This video documents the extensive and recondite process of debugging a high-power, low-noise selective amplifier intended for an ultra-low phase noise oscillator. The primary challenge is a persistent battle against self-oscillations and an unstable, high noise floor. The creator employs a systematic methodology, using a PN2060 phase noise analyzer and a network analyzer to iteratively modify the circuit, improve shielding, and characterize component behavior.
Key events in the debugging process include identifying and mitigating power supply-modulated oscillations, isolating amplifier stages to understand feedback loops, and discovering that a faulty JFET, likely damaged by a power transient, was a major source of noise. After replacing the component, excellent noise performance is temporarily achieved. However, the system exhibits frustrating instability, with the noise floor fluctuating dramatically over time. This leads to the conclusion that the quartz crystals themselves are the likely culprits, possibly due to thermal drift or mismatched characteristics at operating power levels. The project is paused pending a rebuild with a new set of crystals matched at their operational temperature.
Debugging an Ultra-Low Noise Selective Amplifier
0:00 The project begins with a new, more powerful amplifier that is plagued by self-oscillations at various frequencies. Initial attempts to quell them with capacitors were only partially successful.
1:35 Early measurements with a PN2060 phase noise analyzer show a promising noise floor around -192 dBc/Hz, but the readings are unstable and marred by interference.
11:45 Improved shielding using copper foil successfully reduces spurious signals from other lab equipment, cleaning up the measurement environment.
25:10 A systematic diagnostic phase begins, using a network analyzer to isolate amplifier stages and measure their impedance, gain, and phase characteristics to understand and control feedback paths.
31:32 A significant tuning adjustment results in a breakthrough, achieving a near-perfect AC voltage match between the gate and drain of a source-follower FET, which dramatically improves the phase noise measurement.
34:13 A new problem arises: the amplifier becomes unstable and noisy when driven into saturation at higher power levels, as its changing impedance creates positive feedback.
37:18 A catastrophic failure occurs where the amplifier becomes extremely noisy. This is attributed to a component being destroyed by a large electrical transient from discharging a 1.5 Farad capacitor.
40:19 After extensive troubleshooting, including capturing an instability event on a timelapse video, a single JFET among a parallel set is identified as faulty and replaced, restoring the amplifier's low-noise performance.
44:42 The repaired amplifier achieves an excellent noise floor, measuring approximately -136 dBc/Hz at a 1 Hz offset.
49:45 The creator details the complex measurement setup, which uses a reference oscillator and cross-correlation to measure the amplifier's noise floor even when it is lower than that of the reference source.
58:17 Frustration mounts as the high noise problem returns unpredictably. The noise floor fluctuates wildly over hours, sometimes appearing to fix itself before degrading again.
1:05:14 The video concludes with the hypothesis that the quartz crystals are the root cause of the instability. The project is paused until a new set of crystals, which have been carefully selected for matched performance at their precise operating temperature, can be installed.