Alt-BEAM Archive

Message #04402



To: "'beam@sgiblab.sgi.com'" beam@sgiblab.sgi.com
From: Wilf Rigter Wilf.Rigter@powertech.bc.ca
Date: Sun, 13 Jun 1999 12:11:55 -0700
Subject: [alt-beam] Re: bypassing a nv (Correction)


I have had "bad neuron" days like that Bruce so don't bother crawling under
a rock.

first a small clarification:

And furthermore:

> A PNC works by keeping the neuron input clamped to ground ...

should have been, "A PNC works by keeping the neuron INVERTER
clamped to
ground."


I prefer one of the following:

A PNC works by forcing one neuron inverter input high OR a PNC works by
forcing one neuron process ON for a time longer than the propagation time of
all processes circulating around the microcore loop OR a PNC works by
keeping one neuron process active until it is the only process left.

At the risk of confusing the heck out of everyone it is worthwhile to take a
closer look at the microcore engine.

Like all BEAM ideas Nv and Nu neurons seem like the simplest possible
circuits but what goes on "under the hood" is actually far more complex that
an equivalent linear circuit and deserves the term "process" to describe it.


The term Nv is often used in the context of a microcore but there are many
other applications for an Nv. For the case of a microcore consider an Nv we
can call Nv1. It consists of an inverter with a resistor connected from the
input to ground and a capacitor connected from the input of the inverter to
the output of another Nv stage (Nv0). The capacitor is the coupling element
between Nv stages. Since capacitors block DC voltages and can only transmit
a changing voltage the Nv is called a "nervous neuron". A falling voltage
(high to low) on the output of Nv0 causes the Nv1 capacitor to be reset and
a rising voltage of the output of Nv0 will trigger a process in Nv1.

The low to high edge of a process output voltage of Nv0 signals the end of
the Nv0 process coupled to the input of Nv1 and the Nv1 output will fall
signaling the beginning of the Nv1 process coupling a reset signal into the
capacitor of the next Nv stage (Nv2). At the start of the process the
voltage at the input of Nv1 inverter (bias point) is initially high but the
capacitor starts to charge up through the resistor and the voltage at the
Nv1 bias point falls until it drops below the switching threshold of the
input and the output of the Nv1 inverter goes high. This positive going
lagging edge of the Nv1 "process" output waveform signals the end of the Nv1
process and the beginning of the Nv2 process. This continues from one Nv
stage to the next around the microcore loop.

Note that if only one process is present in a microcore, the reset signal at
the start of each process has no effect on the next Nv stage process since
that stage has no process active. The reset signal does perform it's
secondary function to discharge the "process" charge on the coupling
capacitor through the input diode of the next Nv stage which is buried
inside the inverter input. If the microcore is "saturated" with 2 processes
active at a time then the high to low signal at the output of an Nv stage
at the beginning of the process of that Nv does indeed reset any process
still active in the next Nv stage.

Note that only the EDGES of the voltage waveforms on the Nv outputs affect
the next stage. It is the LEVEL and DURATION of the Nv output that affects
the motor rotation and on time.

Note the HIGH to LOW edge on a Nv output is the START of the process for
that Nv but is a RESET signal for the next Nv stage and the LOW to HIGH edge
at the Nv output is the END of the process of that Nv but the TRIGGER for
the next Nv stage.

The general case for a Nv neuron is a gain stage (inverting or
non-inverting) with an RC network at the input with the signal applied to
the capacitor and the resistor connected to 0V,+V or a low impedance voltage
reference above or below the linear input range of the gain stage (ie an
output).

The general case for a Nu neuron is a gain stage (inverting or
non-inverting) with an RC network at the input with the signal applied to
the resistor and the capacitor referenced to 0V,+V or any other low
impedance point.

Finally it can be shown that concepts of Nv and Nu neurons are really quite
ambiguous and that they are really one and the same neuron but stimulated at
different points. For example, the PNC is activated by the edge of the
powersupply voltage changing. This starts a process in the PNC neuron which
ends when the voltage a the bias point of the neuron crosses the trigger
threshold. If the powersupply voltage increases slowly, for example in the
case of a solar powered microcore, the PNC process would never be triggered.
Another example of the fuzzy distinction of a Nv and Nu neuron function is
the slave Bicore which clearly is a phase delay Nu with the series coupling
resistors and capacitors referenced in this case not to ground but to the
slave Bicore outputs.

Take it from me, the key to solving BEAM circuit problems or designing new
BEAM circuits with Nu and Nv neurons is learning to understand what goes on
under the hood.

enjoy

Wilf Rigter mailto:wilf.rigter@powertech.bc.ca
tel: (604)590-7493
fax: (604)590-3411

> -----Original Message-----
> From: Bruce Robinson [SMTP:Bruce_Robinson@bc.sympatico.ca]
> Sent: Sunday, June 13, 1999 9:59 AM
> To: beam@sgiblab.sgi.com; davidperry@mail.geocities.com; Bob Shannon
> Subject: Re: bypassing a nv (Correction)
>
> Bob Shannon wrote:
>
> > I thought a standard Nv has a capacitor at the input, (with the
> resistor
> > to ground) while Nu's have the resistor at the input (with cap to
> > ground)?
>
> You're right, Bob. I was extremely sloppy in my terminology. Apologies
> to all. I MEANT to say:
>
> "Assuming you have a typical Nv with a resistor going to ground at the
> input TO THE INVERTER,"
>
> To compound the error, I carried on in the same vein in a subsequent
> note:
>
> > In a typical 4 Nv microcore, what causes a neuron to fire is when the
> > input to the previous neuron switches from high to low ...
>
> should have been, "In a typical 4 Nv microcore, what causes a neuron to
> fire is when the input to the previous INVERTER switches from high to
> low."
>
> And furthermore:
>
> > A PNC works by keeping the neuron input clamped to ground ...
>
> should have been, "A PNC works by keeping the neuron INVERTER clamped to
> ground."
>
> Again, sorry to everyone I confused. I think I'll crawl into a hole
> somewhere for the next couple of days.
>
> Regards,
> Bruce

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

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



Home