Construction, waveforms and DC alternatives for
sewing-machine drives. A companion to Speed Control of Universal Motors
in Sewing Machines
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 Speed Control of Universal Motors in Sewing
Machines (cited here as “the speed-control paper”), which covers
the physics of the series motor and each method of controlling it. This
paper looks inside the motors themselves. The sources are listed in the
last section; please report errors as issues on the project’s GitHub
repository.
What it covers. The speed-control paper treats the
sewing-machine motor as three letters in an equation: a resistance
,
a motor constant
,
and a speed
.
This paper starts from the metal and copper.
Construction. It shows how a universal motor is
built, and which parts of it make up each term of that equation:
its laminated field iron and field coils;
its wound armature;
its commutator and carbon brushes.
Other motor types. It treats four related motors
the same way, as parallel subjects rather than footnotes:
the wound-field DC motor, the universal motor’s DC relatives;
the brushed permanent-magnet DC motor;
the brushless DC motor;
the induction motor, with the industrial clutch motor and the
variable-frequency drive.
For each it covers how the motor is built, what its equations say, and
how its speed is controlled.
Waveforms. It shows what each control method looks
like on the wire:
the universal motor straight on the line, behind a rheostat, and
chopped by a triac, where the winding’s inductance stretches and smooths
the current;
a permanent-magnet motor on pulse-width modulation;
a brushless motor’s six-step drive.
A practical question. It closes by asking whether a
DC motor exists that has a sewing motor’s power and speed, about 50–100
W at around 6,000 rpm, and also fits the space a sewing motor occupies.
If so, could the MCU controller designed for the VSM Base drive it?
The short answers.
The universal motor is a DC series motor built to survive
AC. Its iron is laminated, its brushes are high-resistance
carbon, and its field and armature share one current. That shared
current is why it runs on AC at all. It is also why its speed sags under
load and why its back-EMF vanishes whenever the current stops.
The inductance is mostly in the field coils, and it
matters. It makes the current lag the line by about 30° at full
speed. Behind a triac, it carries the current past the point where the
line voltage reverses. It also turns each chopped slice of voltage into
a smooth hump of current. At speed the back-EMF acts like extra
resistance and shortens that smoothing. At a crawl the back-EMF is gone,
so the smoothing lasts longer but the torque pulses grow sharper.
Permanent-magnet and brushless motors behave like
voltage-controlled speed sources. Their flux is fixed by
magnets, so speed follows voltage and torque follows current, each on
its own. That makes them far easier to regulate, to hold still, and to
brake. They need DC and, in the brushless case, an electronic
commutator.
The induction motor needs no commutator at all. A
three-phase supply makes a field that rotates by itself, and the rotor’s
bars have their currents induced by the field sweeping past. It runs at
nearly fixed speed off the line, a few percent below the field, and only
a variable-frequency drive changes that. The sewing trade’s clutch motor
was an induction motor with the speed control in a slipping clutch.
Suitable DC motors exist, and they are small and
cheap. Common 24 V brushed motors of the “775” size (42 mm
across) deliver 60 W or more near 7,000 rpm. Brushless motors of 42–57
mm deliver 50–100 W at 3,000–4,000 rpm. Both fit easily within the
envelope of a Singer external motor. Driving one would change the
controller’s power stage completely:
a certified 24 V supply, with no mains anywhere else;
an H-bridge or a three-phase bridge;
current sensing.
The RP2350B, its encoder and its firmware structure would carry over.
Such a controller could do what a triac never can: brake actively and
hold the needle exactly where it is put.
On the simulations. The waveforms here are computed
from simple lumped models with representative values for a 99-class
motor, not measured. They show the shapes and the cause and effect
correctly. The exact numbers belong to the motor test rig, which is
designed to measure
,
and
and to put these waveforms on a scope. Its captures should replace these
figures.
2 The universal motor, in iron and
copper
2.1 What is inside
A sewing machine’s motor, a Singer BZ or BR, or one of the generic
replacements sold for them, is a small two-pole universal motor. Taken
apart, it has four groups of parts.
A two-pole universal motor, drawn
schematically (not to scale): an end view through the armature. The two
field coils are wound on the stator’s poles, and their flux crosses the
armature from pole to pole and returns around the yoke. The armature
conductors carry current one way on the left of the brush axis and the
other way on the right (dots and crosses), which is what makes
torque.The same motor in a side view, in
section: the laminated armature and the end turns of its winding, the
field iron with its coils, the commutator, two carbon brushes on
springs, the fan, the bearings and the pulley.
The stator (field).
The iron. A stack of thin steel laminations,
insulated from one another, with two inward-facing poles. It is
laminated because the flux in it reverses 120 times a second on 60 Hz
mains. A solid iron core would carry large eddy currents, circulating in
the iron itself, and heat up (Electrical
Easy).
The field coils. Each pole carries a coil of
enameled copper wire, wound around its neck, and together the two coils
make the magnetic field. On many small universal motors the two coils
are connected one on each side of the armature. The motor’s circuit then
runs line, field coil, brush, armature, brush, field coil, line, which
keeps the armature away from both supply wires electrically and helps
contain the brushes’ radio noise.
The armature (rotor).
The core. A laminated steel cylinder on the shaft,
with slots along its length.
The winding. Coils of finer wire lie in the slots.
Each coil’s two ends are soldered or welded to two bars of the
commutator.
The commutator and brushes.
The commutator. A ring of copper bars on the end of
the armature, insulated from one another with mica.
The brushes. Two blocks of carbon, each pressed
against the commutator by a spring in a holder. On Singer motors the
holders are reached from outside, under screw caps, because brushes
wear.
The commutator is a rotary switch. As the armature turns, it keeps
reversing the current in each coil just as that coil passes from under
one pole to under the other. The current pattern in the armature
therefore stays fixed in space, flowing one way on one side of the brush
axis and the other way on the other side, while the copper moves through
it. Each conductor in the field feels a force, and every force pushes
the same way around the shaft.
Everything else.
A fan on the shaft, because more than half the
input power becomes heat.
Two bearings. On older motors these are sleeve
bearings fed by oil wicks.
A pulley for the belt.
The housing. On the 99, the motor bolts to a
bracket on the back of the machine and drives the hand wheel by a belt.
