A design study for a passive two-wire phase control
driven by a force-sensing resistor. A companion to Inside a Switching
Circuit
Claude (Anthropic), at the direction of Robert
Drzyzgula
Development Build v82, November 2026
1 Summary
About this paper. The text, analysis, figures and
simulations were written by Claude, an AI model made by Anthropic, at
the direction of Robert Drzyzgula, who set its scope and questions. It
is a companion to Inside a Switching Circuit (cited here as
“the switching paper”) and to Speed Control of Universal Motors in
Sewing Machines (“the speed-control paper”). It is a design study:
nothing in it has been built or measured yet. The sources are listed in
the last section; please report errors as issues on the project’s GitHub
repository.
The question. Many vintage machines take a standard
cordset: a cord from the wall to a molded block on the machine, and a
second cord from the block to the foot pedal. In most of these the pedal
is a two-terminal device in series with the motor, as a carbon pile is.
The pedal gets no neutral and no earth, and with an unpolarized plug
nobody knows which of its two wires is hot. Could a pedal that fits that
cordset unchanged sense the foot’s force with a force-sensing
resistor (FSR) instead of its travel, and do it with no supply and no
microcontroller: a passive circuit, like the KB3151C board but driven by
the FSR?
The answer, on paper, is yes, with one open
question.
The FSR can’t simply replace the KB3151C’s pot. In
that board the pot carries nearly the whole line voltage, about 170 V
peak, and an FSR is made for a few volts (§2).
A clamped timing supply fixes most of that. A
resistor from the line feeds a pair of 56 V zeners, which hold the
timing network’s supply to ±57 V whatever the line does. The FSR, in
series with a fixed resistor, charges the timing capacitor from that.
The diac and triac then work as in any pedal (§3). The FSR sees at most
about 66 V, and only while it is lightly pressed and carrying well under
a milliamp.
Simulated, it works like a pedal. The motor starts
at about 1.6 N on the sensor, reaches 90% of line voltage at about 14 N,
and tops out at 94%. A pedal that passes a fifth of the foot’s force to
the sensor puts that between about 8 N and 70 N of foot force: from
resting the foot on it to pressing firmly (§4).
Two weaknesses the model shows. Like every simple
RC pedal, it starts abruptly, at about a quarter of line voltage. And
the diac’s tolerance (28 to 36 V) moves the whole curve by a factor of
about two, which a trimmer only partly corrects (§4).
The open question is the sensor. Interlink rates
its FSRs by force and current, and describes 0.1 to 5 V as the typical
drive. The circuit’s current is within Interlink’s guidance; its peak
voltage is well above the typical drive. Whether an FSR runs for years
at 60 V of alternating voltage is something to test, not assume. §6 sets
out the test, and what to do if it fails.
What it is not. A passive pedal has no speed
regulation, no needle-position stop and no calibration: the MCU
controller (speed-control paper, §9) does all of those. This design is
the simplest thing that fits a stock cordset and gives a force pedal a
better feel than a slide pot and a weak spring.
2 What the pedal has to work
with
2.1 Two wires, no earth, unknown
polarity
A cordset pedal sits in one leg of the motor circuit. When the pedal
is open, the motor is off and the pedal’s two terminals see the whole
line voltage between them. When it conducts, the motor runs. The pedal
has no other connection, so whatever is inside it must:
run from the voltage across itself, as the KB3151C
does (switching paper, §6). Before the triac fires in each half-cycle,
that is nearly the whole line, because the motor drops almost nothing at
the timing network’s few milliamps. After it fires, it is about a
volt.
be live throughout. Either terminal may be the hot
one. Unpolarized plugs leave it to chance, and even cordsets with
polarized plugs are wired to the pedal at random: in Robert Drzyzgula’s
experience the makers don’t try. Nothing inside is referenced to earth.
Every part in the pedal, the sensor and its tail included, has to be
treated as being at mains potential. The user is protected by the
housing, not by the circuit.
let a little current through when it is off. A
two-wire pedal can’t open completely, as a carbon pile’s contacts do,
because its timing network has to be connected across it. That leakage
flows through the motor, a few milliamps, far too little to turn
it.
