Alt-BEAM Archive

Message #07908



To: "'beam@sgiblab.sgi.com'" beam@sgiblab.sgi.com
From: Wilf Rigter Wilf.Rigter@powertech.bc.ca
Date: Mon, 22 Nov 1999 13:09:43 -0800
Subject: [alt-beam] Re: Cholroplast thingy




Hi Craig,

The "small" caps in my schematic also have a long time constant because I
use very low current green LEDs for photovoltaic cells. The small caps are a
lot cheaper than 1uf monolithic caps and much lower leakage current than
1uf electrolytics. In your simple "feeler/tank" turning arrangement, the
turning time constant depends on the light level which may actually be
desirable since it tracks the solar cell charging rate. This only works if
you use small solar main storage cap ie 4700uf, where each SE trigger
results in a relatively small forward motion but which is repeated
frequently. If you use large super caps, the bot could turn a few circles
before the LED cap charges up again. (g)

It is interesting to observe how placing the caps in parallel or in the
positive feedback path creates different behaviour. Putting the caps in the
feedback path is somewhat desirable since it forces clean switching of the
motor driver. I noticed that with the large parallel caps there can be a lag
in response and when light is equal on both PDs, both drivers come on. With
the smaller feedback caps, the bot is always waggling, and the motors are
pulsewidth modulated when the light is nearly equal. Attached is yet
another version of the PhotoChlorSE circuit which works very well, using the
venerable 2N3904s, good for 100ma motors. The total voltage drop on the
2N3904 transistor is 120mV instead of the 2V drop of two series connected
MPSA12 darlington transistors which should give the motors a little more
Zap. Note the 0.6V clamped voltage feedback from the base of the transistors
was needed to avoid feedback from the sagging voltage on the opamp output as
the main cap discharges. I have, in the meantime, added a simple
feeler/turning circuit with a time out independent of light level which I
will show you next time.

<>

regards

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

> -----Original Message-----
> From: Maynard [SMTP:cybug@home.net]
> Sent: Saturday, November 20, 1999 5:15 PM
> To: beam@sgiblab.sgi.com
> Subject: Re: Cholroplast thingy
>
> Wilf:
>
> Actually, another cool idea that the schematic _doesn't_ show on my page
> is
> that I place a capacitor in parallel with each LED to slow responses to
> transients ( about 1uF ) and the feelers on the front of the robot
> actually
> short out the cap/led tank.
>
> This means that when the feeler grounds out the left led, the right led is
> by default ) larger, and the robot bears right. When the feeler leaves
> it's detected object and releases it's hold on the cap/led , the capacitor
> makes that led voltage charge up VERY slowly, giving the robot a 10 second
> memory of that sides object.
>
> This makes the robot do a wide swing around the object, making excellent
> avoidance.
>
> Simple, and very effective!
>
> Craig
>
> > http://members.home.net/cybug/solarfly.htm
>
> ----- Original Message -----
> From: Wilf Rigter
> To:
> Sent: Saturday, November 20, 1999 1:53 PM
> Subject: RE: Cholroplast thingy
>
>
> > Hey Craig,
> >
> > Yep, you have to use low leakage fet input opamps, especially if you use
> > LEDs for photocells together with small low leakage caps. I would change
> the
> > circuit slightly as shown in the attached to improve efficiency (kind of
> > important with solar engines). With 5V(min) you can easily drive the
> mosfet
> > gates. Not yet tested but uses fewer components, lower forward voltage.
> Note
> > the 0.001 caps are used with the clear LED type (low sensitivity)
> > photodiodes for filtering and snappy positive feedback. Exact cap values
> to
> > be determined empirically.
> >
> > regards
> >
>
>



Attachment: pcse1.gif

Home