Looks like he’s going live tonight with a very fun and very cool project to get a 100W antennae going. This was done a year ago by the second video, but always nice to see a second take on this ancient technology making a comeback just in time for distributed internet (p2p) radio communications!
Coming up to bat - Ground-to-ground (NO LICENSE!) communications, a healthier alternative for the family jewels, extended range, reliable (weather don’t matter) p2p internet! Stay tuned!
►MATERIALS
Transmitter:
ZVS Driver /or/ 555 Driver
Flyback Transformer /or/ Induction Coil
(2x) Brass contacts for Spark Gap
Monopole Antenna
Coherer
(2x) 1/4" Machine Bolts
Iron /or/ Any Ferrous Metal Powder
Plastic /or/ Glass Tube
Receiver:
Oscilloscope
(2x) LM741 Op-Amp
Diode
Various Resistors and Capacitors
LED
(Optional) Relay
https://www.amazon.com/dp/B0C6M29YG8/ref=twister_B0GVRTZ2DW
https://zchg.org/hott/late%20episodes/898%20-%2027-Jun-96%20-%20Capitailism%20vs.%20Nationalism.mp3
Lawfully. The Pen is Mightier.
How to read it
Build order
Things I haven’t verified
Keep your hands off the plug and HV lead while it’s powered.
Shopping list
| Qty | Part | Where it goes |
|---|---|---|
| 1 | Spectra Premium C-615 ignition coil (or Standard UC-12, Echlin IC10, BWD E70) | Section 2 |
| 1 | TIP3055, with an insulating pad and a bolt-on heatsink | Q1 |
| 2 | 2N3904 | Q2, Q3 |
| 3 | 1N4148 | D1, D2, D3 |
| 1 | 100 Ω, ½ W | R1 |
| 1 | 10 Ω | R2 |
| 1 | 10 kΩ | R3 |
| 2 | 4.7 kΩ | mic bias, preamp load |
| 1 | 1 MΩ | preamp feedback |
| 1 | 100 kΩ | bleed |
| 3 | 1 µF film or ceramic | C1, C2, C3 |
| 1 | Electret mic capsule | Section 1 |
| 1 | Small ferrite toroid | Section 2 |
| 1 | Power switch, plus a 5 V supply that can give at least 2 A | Power |
| 1 | PC817, 330 Ω, 1 kΩ, and a Pi with its own USB-C supply | Section 3 only |
You already have the spark plug, plug wire and hookup wire. Get three colors of wire for the toroid windings.
Assembly order
Tell me which step you’re on and what you see, and I’ll help from there.
Build order
Checks
That’s the full build. Do it in this order so you can tell which part is misbehaving.
Parts to add
Build order
The drawing shows these stages:
A few things to know before you build it:
M1 spins the induction motor above its synchronous speed, with a small capacitor bank on its terminals. That makes it a self-excited induction generator. The rotor’s residual magnetism seeds the field, and the capacitors let the AC voltage build up. V3 then rectifies one winding (U to W) into B+. There are no contacts to burn, and the frequency comes from rotor speed, so the saturation problem I hit with the ferrite toroid goes away.
Here is how the drawing works, followed by the points that matter for building it.
Signal path. M1’s supply runs through a speed network with a key. The key shorts out part of that resistance, which shifts the speed and therefore IM1’s frequency, so keying gives FSK. U, V and W go through the 20-turn balun and are tied to rod array A. The star point N goes to rod array B, which is also the black lead. Current runs through the soil between the two arrays. The stage-1 tube preamp stays connected across the same rods at high impedance, so it listens while the node transmits. The node on the other end is a mirror image.
Because the balun forces U, V and W to carry equal current, the output is the zero-sequence part of the three phases, and that is what carries the signal. Cancellation by itself doesn’t raise efficiency. What couples into the soil is the small residual, so the design should create that residual on purpose.
Resonance is a tuning job in the hardware. The balun’s inductance, the electrode and soil resistance, and a series capacitor form a resonant loop at f = 1/(2π√LC). For example, a 10 mH balun needs about 1 mF to resonate at 50 Hz. Paralleled motor-run capacitors are the natural analog part for that. Your identity N = Ω·C²/(m·s) is dimensionally correct for the ohm, but it’s a units identity and doesn’t produce resonance on its own. The script’s Ω (m²/s⁷) is also a different quantity from the ohm. Separately, emergentC1.py multiplies the constants by an assumed 1.05, so its 5% change is an input, not a result.
hdgl_tte_protocol.ino is the digital version of this link. An AD9833 DDS does FSK at ±0.1 Hz around 3.915 Hz with Hamming-coded 4-bit frames. The receive side is a stub, and its comments say “Simulate reception”.
