Alt-BEAM Archive
Message #09107
To: beam@sgiblab.sgi.com
From: Max Inggal tacoman101@yahoo.com
Date: Wed, 12 Jan 2000 14:47:06 -0800 (PST)
Subject: [alt-beam] pancake photovores
I was just thinking could you use one of those cooling
fans from a computer. I took one apart once. All
they are are a coil over a set of magnets. They look
like they'ed be nice. But they take so much voltage
to run. What you guys think?
__________________________________________________
Do You Yahoo!?
Talk to your friends online with Yahoo! Messenger.
http://im.yahoo.com
9108 Wed, 12 Jan 2000 16:57:33 -0600 [alt-beam] Re: BEAM VIDEO was Solar Powered Camera's beam@sgiblab.sgi.com Richard Piotter I'd love to get my hands on single or dual unit schmitt triggered
inverters in surface mount packages. Imagine having a sheet of tiny
boards manufactured that hold a socket for a capacitor, a schmitt
triggered inverter, and a socket that you can plug a resistor into. On
the bottom would be 6 pins. Power +, Power -, Input, Output, RC Bias,
and R source. Power is obvious. So is input and output. The RC Bias and
R source would be the point where the capacitor, resistor and Inverter
input tie together, and the R source is simply the end of the resistor
that normaly attaches to ground or Vcc. Those points allow you to bias
the resistance more or less to change time constants. Switch the
resistor and capacitor for an Nu neuron. A little 6 pin DIP sized
package with a plug on top for a resistor & capacitor would be the ideal
expirimenter's Nv/Nu. I'd make some if I could, but money is the
limiting factor for developing such a fun little toy! :)
For now I'll have to stick with my larger 6 Nv boards wired together. I
had an interesting idea today and I'm going to work on it, but I need to
study how motion can be picked up. The neurons must filter when the
whole screen moves and only watch for localized changes. It also has to
be tracking edges, since shading can differ depending on light and each
object. If differential edge detection were done, you could have a 9x9
or 9x8 grid (depending on the method of differentiation) on an 8x8 grid
of LEDs (and ultimately more neurons for analizing stuff). Actualy, for
most level walking bots, most of the image is going to be horizontal
anyway. I'd say give most of the sensing ability to horizontal
resolution. A 5x7 grid LED character display on it's side very closely
matches the 3x4 proportions of traditional NTSC broadcasts. 4 of them
would be a nice 10x17, or three vertical ones side by side is 15x7, near
the 16x9 widescreen format. Any way you do it though, If the someone
comes up witht he circuitry to generate edge detection, and then
generate motion detection... Well, we could have a realy useful BEAM
based toy to play with!!! :)
I'll have to get some books and study up on the subject.
Later!
George Rix wrote:
>=20
> > Theoreticaly, couldn't you could set up your Nvs and Nus in such a wa=
y
> > that motion across the cameras vield of view would stimulate particul=
ar
> > processes. the more uniform motion, the more it builds up, possibly
> > forming a rudimentary motion follower? It'd require a lot of neurons,
> > and maybe it just wouldn't work well or be feasible, but you have to
> > admit, it's a rudimentary start. BEAM isn't going to be doing any maj=
or
> > "processing" until we up the neuron count anyway! 4 Nvs are nice for
> > walking gaits, but not for following motion. There are limits to smal=
l
> > loops. Wee need more, and we need real networks. Maybe it is a lot
> > harder, maybe it's more dificult to tune, but if it is made to work, =
the
> > results are very rewarding! I may have to make an Nv expirimenters
> > board. Problem is how annoying breadboarded Nv nets can often be.
> Well, a good solution to that would be to buy a hundred or so of those
> little single inverters (but the question IS: do they have schmitt
> triggers?), and then attach your cap and resistor, and then arrange the=
m
> like that.
> I've been drawing up designs for bots that are just neural and nervous =
nets
> connected to a grid of LEDs (but it's not like Wilf's stuff, this was d=
one
> more to see what sort of patterns would develop with larger nervous net=
s),
> but I never did anything about them because of the difficulty of gettin=
g
> parts here. For instance, I ordered a hundred or so 0.22 uF caps over t=
he
> phone, and received that many 0.022 uF caps, which are now not very use=
ful
> unless I up the resistor values tenfold, and I don't have many 20 Meg o=
hm
> resistors hanging around.
