2014年10月5日日曜日

Fringe howling regenerative receiver

The following receiver is a remote-controlled regenerative receiver using hybrid Colpitts-Vackar feedback. A detailed write-up of this receiver will be posted in the future. Currently, at certain frequencies in the upper half of its tuning range, the receiver suffers from fringe howl, a crippling phenomenon that manifests itself as a loud audio oscillation just as the set enters RF oscillation. In a regenerative receiver, the point of just entering RF oscillation is where the detector is most sensitive, so fringe howl -- the unwanted AF oscillation at the most sensitive detector setting -- serves to make the detector almost useless, and is a problem that must be addressed.
Schematic diagram of fringe-howling regenerative receiver. 
The following video illustrates the fringe howl occurring in the receiver prototype. Although the audio and video quality are low, and the prototype circuit layout is not visually pleasing, I felt it important to create an audio-visual recording of the fringe howl phenomenon, because there is (as far as I know) no existing audio-visual documentation of the phenomenon, only written descriptions. 


Before watching the video you may want to review the following labeled screen captures from the video, which explain what you are seeing. 
Video from 00:00 to 00:20: close-up of the control board
Video from 00:20 to 00:27: close-up of the control cable connecting the control board 
Video from 00:27 to 00:50: close-up of the remotely-located receiver board.
Video from 00:50 to 01:14: smooth operation of regeneration control, with no fringe howl, when the antenna is disconnected. Tuning is set to slightly above the middle of the tuning range (4.5 MHz - 16 MHz). 
Video from 01:14 to 01:40: connecting a short antenna to the top of the tank through a 6 pF capacitor. 
Video from 01:40 to 01:48: close-up of the short antenna, a ~30cm piece of wire 
Video from 01:48 to 02:08: fringe howl occurring right at the RF oscillation threshold. Tuning is same as before. Video from 02:08 to 02:30: fringe howl disappears if regeneration is advanced further. 
Video from 02:30 to 02:40: tuning receiver downwards in frequency below the midpoint of the tuning range (4.5 MHz - 16 MHz). Video from 02:40 to 02:57: smooth regeneration control and no fringe howl at lower frequencies.
Video from 02:57 to 03:05: tuning receiver upwards in frequency above the midpoint of the tuning range. Fringe howl re-appears.

Addendum 2014-10-08

Regarding fringe howl: A number of readers have made suggestions as to how to fix the fringe howl, which I am investigating. Again, a full write up of this receiver will follow once the problem has been solved.

Some readers have asked what the V2 component is. For circuit analysis or construction, it can be ignored and considered to be a short circuit. In actuality, it is part of the LTspice circuit simulation that delivers a voltage pulse to the tank inductor as an aid in determining whether the oscillator has enough gain to oscillate with the given circuit parameters (Vcc, emitter resistance, inductor losses, base bias, feedback network). This in turn allows me to optimize the feedback network C1/C2/C5.

Also note that C1 and C2 are unnecessary. Further simulation has shown that the circuit still oscillates when C1 is removed and when C2 is replaced by a short-circuit. C1 and C2 were intended to provide some small amount of Colpitts-style feedback to boost the Vackar-style feedback formed by D1 and C5, but the Colpitts-style feedback is so small in this case that it is actually not needed, according to the simulation. In hardware, removing C1 and replacing C2 with a short circuit yields identical circuit operation.

Regarding the "Vackar-style" feedback network formed by D1 and C5 (with C1 and C2 removed), please see p. 16 of the following document for the derivation of this feedback network: http://www.kearman.com/vladn/hybrid_feedback.pdf. The following LTspice simulation of a simplified version of my circuit shows the open-loop gain behavior of the D1/C5 Vackar-style feedback network vs. the open-loop gain behavior of the C1/C2 Colpitts-style feedback network. As you can see the Vackar-style feedback has decreasing open-loop gain as the tuning capacitor is tuned higher in frequency, whereas the Colpitts-style feedback has increasing open-loop gain as the tuning capacitance is tuned higher in frequency.
Vackar-style vs. Colpitts-style feedback. Notice the difference in open loop gain as the tuning diode is swept across its capacitance range.

Addendum 2014-10-13: solved fringe howl

The fringe howl has been solved.

