This video documents an in-depth experimental investigation into reducing phase noise in a 10 MHz crystal oscillator by using three crystals in parallel. The experimenter systematically builds and refines a multi-stage selective amplifier, moving away from a previous wideband design.
The investigation begins by testing various output configurations, discovering that a high-Q resonator at the output is highly susceptible to noise injection from subsequent stages like limiters and buffers. The design is then pivoted to use a low-Q output resonator, which proves more effective. A significant portion of the video is dedicated to meticulous troubleshooting and optimization, addressing noise introduced by amplifier stages, signal levels, and impedance mismatches. A key finding emerges when comparing two configurations: one with separate amplifiers for each crystal versus one where the crystals and their first-stage amplifiers are simply connected in parallel. The results show no significant performance difference, leading to the conclusion that the simpler parallel configuration is equally effective.
Further troubleshooting uncovers and resolves several issues, including noise from an external power supply and a massive 20 dB noise degradation caused by a single faulty ceramic capacitor. The final, optimized three-crystal oscillator demonstrates significantly improved phase noise and stability compared to the measurement system's reference oscillators, establishing it as a new, superior reference for future experiments.
Investigating a Three-Crystal Parallel Oscillator for Low Phase Noise
0:01 Initial Concept: The video aims to improve on a previous design by using three parallel 10 MHz crystals with selective amplifiers (J310 FETs) instead of wideband ones to lower the overall phase noise.
4:54 Amplifier Design: The system uses a multi-stage amplifier design. The first stage provides a low-impedance (approx. 20 ohms) load for each crystal. The second stage uses two JFETs in parallel, and its output is coupled to a final resonator.
7:20 High-Q Resonator Problems: Initial tests with a high-Q resonator on the output show that it is extremely sensitive. Attaching subsequent stages like a buffer or limiter introduces significant noise, making this approach unworkable.
18:24 Buffer Amplifier Issues: An attempt to isolate the high-Q resonator with a buffer amplifier also fails, as the buffer itself introduces unacceptable levels of noise, demonstrating the difficulty of extracting a signal without disturbing a high-Q circuit.
28:30 A Better Approach: Low-Q Resonator: The design is changed to use a low-Q resonator (Qβ7) loaded with a 470-ohm resistor. This immediately shows a significant reduction in flicker noise.
40:55 Key Optimization: It is discovered that changing the load on the second amplifier's output resonator from 470 ohms to ~235 ohms drastically improves performance, reducing flicker noise by about 6 dB at 1 kHz.
51:12 Power Supply Noise: A recurring noise problem at very low offsets (0.1 Hz) is traced back to an aging external power supply. Adding a large RC filter (8 ohms, 2 Farads) resolves the issue.
1:09:49 Central Experiment: Parallel vs. Separate Amplifiers: A crucial test is performed to see if using separate, isolated amplifiers for each crystal is better than simply connecting the crystals and their corresponding first-stage amplifiers in parallel.
1:10:29 Key Finding: Simpler is Just as Good: The results show no significant performance difference between the separate and parallel configurations. This is a major conclusion, indicating that the much simpler design of paralleling the crystals and first stages is sufficient.
1:16:00 Superior Performance: The final three-crystal oscillator, running in its optimized configuration, demonstrates superior phase noise compared to the test setup's existing Wenzel reference oscillators, especially at offsets from 10 Hz to 1 kHz. Allan deviation measurements confirm its enhanced stability.
1:57:12 A New Problem Arises: While adding fixed capacitors to tune the oscillator precisely to 10 MHz, the phase noise suddenly degrades by a massive 20 dB.
1:59:09 The Culprit Found: The severe noise degradation is traced to a single, faulty 3.3 pF NPO ceramic capacitor used for frequency control coupling.
2:19:30 Faulty Capacitor Replaced: Replacing the faulty ceramic capacitor with a polystyrene type restores the oscillator's excellent performance, with a final noise floor around -177 dBc/Hz.
2:18:31 Final Result: The successfully built and debugged three-crystal oscillator is now significantly better than the previous reference oscillators and will be used to improve the measurement system for future experiments.