Oscillator Instability in a 144 MHz Transverter
A homebrew 144 MHz transverter uses a 116 MHz Colpitts oscillator followed by a buffer, a ×2 multiplier, and a 144 MHz amplifier. The builder reports that the local oscillator (LO) frequency drifts 3 kHz over the first 20 minutes after power-on and then wanders ±400 Hz over the next hour. The oscillator tank uses a ceramic capacitor with a temperature coefficient of ±30 ppm/°C and an air-wound inductor with a Q of 60. A varactor is used for fine tuning with a bias of 4 V from a resistive divider fed from the same 12 V rail as the PA. The PA draws 4 A key-down. The builder also notices a chirp on CW and a broadband noise floor that rises when the PA is keyed.
1. Identify the most likely causes of the warm-up drift and the long-term wander, and rank them.
Read the worked answer
Warm-up drift (3 kHz over 20 min): This is classic thermal settling. At 116 MHz, 3 kHz is 26 ppm - consistent with a ceramic tank capacitor's ±30 ppm/°C coefficient if the capacitor heats by ~1 °C, or with inductor expansion and board flexing. The PA drawing 4 A in the same enclosure creates a thermal gradient that changes the tank temperature. Rank: (1) ceramic capacitor tempco, (2) inductor/board thermal expansion, (3) varactor bias drift as the 12 V rail sags under key-down.
Long-term wander (±400 Hz, ~3.4 ppm): Likely (1) varactor bias noise - the resistive divider is not decoupled and the PA's 4 A current modulates the rail; (2) ambient temperature changes; (3) mechanical vibration. The varactor is the most voltage-sensitive element, so rail noise translates directly to FM.
Chirp on CW: The PA current pulse changes the 12 V rail, which changes the varactor bias, which shifts the LO - classic chirp from insufficient decoupling and shared rail. The rising noise floor when keyed confirms RF feedback or rail modulation.
2. Propose specific design changes, and explain the expected effect of each.
Read the worked answer
- Replace the ceramic tank capacitor with an NP0/C0G type (±30 ppm/°C) or an air/PTFE trimmer. Expected: reduces warm-up drift by roughly an order of magnitude.
- Separate the oscillator supply from the PA rail using a low-dropout regulator or a series pass transistor with a large reservoir capacitor and a ferrite bead. Expected: kills chirp and most rail-induced FM.
- Decouple the varactor bias with a 0.1 µF and 10 nF in parallel to ground at the divider midpoint, and feed the divider from the regulated rail. Expected: removes audio-band noise on the tuning line.
- Thermal isolation: move the oscillator to a corner of the enclosure away from the PA heatsink, or add a small shield can. Expected: reduces the 20-minute thermal settling.
- Buffer the oscillator output with a high-isolation stage (e.g., a dual-gate FET or a common-gate buffer) so that load changes from the multiplier do not pull the frequency. Expected: reduces pulling and improves long-term stability.
- Consider a TCXO-referenced PLL if the application needs better than 1 ppm; a free-running Colpitts will never match a TCXO.
3. How would the builder verify the fixes with test equipment?
Read the worked answer
Use a frequency counter with a stable timebase (GPS-disciplined or OCXO) to log the LO frequency over 60 minutes; plot frequency vs. time and vs. enclosure temperature (thermocouple on the tank). A spectrum analyzer with a near-field probe should be used to check for leakage and to compare the noise floor keyed vs. unkeyed. A directional coupler and oscilloscope on the 12 V rail will show the sag and noise when the PA is keyed; the builder should confirm the sag is <10 mV after the fix. For chirp, use a CW keying test with the spectrum analyzer in zero-span mode at the carrier frequency to see the frequency excursion during the first few milliseconds of key-down. Repeat the 20-minute warm-up test after each change to quantify improvement. Keep a log of temperature and frequency so that residual drift can be attributed to the right mechanism.