Divider design for a 3.3 V ADC, the Teensy 3.0 pad map, expected readings, and known issues.
The build as wired. The D-sub breaks out to screw terminals. Pins 1, 2 and 3 are Y, X and Vref; pins 7 and 8 are +12 V and ground. The 33 Ω sits in the supply leg as a soft fuse, the 100 kΩ / 68 kΩ pairs divide each analogue output down to something the 3.3 V ADC can read, and the taps land on A7, A8 and A9.
diagram.svg alongside it is the editable Inkscape source for those annotations.
joystick pin 7 ── [optional: 33 Ω] ── +12 V (bench supply)
joystick pin 8 ────────────────────── GND (also tied to Teensy GND!)
joystick pin 1 (Y) ──┬── 100 kΩ ──┬──► A7 (pad 21)
│ │
│ 68 kΩ
│ │
│ GND
joystick pin 2 (X) ──┬── 100 kΩ ──┬──► A8 (pad 22)
│ │
│ 68 kΩ
│ │
│ GND
joystick pin 3 (V) ──┬── 100 kΩ ──┬──► A9 (pad 23)
│ │
│ 68 kΩ
│ │
│ GND
3.5mm jack tip ─────────────────► D2 (INPUT_PULLUP, pad 2)
3.5mm jack sleeve ─────────────────► GND
Critical: bench supply GND, joystick pin 8, and Teensy GND must all share a single node. Verify with DMM continuity before powering up.
Official pinout cards: https://www.pjrc.com/teensy/pinout.html (Teensy 3.0 is
card5a / card5b). Not redistributed here; download them separately if you
want them locally.
Front side, component side up, USB at the left. ● marks the pads this project uses:
Vin AGND 3.3V 23 22 21 20 19 18 17 16 15 14 13
○ ● ○ ● ● ● ○ ○ ○ ○ ○ ○ ○ ○
(A9) (A8) (A7) (A6) (A5) (A4) (A3) (A2) (A1) (A0) (LED)
Vref X Y
┌───────────────────────────────────────────────────────────────────────────────┐
USB T E E N S Y 3 . 0 Reset ○
│ MK20DX128VLH5 Program ○
│ GND ○
│ 3.3V ○
└───────────────────────────────────────────────────────────────────────────────┘
● ○ ○ ● ○ ○ ○ ○ ○ ○ ○ ○ ○ ○
GND 0 1 2 3 4 5 6 7 8 9 10 11 12
Analog inputs are not separate pads. A0 to A9 are digital pads 14 to 23
(confirmed in the installed core, cores/teensy3/pins_arduino.h:165:
analogInputToDigitalPin(p) = p + 14). The silkscreen says 14, the sketch says A0.
| Signal | Sketch pin | Teensy pad | Where |
|---|---|---|---|
| Y (joystick pin 1) via divider | A7 |
21 | top row, 6th from left |
| X (joystick pin 2) via divider | A8 |
22 | top row, 5th from left |
| Vref (joystick pin 3) via divider | A9 |
23 | top row, 4th from left |
| MODE (joystick pin 6) | D2 |
2 | bottom row, 4th from left |
| Divider bottom legs, joystick pin 8 | GND | GND or AGND | bottom-left / right edge, or top row 2nd from left |
A7/A8/A9 are three adjacent pads clustered at the left end of the top row. Divider bottom legs return to the shared digital GND, which is correct. AGND is tied to GND internally on Teensy 3.x, so the choice is a noise preference rather than a wiring requirement. Move them to AGND only if the readings turn out noisy.
Note that the 3.3V pad sits two positions left of pad 23, so take care not to
bridge a divider tap into it.
Alternate functions on these pads are all opt-in and unused here: touch sense on 22/23
(needs touchRead()), SPI chip-select, and RX1 on pad 21 as an alternate Serial1
pin. Default Serial1 is pads 0/1 and this sketch uses USB serial, so nothing collides.
Power warnings specific to this board:
- Vin accepts 3.7–5.5 V only. Never connect the 12 V joystick supply to it. Power the Teensy from USB, which is needed for the serial link in any case. The two supplies stay separate and share only ground.
