result will be truncated. But that's
not a big concern.

temtronic · Posted: Sun May 31, 2015 5:13 am

interesting as I'd thought it might not work that way and you'd have to cast to get correct result.

it'll be slow though compared to integer math and that _might_ affect performance.

Jay

Ttelmah · Joined: 11 Mar 2010 Posts: 19504

Yes.

It'd work, because there is a float in every component of the maths. If you parse along the line as the compiler would, at the first operation, there is a float on the right, so float maths will be used. Then the result is a float, so the same is true at the next operation. Similarly inside the bracket, the multiplication will be float, then since this result is float, the addition will also be float, giving a float result. etc. etc...

As you say, very slow though.
I must admit I'd probably be looking to see if I could change the scaling of the values generated (change the Vref....), so that nice binary values can be used, like /4 etc., and integer maths throughout...

BLDD · Joined: 24 May 2015 Posts: 8

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

asmboy · Joined: 20 Nov 2007 Posts: 2128 Location: albany ny

it seems you are using hobby servos of the pulse width control type for fixed position servos and updating them with adc reads.

To make this decently "smooth" in operation you need to know what resolution of both input and output are required.
Do you know for a fact that your servos can accurately resolve 1 millisecond?

If so - then positioning on a stop to stop basis - it could require 13 or 14 bits adc resolution to match the degree of control possible with the servo.

Also a key point : you will N E V E R make working code using the delay_xx()
for more than one servo -
since you can't take a new ADC reading while stuck waiting for the next bit of useful execution.
or anything else for that matter........

I actually had to do JUST what you are attempting for an opthamological
research device and it is very doable.
BUT first you need to learn to use TIMER functions and a "state machine" to
interleave taking adc readings with timing pulses of different lengths , synchronously.
where the stop and starts are not needing to align in any way.

I'd suggest picking a master clock frequency that makes it easy to set up a system timer for continuuous 1 or 2 millisecond rollover and then poll its
I-flag at the 'top' of your tight-loop state machine.

May i assume you are a student?
If so that is the first step on a very ambitious learning curve.

here is a hint: a master clock frequency of 16.384mhz will be more useful for you.

BLDD · Joined: 24 May 2015 Posts: 8

Hello,
I'll upload in youtube a video with my servo working Very Happy

What you recommend to me? switch my pic device and do the project with the adc/pwm hardware or its possible to do with pic16f877A only using software?

asmboy · Joined: 20 Nov 2007 Posts: 2128 Location: albany ny

ckielstra · Joined: 18 Mar 2004 Posts: 3680 Location: The Netherlands

I guess it is possible to control multiple hobby style RC servos from a PIC16F877A but it's not a project for a starting programmer.
Controlling one servo is easy as you have shown in the posted code, but for multiple servos you are hitting processing power limitations with this old processor.

One requirement you didn't mention is how many different servo positions do you want to be able to select?

Problem in creating software PWM for multiple channels is that you have to split the timing and logic into two separate functional blocks and this creates a _lot_ of overhead. The overhead reduces the number of selectable servo positions.
For example see this project in the Code Library. It uses an 48MHz PIC18 to create 8 servo outputs but for each servo can only address 64 positions. Changing the number of channels has no effect on the number of addressable positions.

I don't know what your school project's purpose is and what budget / time you have. Here are some options:
1) Copy your existing design for each motor.
2) Copy your existing design for each motor, but replace the processors by a cheaper PIC10 or PIC12.
3) Do not use software driven PWM but select a PIC model with hardware PWM for each servo. Beware of the PWM resolution. Models with 10-bit PWM will give you only about 51 addressable positions so better choose a PIC with 16-bit PWM resolution. The PIC16F1786 for example has 11 PWM channels with 16-bit resolution. Costs $2.54 when buying one piece from Digikey.com. This is the best technical solution but perhaps not feasible when the chip can not be ordered in your country.
4) Use the project from the Code Library. Note that the number of addressable positions depends on the clock speed. At 48MHz it gives 64 positions. Your PIC16F877a can be clocked up to 25MHz. Accept 32 positions or use another processor.
5) Be brave and try to squeeze multiple software PWM channels into your existing PIC16F877a design. The referenced Code Library project uses interrupts for timing. Interrupts have high overhead. Polling the timer flag instead will create more processing time. Implementation will take more time than the above options, has limitations and gives lower addressable positions but you will learn a lot and can use your existing hardware. We can help you with this but we expect you to put in some effort too.

BLDD · Joined: 24 May 2015 Posts: 8

ckielstra · Joined: 18 Mar 2004 Posts: 3680 Location: The Netherlands

Don't worry about your English. Many people here, including me, are not native English speakers.

Now we understand more of your context it is time for you to tell us more about the requirements for your project.
1) I already asked about the number of positions you want to address. How many positions you require minimum?
2) You have seen that more positions asks for more processing power. The PIC16F877A you are using is very old and not recommended for new designs. Newer models are now being used which are way more powerful, cheaper and consume less current. Is it possible to use another processor?
3) Choosing a chip with hardware PWM will make the program a lot easier but you'll learn less. What do you choose?
4) You are now using the ADC for input. Will there be other inputs or communication as well? (RS-232, buttons, display?)

BLDD · Joined: 24 May 2015 Posts: 8