Others, such as the 15-91 and the 201, have a “potted” motor built into
the casting.
2.2 How the parts make up the
equation
The speed-control paper’s model is
and each term has a physical home in the motor.
The series connection and the lumped
model it reduces to. One current passes through both field coils and the
armature, so the field’s strength is set by the same current the
armature carries. Below, the three terms of the model: the copper’s
resistance, the inductance (mostly the iron-cored field coils), and the
back-EMF.
,
the resistance. This is the copper of both field coils and the
armature winding, plus the brushes and their contact with the
commutator. Carbon brushes are deliberately resistive, because
commutation on AC is harsher than on DC. The resistance limits the
circulating current in the coil being switched, and so limits sparking.
A sewing motor’s total is typically some tens of ohms.
,
the inductance. Most of it is the field coils. They have many
turns around an iron path with only a small air gap, so a modest current
makes a lot of flux. The armature contributes some, partly offset by its
own reaction to the field.
appears only when the current changes, so on DC it does nothing in
steady running, and on AC it is a large part of the impedance.
,
the back-EMF. It is the voltage generated by armature
conductors cutting the field. It is proportional to how fast they move
()
and to how strong the field is. In a series motor the field is made by
the same current
,
so the back-EMF is proportional to both. This one fact sets the motor’s
character:
No current, no back-EMF. When the current stops,
the field collapses and the back-EMF goes with it, apart from a few
volts from the iron’s residual magnetism. This is why an MCU cannot
simply “listen” to a universal motor between triac pulses for its speed
(§3).
Torque goes as
.
Torque is force on conductors carrying
in a field made by
.
When the current reverses, both the field and the armature current
reverse together, so the torque keeps its direction. That is why the
motor runs on AC.
,
the motor constant. It folds together:
the number of armature conductors;
the number of field turns;
the geometry of the magnetic path.
It is constant only while the iron is far from saturation. At high
current (a stall, or a heavy start) the iron saturates, flux stops
rising with current, and torque grows more slowly than
.
2.3 One constant here, two
later
Every brushed motor obeys the same two laws, whatever makes its
field.
Back-EMF:
.
Torque:
.
Here
is the field’s flux,
the armature current, and
a number fixed by how the armature is wound. The same
appears in both, because they are two views of one exchange of energy.
The electrical power the back-EMF absorbs,
,
is the mechanical power the shaft delivers,
.
Both come to
.
Motors with a fixed field, shunt-wound (§4) and permanent-magnet
(§5), have a constant
.
Their
is then one fixed number, and it is written twice, by the job it is
doing:
,
the back-EMF constant, in volts per rad/s:
.
,
the torque constant, in N·m per ampere:
.
In SI units the two are the same number. Datasheets often quote them
in different units, such as volts per 1,000 rpm and ounce-inches per
ampere, which hides that they are equal.
A series motor has no fixed field. Its flux is made by the armature
current itself, so
is proportional to
.
Writing
gives the paper’s model:
Back-EMF:
.
Torque:
.
So the single
used through §2 and §3 is not a third kind of constant.
is the series motor’s
and
at once, a value that rises and falls with the current.
itself has units of volts per (ampere × rad/s), which reduce to henries.
The representative motor’s
is 0.1037 H. At its running current of about 1 A, that makes an
effective
V per rad/s (or N·m per ampere), comparable to a fixed-field motor’s
constant. From §4 on, the motors have fixed fields, and the paper uses
and
.
2.4 Why it runs slower on AC than
on DC
On DC, only
and the back-EMF oppose the supply. On AC the inductance adds a
reactance
,
which is about 57 Ω at 60 Hz for the representative 0.15 H used in this
paper. It uses up part of the supply voltage without doing work. A
universal motor therefore runs noticeably faster on DC of the same
voltage than on AC (Electrical
Easy). This is also why a DC drive such as the KBIC, which rectifies
the line, drives a universal motor as a DC series motor (speed-control
paper, §8).
2.5 What the nameplates say
Sewing-motor nameplates are approximate, but they agree on the
scale:
Current: 0.3–0.6 A on 220–240 V, or about 1 A on
120 V.
Speed: 5,500–7,000 rpm with no load.
Power: outputs of roughly 25–75 W, labeled 1/15 to
1/10 hp, from about 100 W in.
The motor on the bench here draws about 1.3 A (speed-control paper,
§11). The representative motor used in the simulations below runs at
about 7,000 rpm and draws 1.0 A from 120 V, and it delivers a mean
torque of about 0.10 N·m, about 76 W.
3 The universal motor on the
wire
This section shows what a scope would see at the motor’s terminals,
and in its current, under each kind of control. The plots are computed
with the model of §2, using the representative values listed with the
figures:
= 30 Ω,
= 0.15 H,
= 0.104 H, and a load that rises with speed. They come from
tools/make_motor_waveforms.py. On the test rig, the same
traces are taken with the scope behind an isolation transformer, voltage
across the motor’s terminals and current across a small shunt (rig
guide, §7).
3.1 Straight on the line
Simulated: the motor connected straight
to 120 V, 60 Hz. The current (blue) is nearly a sine wave and lags the
voltage (orange) by about 1.3 ms, or 28°. Below, the torque
:
always positive, pulsing at 120 Hz between zero and twice its
mean.
Three things show here:
The current lags the voltage. At full speed the
back-EMF behaves like extra resistance,
,
in series with
.
That gives an effective resistance of about 106 Ω against the
inductance’s 57 Ω of reactance, so the current lags by
.
The power factor is about 0.88. The supply’s volt-amperes exceed its
watts by that much, and a plug-in power meter shows both.
The torque pulses. The armature is pushed in 120 Hz
bursts, hardest at the current’s peaks, with nothing at each zero. At
full speed the armature’s inertia and the machine’s flywheel average
this out, and all that remains is a hum at 120 Hz.
The current is almost a sine. The line is a sine,
and the motor at steady speed is close to a fixed impedance, so the
current is too. Its small distortion comes from the iron’s saturation,
which the model leaves out.
3.2 Behind a rheostat or carbon
pile
Simulated: a 65 Ω rheostat in series,
chosen to bring the motor down to the same speed as the 90° triac below
(about 3,950 rpm). The motor sees a smaller sine, and the pedal turns
about 42 W into heat.