3.3Vis an output, 100 mA max. It supplies the dividers' reference, not the joystick.- Teensy 3.0 is not 5 V tolerant on any pin (unlike 3.1/3.2, which tolerate 5 V on digital pins). This is why every joystick line needs a divider, and why joystick pin 6 must be confirmed as a dry contact before it touches pad 2.
- AREF is an inner pad on the back side, not on the edge. Leave it alone; the sketch
uses
analogReference(DEFAULT)= the 3.3 V rail.
The divider does two jobs.
1. Fit the swing under 3.3 V. Teensy 3.0 is a 3.3 V part and its analog pins are not 5 V tolerant, so the joystick's 4.5–6.8 V output has to be scaled by roughly 0.4. 100k/68k gives 0.400: peak deflection lands at 2.72 V at the current ~11 V supply.
Headroom matters here because Vref tracks about half the supply voltage. Short the 33 Ω soft-fuse and the joystick sees a full 12 V, pushing Vref to ~6.1 V and peak deflection to ~7.35 V, still only 2.95 V at the ADC. A 100k/82k divider (ratio 0.446) gives ~7 % more resolution but hits 3.26 V under that same condition, right at the rail. Only use 82k if the supply stays pinned at 11 V.
2. Keep the load high enough to stay symmetric. The joystick has a 1.8 kΩ series resistor on X/Y but only ~470 Ω on Vref. That mismatch attenuates X/Y slightly more than Vref, biasing dx = X − Vref. At 170 kΩ total load the error is ~0.7 % (about 6 counts of static offset, which the sketch's startup calibration removes anyway). A 10k/6.8k version (same ratio, 17 kΩ load) would push it to ~7 %, well past what calibration should have to absorb. Tradeoff of the high-impedance approach: more noise-susceptible, so use short wires and a ground-plane breadboard.
With Teensy 3.0 (3.3 V reference, 10-bit analogRead), 100k/68k dividers, joystick running at ~11 V supply:
- Vref (A2): ~700 counts (2.26 V at ADC), and should be steady.
- X / Y at rest (A0, A1): ~695 counts, i.e.
dx0/dy0calibrate out at about −6. - Full deflection:
dxordyswings to ±143. - Switch (D2):
0open,1shorted.
Teensy 3.0 supports analogReadResolution(12) for 4× the resolution (Vref ≈ 2800, swing ≈ ±572). It is one line in setup(), commented out in the sketch. The voltages don't change, only the scale.
- High ADC source impedance. The 100k/68k dividers
present ~40.5 kΩ, well above the ~10 kΩ the Teensy 3.0 sample-and-hold wants.
Untreated, each conversion retains part of the previous channel's voltage, so
deflecting the stick visibly drags the Vref reading and the axes look coupled.
Fixed in firmware with
analogReadAveraging(32)plus a discarded read per channel (readSettled()), which cut jitter from ±8 to ±4 counts and removed the correlation. A 100 nF cap from each tap to GND is the proper hardware cure if it ever resurfaces. - Symptoms seen during bring-up and what they meant: all three channels pegged at 1023 = dividers not dividing (tap on the wrong side of the 100 kΩ, or the 68 kΩ legs not reaching the shared ground node). All three reading ~2/3 of expected = joystick supply sagging, check for a brushed-together lead on the breadboard.
- The PAB number
1822988is the most specific identifier for this exact unit. If pinout differs from documented norms in future testing, check Permobil's part registry for that PAB. - The joystick is fully functional and unmodified as of last session. Do not open the housing unless further investigation requires it.
- Power supply was a bench unit set to 12 V with a 33 Ω series resistor in the +12V lead as a soft-fault limiter. The resistor is optional once the unit is verified non-faulty, but it does no harm.
- NEVER apply >12 V or reverse polarity. The joystick has internal regulation but the input stage limits are unknown.
- NEVER apply external voltage to pins 1, 2, or 3. These are joystick outputs. Only the Arduino's high-impedance ADC pin (via divider) should connect to them.
- The 1.8 kΩ series resistors inside the joystick on the X/Y outputs provide modest short-circuit protection on those pins, but Vref (470 Ω series in Bob's schematic, may be similar inside the real joystick) is less protected, so do not short pin 3 to anything.