04298 [alt-beam] BEAMant - description of circuit operation

Alt-BEAM Archive

Message #04298

To: "Mark W. Tilden" mwtilden@math.uwaterloo.ca,
From: Sean Rigter rigter@cafe.net
Date: Sat, 05 Jun 1999 17:25:31 -0700
Subject: [alt-beam] BEAMant - description of circuit operation

BEAMant - Description of circuit operation
wilf rigter 05/06/99 BEAMant (c) M.Tilden

Overview of circuit operation and probable behaviours
-----------------------------------------------------

The core of the BEAMant consists of 2 coupled master Bicore oscillators.
The photo Bicore will oscillate over a wide range of frequency and duty
cycle and the motor Bicore has a relatively fixed frequency with a duty
cycle influenced to a lesser or greater extent by the state of the photo
Bicore. Under some conditions the photo Bicore will have NO influence on
the motor Bicore. which is the reason for describing the BEAMant core as
coupled master Bicores rather then master slave Bicores. (I am not sure
if the term "embedded Bicores" applies here).

The output of the photo Bicore is acting on the bias points of the motor
Bicore and changes the duty cycle of the motor Bicore oscillator but not
it's frequency. This coupled Bicore circuit (Unicore?) overcomes a
serious limitation of the conventional photo Bicore in which both duty
cycle and frequency are affected by light. There are three categories
of circuit behaviour which depend on the light level, and component
selection.

1. In the dark the photo Bicore oscillates many times slower than the
motor Bicore.
2. In medium light the two bicores oscillate with a small ratio between
the two frequencies (ie 2:1 to 1:2)
3. In bright light the photo Bicore frequency is much higher than the
motor Bicore frequency.

In the dark, with the photo Bicore at the lowest frequency, the ant will
turn in clockwise and counter clockwise circles as the motor duty cycle
is alternately changes. then as the light brightens it will alternately
steer left and right.

As the light increases to medium some complex and even chaotic behaviour
can occur as the coupled oscillators do a little dance at low harmonic
ratios. This includes jittery phase locking behaviour with a chaotic
quality . So in the medium light the ant may behave in complex, strange
and unpredictable ways.

In brighter light the photo Bicore frequency rises and with higher
frequency ratios, we see the ant moving in a straight line with moderate
photo tropic behaviour. The high frequency photo Bicore output will be
integrated (smoothed) by the motor Bicore inputs and the duty cycle
information of photo Bicore will be extracted in the form of
differential dc voltages which influence the bias points of the motor
Bicore. This in turn changes the duty cycle but not the frequency of the
motor Bicore.

In order to add a threshold to the point where the BEAMant changes from
travelling in a straight line to turning behaviour, one shots or
monostables were added to the motor Bicore outputs. This is sometimes
called a "dead band" in which motor Bicore duty cycles close to 50% (ie
45%-55%) are ignored. The oneshots are really Nv stages with an "ON"
time which is shorter than the shortest motor Bicore on time which you
want to ignore. In that case the motor will rotate at equal speed if the
motor Bicore duty cycle is near 50% This occurs because the motor
bicore outputs duration's will be slightly longer than the Nv outputs
and small difference will not affect the motor duty cycle. When the
motor Bicore duty cycle is changed to the point where one motor Bicore
output has a shorter duration than the Nv TC then this shorter pulse
will be transmitted unaltered to the output of the corresponding Nv
while the bicore output with the longer duration continues to be
limited by the Nv time constant. The result is that the on time of one
motor output is shorter and the BEAMant will turn into that direction.

This turning is caused by several sources including low light
"searching" where the slow cycle of the photocore is directly expressed
as alternating left/right turns and at very low light clockwise and
counter clockwise turning. At Medium light levels the turning is more
photo tropic but may exhibit unusual waggles and detours ( phase
slipping oscillators), In bright light the behaviour should be quite
predictably photo tropic. The speed of the BEAMant should be constant
except during turns. The so-called XOR Nu circuits will change this
behaviour to turning or reversing or photo phobic in case of a side or
head on collision and depending on the duty cycle of the photo Bicore.

Bicore details of operation
---------------------------

The photo bicore (Pcore) is a dual photo to pulse width generator and
the motor bicore (Mcore) is a differential voltage controlled duty cycle
generator. More precisely the Pcore outputs are complementary signals
whose frequency depends on R/C and the sum of the photo currents AND
whose duty cycle depends on the difference in photo current. The
complementary Pcore outputs are connected to the Mcore with two equal
resistors resulting in coupling currents which are dependent on the
voltage difference of Pcore outputs and the Mcore biaspoint inputs. The
Mcore outputs are complementary signals whose frequency depends on R/C
and on the sum of the Pcore coupling current AND whose duty cycle
depends on the difference of Pcore coupling currents.

The resulting influence of the Pcore on the Mcore output duty cycle is
delightfully complex depending on the frequency ratio of the two coupled
oscillators and the duty cycle of the Pcore.

For the purpose of the BEAMant, frequency ratios are dependent on light
conditions as follows:

1.In low light, the Pcore frequency is much lower than the Mcore and
the ratio is high.
2.In medium light, the frequencies are similar and the frequency ratio
is low
3.In bright light, the Pcore frequency is much higher than the Mcore and
the ratio is high.