2.2 What an FSR is built for
An FSR is two films: one printed with interleaved silver fingers, the
other with a carbon-based ink. Pressing them together shorts the fingers
through the ink, and the harder the press, the lower the resistance.
Interlink’s integration guides (sources 1 and 2) give the numbers this
design depends on:
Resistance. More than 1 MΩ unloaded (the X and UX
guide gives a stand-off resistance above 10 MΩ), falling to about 1 kΩ
at full load. Interlink defines “actuation” as the force that brings it
below 100 kΩ. Between actuation and saturation, the resistance follows
an inverse power law of force.
Range. Base 400 series, 0.2 to 20 N. X series, 0.3
to 50 N. UX series, 0.5 to 150 N.
Drive. “Any voltage that suits your circuit is
fine. From 0.1 V … to 5 V is the typical range.”
Current. Less than 1 mA per square centimeter of
applied force: the area actually loaded, not the sensor’s whole
area.
Repeatability and drift. About ±2% for one part
with a repeatable actuator; about 10% hysteresis between loading and
unloading; a long-term drift under constant load of less than 4–5% per
decade of time.
Overload. Keep the pressure below 10 N/mm² of
actuator area.
Life. Typically beyond a million actuations.
2.3 Why the KB3151C’s pot position
won’t do
The KB3151C charges its timing capacitor from the triac’s MT2 through
the pot (switching paper, §6). Until the diac fires, the capacitor is
below 32 V and MT2 follows the line, so the pot carries the difference:
up to about 140 V in the half-cycles where the triac fires late, and the
full 170 V peak when it doesn’t fire at all, which is most of the time
the pedal is at rest. That is thirty times the FSR’s typical drive,
across silver fingers a fraction of a millimeter apart. An FSR dropped
in where the pot was would work at first. The question is how long.
There is also a gate problem. The KB3151C drives its BT136-600E
through a 1 kΩ resistor, which suits that sensitive-gate triac (10 mA).
The snubberless BTA16-600BW the project is buying needs up to 50 mA, so
its gate resistor must be smaller (§3).
3 The circuit
The proposed pedal: a two-terminal device
in series with the motor, as the KB3151C is. RD and the back-to-back
zeners D1 and D2 make a clamped supply at node N. RS and the FSR charge
C from it; RB bleeds C and sets the force at which the motor starts.
When C reaches the DB3’s breakover, the diac fires the triac through RG.
F1 and the MOV protect it. Every part is at mains
potential.
3.1 How it works
A clamped timing supply. RD (22 kΩ) runs from the
triac’s MT2 side to node N, and two 56 V zeners, back to back, clamp N
to MT1. Until the triac fires, MT2 follows the line, so N follows it too
until it reaches about 57 V (the zener plus the other’s forward drop),
then stays there: a trapezoid instead of a sine (the green trace in §4’s
waveform figure). When the triac fires, MT2 drops to about a volt and N
collapses with it.
This is the classic way to give a phase-control timing network a
steady supply: unijunction-transistor circuits of the 1960s used it
(source 6). Here it does two jobs. It bounds the voltage on everything
after N, the FSR included. And it makes the timing independent of the
line’s exact voltage: a low line no longer means a late firing.
The FSR sets the charging rate. From N, RS (12 kΩ)
and the FSR in series charge the timing capacitor C (0.1 µF). With the
foot off, the FSR is above 10 MΩ: C barely moves and nothing fires.
Pressing lowers the FSR’s resistance, C charges faster, and it reaches
the diac’s breakover earlier in each half-cycle, so the triac fires
earlier and the motor gets more of the cycle.
RS sets the top end. At full force the FSR falls to
about 1 kΩ, so RS is most of the charging resistance there. It limits
the FSR’s current, and it decides how quickly the curve approaches full
speed: the larger RS, the more of the FSR’s range is spread across the
top half of the speed range (§4).
RB sets where the motor starts. RB (100 kΩ plus a 1
MΩ trimmer) across C drains it continuously. C can only reach the diac’s
breakover when the charging current outruns RB, so RB sets the smallest
force that starts the motor. It also trims out part of the diac’s
tolerance (§4).
