On Saturday 16 August 2014 14:33:56 Philipp Burch did opine
And Gene did reply:
> Hi Gene!
> 
> Replying on a tiny note you made:
> 
> On 16.08.2014 16:55, Gene Heskett wrote:
> > ...
> > This causes the Mesa SpinX1 controlled fwd/reverse relays to switch,
> > throwing the full power of the controller into the motor trying to
> > reverse it.  The controller, a $185 Gemini, survived only 3 such
> > events before it punched the IRFP250N hexfet, a 200 volt 30 amp
> > device.  So that is an error condition that I'll use to defeat the 2
> > input and's setup in lut5's, releasing the 3rd relay, which if
> > energized, shorts out a 50 ohm 400 watt resistor in series with the
> > motor.  And I'll do it based on the man page comment about leaving
> > unused inputs open by feeding this true if dir is missmatched from
> > an abs.piddir module into in-2 of both the forward and reverse path
> > logic, which if the man page comment is true should cause a false
> > output.
> > ...
> 
> Actually, it isn't a special feature of this component that an unused
> input going true sets the output to false. This is just depending on
> your function and assumes that you insert zeros in the unused fields of
> your truth table.
> If you take my other examples for inverter and nand2, then you will see
> this behaviour only in the former:
> 
> setp inv-1.function 0x00000001    # Single inverter
> setp nand-1.function 0xfffffff7   # Two-input NAND
> 
> For the inverter, if all inputs are zero, then you have state 0 for the
> truth table, so you get out = 1 (as bit 0 in the function is 1). In any
> other state, no matter if it is because in-0 is one or any other input,
> you will hit one of the zeros in the function word and therefore get a
> zero output.
> But the NAND is different: When calculating the function parameter
> value, I first thought how an AND with two inputs has to look. Well,
> easy enough, the output should be one if in-0 and in-1 are one and zero
> in all other cases. Thats state 3, so bit 3 must be 1 and the rest
> zero. Makes 0x00000008. A NAND is simply the inverted version of the
> AND, so all I did then was to invert the whole function value, giving
> 0xfffffff7. But this will obviously generate a one also if any of the
> other inputs is one. You only get the advertised behaviour with a
> function value of 0x00000007. If the NAND2 should function without
> considering any of the other inputs, the word would be 0x77777777. I
> hope that bit patterns are correct, have just put them together in my
> head.
> 
> So when checking if your code is actually fail-safe, you should
> consider all truth tables completely and make sure that the output
> bits for your unused states really generate a safe condition if the
> inputs are bogus.
> 
> Oh, just as a side note: The source of the lut5 component (in
> src/hal/components/lut5.comp) is worth a look. That's 12 lines of
> comp-related code, 126 lines of description and only 10 lines for the
> actual functionality :)

You obviously understand that puppy way better than I do.  Nevertheless, 
when I managed to get it working like it should, and did the final hook up 
from the controller to the motor, only to find that the controller is 
still toasted.  So, I guess I buy yet another like those 2 and bow to the 
east.  Unforch, I have not found a 4 quadrant with enough amps, so I guess 
I am still stuck synthesizing it in hal, but without the spindle 
positioning the 4 quadrant could do.

I'll call Monday and get another headed my way, and make arrangements to 
get these two failures repaired.  It looks like I may need to keep spares 
handy.  I think I'll also order a 5 pack of that much higher voltage 
hexfet & change it out before the puny one blows once the spares get back 
repaired.

> 
> Cheers,
> Philipp


Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page <http://geneslinuxbox.net:6309/gene>
US V Castleman, SCOTUS, Mar 2014 is grounds for Impeaching SCOTUS

------------------------------------------------------------------------------
_______________________________________________
Emc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-users

Reply via email to