Ed Nisley's Blog: Shop notes, electronics, firmware, machinery, 3D printing, laser cuttery, and curiosities. Contents: 100% human thinking, 0% AI slop.
The first sensor bracket came from the scrap pile, but showed that it would produce 1/rev pulses from the motor shaft pulley. The positioning wasn’t quite right, so I made another bracket that put the TCRT5000 sensor at right angles to the pulley:
TCTR5000 Motor RPM Sensor – end view
All of the sensors have a rakish tilt over their PCB, so at some point I must resolder them:
TCTR5000 Motor RPM Sensor – side view
It might not matter, as the phototransistor on the left peers directly at the pulley, with the LED on the right acting as a floodlight.
“Made another bracket” sounds like the metal sprang fully formed from the concept. Herewith, the early contestants atop a sketch and the flat layout for The Ultimate Bracket:
Motor RPM Sensor Brackets
A closer look at that final dimension sketch, because I’ll need it again:
RPM Bracket Dimensions
The vertical size of the center section (12 mm) sets the perpendicular distance of the sensor from the shaft. The horizontal size (14 mm) controls the pulley-to-sensor spacing.
The horizontal distance from the center section to the hole on the right (10 mm) adjusts the sensor spacing parallel to the shaft.
I cut the overall rectangle with tin snips, drilled & cleaned the holes, applied a nibbling tool to the details, trimmed the corners, filed off sharp edges & spines, and it was all good.
The doodles for the first few attempts, as I don’t want to repeat those mistakes:
Bracket Doodles
All in all, a few more hours of Quality Shop Time than I expected…
A quick-and-dirty bracket (made from a leftover strip in the pile of chassis clips) affixed an IR reflective sensor (based on the ubiquitous TCRT5000 module) to the sewing machine motor:
TCRT5000 sensor on motor
That’s scribbling black Sharpie around the retroreflective tape for the laser tachometer, which worked just about as poorly as you’d expect. Retroreflective tape, by definition, reflects the light directly back at the LED, but in this case you want it bounced to the photosensor.
An IR view shows the geometry and highlights the LED:
TCRT5000 sensor – IR view
The TCRT5000 datasheet suggests that the peak operating distance is 2.5 mm, roughly attained by tinkering with the bracket. The datasheet graph shows that anything between 1 and 5 mm should be just fine:
IR Reflective Sensor module – TCRT5000 – response vs distance
Apply stainless steel tape around half the circumference
Burnish flat
Which looks pretty good:
Kenmore 158 motor pulley – black-silver
The stainless tape butts up against the setscrew:
Kenmore 158 motor pulley – black-silver at setscrew
Adjusting the sensitivity midway between the point where the output is low (OFF) over the black and high (ON) over the tape seems reasonable.
Running at the slowest possible speed produces this pulse train:
Motor sense – min speed
The motor at 19 rev/s = 1140 RPM corresponds to about 2 rev/s of the sewing machine shaft= 2 stitch/s. Slower than, that, the pedal won’t go in simple open-loop mode.
The setscrew causes those “glitches” on the rising edge. They look like this at a faster sweep:
Motor sense – min speed – setscrew
At maximum speed, the setscrew doesn’t show up:
Motor sense – max speed
The motor at 174 rev/s = 10440 RPM would do 1000 stitch/s, but that’s just crazy talk: it runs at that speed with the handwheel clutch disengaged and the motor driving only the bobbin winder. I was holding the machine down with the shaft engaged and all the gimcrackery flailing around during that shot.
The sensor board may have an internal glitch filter, but it’s hard to say: the eBay description has broken links to the circuit documentation.
I could grind the setscrew flush with the pulley OD and cover it with tape, but that seems unreasonable. Fixing the glitch in firmware shouldn’t be too difficult: ignore a rising edge that occurs less than, say, 1/4 of the previous period following the previous edge.
Perhaps buffing half the pulley’s circumference to a reasonable shine (minus the bluing) would eliminate the need for the stainless steel tape.
Iterating the bluing operation / scrubbing with steel wool should produce a darker black, although two passes yields a nice flat black.
The relay at the top connects the AC hot line to the rest of the circuitry, with a feeble red LED to show when it’s live:
AC Power Interface
The driver lives on the Low Voltage Interface board:
LV Power Interface – AC Relay driver
The GX270’s front-panel hard drive LED now serves to indicate when the AC power goes live.
I’d originally intended to turn the AC on when the Arduino gains control, but after seeing those pictures, I think it’ll remain disabled unless there’s a call for motor motion.
The interlock switch closes when the case opens, grounding the transistor base and disconnecting the AC power.
Of course, you can cheat by simply unplugging the switch, so it’s not failsafe. If you want failsafe, you need a normally closed switch in series with the collector; that’s not what Dell used as a chassis intrusion switch. That’s my story and I’m sticking with it.
Although I plan to servo the motor speed to the pedal position, a quick open-loop test seems in order. The motor requires nigh onto half an amp before it can spin the sewing machine shaft, so this chunk of Arduino code scales-and-offsets ten bits of pedal position voltage into twelve bits of DAC output that produce a corresponding current limit for the motor winding:
Putting that in the Arduino’s main loop and holding the pedal down produces this pleasant result:
Current Sense Amp vs Tek – 200 mA-div
The current sense amp output in the top trace is scaled at 525 mA/V = 525 mA/div and the bottom trace is from the Tek current probe at 200 mA/div. Fiddling with the scope’s gain & offset exactly overlays the two traces and they remain overlaid through the full pedal travel, so the ferrite toroid isn’t saturating and the output remains nicely linear.