A series resistance leaves every waveform a sine; it only makes them
smaller. The motor sees a reduced sinusoidal voltage, and the current
lags it a little less, because the circuit is now more resistive. What
the scope does not show is the cost. At this setting the pedal
dissipates about 42 W as heat in the foot control, more than the 28 W or
so that the motor delivers. That is the heat the speed-control paper
describes in §3. It also explains the carbon pile in that paper’s §11,
which grew hot when held just short of starting. With the motor stalled
there is no back-EMF, so most of the power goes into the pile.
3.3 Chopped by a triac
Simulated: a triac fired 45°, 90° and
120° into each half-cycle. The gray trace is the line, the orange trace
the voltage at the motor, and the blue trace the motor current. In the
middle panel, the dashed line shows what a plain resistor’s current
would look like: a slice of sine that stops dead at the voltage
zero.
How a triac switches. (The companion paper
Inside a Switching Circuit covers the devices themselves.) A
triac is a pair of thyristors (SCRs) back to back in one package, so it
conducts in either direction. Its gate can turn it on but cannot turn it
off. A short pulse of gate current starts it conducting. Once the load
current has passed the triac’s latching current, it holds
itself on and the gate has no further say. It turns off only when the
current through it falls below its holding current, a few
milliamps. In practice that means when the load current crosses zero,
and then it waits for the next gate pulse. So the only control a triac
offers is when to switch on in each half-cycle. This is
leading-edge (forward-phase) control:
The line crosses zero. The triac is off, and stays
off.
The firing angle
.
The gate is pulsed, and the triac switches the full line on.
is measured from the voltage zero, so a larger
means a later start and less power.
The current reaches zero. The triac turns itself
off.
The other scheme, trailing-edge (reverse-phase)
control, switches on at the zero and cuts off part-way through. It needs
a switch that its gate can turn off, a MOSFET or an IGBT, and it is what
some dimmers for LED lamps use. It suits a motor badly: cutting the
current off sharply in an inductive load makes the inductance answer
with a voltage spike
().
This is why dimmers sold for motors and transformers are
leading-edge.
The triac switches the full line on at the firing angle and leaves it
on until the current falls to zero. The plots show what the inductance
does with that.
The voltage is a sliced sine. Before firing, the
motor sees almost nothing. The line is held off by the triac, and the
motor’s own terminals show only a few volts of residual-field back-EMF.
At
the motor snaps to the line voltage and follows it.
The current is a smooth hump, not a slice. A
resistor would draw a current shaped exactly like the voltage: a
vertical step at
,
then a sine segment ending at the zero (the dashed trace). The motor’s
inductance forbids the step. The current must start from zero and rise
at a rate set by
,
so it builds gradually after firing.
The current outlasts the voltage. Once the line
passes its peak and starts falling, the inductance keeps the current
flowing. That stored energy drives the current on past the voltage
zero, into the start of the next half-cycle. For a short time the
motor’s voltage is negative while its current is still positive, and the
motor returns some of its stored magnetic energy to the line. Only when
the current reaches zero does the triac turn off. This is the
“conduction stretched past the voltage zero” that the speed-control
paper mentions in §5.
Speed changes how much the current is smoothed. At
high speed the back-EMF adds about 76 Ω of effective resistance. The
current’s time constant is then
ms, short against the 8.3 ms half-cycle, and the current follows the
chopped voltage fairly closely (top panel). At a crawl the back-EMF has
almost gone, and the time constant approaches
ms. Each pulse then becomes a slow, rounded hump that rises late and
decays late (bottom panel).
At low speed the torque comes in pulses. At 120°
the motor gets a short hump of current twice per cycle and nothing
between. Each hump gives a squared kick of torque. With the inertia of a
sewing machine, this representative motor still turns at about 970 rpm.
Five degrees later it is down to about 450 rpm, and by 135° this load
stalls it outright. The narrow band between running and stalling is the
“cogging” region of the speed-control paper, §5. It is also why an
open-loop triac pedal is twitchy at the bottom of its travel.
Two practical consequences for the MCU controller:
The triac must be fired with a pulse long enough to
latch. At large firing angles the current rises slowly from
zero. If the gate pulse ends before the current exceeds the triac’s
latching current, the triac drops out and the half-cycle is lost. AN1662
uses a gate pulse of about 100 µs for this reason, and so does the
concept circuit.
Back-EMF between pulses is tiny. While the triac is
off there is no current, so there is no field and almost no back-EMF. A
universal motor cannot simply be listened to for its speed, as a
permanent-magnet motor can (§5). This is why the concept controller uses
a hand-wheel sensor.
What else the scope shows. The model leaves out two
effects that a real trace has:
Commutation spikes. Each time a brush leaves a
commutator bar, the coil it was shorting is opened with current in it.
This makes small sharp spikes on the voltage, and the radio noise that
the split field coils and the controller’s snubber help contain.
A slower start. A real triac fires only once its
gate current is enough. Near
= 180° it may not fire at all, and the motor simply stops.
4 Wound-field DC motors: the
universal motor’s relatives
A universal motor is one member of a family. Every brushed motor with
a wound field is built from the same parts: field iron and coils, a
slotted armature, a commutator and brushes. They differ in how
the field coil is connected to the armature, and that one
choice sets the speed–torque character. Before permanent magnets became
strong and cheap, these were the DC motors, and they still drive
treadmills, cranes and machine tools.
Series-wound. The field in series with the
armature, carrying the same current. Many turns of heavy wire, because
the field carries the full load current. This is the universal motor’s
connection. Its torque goes as
,
its speed falls steeply with load, and it races when unloaded. On DC, a
series motor with no load can over-speed until it damages itself, which
is why large ones are never run without a load.
Shunt-wound. The field across the supply, in
parallel with the armature, with many turns of fine wire carrying a
small, steady current. The flux is then nearly constant whatever the
load. Back-EMF is proportional to speed alone,
,
and torque to armature current alone,
(the two constants are explained in §2). The speed falls only by the
small
drop as load rises: the “nearly constant speed” motor of machine
shops.
Separately excited. A shunt motor whose field has
its own supply. This makes two independent controls: armature voltage
sets speed up to base speed, and weakening the field raises speed above
it, trading torque for speed.
