RF Noise Generator

Design

• Useful as a filter test signal source
  • Pass filters
  • Crystal filters
• From Experimental Methods in RF Design
  • Fig 7.72
  • Described as "not flat"
• "Junk box" parts
  • (2) 2N3904 transistors
  • (2) 5V (nominal) zener diodes
• Noise figure -50 dBm @ 10 MHz

Schematic

First Prototype

• Unique nodes in red
• Prototype on single sided copper clad PCB
• KiCAD zener packages have backwards silkscreen (will need to fix on OshPark PCBs)

Meaurements

• 12V, 43mA current draw

tinySA Measurements

• 1-30 MHz
• Power Off

• Power On

• 100 KHz-1 MHz
• Power Off

• Power On
• Lot more energy at ~400 KHz
• -30 dB

Rigol DS1054Z Measurements

• • 50 MHz scope with mods
  • UltraScope settings
    • SCAL 1.00 V
    • H 200nS
  • FFT settings
    • CF = 16 MHz
    • 2 MHz start, 50 MHz stop
• RBW 416.6 KHz
• Background noise
• Generator off
• Average ~ -64 dB

• Noise Generator On
• Falls off with frequency, but better than -50 dB

• Measured performance per EMiRFD (predicted - 50 dB)

Better Design #2

• Replace Zener as noise source with 2N3904 B-E junction reverse biased
• C1 is connected Emitter of Q1 to GND instead of +12V to reduce power supply noise
• Higher noise than EMiRFD design
  • Usable from 1-10 MHz
  • -40 dBm at 1 MHz
  • -80 dBm at 30 MHz
• Nearly zero power supply current

Better Design #2 - LTSPice Simulation

• Falls off in a straight line (in dBm) with frequency
  • 10 dB at 1 MHz
  • 0 dB at 10 MHz
  • -10dB at 100 MHz

tinySA Measurements

• Off

• On, 1-30 MHz
  • Consistent with the LTSpice simulation

Better Design #3 - Cascode Amp with output buffer

• Previous design is probably good into receiver but not as good for tinySA tracking generator
  • Want to be able to drive passive/crystal filters for testing
  • Can already use NanoVNA but I'd prefer the tinySA due to lower noise floor
• Goal
  • Flat response across 1-30 MHz
  • Good parts of Design #2
  • Transistor seems to be a better noise source than a zener
• Idea - replace amp stage with cascode output stage followed by emitter follower for better match to 50 ohm load

Example cascode design

• Replace R1 with 0.1uF in parallel with 470K (from the the previous)
• Remove R2

Better Design #3 - LTSpice Simulation

• Output stage is high gain, but limits bandwidth

Better Design #4

• #3 output stage was limiting
  • Remove output stage

• < 3 dB down at 100 MHz

Build / Enclosure (Original board design)

• Used very old, but high quality chassis mount BNC connector
  • Silver connector has a nice patina
• Built onto Single Side PCB material
  • Acts as ground plane
  • Pads cut from PCB material and superglued down onto the Base PCB
• Hot glued down into plastic enclosure

Assembly Sheet

RF Noise Generator Rev 1