Radiant Tech Note
- James Wiebe
- 12 minutes ago
- 5 min read
Radiant Tech Note: One Node, Three Masters – Sharing a Gyro Line with an Altimeter and an MCU
How we taught one humble analog node to be a gyro output, an MSP430 ADC input, and an I²C clock line… all at once.
1. Background: When a “Simple” Turn Coordinator Grows Up
Radiant’s SafeTurn started life as a portable digital turn coordinator with an inclinometer (“the ball”). Simple idea:
Solid-state yaw gyro
Nice LCD
Battery power
No vacuum, no spinning iron, no drama.
Then reality hit: pilots loved it, and the obvious next step was altitude and VSI in the same box.
We already had that technology in SafePanel (our mini glass panel) – a MPL3115 pressure sensor feeding altitude and VSI logic, running on a tight little MSP430G2553.
So the mandate became:
“Make SafeTurn do what SafePanel does, on the same 20-pin MSP430, on a board that’s already flying, and don’t break the gyro.”
That’s where the fun started.
2. The Problem: No Pins Left, But We Need I²C
SafeTurn uses the MSP430G2553’s 20 pins very aggressively:
LCD data + control
Buttons / analog switch input
PWM dimming
Yaw gyro analog input
Power / housekeeping
On the SafePanel board, the MPL3115 pressure sensor had its own clean I²C pins. On SafeTurn… it didn’t. There was no spare pin sitting around saying “Hi, I’m SCL.”
Instead, we had a single, very important analog node:
Gyro analog output
Through a 1.13 kΩ resistor
Into the MSP430 ADC (P1.0 / A0)
On SafePanel we also had a 0.47 µF cap on that node, forming a nice ~300 Hz low-pass:
R = 1.13 kΩ, C = 0.47 µF
( f_c \approx 300 \text{ Hz} )
Beautiful for smoothing yaw. Terrible if you now want that node to double as a digital clock line.
But on SafeTurn, that’s exactly what we decided to do:
The gyro analog node becomes MPL3115 SCL.
One node. Three masters:
Gyro
MSP430 ADC
MPL3115 SCL
3. First Step: Making an Analog Node Talk Digital
On the firmware side, we taught the MSP430 to “shape-shift” that pin:
Most of the time:
P1.0 is analog input (ADC10 enabled), reading gyro.
During MPL transactions:
We temporarily clear ADC10’s analog enable for A0,
Reconfigure P1.0 as a digital output,
Bit-bang SCL for the MPL3115,
Then put it back to high-Z + analog when we’re done.
That part actually works surprisingly well. But the analog world noticed:
With the MPL3115 hanging on the node,
And the original 0.47 µF cap removed (it was too much load for SCL),
The gyro “feel” changed.
We gained Alt/VSI, but lost some of that analog filtering that made the ball and yaw so clean and calm.
Time to get clever with passives.
4. The Hardware Hack: A 10 kΩ Lifeline and a New RC
We ended up with this topology:
Gyro → 1.13 kΩ → “node”
Node goes to:
MSP430 P1.0 (ADC input / SCL driver)
0.1 µF to ground
10 kΩ → MPL3115 SCL (pin 8)
Let’s unpack what that does.
4.1 Re-introducing Analog Filtering
The new RC at the node is:
R = 1.13 kΩ
C = 0.1 µF
Low-pass corner:
( \tau = R \cdot C \approx 0.113 \text{ ms} )
( f_c \approx 1.4 \text{ kHz} )
So compared to SafePanel’s original 300 Hz filter, we now have a gentler ~1.4 kHz filter:
Still far above any real yaw motion (0–2 Hz)
Very effective at trimming high-frequency hash, EMI, and “this cannot be real yaw” components
Light enough to allow SCL edges to move the node without being smothered
When we experimented with going back to 0.47 µF on this shared node, SCL stopped behaving. Somewhere between 0.1 µF and 0.47 µF, the bit-banged I²C just gave up – not enough edge speed, too much capacitive load.
0.1 µF turned out to be the sweet spot.
4.2 Decoupling the MPL3115: The 10 kΩ Resistor
The 10 kΩ between node and MPL3115 SCL does two important things:
Reduces MPL’s influence on the nodeAny leakage, ESD clamp behavior, or internal quirks on the MPL’s SCL pin are now seen through 10 kΩ, so their DC and AC impact on the gyro node is tiny.
Softens any fast junk coming from the sensor sideEven though SCL is an input, transients and internal switching events at the sensor are now loosely coupled into our analog world.
From the digital vantage point, 10 kΩ into the MPL’s SCL input is no problem at all; the MPL pin capacitance is tiny, so its own RC is way up in the MHz range.
The big RC that shapes SCL is the 0.1 µF at the node – and we verified in the lab that our bit-bang timing has plenty of margin there.
5. Scheduling Around the Noise: Firmware as Part of the Filter
Analog tricks are only half the story. The other half is when we do things.
The MSP430 firmware now:
Ensures P1.0 is analog input when the ADC is sampling yaw.
Only grabs P1.0 as a digital SCL line for short bursts when we talk to the MPL3115 (e.g., in a 1 Hz Alt/VSI task).
Avoids painting critical UI or sampling gyro right in the middle of SCL bursts.
Applies digital low-pass filtering on yaw over timescales much longer than any SCL wiggle.
So the final “filter” is really multi-layer:
Gyro’s internal analog bandwidth
External RC at the node (1.13 kΩ / 0.1 µF)
ADC sampling scheme
Digital smoothing in the MSP430
The result in the cockpit:
Yaw and the “ball” remain calm and natural.
The MPL3115 happily talks I²C over that same node.
SafeTurn gains Alt/VSI without needing a new PCB spin or a bigger MCU.
6. AI on the Bench: Not Just for Buzzwords
One fun part of this project: we didn’t reason this out in a vacuum.
Alongside the usual suspects on the bench – scope, meter, soldering iron – we also leaned on an AI-assisted design tool for “what if?” exploration:
“What happens to the time constant if we move the cap here instead of there?”
“If we add 10 kΩ between the node and SCL, will we still make timing at the sensor?”
“What’s the cutoff difference between 0.1 µF and 0.47 µF with this series resistor?”
The AI didn’t replace bench work (we still had to solder, scope, and fly it), but it did compress the ideation and back-of-the-envelope math phase dramatically. In minutes we could evaluate several candidate topologies and convergence points, then focus lab time on promising ones.
In other words:
The oscilloscope still has the final vote,but the AI now helps pick which experiments are worth running.
7. Why This Matters (Beyond One Product)
For electronic and avionics enthusiasts, this is the kind of engineering that rarely shows up in textbooks:
One physical node doing double duty (or triple) across analog and digital domains
Passives used as “relationship managers” between devices, not just static filters
Firmware that actively participates in system-level behavior:
Reclaiming pins on the fly
Scheduling around noise
Using digital filtering to complement analog choices
For Radiant, this work:
Let SafeTurn grow into a credible portable backup Alt/VSI device without a board respin.
Kept the MSP430G2553 front and center – a tiny MCU doing a lot of real work.
Fed directly into our code base for RTCC (panel-mount turn coordinator) and SafePanel (mini glass panel), giving all three products the benefit of the same hard-won lessons on that one little node.
We like that. It’s efficient, it’s elegant, and it’s the kind of engineering we enjoy: taking small, constrained hardware and making it do big things.
Radiant Technology Company – designing and building compact avionics for real cockpits, with real constraints, and real pilots in mind.
Comments