Firing. When C reaches the DB3’s breakover (32 V
nominal, 28 to 36 V across parts), the diac snaps back by about 10 V and
dumps that charge into the triac’s gate through RG. With RG at 100 Ω,
the pulse peaks near 90 mA for about 10 µs. That comfortably exceeds the
BTA16-600BW’s 50 mA maximum gate trigger current. The KB3151C’s 1 kΩ
would give about 9 mA, enough for its sensitive-gate BT136-600E but not
for a snubberless triac. A diac fires the triac in quadrants I and III
only, which is all a snubberless “W” triac needs (switching paper,
§3).
Protection. F1 (T2A, slow-blow for the motor’s
inrush) is in series with the pedal, so it carries the motor current.
The MOV (the 241KD14 the project is buying) sits across the pedal on the
inner side of the fuse, so a MOV that fails short blows F1 instead of
burning. It clamps surges to about 395 V, well inside the triac’s 600
V.
No snubber. The snubberless BTA16-600BW is rated to
commutate an inductive load without one (switching paper, §5). Pads for
a 100 Ω resistor and a 47 nF X2 capacitor across the triac are worth
leaving on the board in case the bench shows otherwise.
3.2 Parts
Part
Value
Notes
TR1
BTA16-600BW (or -600BWRG)
Snubberless, insulated tab; small clip heatsink
DB3
Diac, 32 V nominal
28–36 V spread; see §4
RG
100 Ω, 0.25 W
Gate pulse about 90 mA
C
0.1 µF film, 100 V or more
Sees about ±57 V; not across the line, so it needn’t be X2
RS
12 kΩ, 0.25 W
Range; 6.8–18 kΩ explored in §4
FSR
Interlink UX 406
About 34 mm square active area; the larger area matters (§4)
RB
100 kΩ fixed plus 1 MΩ trimmer
Start force
D1, D2
56 V, 1 W zeners (1N4758A), back to back
Each carries a few mA
RD
22 kΩ, 1 W, rated 350 V or more
0.23 W with the pedal at rest
RV1
MOV, 14 mm, 150 VAC (241KD14)
Across the pedal, after F1
F1
T2A, 5 × 20 mm, in a holder
In series with the pedal
3.3 What it costs at rest
With the foot off, the pedal still passes the clamp’s current. The
simulation gives 3.3 mA RMS through the motor, and 0.23 W in RD. The
motor won’t turn on 3 mA. RD wants a 1 W part, for margin, and for the
voltage: it carries up to about 115 V.
At full force the triac fires at about 49°, the motor gets 94% of the
line, and RD dissipates almost nothing, because the voltage across the
pedal is gone for most of each half-cycle. The triac itself dissipates
about 1.5 W at a 99K’s current, as in any triac pedal; the heatsink
already chosen for the knee controller is enough.
4 The circuit simulated
4.1 The model
The figures in this section come from
tools/make_fsr_pedal_figures.py, the same kind of
half-cycle model the switching paper uses for the KB3151C (its §4 and
§6). Its assumptions:
The line is 120 V, 60 Hz (170 V peak). Until the
triac fires, the pedal sees the whole line; after, about 1 V. The load
is resistive, so the triac turns off at each voltage zero. A motor’s
inductance carries the current a little past the zero, which this
ignores.
The clamp is ideal at ±56.7 V.
The diac fires at 32 V and drops C to 22 V, as in
the switching paper. Its 28–36 V spread is explored separately.
The FSR follows a power law,
.
For the UX series,
N and
,
which gives about 1 kΩ at 150 N. That curve is an
assumption, fitted to Interlink’s stated range and its
“about 1 kΩ at full load”; Interlink publishes the shape of the curve,
not a formula. The sensors on order will replace it with measurements
(§6).
Hysteresis. C’s state carries from half-cycle to
half-cycle, so the model shows the start and stop forces as the force
rises and falls. The FSR’s own hysteresis, about 10%, is not modeled; it
adds to what the figures show.
“Motor voltage” is the RMS voltage a resistive load
would see, as a percentage of the line: a stand-in for speed, which also
depends on load.
4.2 One operating point
At 4 N on the sensor, the FSR is about 19
kΩ. Node N (green) rises with the line, flattens at 57 V, and collapses
when the triac fires. C (blue) charges from it to the diac’s 32 V at
about 84°, then drops to 22 V and drains through RS, the FSR and RB for
the rest of the half-cycle. The motor gets the shaded part of each
half-cycle.