Compound. Both a series and a shunt winding,
blending the series motor’s starting torque with the shunt motor’s
steadier speed.
Speed control. For a shunt or separately excited
motor, the controller sets the armature voltage.
Industrial drives do this with a controlled rectifier, SCRs fired in
phase like a triac but feeding a DC motor through a bridge. That is the
KBIC’s circuit (speed-control paper, §8). Because the flux is constant,
the armature voltage minus its
drop is a direct measure of speed. That is why “IR compensation”
regulates a shunt motor well, and a series motor only to a first
approximation.
On the wire. A DC motor on a controlled rectifier
sees the rectified line’s humps, with chopped leading edges. Its
armature and field inductance smooth the current much as they do for the
universal motor on the triac (§3). With enough inductance, the current
flows continuously from one hump to the next instead of in pulses.
Why none of this is in a sewing machine.
The shunt field. A shunt field wound for 120 V DC
is many turns of fine wire and dissipates power continuously. The motor
needs a DC supply and a rectifier either way.
The series connection already works. The universal
motor runs directly on AC, gives high torque from a standstill, and
costs little. For a machine that only ever turned one way, at the speed
a foot asked for, that was enough.
Permanent magnets now give the shunt motor’s constant flux with no
field winding at all, and that is the next section.
Speed against load torque, normalized to
full speed at full load, at 100%, 70% and 40% of full voltage. Left: a
series (universal) motor, whose speed collapses as load rises and soars
as it falls. Right: a constant-flux motor, permanent-magnet or shunt,
whose speed is set by its voltage and falls only slightly with load.
Illustrative, from the two motors’ ideal equations.
5 Brushed permanent-magnet DC
motors
5.1 Construction
A brushed permanent-magnet (PM) motor is a shunt motor whose field
coils have been replaced by magnets.
The stator. A steel can that serves as the magnetic
return path, with two (or four) curved magnets glued inside it: ceramic
ferrite in cheap motors, neodymium in compact, powerful ones.
The rest. The armature, commutator and brushes are
the same as in the universal motor, only smaller for the same power. The
motor has nothing else.
The common “775” motor, sold for drills, pumps and robots, is a good
example. It is a 42 mm can, 66.5 mm long, with a 5 mm shaft.
A brushed permanent-magnet motor,
schematically. The magnets set the flux, which crosses the armature
exactly as in the universal motor, but the armature current no longer
makes it. The lumped model has the same three terms, with a back-EMF
that depends on speed alone.
5.2 What the equations say
With the flux fixed,
and in SI units
(N·m per ampere equals volts per rad/s). Three consequences matter for a
sewing machine:
Speed follows voltage; torque follows current. At
steady speed,
:
a straight line falling only by the resistive drop as load rises (§4,
right-hand plot). A drop that is 10% of the supply voltage at full load
costs about 10% of the speed. For a universal motor at a crawl it costs
nearly everything.
Torque at zero speed is just current. A controller
that sets the current sets the torque directly, at any speed including
zero. That is what makes a PM motor a servo: it can push the needle
through a seam at one stitch a second, or hold it still.
The back-EMF is always there. It is made by
magnets, so it does not vanish when the current stops. Switch the drive
off for a moment and the terminals show
,
the speed, directly. The speed-control paper’s back-EMF sampling (its
§9), weak and noisy on a universal motor, is easy and accurate on a PM
motor.
They also run backward when the supply is reversed, and they act as
generators when driven, which is the basis of braking.
5.3 Speed control: the chopper and
the H-bridge
A PM motor needs DC, and its speed is set by the average voltage
across it. A controller makes that average by pulse-width
modulation (PWM): it switches the full supply on and off many
thousands of times a second, and the share of time it is on, the duty
cycle, sets the average.
Outline drive circuits. Left: a triac in
series with a universal motor on the mains. Center: a chopper for a PM
motor, with one switch and a freewheeling diode, for one direction.
Right: an H-bridge, with four switches for forward, reverse and
braking.
The chopper. One MOSFET in series with the motor,
and a diode across the motor. When the MOSFET turns off, the motor’s
inductance keeps the current flowing, and it circulates through the
diode. Without the diode that current would have nowhere to go, and the
voltage spike would destroy the switch.
The H-bridge. Four switches, two on each side of
the motor. Turning on opposite corners drives it one way, the other pair
drives it the other, and turning on both low-side switches shorts the
motor. That short is dynamic braking: the motor, now a
generator, brakes itself hard on its own back-EMF. A sewing machine that
must stop with the needle in the right place wants exactly this.
5.4 On the wire
Simulated: a 24 V, 775-size motor (R =
0.98 Ω, an assumed L = 0.2 mH) on a chopper at 50% duty, turning at
about 3,500 rpm. At 2 kHz (left) the switch period is longer than the
motor’s 0.2 ms electrical time constant. The current rises steeply, then
falls to zero partway through each off-time, and the terminals show the
back-EMF (the 10 V steps) until the next pulse. At 20 kHz (right) the
current only ripples by about 1.6 A around its mean.
The voltage at the motor is a square wave between the supply and
slightly below zero (the diode’s drop). The current is what the
inductance makes of it.
The ripple depends on how the switching period compares with
.
Here the motor’s electrical time constant is only about 0.2 ms, against
a universal motor’s several milliseconds. At 20 kHz the switching period
is 50 µs, short against that, so the current rises and falls only a
little in each period: the ripple is about
A.
At 2 kHz the ripple is ten times larger, and the current
goes discontinuous. The current is zero for part of each
period, which wastes heat in the copper and makes the motor sing at the
switching frequency.
So PWM for motors runs at 20 kHz or more. That is
above hearing, and with small ripple. The RP2350’s PWM slices do this
with no effort.
Discontinuous current changes the average. While
the current is zero, the terminals show the back-EMF, not 0 V, so the
average voltage, and with it the speed, depends on the load. At 20 kHz
the current stays continuous and the average is simply the duty cycle
times the supply.
Hearing it: the cordless drill. A cordless drill
whose trigger is held just short of turning the chuck often lets out a
high-pitched whine. That whine is the switching frequency.
Where the sound comes from. Each pulse of current
pulls on the windings and the iron, and they flex very slightly at the
switching rate. The motor becomes a small loudspeaker playing a tone at
the PWM frequency.
