Alt-BEAM Archive

Message #05295



To: beam@sgiblab.sgi.com
From: Sean Rigter rigter@cafe.net
Date: Wed, 14 Jul 1999 21:30:05 -0700
Subject: [alt-beam] Re: automatic pnc


Hello all,

Hey, I tried it and that 2 diode PNC works great! The ucore never saturates and
starts instantly with a single process every time. How's that for evolution! The
Pulse Injection Circuit also does the job. To see how the PIC works, short one
bias point to 0V on a normal ucore and you kill all processes but with the PIC
installed, the ucore simply restarts when you remove the short. The LED AND
gate is a little marginal using my stock red LEDs which have a Vf of 1.70V and
that is a little close to the lower threshold of the HC14 working at +5V but is
just fine at 6V. My green LEDs (1.85V) don't work at all and I'm scanning the
databooks for low Vf (<1.60V) LEDs (anyone?) . Don't forget this idea of using
a single current limiting resistor for ucore indication LEDs works for any
other application not just the PIC and is a quick way to save some components.
Alternatively you don't have to use LEDs for the PIC AND gate, if you want to
keep the current drain down for SE apps toss the LEDS and use four 1N4148 diodes
with a 100K resistor. That way you get about 400mV drop and little effect on
the Nv time constant. The PIC may seem to be of limited interest (ie used
only if you have an application which suffers from process death) but process
injection as a concept may have far wider applications as we may discover in the
fullness of time.. PIC? any other name? anyone?

enjoy

Wilf Rigter wrote:

> Hello David,
>
> I feel a little guilty knowing that Bruce has invested significant effort in
> his cold start circuit design but it all part of the learning game so here
> is a some friendly competition:
>
> Although not tested, I think what we have here is an instant auto Pulse
> Neutralizing Circuit (PNC) which will force a single process at power up and
> any other time at the cost of two 1N4148 diodes. It makes the assumption
> that in a 4 Nv ucore you can have zero, one or two processes active at one
> time. Since the PNC guards against the 2 process condition and since these
> only occur as process pairs in 1 and 3 or 2 and 4, then adding the diodes
> will neutralize the twin processes in 2 or 4 whenever a process is active in
> 1 or 3.
>
> If you wish to guard against the more unusual zero process condition, you
> can combine a Pulse Injection Circuit (PIC ( yikes! no flame wars please!))
> and the function of 4 red LEDs for indication. This is done with the
> addition of a capacitor and 1K resistor to inject a process into the ucore
> should all processes for some reason die out (unusual). The PIC is actually
> a 4 input AND gate made from the 4 LEDs together with the 1K pull up
> resistor. When all processes die off all four Nv output will go high and the
> output of the LED AND gate will also go high pulling the 1M resistors of the
> Nv input high. depending on the value of each Nv time constant one Nv input
> will cross the positive trigger level and the corresponding output will go
> low effectively inject a process into the core. Simultaneously the AND gate
> output drops since there are no longer four high input and the process
> starts to circulate around the core with each LED lighting to indicate the
> Nv which is currently active.
>
> The additional 0.1 capacitor is there to deglitch the circuit Nv but is
> probably not be necessary. Using a single 1K resistor instead of the normal
> 4 resistors for LED indication effectively cancels out the 2 additional
> diodes and the capacitor and so it's fair to say that the PNC and PIC
> functions comes for free. In addition, the time constants of the Nvs are
> increased by a factor of 2 or more, since the 1M Nv resistors are referenced
> to the forward voltage drops of the LEDs (should be about +1.2V) making it
> practical to use smaller resistors or capacitors. The obvious applications
> of voltage controlled Nv time constants was covered earlier in my ETC and
> Servocore articles. The only restrictions for PNC/PIC is a minimum 5V
> powersupply, a 74HC14 (not 74HCT14) chip and relatively low forward voltage
> drop red LEDs. Be sure to terminate the unused inputs of the spare
> inverters to 0V if not used for some other functions (like a 6Nv core).
>
> This circuit comes quite close to my BEAM design target of "a linear
> INCREASE in functionality with a linear REDUCTION in complexity" ie the
> more I throw out, the more it does, which of course I hope is scalable to
> "zero complexity=infinite functionality" (reductio ad absurdum =) .
>
> enjoy
>
> Wilf Rigter mailto:wilf.rigter@powertech.bc.ca
> tel: (604)590-7493
> fax: (604)590-3411
>
> <>
>
> > -----Original Message-----
> > From: David Perry [SMTP:davidperry@geocities.com]
> > Sent: Tuesday, July 13, 1999 10:54 AM
> > To: beam mailing list
> > Subject: Re: automatic pnc
> >
> > great!
> > Thanks heaps, but i have a few questions.
> > Where do the diode outputs connect to, the input of the microcore
> > inverters?
> > What's that 3rd inverter along, is it the same as the others? Do i just
> > use
> > another 74hc14 for all of them?
> > Great, this will really help, thanks again!!!
> >
> > David Perry
> >
> >
> > Oh and Ben Morgan, when you sent that message EVERYONE GETS IT thats
> > hundreds of people that get an even more cluttered up mailbox.
> >
> > -----Original Message-----
> > From: Bruce Robinson
> > To: David Perry
> > Date: Wednesday, 14 July 1999 11:40
> > Subject: Re: automatic pnc
> >
> >
> > >> David Perry wrote:
> > >>
> > >> But i still can't work out how to make a circuit to automatically
> > >> introduce pulses for one of the following ways:-
> > >> *detect with logic circuits when the microcore has died out and
> > >> restart it OR
> > >> *just introduce a pulse when the circuit starts which i don't really
> > >> want beacuse i don't know why the circuits died
> > >
> > >Hi, David. Here's my cold-start circuit which I posted a while back. It
> > >starts your microcore up with one pulse, period. No waiting while extra
> > >pulses die out, or any of that stuff.
> > >
> > >This one is 100% guaranteed. I have been experimenting with another
> > >circuit which does the same thing, but only uses two inverters. It's
> > >worked perfectly for me so far, just have to get it into GIF format.
> > >
> > >Finally, I think I can adapt this circuit to figure out if your
> > >microcore has gone dead. Essentially, there would be one neuron with a
> > >very long delay, that would be triggered by one of the neurons in your
> > >microcore. As long as a pulse kept coming around to trigger it, the
> > >delay neuron would keep getting reset. However, if all your pulses died
> > >out, the delay neuron would eventually time out and send out a signal,
> > >that could be used to start up your microcore again. This one's going to
> > >take some thought, because I don't know how it will behave on startup.
> > >
> > >More to follow.
> > >
> > >Regards,
> > >Bruce
>
> ------------------------------------------------------------------------
> Name: pncpic.gif
> pncpic.gif Type: GIF Image (image/gif)
> Encoding: base64


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

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



Home