The figure shows what the clamp buys. Node N never exceeds 57 V, so
neither does anything downstream of it. The highest voltage the FSR sees
is when it is barely pressed and C swings from one polarity to the
other: about 66 V (§4.5).
4.3 Force against speed
Motor voltage against force on the
sensor, for three values of RS. The top axis shows the foot force if the
pedal passes a fifth of it to the sensor (§5). Dashed lines show the
curve when easing off, where it differs.
With the nominal values (RS = 12 kΩ, RB = 330 kΩ, DB3 at 32 V):
Force on the sensor
On a 5:1 pedal
Motor starts
1.6 N, jumping to about 23%
8 N
Motor stops when easing off
1.5 N
7 N
90% of line voltage
13.6 N
68 N
Top
94% at 150 N
Three things to see in the figure:
The curve is roughly logarithmic in force. Each
doubling of force adds a similar step of speed through the middle of the
range, the carbon pile’s virtue the speed-control paper (its §4) wanted
to keep. It falls out of the FSR’s power law and the RC timing together,
with no shaping circuit.
RS spreads or compresses the top. With RS at 6.8 kΩ
the curve reaches 90% by 7 N; at 18 kΩ, not until 127 N, and the top is
lower (90%). RS is the “how hard do I have to push for full speed”
setting.
The start is abrupt, and slightly hysteretic. The
motor starts at about a quarter of line voltage rather than from zero,
and once started keeps going down to a slightly lower force. This is the
same snap-on the switching paper (§4) explains for any single-RC pedal:
C’s leftover charge from the last firing helps the next one. It is
milder here than the KB3151C’s, about 8% in force against the KB3151C’s
15% in resistance, but it is there. A second RC stage, the usual cure in
better dimmers, would soften it at the cost of a few more parts and
another round of simulation.
4.4 The diac’s tolerance
The same curve with the DB3 at the ends
of its tolerance, gray dashed and dotted, and with RB trimmed to
compensate, in color.
A DB3 may break over anywhere from 28 to 36 V. Because C charges
toward a fixed 57 V, a higher breakover takes disproportionately longer
to reach, and the curve moves a long way:
DB3
RB
Starts at
90% at
28 V
330 kΩ
1.2 N
7.7 N
32 V
330 kΩ
1.6 N
13.6 N
36 V
330 kΩ
2.2 N
37 N
28 V
100 kΩ
1.5 N
8.3 N
36 V
1.1 MΩ
2.0 N
31 N
The RB trimmer brings the starting force back into line, but it can’t
correct the top of the curve, which depends on RS and the breakover
together. Two practical answers: measure a few DB3s (a variable supply
through a 100 kΩ resistor shows the snap) and keep the ones near 32 V;
or make RS a trimmer as well, so each pedal is set at both ends. A
higher clamp voltage would make the circuit less sensitive to the diac,
but it raises the voltage on the FSR, the one number this design most
needs to keep down.
4.5 The stress on the sensor
Peak voltage across, and peak current
through, the FSR as the force rises (RS = 12 kΩ), against Interlink’s
guidance: the typical drive, and 1 mA per square centimeter of loaded
area for a fully loaded 402 and for a 406 under a 20 mm
puck.
Current. The FSR’s current peaks at 2.5 mA at full
force. That is twice what a 402 may carry fully loaded (1.27 cm²), but
within the guidance for a 406 pressed through a 20 mm puck (3.1 cm²) or
anything larger. This is the main reason the design calls for the
406.
Voltage. The voltage is highest, about 62–66 V,
when the sensor is barely loaded and its resistance is high, and falls
to a few volts at full force. The current at those voltages is under a
milliamp, so the power in the sensor is small: tens of milliwatts at
most.
Polarity. The voltage alternates every half-cycle.
That matters, because the failure most feared for silver conductors
under voltage, silver migration, is driven by a steady DC voltage across
the gap in the presence of moisture. Alternating voltage greatly reduces
it, and the pedal is dry.
Even so, 66 V is well above the 5 V Interlink calls typical, and
Interlink does not publish a maximum. That is the open question in this
design, and §6 proposes how to answer it.
