[ESjaavik]

Greg: Yes, it's quite common that servo drives take step/dir input. All I have does that. Internally it counts the desired-position value up/down based on step/dir. This means the drive must do the torque, velocity and position loops computations internally. With a voltage input drive often doing just the first or the two first.

And the step/dir servo can take pulses at high speeds. If the motor is not able to follow, it will not loose the end position. It will just loose position on parts of it's journey there. Not like a real stepper where a lost step will never be regained. (Unless by chance it looses the same number of steps going in the other direction.)

Here is a simple approach to the main components, except step/dir input: http://elm-chan.org/works/smc/report_e.html

[Greg Fill]

Thanks that makes sense ES. The motor will resolve its final position based on commanded number of steps, direction and the encoder feedback. I guess there is probably no need to set an acceleration profile in the computer front end, you just feed the step pulses as fast as the controller can accept. In my case probably only 80 KHz or 160 kHz. So to say it another way the controller uses the constantly arriving step pulses to dynamically set a position to reach and uses the PID algorithm to get there. Is there any type of feedback the controllers and PCs use so that the front end controller will stop sending step pulses until the commanded position is reached?

ie goto x = 3, y = 2 then wait until it gets there before sending x = 4, y = 1

I will read the document you mentioned.

[tachus42]

Ok source code for my servo controller in the attached file.
It's probably not very clear whats going on at a first look - i didn't comment the code with other viewers in mind. Basically all the important stuff is handled by the interupt handlers at the top of pid.c. The SIG_OVERFLOW0 handler gets excuted at a 40kHz rate and reads the encoders and updates a 8 bit postion.
The SIG_OVERFLOW2 handler runs at 400 Hz or every 2.5ms and transfers the 8 bit encoder and target counts into their 32 bit equivalents and then calls DoPID where the PID algorythm takes place and the PWM drive to the motors is updated.

The SIG_INTERRUPT0-2 interrupts are excuted on the falling edge of a step pulse and read the corresponding direction input and update 8 bit target postion.

The rest of the code Initialises the timers and Interrupts etc and Drives the LCD reads the keypad does the menu etc.

[CNCAddict]

tachus42,
Do you want me to put this on my webpage so it can be referred to more easily? I am trying to learn as much as I can about AVR's, and this looks like a fun project.

David B.

www4.ncsu.edu/~dsbloomf

[tachus42]

David feel free to put it up on page jusr credit me, Stephen Dredge ( not that i think the code is that worthy)
The code was compiled with the gcc-avr http://winavr.sourceforge.net/
And I use pony prog to program the avr http://www.lancos.com/prog.htm with the (STK200/300) parallel port interface descibed on the pony prog page.
The LCD Port will connect to a standard 2 * 16 LCD with a ribbon cable and the keypad is a 4*4 matrix.

[ESjaavik]

It's probably not very clear whats going on at a first look - i didn't comment the code with other viewers in mind.

I've had the doubtful pleasure of reopening my own code years later. That taught me to try to comment for other viewers. :frown:
You might say I'm not the same as I used to be.

Thanks for sharing the code. What device do you use?
Myself I have a fairly full setup for older AVR's (STK500 + ICE200). But the only design that came past the qty. one breadboard stage was with ATtiny15 where the ICE will do nothing.

Would it be possible to see the schematics in .PDF .GIF or some other easily viewable format?

[tachus42]

I used a Mega 162 for this design.

Jpg's attached the schematic's aren't very clear but enough to give the idea, If you want a closer look google for and grab eagle - there is a free version that has become a bit of a defacto standard.

I think that stepper emulation is the only sensible way to make a roll your own servo controller unless you really want to be a smarty pants. A modern pc is going to be able to a much better job of motion control than a puny microcontroller can.

[CNCAddict]

ok, it's all up on my webpage now for easy viewing. Tachus, if you have any pictures of your assembled board it would really help out. Right now the page looks pretty anemic

http://www4.ncsu.edu/~dsbloomf/Steve...vo%20Drive.htm

[gabrielus]

do you have any documentacion about this hardware?
and what is the software can i use with this hardware?

[ESjaavik]

Thanks that makes sense ES. The motor will resolve its final position based on commanded number of steps, direction and the encoder feedback. I guess there is probably no need to set an acceleration profile in the computer front end, you just feed the step pulses as fast as the controller can accept.

Yes, and no. Let's say you want to mill a large circle. Then you feed in all the pulses as fast as possible. The motors will start out in the right direction. But very soon the 'puter will be done feeding in all the pulses and the desired position will again be where it started. But the slow motors/mechanics will all the time try to move towards the desired position. So you will not get a circle, but just a small doodle.
That's the "no". You need to wait out at least some of each incremental movement.
The "yes" is that each incremental movement can be pulsed out as fast as your controller can receive them. As the pulses are not used to control the motor, there is little constraint in the timing. This makes it possible to put less realtime constraints on the PC doing the machine control task. It can output a bunch of pulses in succession, and do other computational tasks interrupting the pulse stream.

In my case probably only 80 KHz or 160 kHz. So to say it another way the controller uses the constantly arriving step pulses to dynamically set a position to reach and uses the PID algorithm to get there. Is there any type of feedback the controllers and PCs use so that the front end controller will stop sending step pulses until the commanded position is reached?

ie goto x = 3, y = 2 then wait until it gets there before sending x = 4, y = 1

The feedback is up to you when you design the drive. You can implement a "move done" as a status output to the PC. But I use them in stepper-emulating mode. That means the PC thinks it feeds a stepper drive. And it calculates each move as for a stepper. But as I wrote, without the timing constraints down to the spacing of each step pulse. If using a move-done output on the drive, the control software must handle this. I don't know about others, but neither EMC nor Mach2 does this.

One output you should have is a following error output. The drive knows the commanded and the current position. When the difference is greater than a configurable amount, you output a following error. The PC would typically use this as a fault input. It could mean that your axis have met a solid object along the path that should not have been there. Some following error must be allowed, as a PID loop with no error is an infinite improbability.

[zekk_2]

Hi
There seems to be somthing missing from this cirut. Where is the 4 X Zenadiodes hooked up to??(left of 4 X U3)?
regards
Zekk_2

[zekk_2]

Greg: Yes, it's quite common that servo drives take step/dir input. All I have does that. Internally it counts the desired-position value up/down based on step/dir. This means the drive must do the torque, velocity and position loops computations internally. With a voltage input drive often doing just the first or the two first.

And the step/dir servo can take pulses at high speeds. If the motor is not able to follow, it will not loose the end position. It will just loose position on parts of it's journey there. Not like a real stepper where a lost step will never be regained. (Unless by chance it looses the same number of steps going in the other direction.)

Here is a simple approach to the main components, except step/dir input: http://elm-chan.org/works/smc/report_e.html

Hi Einar
Ref. your link to :Here is a simple approach to the main components, except step/dir input: http://elm-chan.org/works/smc/report_e.html

Can you download the software that is posted on http://elm-chan.org/works/smc/report_e.html via the serial I/F on the schematic, or do you need a separat burner to download the software??
Regards
Zekk_2