This video documents a detailed series of experiments aimed at designing and building a high-stability, low-noise 10 MHz oscillator using three parallel quartz crystals. The experimenter systematically investigates various aspects of the design, starting with measuring and optimizing the Q-factor of the crystal assembly by developing methods to cancel the crystals' inherent parallel capacitance.
A significant portion of the video is dedicated to testing and comparing three different BJT transistors (BFQ69, BFR91A, BFR134) for the amplifier stages. Performance is meticulously evaluated based on Q-factor, input/output impedance, signal-to-noise ratio, and phase noise under different loading and power conditions. After identifying and correcting several design and measurement errors, the BFQ69 is selected for its superior overall performance.
The final experiments involve adding subsequent amplifier stages and an amplitude limiter to close the feedback loop and form an oscillator. This reveals a critical design flaw: wideband noise from the amplifier chain is converted into phase noise by the limiter, degrading performance. The video concludes that the wideband amplifier approach is unsuitable, setting the stage for a future experiment using narrowband amplifiers.
Building a Low-Noise Oscillator: Experiments with Parallel Crystals and Amplifiers
00:00:04 - Initial Concept: The experiment begins with a plan to create a 10 MHz oscillator using three crystals in parallel. The initial tests model this setup and measure the quality factor (Q) of a single crystal, finding it to be a high 1.2 million.
00:01:47 - Compensating for Parallel Capacitance: The crystal's inherent parallel capacitance creates an unwanted resonance. The first attempt to cancel this with a parallel inductor proves flawed, as the inductor's own losses degrade the crystal's Q.
00:15:56 - A Better Solution: A new method is implemented to cancel the capacitance by feeding an antiphase signal through a capacitive voltage divider. This successfully restores the high Q-factor without introducing extra losses.
00:19:18 - Transistor Comparison: Three different transistors (BFQ69, BFR134, BFR91A) are tested in the amplifier stage. After extensive measurements of Q, noise, and impedance ratios, the BFQ69 is found to have the best overall characteristics.
00:22:04 - Power Dissipation Check: The power dissipated in the crystals is calculated to ensure it does not exceed the maximum rating of 100 microwatts. After correcting a measurement error, the power is confirmed to be at a safe level of 116 microwatts per crystal.
00:34:51 - Power Level vs. Noise: A key finding is that reducing the amplifier's output power by 6 dB dramatically improves the close-in phase noise, lowering the noise floor by approximately 9 dB at a 1 Hz offset.
00:50:47 - Load Impedance Trade-offs: The experiment explores the compromise between close-range and wide-range noise. A lower load impedance improves the wide-range noise floor but degrades the Q-factor, which is critical for close-range phase noise. A load of around 10-12 ohms is identified as a reasonable compromise.
01:00:13 - Design Error and Correction: An incorrect output tap in a second amplifier stage is found to be severely degrading the Q. Correcting this wiring mistake restores the Q to the expected high value of 1.14 million.
01:22:21 - Creating the Oscillator: An amplitude limiter is added to the amplifier chain to create a closed-loop oscillator. However, the resulting performance is degraded at very close offsets (1 Hz) compared to the reference source.
01:25:25 - Asymmetric Noise: The oscillator exhibits asymmetric noise sidebands, indicating a strong correlation between amplitude (AM) and phase (PM) noise, pointing to a common noise source within the new circuit.
01:33:10 - Critical Conclusion: The Problem with Wideband Noise: The experimenter concludes that the wideband noise from the amplifier chain is being converted into phase noise at the amplitude limiter. This "noise folding" is a critical design flaw that explains the poor performance.
01:33:55 - Next Steps: The video concludes that wideband amplifiers are a bad idea for this application. The next video will explore a new design using narrowband amplifiers to prevent wideband noise from reaching the limiter.