This video details experimental work on designing a multi-crystal oscillator at 10 MHz, focusing on optimizing its Q factor and phase noise performance. The creator characterizes individual crystals, measures their impedance and Q, and explores methods to compensate for parallel capacitance using inductors, highlighting the critical importance of coil quality. Multiple transistor types (BFQ69, BFR134, BFR91A) are tested for their impact on amplifier performance, signal-to-noise ratio, and phase stability, revealing the BFQ69 generally performs best for crystal ratio and signal-to-noise. Significant challenges are encountered, including measurement system quirks, temperature-induced phase drift, and unexpected microwave oscillations. A key discovery is that wideband amplifiers preceding an amplitude limiter introduce substantial, unwanted phase noise, diminishing the overall performance despite high Q measurements. The video concludes with the realization that this fundamental design choice is detrimental and necessitates a shift to narrowband amplifiers for future iterations of the oscillator.
Summarizing the Multi-Crystal Oscillator Experiment
0:00 Initial Crystal Characterization: Experiments begin with three 10 MHz crystals in an oscillator setup, initially measuring their Q factor (around 1.2 million, slightly better than specification) and parallel resonance/impedance.
1:50 Parallel Resonance Compensation: To address the parallel capacitance of the crystals (1.8 kΩ impedance), inductors are introduced. Initial attempts with an inadequate coil (thin wire, many turns) lead to a Q drop (from 1.2M to 1.16M).
3:16 Improved Inductor Design: A new, higher-Q tunable coil with thicker copper and a 4C65 ferrite toroidal core is implemented, successfully eliminating parallel resonance and restoring a symmetric filter transmission with minimal Q loss (1.195M).
5:38 Amplifier Design and Testing: The first stage of a selective amplifier is introduced, designed to eventually become an oscillator. Three different transistors (BFQ69, BFR134, BFR91A) are tested to compare their performance regarding output impedance, input impedance, Q, signal level, and noise floor.
7:20 Crystal Parallel Operation Theory: The conceptual schematic explains how crystals in parallel act as combined inductors, allowing a common resonance and managing slight differences in crystal characteristics.
8:50 Q Measurement Refinement: The creator discovers and corrects a misconfiguration in the network analyzer's bandwidth setting, leading to Q measurements that now perfectly align with manufacturer specifications.
11:40 Transistor Performance Comparison (Pre-Correction): Initial comparisons show BFQ69 to be best for the ratio between output impedance at center frequency and wideband, and slightly better in signal-to-noise ratio.
13:50 Inductor Placement Impact: It's discovered that placing inductors across the crystals, rather than in series, leads to significant losses due to the high impedance of the parallel resonance circuit, causing Q degradation.
15:50 Alternative Capacitance Tuning: An external, adjustable capacitive voltage divider is adopted to tune out parallel capacitances, restoring the Q to expected levels (~1.06M) and achieving a symmetric frequency response.
17:00 Power Dissipation and Oscillations: The unit's input impedance is measured, and maximum power dissipation in the crystals is calculated. Measures are taken to prevent microwave/UHF oscillations (e.g., 47-ohm resistor with thin wire, 8.2pF to ground).
22:20 Unexpected Resonances: Output impedance measurements reveal unusual secondary resonances and minima, indicating complex interactions within the circuit.
24:40 Combining Amplifier Outputs: Connecting the three amplifier outputs in parallel using T-connectors significantly degrades Q (down to 910,000) and introduces substantial capacitive load.
27:30 Initial Phase Noise Measurements: Early phase noise tests show interference pulses and phase noise stronger than amplitude noise. Issues with BNC connectors causing unstable readings are identified and addressed.
38:50 Phase Shift and Temperature Effects: Phase shift through the amplifiers is measured, showing a dependency on temperature (approx. 1°C change causes a significant phase shift over time) and supply voltage changes.
44:20 Comprehensive Transistor Summary: A detailed table summarizes all findings for the transistors (BFQ69, BFR134, BFR91A), highlighting BFQ69's overall superior performance in most metrics, especially crystal ratio and wideband signal-to-noise, while noting that phase noise at 1 Hz offset doesn't differ significantly across transistors.
47:15 Crystal Common Resonance Confirmed: Experiments with three very different crystals confirm that they do exhibit a common resonance when paralleled, indicating the amplifier design is the source of performance issues.
49:10 Amplifier Loading Optimization: The problem of loading the sum of three amplifiers with 25 ohms is identified as problematic, suggesting a load of 25/3 ohms is needed for optimal impedance.
53:20 Second Amplifier Stage & Noise: A second amplifier stage is added, but initial phase noise measurements show a degradation in performance, with the wide noise floor becoming much higher than hoped.
1:00:00 Output Connection Mistake Corrected: A critical mistake in the output connection of the amplifier stage, leading to poor Q, is identified and corrected, immediately improving Q to 1.14M and output impedance.
1:01:00 Ferrite Core Impact and Microwave Oscillations: A magnet test reveals distortion when moved near the ferrite core. Microwave oscillations at 1.7 GHz are detected, indicating instability.
1:03:00 Diagram Correction and Oscillation Fix: The schematic is corrected, and proper output connection eliminates the microwave oscillations, leading to a higher Q (1.158M) and lower phase noise in some regions.
1:11:15 Unexplained Noise & Measurement Limits: Despite improvements, unexplained noise, especially in the 20-100 Hz range, persists. The creator notes limitations of measurement tools in accurately capturing this region.
1:16:10 Amplifier Stages Adding Noise: Analysis suggests the added amplifier stages, particularly the wideband ones, are introducing amplitude noise that translates into phase noise, especially above 10 Hz.
1:19:00 Third Amplifier Stage & Voltage Reduction: A third amplifier stage is added, and the supply voltage is reduced (from 16V to 8V), leading to a small but clear improvement in noise floor at 50 Hz.
1:22:00 Limiter Introduction & Degradation: An amplitude limiter is added, and the system is run as an oscillator. This addition significantly degrades wideband noise, and introduces strong correlation between amplitude and phase noise, with the spectrum becoming asymmetric.
1:28:40 Limiter Issues Diagnosis: The limiter is found to be adding substantial noise. Microwave probe tests confirm no oscillations below 10 GHz, ruling out that as the cause of limiter-induced noise.
1:31:00 Key Takeaway: Wideband Amplifiers are Problematic: The final conclusion is that using wideband amplifiers in front of a wideband amplitude limiter is a "bad idea," as noise peaks from the wideband signal add to the 10 MHz signal, causing zero-crossing shifts and introducing phase noise.
1:33:55 Future Work: The creator decides to stop the current experiment and plans to replace the wideband amplifiers with narrowband amplifiers in the next video.