That raises a design issue for the motor version. At a 50 Hz carrier, ±0.1 Hz is about 0.2% of speed, roughly 3 rpm on a 4-pole machine. A 5V DC motor behind a series resistor won’t hold that, so M1 needs tach feedback or a mechanical governor. A 3.9 Hz carrier is worse. An induction generator at that frequency barely self-excites, because voltage scales with frequency and the excitation capacitance grows as 1/f².
Should the carrier be the motor’s own ~50 Hz, with the Schumann sub-harmonics as the modulation as in our waterfall sim, or the 3.9 Hz range from the Arduino?
The numbers below are hand arithmetic, not a simulation, and every part value is a starting point to tune on the bench.
| Item | Value | Why |
|---|---|---|
| Carrier | about 40 Hz | It sits 20 Hz from 60 Hz mains, so the loop’s Q rejects hum by roughly 14 dB. |
| IM1 speed | about 1250 rpm | A 4-pole machine generates slightly below its shaft frequency. |
| Excitation | 3 × 20 µF star | C = I_m/(ωV) with I_m ≈ 0.4 A at 89 V per phase. |
| Extractor | 120 V : 6.3 V ×3 | Balanced gives 0 V. Key flipped gives about 9.4 V (−2 × 4.7 V). |
| Soil loop | L ≈ 0.5 H, C_s ≈ 33 µF | Resonates at 39 Hz. With about 34 Ω total it carries about 0.28 A (about 2.7 W) at Q ≈ 4. |
| Rod arrays | 4 × 8 ft each, about 10 m apart | About 12 Ω per array in 100 Ω·m soil. |
| B+ | T_p, 6X4, 22 µF, 7.5 kΩ, VR90 | About 190 V unregulated, 90 V regulated. The 1S5 plate limit is 90 V. |
| Receive gain | Q × T_r (1:19) × V1 (about 12) | About 0.9 mV of field at 100 m becomes about 0.7 V at V2’s grid. |
M1 needs a 12 V supply rather than 5V. Spinning the generator costs roughly 25 to 40 W, which is 8 A or more at 5V.
First light
Weak points
I dropped the spark and ignition-coil stage for this node. The second-node cooperation (each end modulating the other) would be the next step: V5’s output closes a keyer contact so a node can repeat what it hears.
This stage lets a node repeat what it hears. V5 is wired as a relay tube, so its starter fires when a signal is detected, and the current through the relay coil pulls the relay. The same relay’s contacts flip the extractor key and mute the receiver, so the node transmits what it just received.
How it works. The V5 anode runs off the generator’s AC through the relay coil, so the tube shuts off every cycle and can’t latch. Conduction follows the starter, so V5 conducts only while a signal is present. The 4.7 µF capacitor smooths the half-wave pulses so the plate relay holds. RY1’s DPDT contacts flip Tc in the extractor, which makes the node transmit. Its mute contact shorts T_r’s primary while RY1 is pulled, so the node doesn’t hear itself.
The two cams. M2’s cams give the node a time slot and a manual key.
What to watch
A massive engineering problem with any signal repeater is local loop feedback (acoustic screeching in microphones, or latching in radio relays). If the node transmits a massive current into the dirt, its own ultra-sensitive receiver inputs (X1 and X2) will instantly detect it, keeping the node permanently jammed on.
Lets EMBRACE the feedback, and see what happens, minding the below as our physics.
Removing the mute contact needs more than deleting a contact. When the node transmits, C_s carries about 47 V, and T_r at 1:19 would put roughly 900 V on V1’s grid. So the mute was also protecting the tube, not just preventing feedback. What the open loop does after that depends on the sign of the feedback:
I’d build the second one. V5 becomes the oscillator: HV charges C_c through R_c, the tube fires and discharges C_c through the relay coil, and the relay pings the soil. The detector output on the starter node is the coupling. A neighbor’s ping raises the starter bias, so it fires sooner.
Make these two edits to the earlier drawings:
Starting values are R_c = 470 kΩ and C_c = 2.2 µF. They are chosen so the node pings about once per second (T ≈ 0.98·R_c·C_c). If V5 won’t extinguish between pings, raise R_c. If the relay doesn’t pull, raise C_c. These assume a 190 V HV rail, a transfer voltage of about 130 V and extinction near 30 V. I haven’t checked the 0A4G-type datasheet for those, so verify them before building.
The self-feedback is harmless here. A node’s own ping raises its starter bias only after it has fired, while C_c is still recharging, so it can’t re-fire early. At most it shortens the period slightly.
What to test
Log ping times by measuring across a spare rod pair or across the RY1 coil, so the measurement doesn’t load the loop.
Weak points
