This is my second post and i am beginner in stepper motor field. I am interested to hear your opinions what are the must features of modern stepper motor driver ?
Some come to my mind :
Fast decay with synchronous rectification, slow decay, mixed decay,
back EMF and motor current sensing,
PWM off time adjustment based on motor rpm, microstepping ,
sine/cosine current profile adjustment ,
holding torques ripple control
active resonance damping
idle power mode
What can be added to above list and how big is the impact of items to motor performance ?
Thanks in advance
if we can put everything into soft it will be cheap . I guess newly arm chips starting from 3-4 USD running at 60 MHz will allow that .
Most of the features, you correctly selected, are already implemented into a 20 MHz ATtiny2313 with only 2k flash. Look at this thread: http://www.cnczone.com/forums/showthread.php?t=25361 (unipolar motor microstepping)
The code is not optimized (assembly) yet, but it is working at 45,000 KHz step input (tested). Some of the features don't apply to this design because it is an Unipolar Driver, except for the synchronous rectification that could also be implemented in hardware without adding too much cost.
You don't really need too much processing power, unless you are trying to implement a closed loop control system.
I so this thread 2 hour before you posted here.)
Nice job. I will try to study it .
I think you don't need complete assembler implementation. Major part can be written in C and critical part such as interrupt handling in assembly if required . But i guess you know this already.
The first thing i so there - is chopping is done outside of micro . Yes, you make it synch from micro. Would it be better to synch by discharging 555's capacitor directly through additional transistor, so 555's generation sequence will be in phase with micro orders? But i dont really get into the synch issue details yet.
And i suppose there would be great chance of chopper wont be ON second time by 555 if microstepping frequency will exceed that of 555's (set as stated in schematics to something around 19 kHz).
Yet may be you don't need variable resistors for current regulation symmetry - if comparators outputs will be available to micro and 555 can be disabled upon request you can set the VREF channel A first to some value from micro then enable chopper and count time when comparator will be triggered. Applying same sequence to channel B will give difference in time and that is channel current measurement mismatch which can be corrected when working with microsteps .
And may be it is possible to reduce power down current in software , by introducing 2dimensional lookup table for current reference output towards comparators . One index is mode (power down or normal) another is current values .
So, when mode will be changed upon execution of watchdog interrupt - timer can be updated with values for current step taken from power down mode index of lookup table .
Above is just IMHO.
BUt i am probably wrong regarding current regulation - you need to keep PWM fast to change reference voltage fast for 45 kHz microstepping, to get it established soon at step change . Then top value for TCNT needs to be low and that means no chance to correct current tables in time due to low resolution of fast PWM.
But that is stimulus to look towards more port available avr series to implement references via r-2r matrix. IMHO.
Read the complete thread, then we talk.
At about 200 to 250 Full-Steps/sec the reference waveform changes from sine-cosine into the intermediate waveform we call Maximum-torque and then at over 300 Full-Steps/sec it changes to a DC level (maximum current or Full-Step reference). So, at 45 KHz there is no variable reference waveform even when we are using 1/16 micro-stepping.
What it means is; although we are counting every single step pulse at the input as, lets say, 1/16 micro-step; The output reference waveform is a Full Step waveform. When decelerating the opposite occurs. No steps are lost or added because we continue counting 64 micro-steps per full step on the coil sequence switching (in case of 1/16 micro-stepping)
So, in practice, at the speed levels when the time constant of the motor coil will limit the coil current from following the reference waveform, we change the reference waveform, at those speeds micro-stepping is not important, it is already done with its movement smoothing function..
Micro-stepping is intended for smooth movement at low speeds only.
Another thing, The 555 and related components for the PWM oscillator are not being used with the ATTiny2313 translator. Read the note on top of the translator's schematics. What we use here is fixed frequency chopping. The oscillator is a 40 Khz PWM generator (using timer0), the output compare value is set to the required value to give the desired "blanking time" for the comparators (1 uSec).
About the assembler implementation;
As the code is now (compiled basic), the interrupt routine for Int0 takes 11.40 uSec (worst case). There are no other tasks working on that controller, so the maximum theoretical step-rate is supposed to be over 75 Khz. I haven't tested it yet. Nevertheless ,as the important code is on that interrupt routine, what is left is only initialization code that is also easy to program in assembly. The minimum tested step pulse width is 1 uSec.
I tried that too, but then, it will be necessary to restore the right current level at the beginning of the Step interrupt routine before doing the change to the corresponding new levels, and those instructions took a few microseconds to run. It was easier and faster to change one output pin value and let the user decide what current reduction he/she wants (if any) by replacing only a couple of resistors and/or a jumper.And may be it is possible to reduce power down current in software
Another issue related to this reduction was the fact that I wanted the user to select the feature by means of a jumper, and I had only one pin left.
Last edited by kreutz; 01-18-2007 at 10:39 PM.