> Anyhow, one of my paper designs was just to have four neurons arranged =
in a
> tetrahedron.
> Compared to a loop of four neurons, I'd expect that this thing would be
> quite powerful. However, I'm fresh out of caps...
> And then, I got to thinking about having thirteen neurons arranged in a=
n
> icosahedron (the thirteenth one is in the centre), which would give you=
TONS
> of interconnections, and might prove EXTREMELY interesting.
> But again, no caps.
> Hmm.
> > I have a big interest in Neural Nets, and I was studying them way bac=
k
> > in high school. The real networks are ones that branch. They can do t=
he
> > decision and logic that simple loops can't realy touch. BEAM has
> > potential, and low res "BEAM TV" is one heck of a start. I have
> > extensive knowledge about computer, neural, and TV technology. It's f=
un
> > combining them!!! I'll have to play around with this little toy! 64 L=
EDs
> > though is sure plenty! I'll need to start digging them up! The timing
> > for the Nvs would have to be timed properly too. I'll have to figure
> > that out.
> Yeah, well, I have plenty of LEDs, but none are of the same value...
> > Remember, just cause it's not cost effective doesn't mean it shouldn'=
t
> > be done! Most new things aren't cost effective! They need time to
> > develop and be understood well enough to do something.
> Precisely. And guess what:
> I have a job now, so caps are coming... =3D)
>
> > I wonder if I can set up that old Neural Networks program on my Mac
> > emulate BEAM Nvs and Nus. At least imitate them. I wish I knew a good
> > Mac programer with an interest in BEAM. It'd be nice to have an BEAM
> > emulator for expirimentation. Heck, I'd even settle for a PeeCee
> > version! I could just emulate it! :)
> Hey, Richard, I'm working on my programming skills, so call me in three=
or
> four years =3D)
> > Hopefuly, we'll see BEAM bots finaly chasing our cats (or dogs, or
> > whatever you got) instead of the other way around! Hehe! :)
> Actually, my cats are scared to death when I turn on Radon (he's my onl=
y
> working BEAM bot: a non-solar photohead).
> But then, they run for it when my guinea pig Murray sneezes.
> Cats are hilarious =3D)
>=20
> BTW, does anyone know where those single inverters can be found?
>=20
> Dave: do you think Solarbotics might start stocking those... =3D)
>=20
> Well, there's my rant for today, so I'm back to lurking...
>=20
> Peace out!
>=20
> Rob Rix
>=20
> Don't take a walk, climb a tree =D0 an ancient truism, invented last we=
ek.
--=20
Richard Piotter The Richfiles Robotics & TI web page:
richfile@rconnect.com http://richfiles.calc.org
-- Make Money by Simply Surfing the Net or responding to E-Mail!!!
-- Click below!!!
http://www.alladvantage.com/go.asp?refid=3DATL147
http://www.spedia.net/cgi-bin/dir/tz.cgi?run=3Dshow_svc&fl=3D8&vid=3D3296=
30
http://www.utopiad.com/member/getPaid?refId=3D284400
http://www.itadsup.com/index.cgi?ref=3Drichfiles
http://www.desktophorizon.com/index.html?referrer=3Drichfiles
http://www.gotoworld.com/getpaid/default.asp?rid=3D1040486426
http://www.dotad.com/Default.asp?ID=3D474867
http://www.moneyformail.com/home.asp?ref=3Drichfiles
9109 Thursday, 13 January 2000 1:37 Photovore Performance? beam@sgiblab.sgi.com Dye, Roy
>What performance, in terms of time to charge and popping distance, could
>one expect from a typical solar powered photovore?
>
>I have just built my first photovore utilizing a 1381SE, 3766 solar
>cell, and two pager motors from Solarbotics. It takes about 20s on my
>workbench to charge and it moves about an inch per pop. In direct noon
>sunlight, it fires about every 2 seconds. Is this typical?
>
>Roy
>
>
Home