Why it is loudest at the edge of the trigger. With
the rotor stopped there is no fan, gear or brush noise to cover the
tone. The duty cycle is also small, so the current is discontinuous, in
short pulses that start and stop sharply. That makes a harsher, louder
tone than a current rippling around a steady mean.
Why the tool sings at all. Many cordless tools
switch at a few kilohertz to around 10 kHz, where it can be heard.
Slower switching wastes less power in the transistors, and the parts
cost less. Tools that switch above 16–20 kHz are silent.
The same holds for any battery tool. A cordless
router or saw, brushed or brushless, sings in the same way, since a
brushless controller chops its three phases with PWM too. A corded tool
on a triac control is different: its hum is at twice the line frequency
(120 Hz on 60 Hz mains) and its multiples.
The tone can be measured. A spectrum-analyzer app on a phone, held
near the tool, shows a sharp peak at the switching frequency. There are
two cautions:
Harmonics. Peaks may also show at 2×, 3× and so on.
A part whose force goes as the square of the current can make 2× the
strongest. If the peaks are evenly spaced, the spacing is the switching
frequency.
A changing frequency. Some cheap controllers change
frequency with the trigger’s position, so the pitch moves as the trigger
is squeezed.
The certain answer is a scope across the motor’s leads, which shows
the frequency and the duty cycle at once, at a battery voltage that is
safe to probe.
Measured: a drill and an impact driver. Two DeWalt
20 V Max tools were measured this way in a quiet room, with a phone’s
microphone and a spectrum-analyzer app:
A DCD780 drill driver, 0–600 and 0–2,000 rpm in its
two gears.
A DCF885 impact driver, 0–2,800 rpm and 0–3,200
impacts per minute.
Both are brushed models: replacement carbon brushes are sold for both
(iFixit,
vhbw).
Each is a PM motor on a single PWM switch, the circuit of this section.
Each tool was recorded three times: the room alone, the trigger just in,
and the trigger fully in.
Measured: sound spectra of a DeWalt
DCD780 drill (top) and DCF885 impact driver (bottom), from an iPhone
running an FFT analyzer app, each with the room alone (gray), the
trigger just in (blue) and the trigger fully in (orange). The dotted
lines mark the narrow lines described in the text. Read back from the
app’s screenshots and replotted.
The drill sings at about 3.85 kHz. With the trigger
just in, the spectrum shows a sharp line there, some 20 dB above its
surroundings. It has harmonics at 7.7 kHz and 15.4 kHz, twice and four
times the fundamental. The third harmonic, near 11.5 kHz, is faint.
It never stops chopping. The line is still there
with the trigger fully in, so this drill tops out short of 100% duty. It
has moved down to about 3.3 kHz, so the controller changes its frequency
near full power, whether by design or as a side effect of how it
controls current. The drill’s two-position speed switch is mechanical,
selecting first or second gear, and it changed nothing in the
spectrum.
The impact driver shows 7.7 kHz and 15.4 kHz. These
are narrow lines that hold their frequency as the trigger moves, the
signature of switching. There is no clean line at 3.85 kHz, only a broad
hump from 2 to 4 kHz that grows with speed, and is more likely
mechanical. So either this tool switches at 7.7 kHz, or it switches at
3.85 kHz and its fundamental is lost in the hump. A scope across its
motor would settle it.
The speed can be heard too. In the impact driver a
tall, narrow line rises from about 320 Hz to 380 Hz as the trigger goes
in, and the drill shows one near 320 Hz at full trigger. A line that
moves with speed is most likely the motor’s own rotation, felt through
any small imbalance: 320–380 turns a second is 19,000–23,000 rpm. That
is a usual speed for a small brushed motor, and a planetary gearbox
brings it down to the 2,000–2,800 rpm on the nameplates. Here the
spectrum doubles as a tachometer.
A faint third harmonic has a likely cause. A train of pulses at duty
cycle
has harmonics whose sizes go as
.
At
near 1/3, the third falls to nothing while the second and fourth stay
strong. The motor’s mechanical response and the phone’s microphone shape
the levels too, so this is a likely cause, not a proof.
What it means for control. A PM motor drive usually
has two loops:
A fast current loop. It measures the current with a
shunt resistor, once per PWM period, and sets the duty cycle to hold the
commanded current, which is the commanded torque.
A slower speed loop around it. It commands the
current.
A third, position loop can sit outside both. This is how industrial
servo drives are built, and an RP2350 can run all three.
6 Brushless DC motors
6.1 Construction
Turn the PM motor inside out:
The magnets are on the rotor. They turn with the
shaft.
The windings are on the stator. They stay fixed and
are cooled directly through the frame.
The commutator goes away. There are no brushes to
wear, spark or shed dust. Its job, switching current between coils at
the right moment, is done by a controller with transistors.
The stator is wound as three phases, usually connected in a star.
Left: a brushless inrunner,
schematically. Six stator teeth are wound as three phases (A, B, C, each
on two opposite teeth), with a four-pole magnet rotor inside. Three Hall
sensors report the rotor’s position. Right: the three-phase bridge that
drives it, six switches, one pair per phase.
Two arrangements are common:
The inrunner. The rotor turns inside the stator, as
drawn. It is the form of industrial motors, including the 42 mm and 57
mm motors of §9.
The outrunner. The rotor is a cup of magnets
turning outside a stationary wound core. This is the form of
drone and model-aircraft motors: more torque for their size, but an
exposed rotating can.
Poles and Hall sensors. Motors usually have 4–14
magnet poles. With
pole pairs, the current pattern must go round its cycle
times per mechanical revolution. Three Hall-effect sensors, set 120
electrical degrees apart, divide each electrical cycle into six states,
so an 8-pole motor (4 pole pairs) reports 24 distinct positions per
turn. That point returns in §9.
The ratings. Brushless motors are rated by:
Voltage and speed, or by
,
rpm per volt, which is just
in other units.
Rated torque, rated current, and a peak
torque, often about three times the rated torque, available for
short periods.
6.2 Driving it
A brushless motor cannot run without its controller. Connect it to a
supply and it sits still. The controller’s three-phase bridge does the
commutator’s work, in one of two ways.
