Alt-BEAM Archive

Message #04757



To: beam@sgiblab.sgi.com
From: Bruce Robinson Bruce_Robinson@bc.sympatico.ca
Date: Mon, 21 Jun 1999 19:09:11 -0700
Subject: [alt-beam] Re: Feedback explained?


Ben Hitchcock wrote:
>
> True, for CONSTANT battery voltages. But what happens if the walker is
> happily walking away, and in the middle of a step. Say the motor voltage
> is 5 volts. Also say that the normal unloaded time constant is, say, 2
> seconds. Also, to make things a bit simpler, let's say that the trigger
> voltage is 2.5 volts, and that the cap charges from zero.
>
> Now let's freezeframe at the point where it started a step one second
> ago: the cap voltage is a bit above (because the charge curve is
> exponential) 1.25 volts. Let's say about 1.4 for the sake of argument.
> So you still have 1 second to go before the inverter will trigger.

No argument so far, 1.4 is close enough.

> Now grab the leg so it can't move. The battery voltage will decrease due
> to the increased load, right? Let's say it decreases to 2.8 volts.
> (That's a bit exaggerated, but it serves to illustrate a point.)

I'll accept the exaggeration, for now.

> Now, what's half the supply voltage? 2.8 / 2 = 1.4 V.

Yep - the threshold is half the supply voltage, so it drops to 1.4 V.

> What does this mean for our little walker? It means that the inverter
> will switch STRAIGHT AWAY instead of waiting for another second.

Whoops, let's go back and clear something up. Why's the cap charging?
Because when we switched on this Nv, we applied the equivalent of supply
voltage -- 5 volts -- to the input of the Nv, which is the cap. The cap
had zero charge, so the inverter "sees" 5 - 0 = 5 volts. Now, 1 second
later, we've charged the cap up to 1.4 volts, so the inverter "sees" 5 -
1.4 = 3.6 volts. As you'll see in a minute, this doesn't change your
argument. I make just wanted to make it clear that we're looking at the
LOW-going threshold.

So now, you dropped the voltage to 2.8, the threshold falls to 1.4, the
cap INPUT also falls to 2.8 (we assume that whatever triggered this Nv
is also proportional to the supply voltage). And the voltage the
inverter "sees" is 2.8 - 1.4 = 1.4 volts, which as you say, causes the
inverter to switch.

> So the time constant is decreased.

So you've proven me wrong. Why? Because I assumed (not very wisely) that
the leg would get caught at the start of the cycle, thereby dragging
down all the related voltages before the cap got any charge.

But there is still something to be learned from this, and while I'll
concede there may be SOME feedback affect, I suspect it has been greatly
exaggerated.

> As far as I understand it, the time constant changes when the battery
> voltage changes, but not when the battery voltage is constant. So if you
> hit an obstacle at the end of a step, you won't be able to see any change
> at all. But if you hit one halfway through, then you will.

The effect of the feedback varies with the time into the cycle and the
amount of voltage drop. You picked the combination that puts the worst
light on things.

The earlier in it's cycle the leg gets caught, the less effect it has on
the Nv cycle time. The later it gets caught, the less significant the
feedback effect (because it's nearly finished it's cycle. So there is
some particular time when the affect on cycle time is greatest.

It turns out (using real numbers) that the less the voltage drops, the
later in the cycle this "greatest effect" occurs, AND the less
significant effect it has.

> That's my understanding, at least, and there are probably flaws in it -
> feel free to point them out.

Nope, I'll go along with you.

But the thing for us all to remember is that this feedback effect is not
particularly consistant. It depends on:

- where in the cycle the leg gets caught.
- how much the voltage in the Nv net gets pulled down (and this
depends
on how you construct your walker).
- how fully charged the battery is (low charge means greater voltage
pulldown).
- how "stuck" the leg is -- some movment vs. no movement.

If the feedback effect gave our walker a particular advantage by
occuring in the middle of a walking cycle, or when the battery charge
was low, I'd say, "Great, lets use it." Personally, I'd like to develop
some other, more reliable methods of modifying leg motion.

Thanks for putting me straight, Ben.

Bruce

------------------------------------------------------------------------

eGroups.com home: http://www.egroups.com/group/alt-beam
http://www.egroups.com - Simplifying group communications



Home