This video, the 108th in a series on oscillator sideband noise, details continued experiments with multiple crystals in parallel. The presenter troubleshoots persistent noise issues, initially misattributing them to the experimental unit but later tracing them to the test oscillator, faulty tuning capacitors, and poor connector contacts. Key efforts include replacing and cleaning tuning capacitors, designing a frequency shifter for better low-offset measurements, and redesigning RF transformers with low-loss polyethylene-insulated wire. A critical discovery reveals that feeding amplifiers in a correlation measurement system from a common power supply significantly degrades low-offset noise performance, leading to a substantial improvement when separate supplies are used. The video demonstrates the challenges of achieving ultra-low phase noise measurements, highlighting discrepancies between Linrad software and a commercial PN 2060 analyzer, and the impact of seemingly minor component and setup choices.
0:01 Test Oscillator & Initial Unit Performance: The presenter switches to a more stable "ventil oscillator" after his primary test oscillator becomes noisy. Initial tests show the experimental unit (many crystals in parallel) adds no noise at 1 Hz offset, indicating the previous noise was from the test oscillator.
1:34 Noise Filtering & Linrad/PN 2060 Discrepancy: The crystal array provides some noise filtering at 10 Hz. Linrad shows very low noise floors (e.g., -182 dBc/Hz at 1 kHz, -197 dBc/Hz flat noise floor), often outperforming the specified capabilities of the commercial PN 2060 analyzer.
5:17 Troubleshooting New Noise Sources: A new noise source causes a 3 dB degradation. Extensive troubleshooting points to poor connections in BNC connectors and attenuators within the "box" (experimental unit), rather than the test system itself.
10:08 Tuning Capacitor Issues & Replacement: Cleaning the main tuning capacitor (a known source of noise) initially improves performance significantly. However, occasional wideband glitches persist, leading to the replacement of all tuning capacitors with fixed polystyrene capacitors to reduce noise from problematic movable contacts.
14:05 Custom Frequency Shifter for Rubidium Source: To allow for better low-offset noise measurements (up to 30 Hz), a custom frequency shifter is built. This device shifts a Rubidium source signal by 24 Hz, resulting in a zero-frequency offset of 37 Hz.
18:30 Long-Term Measurement Insights: After nearly 8 hours, Linrad measurements show better performance than the PN 2060, suggesting the commercial analyzer's specifications might be optimistic. The video also reveals intermittent, strong "events" (noise bursts) that significantly degrade measurement statistics over time.
22:00 Source Follower Operating Point & Connector Sensitivity: Investigation into the source follower's operating point (transistor current) initially shows a 10 dB degradation, which is later traced back to sensitive BNC connector contacts. Repeated issues with connectors underscore their critical role in low-noise systems.
34:04 RF Transformer Redesign: Suspecting the original RF transformer's construction (thick enameled wire potentially scratched by ferrite), a new transformer is designed using thinner, polyethylene-insulated copper wire from telephone cable, aiming for lower loss and higher Q.
44:56 "Events" & Correlation Spectra: Time-lapse videos illustrate the slow convergence of correlation spectra and highlight sudden "events" where the noise floor jumps by many decibels in both channels, severely impacting measurement accuracy.
52:33 Critical Power Supply Correction: A crucial error is identified and corrected in the PN 2060 setup: two input amplifiers were sharing a common power supply, introducing common-mode noise. By feeding these amplifiers from separate power supplies, the PN 2060's performance at 10 Hz dramatically improves to -152 dBc/Hz, now closely matching Linrad's readings.
Key Takeaways (from video description):
Do not use trimmer capacitors.
Use polyethylene insulated copper wire for RF transformers.
In a correlation system, use separate power supplies for amplifiers in the two legs after the hybrid.