Six-step (trapezoidal) drive. At any moment two of
the three phases carry current, one in and one out, and the third
floats. The Hall sensors say which 60° step the rotor is in, and the
controller switches to the next pair as each step is reached. Speed is
set by PWM on the active switches, as for a PM motor.
Idealized six-step drive. Top three rows:
each phase’s back-EMF (trapezoidal, flat for 120°) and the current block
the driver applies while that phase’s back-EMF is flat. Fourth row: the
three Hall signals, whose edges fall at the commutation points. Bottom:
which two phases carry current in each 60° step.
Field-oriented control (FOC). The controller drives
all three phases with sinusoidal currents, placed at exactly 90
electrical degrees to the rotor’s field. That gives the most torque per
ampere and almost no torque ripple. It needs the rotor’s angle
continuously, from an encoder or magnetic angle sensor, or estimated
from the motor’s own voltages, and more arithmetic each PWM period. A
Cortex-M33 with a floating-point unit, like the RP2350’s, does it
comfortably.
On the wire. In six-step drive:
The phase voltages are PWM-chopped blocks. Each
phase is driven high for 120°, floats for 60°, driven low for 120°, then
floats again. During the floating intervals the scope shows the back-EMF
rising or falling through zero. Sensorless controllers watch for that
zero crossing in place of Hall sensors.
The phase currents are rounded trapezoids. Each
rises and falls at every step at a rate set by the winding’s inductance,
and carries a PWM ripple like the PM motor’s.
Under FOC the currents become clean sines with a little PWM
ripple.
What it gives a sewing machine. It gives everything
the PM motor does:
torque from current at any speed;
active braking;
holding still.
On top of that it gives:
no brushes to wear or spark;
higher efficiency, so less heat;
quiet running at 20 kHz PWM;
in Hall sensors that come built in, a measure of the motor’s
position.
Brushless servo motors replaced the clutch motor (§7) on industrial
sewing machines for these reasons. They are typically a few hundred
watts, mounted under the table, with the controller in the same housing
(Leather
Machine Co.).
7 Induction motors: AC with nothing
to wear
The motors so far all have something that switches the current to
keep the torque going one way: a commutator and brushes, or transistors.
The induction motor has neither. It is the motor of fans, pumps,
compressors, table saws and almost every factory floor, and on the
sewing floor it was the clutch motor that the brushless servo replaced
(§6).
7.1 Construction
The stator. Laminated iron with slots, carrying a
three-phase winding. It is the same kind of stator as a brushless
motor’s (§6).
The rotor. A squirrel cage: a laminated
core with bars of aluminium or copper running through its slots, joined
at each end by a ring that shorts them all together. Taken out of the
iron, the bars and rings look like a hamster’s exercise wheel, hence the
name.
The rotor has no magnets, no brushes, no commutator and no electrical
connection of any kind. Its only wearing parts are its two bearings.
7.2 The rotating field
Three windings set 120° apart around the stator, carrying three
currents 120° apart in time, make a field of constant strength that
turns. Nothing moves to make it rotate. The rotation comes from the
supply itself, as each phase’s current peaks in turn. The field turns at
the synchronous speed
with
the supply frequency and
the number of poles. On 60 Hz a two-pole motor’s field turns at 3,600
rpm, a four-pole motor’s at 1,800, and a six-pole motor’s at 1,200.
This is what a brushless controller does with its six transistors
(§6): it makes a rotating field. The difference is that the brushless
controller times the field to the rotor’s position, which it reads from
Hall sensors. The induction motor’s field just turns at line frequency,
and the rotor follows as best it can.
7.3 Slip and torque
If the rotor turned at the synchronous speed, the field would stand
still relative to the bars, nothing would be induced in them, and there
would be no torque. So the rotor always runs a little slower than the
field. The difference is the slip:
The field sweeps past the bars at the slip speed. That induces
currents in them, the shorting rings let those currents flow, and the
currents in the field make torque. A larger load slows the rotor a
little more, which raises the slip, the induced current and the torque,
until the torque matches the load. At full load the slip is typically
2–5%, which is why a four-pole motor’s nameplate says about 1,725 rpm,
not 1,800.
A representative four-pole induction
motor (computed with the Kloss approximation: breakdown torque 2.8 times
full load, at 20% slip). On the 60 Hz line (black) it starts with about
full-load torque, rises to its breakdown torque, and settles on the
steep upper part of the curve, where speed barely changes with load. On
a VFD (gray), the same curve slides down with the synchronous
speed.
The curve has three landmarks:
Starting (locked-rotor) torque. With the rotor
standing still, the slip is 100%. The rotor currents are large but lag
badly, so the torque is only moderate. The current drawn is large too,
typically five to seven times full-load current. That is why big
induction motors dim the lights when they start.
Breakdown torque. The most the motor can give,
typically two to three times full load. Load it beyond this and the
motor slides down the curve and stalls.
The running region. Between full load and no load,
the curve is nearly flat. The speed falls only by the slip, a few
percent, so the motor behaves much like a shunt or PM motor (§4, §5). It
runs at nearly the same speed whatever the load.
7.4 Single-phase motors
A single winding on single-phase AC makes a field that pulses back
and forth but does not rotate. Such a motor, once turning, keeps turning
in either direction, but it has no starting torque at all. The cure is a
second, auxiliary winding, set 90° around the stator from the
main one, with its current shifted in time:
Split-phase. The auxiliary winding is wound with
finer wire, so its current leads the main winding’s a little.
Capacitor-start. A capacitor in series with the
auxiliary winding shifts its current much further, for a strong start. A
centrifugal switch cuts the auxiliary winding out once the motor is near
speed, which is the click heard as a table saw or a bench grinder spins
down.
Permanent split capacitor (PSC). A smaller
capacitor stays in circuit all the time. It starts weakly but runs
quietly, as in fans.
7.5 The clutch motor
The induction motor’s fixed speed is its strength in a factory and
its weakness at a sewing machine. The industrial clutch motor works
around it mechanically:
The motor runs all the time, at its fixed speed,
commonly about 1,725 or 3,450 rpm, driving a heavy flywheel. Clutch
motors are made both single-phase and three-phase.