5 The pedal around it
5.1 Dividing the force
A foot resting on a pedal pushes with something like 5 to 10 N, and a
firm press is several times that. The simulated curve runs from about
1.6 N to 14 N on the sensor, so the pedal should pass the sensor roughly
a fifth of the foot’s force. Two ways to do it:
A spring in parallel. The tread rests on a spring
that carries most of the load, and presses the sensor through a puck
with the rest. The spring also gives the pedal travel and feel: an FSR
itself moves about 0.15 mm, so without a spring the pedal is a stiff
button. This is the arrangement Interlink recommends for designs that
need travel (“use soft rubber or a spring as part of the
actuator”).
A lever. The tread pivots, and the sensor sits
closer to the pivot than the foot does. A 5:1 lever gives the 5:1
division directly. It is simpler, with less travel.
Either way, three things protect the sensor:
A puck to spread the load. A rigid disc about 20 mm
across with a thin rubber face, pressing on the 406’s center. It sets
the loaded area that the current limit depends on (§4.5), keeps the
pressure far below Interlink’s 10 N/mm² damage threshold, and makes the
response repeatable.
A hard stop. The tread should bottom out on the
housing before the sensor sees much more than the force for full speed,
so a stamp on the pedal can’t crush it.
A flat, rigid seat for the sensor, with its vent,
at the end of the tail, left open. Interlink warns that a blocked vent
makes the response non-repeatable.
The UX’s range helps here. Because it reads to 150 N, the force
division needn’t be precise: a pedal that passes a quarter or a third of
the foot’s force still stays within the sensor’s range, and RS and RB
move the curve to suit.
5.2 Everything is live
Every part of this pedal is at mains potential, the FSR, its tail and
the puck’s underside included. The user is protected by the housing
alone, as in a carbon-pile pedal, so the housing has to provide what
standards call reinforced insulation:
Only insulating material where a foot or hand can
reach: a printed or molded tread, no metal screws or plates
exposed, and fasteners recessed below insulating plugs.
An insulating layer between the tread and the
sensor. The puck presses on the FSR’s top film. That film is
thin polyester, not an insulation barrier, so the puck itself must be
plastic, and the tread plastic above it.
Creepage and clearance. The spacing between live
parts and the outside of the housing, along surfaces and through air,
should follow what IEC 60335-1 (household appliances) requires for
reinforced insulation at 120 V, several millimeters. Printed parts make
this easy to design in.
Strain relief on the cord, as on any pedal, so a
yank can’t pull live conductors against the housing or out of the
board.
Heat. About 1.5 W in the triac and 0.2 W in RD, in
a closed box under a foot. The clip heatsink and a few cubic inches of
air are enough, but the triac shouldn’t sit against the tread.
Because every part is live, nothing inside needs isolating from
anything else. That is what lets this design skip the vactrol: the FSR
can be wired straight into the circuit, as long as the voltage across it
is acceptable (§6).
5.3 Fitting the cordset
The pedal needs only its two terminals, so it fits any cordset that
wires the pedal in series with the motor. Most do, the Kenmore’s among
them as far as this paper knows, but check before building: some
cordsets carry a third conductor to the pedal, for a lamp or a two-speed
switch, and that conductor must not be connected to anything in this
design. A pedal from the same cordset family, opened and traced, answers
it in a minute.
6 Testing it, and the open
questions
6.1 Testing order
Build and test in this order, so that each step answers one
question:
Measure the sensors. Before any mains wiring,
measure each FSR’s resistance against force: the three pre-terminated
402s (base, X and UX) and a UX 406. A kitchen scale makes a good force
gauge: put the sensor on the scale, press it through the puck, and read
the force from the scale and the resistance from a meter. Ten or fifteen
points from actuation to full load give the curve that replaces the
model’s assumed one. Rerun tools/make_fsr_pedal_figures.py
with the fitted values.
Measure the DB3s. Raise a variable DC supply slowly
through a 100 kΩ resistor across each diac and note the voltage where it
snaps. Keep the ones near 32 V.
The voltage test on the sensor. This is the test
that decides the design (below).
The circuit on the bench, with a lamp. Build it on
the isolation transformer (the Hammond 171B), with a 60 to 100 W
incandescent bulb as the load. A bulb is close to the model’s resistive
load, so the scope should show §4’s waveforms, and the firing angle
against force can be checked directly against the simulation. Look at
node N, C and the gate pulse.
