Trajectory generator? Servo

[imET]

Hello,
I'm newbie in electronics and programming,
but highly motivated , please bear with me.
it took me a while to understand
data types are different between
compilers.
does anyone have example
of trapezoidal motion
calculation for a dc motor
position control(trajectory generation)
base on encoder feedback.
I'd like to know how I could accelerate/deccelerate.
I'm trying to recreate
microchip an696a but can't
understand(yet) the shifting and rotate
that was done in the appnote.
I just want something simpler, if exist.

Thanks.

[Ttelmah]

Unfortunately, servo control, is quite a complex area, and few systems will actually work acceptably with simpler algorithms.
Consider 'splitting' your learning. First control a motor. Use PWM, and just crack how to adjust it's speed, perhaps using buttons for the input. Then crack getting a position feedback from a shaft, using quadrature encoders. You now know how to read 'where' a motor shaft is, and adjust it's motion. You can then try a very simple approach, of just changing the speed to move 'towards' a target position. You will then learn about the big problem of inertia. By now, you will start to have an idea of the problems of motion control.
Now consider ignoring the servo motor for a while, and look at acceleration profiling on stepper motors. This allows you to ignore the servo algorithm, and just look at the actual 'motion'. There is a good algorithm for an acceleration curve, on the MicroMouse site. The table for acceleration here, can be coded as a look-up table. Once you understand this, go back to the servo algorithm, and now, when asked to move a distance, you can use the values in the lookup table, to adjust the target position and give the required acceleration ramp.
The problem is that there are multiple 'parts' to making a servo system work. You have the basic (probably PID based) motion code for the motor itself, and then on top of this, any acceleration profiling as well. All this has to be handled _quickly_, if the system is to work. If you try to go to 'simple', you have to lose part of the actual control, or become too slow to work properly. The servo system outlined in AN696, is about as simple as it gets!. However going to a stepper based system, for the 'learning' phase, allows you to seperate the motion profiling from the control algorithm, and possibly understand more easily, just what is going on.

Best Wishes

[imET]

[Ttemah]

OK. You have got quite a long way already. Think about it like this. If you have the motor at one position, and tell it to move to another spot a log way away, the servo algorithm, will 'see' a massive positional error, and slam the power on 'full bore'. The motor will then try to accelerate very hard indeed, and may even cause damage. At the other end, a similar problem will apply as you slow down.
Instead what you do, is look at how far you want to move, and move the desired position in a series of steps towards this target. If these steps start very small, then get larger, then get smaller again as you approach the target, the servo algorithm will 'chase' this moving target, but not have to deal the single large massive error, or the large acceleration this deals with. If you limit the maximum step to the fastest you want the motor to move, you have the ramped acceleration profile.
What the code does is add numbers in the required direction, to the desired position, till a maximum required speed is reached (velocity limit). At the point where maximum speed is reached, or the half way point whichever comes sooner), a 'note' is made of the offset from the start, and when this distance from the destination is reached, the same algorithm is run the other way (moving the target successively smaller amounts), till the final destination is reached.
The attached below, is my own ramping code used for this on a servo sytem.

[imET]

thanks so much,
this considerably helped me understand
how things work in the an696a appnote.
I'll digest this some more and see how it goes.

[dani]

hello,

<qoute>
Think about it like this. If you have the motor at one position, and tell it to move to another spot a log way away</qoute>

What if we want to drive servo motor with step and direction input comming realtime from parallel prot through some cnc software.

thanks

Dani[/quote]

[Ttelmah]

Two seperate parts to this.
What you do, is build the servo control algorithm, as already outlined. Then seperately, use hardware (probably one of the CCP modules), operated by the 'step' signal from the PC, to determine the position input to the servo algorithm. Step/direction, is slightly awkward with the CCP modules on the PIC, since these do not support a 'direction' control. Normally, this can be done by using both CCPs', and external logic, so one becomes the 'up' counter, and the other the 'down'. On the simple servo system, this can't be done, since one CCP is already used for the PWM. It is actually much simpler to let the PIC do it all, and just send it movement commands via RS232, or USB. This is what I have done in the past for such systems, and is what most commercial systems do. If step/direction, must be used, then the simplest way, is to use an external counter (typically an 8bit up/down counter), interfacing it's parallel output to one port of the PIC. Then use it's 'carry/borrow' bits to feed the hardware interrupt on the PIC, to handle the extra count bits. I'd suggest using an external quadrature counter, like the LS7166, or a PIC with built in quadrature support, otherwise the workload in the processor will start to become unacceptable.
It is worth saying, that CNC software using the parallel port, is a 'dangerous' way to go for the future. The lifespan of this port is already visibly shortening, and the better motion control units are already offering alternative interfaces. For a 'homebrew' system, the port is acceptable, but for commercial applications, it would be well worth considering alternatives.

Best Wishes

[dani]

Hi,

This is true that the parallel port in modern pc mother boards is being replaced by USB etc.

There is another fact is that CNC software like MACH3 which controls stepper motor board directly with its step and direction output are cost effective solution to make a good retrofit for cnc machines.

Now using stepper , It is ok with parallel port step and direction output.

But with bldc servo motors on cnc machine... I we can convert these step/direction output for servo motor the quick retrofit is almost ready to serve lots of people.

At the moment finding Mother board with Parallel prot is not too difficult. For future we may switch to USB.


So thats y i want to work on PIC based STEP/DIR Servo Controller. Can any one help me in this regard...
your views plz...?

[adrian]

Another problem with using a PC is that if you accidentally blow up your parallel port, then you need to replace the motherboard. One option is to use a 'sacrificial' PCI card with a parallel port.

[Ttelmah]

Seriously, 90% of the 'work' in Mach, is generating the acceleration profiles, and outputting the data for steppers. this is completely uneccessary for servos. They are capable of ramping their own accelerations. I did a system like this about ten years ago, and at the time used a couple of Z80's, with their CTC chips. The motor controllers are just fed the maximum speed, and acceleration rates, on a serial bus, pulled from the G codes', and are then just told to move. Far easier to do, than trying to make a servo behave like a stepper, and smoother too. The PC code to take the required data from the G code, was only a few tens of hours of programming, and still runs in a few systems today. Trying to make a servo behave like a stepper, just so you can drive it from Mach, is a 'sledgehammer/nut' solution.
Also, you may well find, that Art, is quite happy to do you a version of Mach, which outputs just position commands in serial for each motor. He is friendly to talk to, and has done custom versions like this for other motion solutions.

Best Wishes