The treadle works a clutch. Pressing it pushes a
cork-faced clutch plate against the flywheel, and the plate drives the
belt to the machine. Letting it up pulls the plate back against a
brake.
Speed control is slip in the clutch. A skilled
operator feathers it. A beginner gets a machine that lunges to full
speed, and a slow, steady crawl is hard for anyone.
The motor hums and draws power all day, sewing or not. The brushless
servo that replaced it (§6) is quiet, off between seams, holds a set
speed, and can stop with the needle up or down. The clutch motor’s speed
control was in the operator’s foot and in the cork, never in the
motor.
7.6 Speed control: the
variable-frequency drive
Since the speed follows the supply’s frequency, the way to change the
speed is to change the frequency. That is what a
variable-frequency drive (VFD) does:
A rectifier turns the mains into DC.
A DC bus, a bank of capacitors, smooths it.
A three-phase inverter, six transistors on PWM
(2–16 kHz), turns the DC back into three-phase AC at whatever frequency
is wanted.
The inverter is the same three-phase bridge that drives a brushless
motor (§6). The difference is in what times it. A brushless controller
follows the rotor’s position. A plain VFD just generates the frequency
it is told to.
Volts per hertz. A winding’s reactance falls with
frequency, so at a lower frequency the VFD must also lower the voltage,
in proportion, to keep the same flux. It holds
constant, which keeps the curve’s shape (the gray curves in the figure).
Near zero frequency the winding’s resistance takes over, and the drive
adds a little voltage boost to keep the torque up.
On the wire. The voltage at the motor is PWM pulses
at the switching frequency. The windings’ inductance smooths them, as in
a PM motor on PWM (§5), so the current is nearly a sine wave at the
output frequency.
Vector control. A drive that models the rotor’s
currents, often with an encoder, can control torque directly, much as
field-oriented control does for a brushless motor (§6). Such a drive can
hold full torque at a standstill. Without it, an induction motor cannot
hold a load still.
Why it is not the answer for the VSM Base. An
induction motor of a sewing motor’s power would be larger and heavier
than the brushless motors of §9 for the same output, and needs its VFD
for any speed control. A three-phase motor needs a three-phase drive.
The drives sold for single-phase motors are few, and a capacitor-start
motor, with its centrifugal switch, does not work on one at all. The
brushless motor does the same job, is smaller, and its controller
already knows where the rotor is.
8 The five motors side by side
Universal (series)
Wound-field DC (shunt)
Brushed PM DC
Brushless DC
Induction (squirrel cage)
Field made by
the load current
a separate field winding
magnets on the stator
magnets on the rotor
the stator winding; rotor currents induced
Commutation
commutator and brushes
commutator and brushes
commutator and brushes
transistors, from Hall sensors or an encoder
none: the supply rotates the field
Runs on
AC or DC, straight off the line
DC (rectified mains)
DC only
DC, through its controller only
AC, straight off the line; single-phase needs a start winding
Torque
:
high at a stall
rises with slip; moderate at a stall
Speed with load
falls steeply; races unloaded
nearly constant
nearly constant
nearly constant
nearly constant (a few % slip)
Speed control
voltage: rheostat, triac, variac
armature voltage (SCR drive), field weakening
PWM chopper or H-bridge
PWM on a three-phase bridge
frequency: a VFD
Speed feedback
needs a sensor; back-EMF only while current flows
armature voltage minus
back-EMF read directly
Hall sensors or encoder, built in
none needed on the line; an encoder for vector control
Braking, holding
coasts; cannot hold
dynamic braking
dynamic braking; holds with current
dynamic braking; holds with current
braking through a VFD; holds only with vector control
Wear
brushes, commutator
brushes, commutator
brushes, commutator
bearings only
bearings only
Electrical noise
brush sparking
brush sparking
brush sparking, PWM
PWM
none on the line; PWM on a VFD
Without electronics
works (with a pedal)
works on DC
works on DC (no speed control)
does not run
works, at one speed
Read across, the table says why the universal motor won the sewing
machine and why it is no longer the obvious choice.
Why it won. It is the only one that runs straight
off the wall with nothing but a pedal. It gives plenty of torque from a
standstill. It was cheap to make in 1920 and still is.
What it costs. Speed that depends on load, no
braking, no holding, and brushes. Its low-speed weakness is the subject
of the speed-control paper.
What changed. Strong magnets, cheap power
transistors and microcontrollers moved the hard part of the motor into
electronics. The brushless servo motor that replaced the clutch motor on
industrial machines is that change carried through. The question for the
VSM Base is whether the same change fits in a domestic machine’s motor
bracket.
9 Could a DC motor drive a vintage
machine?
9.1 What the job asks for
The universal motors that drive these machines set the target
(§2):
Power: an output of roughly 40–80 W.
Speed: about 6,000–7,000 rpm, so a running torque
near 0.06–0.11 N·m.
Reserve torque: a good deal more for a few moments,
for a heavy seam or a cold start in old oil.
The space. A Singer external motor’s body, as
measured for the test rig’s cradle, is about 51 × 64 mm in section and
114 mm long. It hangs from a bracket on the back of the machine and
drives the hand wheel through a belt.
Speed itself is negotiable: the belt’s ratio sets it. A motor with
twice the torque at half the speed, given a motor pulley twice the size,
turns the hand wheel just as well. Power is what has to
match.
9.2 Candidates
All of these run on 24 V DC. The first three fit well inside the
Singer motor’s envelope. The 57 mm brushless motor fits its length and
its 64 mm side, but is a little wider than its 51 mm side.
Specifications are as published by the sellers cited. Ratings for
generic motors vary between sellers and should be checked on the
bench.