The range of Mcore duty cycle depends on the ratio of the Pcore coupling
resistors and the Mcore timing resistor.

This gives rise to 3 distinctively different modes of operation:

1. In low light the Mcore duty cycle depends on the DC state of the
Pcore outputs. ie once every few seconds the Pcore changes state and the
duty cycle of the Mcore changes state correspondingly e.g. 25% to 75%.to
25% to 75% etc

2. In medium light the Mcore duty cycle strongly depends on the phase
difference of the Pcore and Mcore signals. At very low frequency
ratios, the Mcore duty cycle will be modulated by the difference
frequency gradually shifting between 25% and 75 % and back to 25% etc.
If the 2 frequencies are nearly equal the Mcore may phase lock to the
Pcore with the Mcore duty cycle dependent on the leading or lagging
phase difference.

3. In bright light, when the Pcore frequency is high, the Mcore duty
cycle will depend only on the Pcore duty cycle ie the difference in
photo currents.

Oneshot Nv Stage/PWM motor drivers - detail of operation

The Nv motor driver output stages are "oneshots" which is what Tilden
calls them on his neural network drawing. These are not conventional
oneshots however since their time constant is not independent of the
input signal. For the purpose of this discussion I will call these
"resettable oneshots" simply Nv. For a given rising edge on the input,
the Nv output goes low for a period determined by RC but this period is
shortened if a falling edge occurs on the input before the Nv times out
and the output is reset on that falling edge.

More than one behaviour is possible depending on various ratios of time
constants. The most obvious function of the Nv is as a duty cycle
threshold detector providing a "dead band" between driving in a straight
line and turning. Since the direction of a 2 motor platform is sensitive
to the motor bicore (Mcore) duty cycle, tuning would be required to make
it go in a straight direction under "neutral" conditions.

In practice, bicores are asymmetrical and coupled oscillators are not
particularly stable. Therefore uncontrolled parameters (ie threshold,
temperature, frequency) can shift duty cycle which would detune a
"straight direction" adjustment. This can be corrected by adding a duty
cycle dead band between the mcore and the motor large enough to ignore
small duty cycle variations.

This is done with a "resettable oneshot" Nv on the motor bicore
outputs with Nv R/C time constant (TC(Nv)) adjusted for a "full
pulse width" which is shorter than minimum Mcore output pulse width
that you want to ignore. Now any pulses from the Mcore that are
greater than the TC(Nv) will always be limited to the full pulse width
of the Nv but pulses that are shorter than TC(Nv) will reset the output
of the Nv on the falling edge and are therefore equal to the Mcore pulse
width.

Other ratios between the time constants of the motor bicore and
the Nv oneshot exist which give rise to a different function but I
believe that this description was the reason for including a Nv in
the motor Bicore outputs.

XOR Nu sensor/direction motor drivers - details of operation

The term XOR, as it is applied here, was derived from the boolean
eXcluxive OR logic. In this case,it means a motor will turn only if one
terminal is high and the other low. One side of each motor is driven by
the one shot Nv motor drivers and the other side of each motor is
connected to the output a Nu motor driver.

Together each Nv and Nu driver form a bridge with the Nv side providing
fixed frequency/PWM pulses and the Nu side providing a negative or
positive supply reference for motor reversing. When the motor reference
supply is high it causes the motors to rotate in the forward direction
and when Nu output is low, it causes the motor to rotate in the reverse
direction.

Each Nu stage input is connected to a tactile sensor which detects
collisions on opposite side of the motor to which it is connected as
well as full frontal collisions. The function of the Nu stages is to
stretch a tactile input (determined by TC(Nu)) and reverse the motor
supply reference voltage for the duration of the Nu pulse width

When a collision occurs on one side the Nu, the supply reference for
the motor on the other side is reversed and for the duration of the
TC(Nu) that motor will now rotate in the reverse direction with each
active Nv PWM pulse width determined by the normally inactive "off
time" for that motor.

Assuming high light conditions, the BEAMant behaviour is reasonably
predictable, and side collisions cause the BEAMant to pivot or spiral
on it's axis. Unless the second tactile sensor is activated it will
continue to spiral until the Nu times out. The photo effect for the time
that the motor is reversed becomes photo phobic for the reversing motor
and photo tropic for the forward motor.

If both tactile sensors are activated the BEAMant will exhibit it's
normal waggling behaviour in reverse at a faster than normal speed and
the overall photo effect will be photo phobic. Under medium and low
light conditions behaviour is more complex and (for me) difficult to
describe in detail (you'll just have to build a BEAMant and tell
me).

Well this is my best understanding of the BEAMant circuit operation and
predicted BEAMant behaviour. Has anyone seen these critters in action? I
look forward to hear examples of the real BEAMant behaviour both Mark
Tildens's and of course your own experiments with this design.

I think there will be immediate applications for the coupled master
Bicore (Unicore?) for walkers and heads.

Since the frequency of the motor bicore is more or less fixed, it solves
several design obstacles related to the conventional walker and head
photo Bicores including the zero power problem for a Bicore Head.

enjoy

Wilf Rigter mailto:wilf.rigter@powertech.bc.ca

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

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

Home