announcing SiMAD - a new algorithm to control N S&D channels with N+1 signals using existing S&D drivers

[Doug Fortune <pentam-at-home.com>]

I'm happy to announce that last fall I developed 
a new software algorithm (I don't think this has 
been described anywhere else), and I named it:

SiMAD - Simultaneous Multi-Axis Direction 

This software-only technique amazingly allows the 
control of N Step&Direction channels with only 
N+1 bits (ex: 6 axis hexapod control with 7 bits). 

Even more amazingly, the SiMAD technique allows 
the use of existing Step & Direction motor drivers. 

This is important for those of us who have a limited
number of output bits (such as in a parallel port which 
is commonly configured for only 12 output bits).  The 
bits freed up can be used to add more functionality 
(ie more relays to control coolant, vacuum, etc). 

Therefore, I am hoping some day EMC will encorporate 
this algorithm to make more efficient use of the 
parallel port. Huge multinational corporations are not 
exempt from paying licensing fees of course!  

Here is a brief review of the commonly used techniques:

#0 Phase Control 
   Just for steppers, this uses 4 bits per channel
   an is very limiting (3 channels max, or 2 channels 
   with 4 aux output bits). 

#1 ClockUP, ClockDOWN  
   This uses 2 bits per channel, and needs a CW/CCW 
   controller. 

#2 Step & Direction 
   This also uses two bits per channel and uses the 
   standard Step & Direction motor controller. 

and now SiMAD which uses 1 bit (called a Direction bit) 
which is distributed in parallel to all the S&D controllers, 
and is actually hooked up to their STEP input terminals
(giving the Multi-Axis-Direction moniker); 

and an additional 1 bit per channel (each output bit is 
connected to the S&D controllers DIRECTION input terminal).

The only downside of the technique is that simultaneous 
CW and CCW motor motion is not allowed.  But that is academic 
because CW and CCW commands can be given microseconds 
(or even nanoseconds) apart and thus essentially are moving 
simultaneously mechanically. 

Comparing 4 and 6 channel Step&Direction and SiMAD:
     S&D                  SiMAD    
4 channels            4 channels  
  uses 8 bits           uses 5 bits  
4 aux output bits     7 aux output bits  

            OR 

6 channels (hexapod)  6 channels (hexapod)  
  uses 12 bits          uses 7 bits  
0 aux output bits     5 aux output bits   



 -----------------------------------------------------
    technical stuff below just for experts...

Basically, there is only one DIRECTION bit which is hooked 
up to all the S&D drivers DIRECTION input (why didn't anyone 
think of this sooner?). 

DIR (bit0)    .. hooks to all DIR input drivers
X   (bit1)    .. hooks to X Step input driver 
Y   (bit2)    .. hooks to Y Step input driver
.     .
.     .

(just draw a picture on the left with the bits, on the right with 
N axes of S&D drivers.....) 

so here comes an X step CW (DIR high) 
       _____________________
DIR ...                     |____________________
       _______        ___________________________
X   ...       |______|
             step
            

         and here comes a Y step CCW (DIR low) 
       _____________________         ____________
Y   ...                     |_______|
                          step



 oh look, Z stepped CW with X 
       _______        ___________________________
Z   ...       |______|
             step



OK, the downside of the process is that, if you 
are stepping multiple axes at the same time, they 
have to be in the same direction.  If you want two 
axes to go different directions (say X CW and Y CCW) 
then you have to do them serially, but thats no big 
deal at all.  In fact before with two cards, you 
could never step all the axes together anyway (you'd 
do one pport, then the other pport). 

The upside is that we free up some output pins! 
Most of us use 12 out and 5 in.  That allows us to: 

 - 4 axes out (1+4=5) PLUS  7 bits out  PLUS 5 bits in. 

or, as most everyone uses 8 bits (for 4 axes) 

 - 7 axes out (1+7=8) PLUS  4 bits out  PLUS 5 bits in. 
 
or, for a hexapod (6 axes)

 - 6 axes out (1+6=7) PLUS 5 bits out PLUS 5 bits in. 

OK, now Gecko drives have a funny requirement that 
the signals are UP for a minimum of 4000ns and down 
for a minimum of 500 ns, and stepping happens on the 
falling edge of the step bit.

Here is a more detailed microscopic explanation: 

     ________________________ ______________________    
D:__|                        ^                      |________  
  
      1: prev state          2: processor           5: next   
       was set to CW            wakes up,           pulse (some 
       because we knew        keeps signal hi       time from now) 
        in advance                                  is going to be 
      (thus it satisfies                            CCW, so lower 
        setup time of 4uS )                         the signal in 
                                                    preparation 
    _________________________                      _______  
X:_|                         ^____________________|        
                                                   4: raise line    
     1: prev state for the    3: upon waking up jam
        axes was left at hi   the signal down which 
                             which causes a Gecko step 
                             and keep it down 500ns 
                             (ie go check the pport 
                              for changed in-bits).

                             4: when done that, 500ns 
                              must have elapsed, so 
                              raise the X line again...
                              
                             5: look ahead, the next pulse 
                              (say Z axis) is CCW, so lower 
                              the C signal in preparation 
                              (so it has a setup time of 4uS) 
 

This more detailed view addresses both the 500ns & 4uS requirements, 
and also the simultaneous CW and CCW steps (ie you can't, simultaneous
same-direction steps CAN happen, but opposite direction steps have 
to be done at a DIFFERENT time).  Thus we try to leave all the 
axis (X,Y,Z,A,B,C) signals high between interrupts (because, upon 
waking up, we just haul them down immediately and that makes a step, 
and then wait a mere 500ns delay, and then we can bring the signal 
back up, and then leave the interrupt). 


This seems so simple, doesn't it?  In fact, it did not 
come to me in a brainstorm, but by the math (this is not to 
scale, the clocks are too close, and X is low for explanation 
purposes, in the real world it should be left high):


     _   _   _   _   _   _   _   _   _   _   _   _   _   _   _
C: _| |_| |_| |_| |_| |_| |_| |_| |_| |_| |_| |_| |_| |_| |_| |_

                     _   _   _                 _   _
X: _________________| |_| |_| |_______________| |_|
|____________________
    do nothing       forward 3    nothing      back 2
                      because                 because
                   X is in sync             the signal is 
                  with the clock         inverse with the clock


Step = X*C + X*notC  = X*(C + notC) = X*1 = X 
Dir  = X*C + not(X * notC) = X*C + notX + C = C*(1+X) + notX = C+notX 

but we do nothing when (notX) so essentially Dir = C 

conclusion for 4 Axes (XYZA) and 1 clock channel C: it only takes 5 bits

X:  Step = X           see above diagram, what could be simpler! 
    Dir  = C 

Y:  Step = Y 
    Dir  = C

Z:  Step = Z
    Dir  = C 

A:  Step = A 
    Dir  = C 

Note that we can use standard "Step and Direction" motor drivers, 
except now the clock pulse stream goes to the motor driver "Direction" 
input, and the axis pulse goes to the "Step" driver input (exactly 
backwards from normal). 


regards
Doug Fortune 
Calgary Canada 
http://www.cncKITS.com