Motor
Type
Size
Ratings at 24 V
Approx. price
“775” brushed motor
Brushed PM
42 mm dia. × 66.5 mm, 5 mm shaft
8,300 rpm no load; 0.082 N·m and 3.5 A at 7,354 rpm (about 63 W);
0.72 N·m at a stall (RobotShop)
$12
“50SRZ” brushed motor
Brushed PM
50 mm dia. × about 100 mm, 6 mm shaft
Sold in 2,000–8,000 rpm versions; power ratings vary by seller (eBay listing)
$55
42BLS02
Brushless, 8 poles
42 mm square flange × 61 mm
4,000 rpm; 0.125 N·m and 3.3 A rated (52.5 W); 0.38 N·m peak (Robu)
$25
57BLY100, with driver
Brushless, 4 poles, Hall sensors
57 mm dia. × 87 mm, 8 mm shaft
3,000 rpm; 0.33 N·m and 5 A rated (103 W) (RobotShop)
$77 with driver
There is also evidence from practice:
A Janome 415 conversion. Robert Scheibe replaced
the motor of a Janome 415 with a 24 V, 50 mm brushed motor of the SRZ
family (a Tsukasa TS-50SRZ-R: 5,000 rpm, 50 × 101.5 mm, rated 10.8–50.4
W). He drove it with an off-the-shelf PWM controller and an old guitar
volume pedal, through a toothed belt. He reports much better control at
low speed, especially when starting difficult seams (Scheibe).
A brushless attempt. A SimpleFOC forum thread on
brushless drives for sewing machines set targets at the hand wheel of
2,000 rpm and 0.7 N·m (about 140 W), and a smooth crawl below 8 rpm with
over 1 N·m. That calls for a substantial motor. The thread recommends
field-oriented control with a magnetic angle sensor, and reports no
finished build (SimpleFOC).
Industrial machines. Brushless servo motors are the
standard on industrial machines (§6), but at a few hundred watts they
are far too large for a domestic machine’s bracket.
So the answer to the first question is yes. Small,
cheap DC motors with a sewing motor’s power exist, and they fit its
space. A brushed 775 at 24 V is the simplest candidate. A 42 mm or 57 mm
brushless motor is the better one.
9.3 What the controller would
become
The MCU controller in the speed-control paper is built around a triac
because the motor runs on mains AC. A DC motor changes its power stage
completely and leaves its brain much as it is. A concept schematic of
the brushed version, a 775 on an off-the-shelf H-bridge, is in the
speed-control paper’s §11 (“A DC-motor variant”) and in the repository
at hardware/mcu-controller-dc/.
The supply becomes a certified 24 V adapter. A 24
V, 5 A (120 W) supply with a safety listing does all of the isolation
from the mains, inside a part that is bought, not built. The triac, the
opto-triac, the zero-cross detector, the relay and the mains wiring all
disappear. Every wire the builder touches is at 24 V. For a published
hobby design, that may be the largest single benefit.
The power stage becomes a bridge.
For a brushed motor: an H-bridge (four MOSFETs and a gate
driver, or an integrated H-bridge rated for 10 A or more).
For a brushless motor: a three-phase bridge (six MOSFETs
and a three-phase gate-driver IC, such as one of TI’s DRV83xx family),
with the motor’s three Hall sensors wired to three GPIOs.
Either way: a shunt resistor and amplifier measure the
current into one of the RP2350B’s spare analog inputs.
The firmware gains a current loop. The current loop
runs at the PWM rate (20 kHz or more), with the speed loop around it,
and a position loop around that for needle stops. All three fit
comfortably in an RP2350. The pedal, knee and heel inputs, the
speed-limit knob and the stop teaching carry over unchanged.
Needle stops become exact. A triac controller can
only cut power early and let the machine coast to its stop, learning the
coast as it goes (speed-control paper, §11). A bridge can brake the
motor electrically, then hold the needle where it is with current: true
servo positioning.
The encoder may become optional. A brushless
motor’s Hall sensors report 6 states per pole pair per turn, 24 per turn
for the 8-pole 42BLS02. Say the motor turns three to five times per
stitch, as a 6,000 rpm motor driving 1,200–2,000 stitches a minute must.
That is 70–120 positions per stitch, comparable to the optical encoder’s
144. Round and V-belts slip a little, so the hand-wheel index would
still be needed to re-anchor the count once a stitch, but the expensive
part of the encoder would not.
9.4 What it costs
Authenticity. For a collector, a vintage Singer
with a modern motor is a different object. A bolt-on motor on an adapter
bracket, with the original motor kept, keeps the change reversible.
Machines with a potted motor built into the casting (the 15-91, the 201)
are out of reach.
Mechanics.
Fitting: an adapter to the machine’s motor bracket, and a
pulley sized to the new motor’s speed for the machine’s belt.
Bench testing: the test rig’s cradle could take a 42 mm
motor with a printed adapter.
A motor that needs its controller. The original
motor runs from any pedal; a DC motor runs only from its electronics. A
brushless motor does not run at all without its controller.
Brushes, still, for a brushed motor. A 775 costs
little enough to treat as a wear part. A brushless motor has no
brushes.
Electrical noise of a new kind. PWM at 20 kHz, with
short, twisted motor leads and a filtered supply, replaces the brush
hash of the universal motor.
9.5 Recommendation
A DC-motor variant of the controller is worth keeping open, in the
same spirit as the parked ideas in the speed-control paper’s §11.
Start on the bench, not in the machine.
Measure the candidate: run a 775 and a 42BLS02 from a 24 V
supply with off-the-shelf drivers, and measure speed, current and torque
with the rig’s Prony brake, as for the universal motor.
Try it on a machine: put the better one on a machine on a
temporary bracket.
If that goes well, design a low-voltage power stage for the
RP2350B board, H-bridge first, three-phase second. Keep the
triac design for machines that keep their original motors.
The two would share their sensors, their pedals and most of their
firmware.
A. Hughes and B. Drury, Electric Motors and Drives:
Fundamentals, Types and Applications, 5th ed., Newnes, 2019.
A. E. Fitzgerald, C. Kingsley and S. D. Umans, Electric
Machinery, 7th ed., McGraw-Hill, 2014.
The motor model, its equations and the speed-control methods are
those of the companion paper, Speed Control of Universal Motors in
Sewing Machines, and its sources. The simulations and drawings are
generated by tools/make_motor_waveforms.py and
tools/make_motor_drawings.py in the project repository. The
cordless-tool spectra in §5 are Robert Drzyzgula’s measurements, phone
screenshots kept in docs/images/motors/measured/, read back
and replotted by tools/make_tool_spectra.py.
The construction and theory of the motors themselves, §2 and §4–§7,
are standard textbook material, given here without citation line by
line. Hughes and Drury is the readable practitioner’s account, drives
included; Fitzgerald, Kingsley and Umans is the standard engineering
text.