The flat tops in that picture show the ET227 transistor limiting the motor current to 600 mA, exactly the way it should.
Of course, the LM324 has a GBW = 1 MHz and, with a gain of three, a bandwidth of barely 300 kHz, so there’s a distinct lack of fuzz on that trace compared to the Tek probe’s 10 MHz bandwidth.
It’s easy to hold the sewing machine at a constant speed with a constant load, but touching the handwheel stalls the motor at a constant pedal position. Similarly, releasing the handwheel causes a runaway, unless I let up on the pedal fairly quickly.
Setting the Tek probe to 500 mA/div and triggering on a somewhat higher current while stomping on the pedal and grabbing the sewing machine’s handwheel shows the current increasing with the motor under heavier load:
Model 158 – Current sense vs Tek 500 mA-div
The current limit reaches just under 2 A, over on the right side, for both traces.
So the hardware works pretty much the way it should.
Given that the motor current amounts to maybe 3 A, absolute maximum, with a locked rotor, those skinny wires on the slit ferrite toroid won’t pose a problem:
HV Interface board – detail
I really like how that 3D printed armor worked out, even if I still haven’t poured any goop around the windings to lock them down; they’re held in by raw faith and friction.
The current sense circuitry appears along the bottom of the AC Power Interface schematic:
AC Power Interface
The differential amplifier lives on the Low Voltage Interface board, forming the clump of parts just in front of the LM324 op amp on the left:
Low Voltage Interface Board – detail
Which has the usual handful of components required to get anything done in the analog realm:
Current Sense Amp – schematic
The power supplies come directly from the ATX connector. I’m ignoring the whole decoupling issue, because the supplies have essentially no load (and it’s all DC, anyway).
The trim pot sets the offset voltage to bring the Hall effect sensors’s VCC/2 offset down close to zero; the 100 mV figure is nominal, not actual, but should be a bit over 0 V to allow for a wee bit o’ drift. This time around, I’ll measure and subtract the actual offset, rather than (try to) auto-zero it.
The voltage gain runs just under 3, set by 1% resistors from the heap. The overall gain works out to about 1.9 V/A or 525 mA/V, setting the high end at 5 V to a scant 2.6 A. Subject to actual calibration with more attention to detail, that’s close enough; we’re more interested in the around-an-amp range where the motor operates under nominal load.
The nose-to-tail Schottky diodes clamp the op amp output to avoid annoying the Arduino’s ADC input. It has protection circuitry, too, but there’s no point in stressing it.
Because the ET227 transistor operates at power line voltages through a full wave rectifier, the base drive circuit requires an optoisolator. The ET227 is a low-gain device with hFE < 10, so it takes about 100 mA of base drive to control an amp of motor current, soooo the optoisolator needs a current amplifier.
I used an MJE2955T PNP transistor, with the emitter powered from an isolated +5 V supply to let the optoisolator pull current from the base. You could use an NPN transistor as a Darlington amp, but wiring the collectors together means the driver dissipates way too much power; the PNP seemed all-around easier.
That circuitry sprawls across the middle of the schematic:
AC Power Interface
The ET227 base runs at about 900 mV, so the MJE2955 PNP transistor will dissipate half a watt and needs a little heatsink, seen over on the right (with the hulking ET227 heatsink at the edge):
HV Interface board – detail
With all those parts safely secured, I ran some end-to-end current measurements from the optoisolator’s LED to the ET227’s collector current, with a safe 10 VDC applied to the collector:
ET227 – base drive – optoisolators
It’s worth noting that the two optoisolators have different pinouts. The DIP socket has wiring for both of ’em, so I could swap the two without rewiring the board. No, I didn’t notice that the first time around.
The curves are nicely linear above 250 mA, which is about what you’d expect for bipolar transistors driven from a current source. Below that, the current into the 13 Ω base-emitter resistor starts to overwhelm the actual base junction current and makes the curves all bendy. Given that the motor doesn’t start spinning the sewing machine with less than half an amp, that region doesn’t matter.
It’s also worth noting that the ET227 normally sees tens of amps (!) into the base terminal to control up to 200 A pulsed collector current with up to 1 kV collector voltage. That puppy loafs along here…
The ratio between the isolator gains doesn’t match the ratio between the spec sheet values, so maybe they’re mismarked or I (once again) have an outlier. In any event, there’s no point in getting too fussy, because the transistor gains depend strongly on temperature. I picked the lower-gain SFH6106-2 for more headroom, but it probably doesn’t make much difference.
The voltage-to-current circuitry driving the optoisolator’s LED lives on the Low Voltage Interface board, with the MCP4725 DAC breakout board above the Arduino Pro Mini and the rest just beyond the LM324 op amp over on the left:
Low Voltage Interface Board – detail
There’s nothing much to it:
Current Control DAC and Driver – schematic
I finally broke down and got some of Adafruit’s nice MCP4725 I2C DAC breakout boards: 12 bits, rail-to-rail output, no PWM ripple. What’s not to like?
R409 scales the gain so that +5 V tops out around 1.5 mA, which should deliver a collector current around 3 A: far more than seems absolutely necessary. R408 lets the op amp develop some voltage while trickling a few dozen microamps into the 2N3904’s base; the hFE runs around 50, so the error due to base current amounts to maybe 2% and, remember, the final current depends on the temperature anyway.
The Smell of Molten Projects in the Morning 