The motor on the rig. Then the motor test rig, with
the motor as the load: check that the triac commutates cleanly without a
snubber, and how the start feels with a real motor. A universal motor
needs some voltage before it turns at all, which may make the abrupt
start less noticeable, or more.
The pedal. Only then the housing, the spring or
lever and the puck.
6.2 The voltage test
The open question is whether an FSR tolerates this circuit’s voltage
for years: up to 66 V, alternating, at well under a milliamp. Interlink
calls 0.1 to 5 V typical but publishes no maximum. The risks are slow
ones: silver migration between the fingers, or a change in the ink,
either of which would show up as drift or a falling unloaded
resistance.
A test that answers it:
The setup. Two UX 406s (or a 406 and a spare 402)
under fixed weights, one at a light load (where the voltage is highest)
and one at rest, unloaded. Drive each from a 60 V AC source through a
resistor sized to give the circuit’s current: a small transformer
secondary will do, or the rig’s isolation transformer through a divider.
Keep a third, identical sensor unpowered, under the same weight, as a
control.
The measurement. Each day or two, disconnect the
drive and measure the resistance at the same load, and the unloaded
resistance.
The duration. A few weeks is a start: a thousand
hours is a respectable soak.
Pass. The powered sensors drift no more than the
unpowered control, and the unloaded resistance stays above 10 MΩ.
If the sensors pass, design (a) stands as drawn. If they drift, the
fallbacks are, in order:
Lower the voltage. A lower clamp (back-to-back 43 V
zeners, with DB3s selected below 32 V) cuts the sensor’s peak voltage by
a quarter, at the cost of more sensitivity to the diac.
Move the sensor out of the high-voltage position.
Put the FSR across a smaller part of the timing resistance, with a fixed
resistor taking most of the voltage. This narrows the force range the
sensor controls.
Isolate it. Design (b) from the earlier discussion:
a vactrol whose photocell sits in the timing network, with the FSR
driving its LED at 5 V. It needs a supply in the pedal, which a two-wire
pedal has to harvest, and a vactrol whose cell is rated for the voltage
it sees.
6.3 Other open questions
The abrupt start. Whether a quarter of line voltage
at turn-on feels wrong with a real motor, and if so, whether a second RC
stage fixes it.
The real FSR curve. The model’s power law is a fit
to two numbers. The measured curves will move the force window, and RS,
RB and the pedal’s force division will move with them.
The 402 alternatives. The base and X sensors reach
full scale sooner, which would move the window down and make the pedal’s
force division larger. The 406’s area still makes it the better choice
for the current (§4.5).
Temperature. FSRs and diacs both drift with
temperature, and a pedal near a heater or in a cold garage will feel
different. The model ignores it.
7 Sources
Interlink Electronics, FSR 400 Series Integration Guide:
construction, resistance against force, interface circuits, current
limit, usage notes. Available from Interlink’s web site.
Interlink Electronics, FSR X & UX Series Integration
Guide: the same for the extended-range sensors, with their
performance table (stand-off resistance, hysteresis, drift,
repeatability) and the FAQ quoted in §2 (drive voltage, resistance
range, overload, lifetime).
DB3, DB4 and DC34 diacs, STMicroelectronics data sheet (breakover
28–36 V, breakback at least 5 V).
241KD14 metal-oxide varistor, Yageo 14D series data sheet.
General Electric, SCR Manual, 6th ed., 1979: phase control,
RC timing networks and zener-clamped timing supplies.
Inside a Switching Circuit, the companion paper in this
repository (docs/switching/): §3 on triac quadrants, §4 on
phase control and snap-on, §5 on snubbers, §6 on the KB3151C.
Speed Control of Universal Motors in Sewing Machines, in
this repository (docs/motor-control/): §4 on the carbon
pile, §9 on the MCU controller.
IEC 60335-1, Household and similar electrical appliances:
Safety, Part 1: General requirements, for insulation classes,
creepage and clearance.
The simulation and the figures are generated by
tools/make_fsr_pedal_figures.py in the project repository.
The schematic is drawn by the same script; there is no KiCad source
yet.