Measures tried

  1. Rewire DC control cable to physically move it far away from the hot end of the tank inductor. Result: Fringe howl still occurs.
  2. Remove C6 (RF decoupling for input of AF amp). Result: Fringe howl still occurs.
  3. Increase C20 (Vcc decoupling) to 120uF. Result: Fringe howl still occurs.
  4. Increase R14 (series resistor from emitter to AF amp) from 1k to 10k. Result: Reduced volume. Howling stops, but there is still a seashell sound indicating instability.
  5. Decrease R14 (series resistor from emitter to AF amp) from 10k to 0k. Result: Loud volume, but fringe howl observed at wider range of frequencies.
  6. Add emitter follower stage between detector and AF amp. Remove R14. New emitter follower base connects through 100 nF capacitor to Q1 emitter. Base bias set to Vcc/2 with 20k/20k divider. Emitter resistor set to 1k. Output taken from emitter resistor and fed into C3 (AF amp). Result: Fringe howl still occurs.
  7. Add 10k resistor between Q1 emitter and emitter follower base. Result: Reduced volume. Seashell sound still observed.
  8. Replace emitter follower with DC-coupled common-base buffer. New buffer emitter connects directly to Q1 emitter. Base goes through 100k to collector. Collector goes through 4.7k to decoupled Vcc at R15. AF output taken from collector and fed into AF amp at C3. Result: Fringe howl still occurs.
  9. Change DC-coupled common-base buffer to AC-coupled common-base buffer: Add new 1k emitter resistor to buffer, and connect buffer emitter through 100nF to Q1 detector emitter. Result: Fringe howl still occurs.
  10. Add 10k resistance between common-base buffer and Q1 detector emitter. Result: Fringe howl still occurs.
  11. Remove C7, C8, C9 (RF decoupling for intermediate stages of AF amp). Result: Fringe howl still occurs.
  12. Connect external LM386 amp to output of common-base buffer (at collector 4.7k load resistor). Result: Fringe howl still occurs.
  13. Connect external LM386 amp to Q1 detector emitter directly. Result: AF volume too low to determine whether or not fringe howl occurs.
  14. Remove common-base buffer, reconnect original AF amp (but still with C6, C7, C8, and C9 removed), allow fringe howl to occur, and monitor oscillator signal on nearby receiver. Result: fringe howl heard in radiated signal on monitoring receiver.
  15. Disconnect AF amp when fringe howl is occurring and observe radiated signal on nearby receiver. Result: fringe howl stops in radiated signal on monitoring receiver, indicating it is some interaction between the AF amp and the detector that is causing the howling.
  16. Reconnect AF amp and replace R13 (varactor DC bias at hot end) with 10 mH choke. Result: fringe howl still observed, but the generated AF oscillation has become lower in frequency.
  17. Add 10k in parallel with R12 (varactor DC bias at cold end) effectively reducing R12 to ~10k. Result: fringe howl still occurs in the same manner as #16.
  18. Change C5 (Vackar feedback capacitor) to 820 pF. Result: fringe howl still occurs in the same manner as #16.
  19. Change oscillator from Vackar-style to Hartley: remove C5, connect varactor anode directly to ground, connect Q1 emitter through 100 nF to tap on L3 (5 turns from cold end). With the Hartley feedback topology, the varactor anode directly grounded, and the varactor cathode biased through a 10 mH choke, most of the original RC networks in the detector itself have now been removed. Result: fringe howl still occurs, though AF frequency has changed. Fringe howl also now occurs with antenna disconnected.
  20. Remove C20 and attempt to take AF output off of collector resistor R15. Result: a constant high-pitched AF oscillation regardless of regeneration setting, indicating AF amp instability.
  21. Connect C3 (AF amp input capacitor) to ground. Result: quiet hissing from AF amp.
  22. Connect C3 (AF amp input capacitor) to decoupled Vcc at R15. Result: a high-pitched AF oscillation indicating insufficient AF amp power supply decoupling.
  23. Add additional 100 ohm resistor and 22uF capacitor (R16 and C23 in new schematic) to decouple Q3 power supply. Reconnect C6, C7, C8, C9. Result: high pitched AF oscillation of #22 stops. However, fringe howl is still observed.
  24. Add additional 100 ohm resistor and 22uF capacitor (R17 and C24 in new schematic) to decouple Q4 power supply. Result: fringe howl is still observed.
  25. Use decoupled AF amp with an older regenerative receiver using a common-collector oscillator, a 176 uH choke as the emitter RF load, and a 100 ohm resistor as the collector AF load. C3 from AF amp is connected to the detector's collector load resistor. Tank inductor is a ferrite rod antenna; no external antenna is connected. Result: no fringe howl observed, but a slight seashell sound is present at some frequencies.
  26. Sleep one night on the problem. In the morning, I think the problem may be the AF amplifier causing the Vcc voltage to vary, and the regeneration control voltage, being taken from the non-decoupled Vcc supply, may be varying in sync with the amplifier-caused Vcc voltage swings, which could cause motorboating: as RF oscillation starts, the detector noise comes up, causing the AF amp to draw more current, causing Vcc to drop, causing the regeneration voltage (taken from the non-decouled Vcc supply) to drop, causing a drop in detector noise, causing a drop in the AF amp current, causing a rise in Vcc, allowing the regeneration voltage to again rise.
  27. Rebuild new regenerative receiver to remove the temporary Hartley feedback of #19 and to again use the original Vackar-style feedback topology. Remove varactor bias choke and again use resistors to supply varactor bias. Reconnect AF amp back to the current regenerative receiver, but supply VR1 voltage (regeneration voltage applied to base) not directly from V4 but instead from the decoupled Vcc supply at R15. Result: fringe howl can no longer be observed in the new receiver when using either a whip antenna or a large antenna connected to C22. However, a slight seashell sound seems to be present around the middle of the tuning range. It is still possible to induce fringe howl by connecting a large antenna (a metal door frame) directly to the top of the tank L3, bypassing the 6 pF antenna coupling capacitor C22.
  28. Rewire older regenerative receiver of #25 to use decoupled Vcc supply for the regeneration control voltage. Result: no fringe howl observed in older receiver, but seashell sound is still present. Connecting whip antenna to the tank inductor (ferrite rod antenna) causes fringe howl at certain frequencies. Even with maximum decoupling the fringe howl could not be completely eliminated from the old circuit, so the decoupling solution -- i.e. decoupling of AF amplifier supply voltage, of detector supply voltage, and of detector regeneration voltage -- is not enough.
  29.  Reduce VR1 (regeneration control potentiometer) from 10k potentiometer to 1k potentiometer. Result: some reduction in volume, but no fringe howl observed in any case in either old receiver or new receiver, under all antenna cases (no antenna, short antenna, large antenna, connected directly to tank or through coupling capacitor C22). The reduction in volume was unexpected, but may be due to increased current through the detector transistor's base, leading to decreased AF detection efficiency.

Summary of solution to fringe howl

  • Ensure AF amp is stable by implementing Vcc decoupling. Test AF amp stability by connecting amp input capacitor to +Vcc rail; if it oscillates, more Vcc decoupling is needed.
  • Ensure the detector supply voltages, including the voltage used to control regeneration, come from the decoupled Vcc supply to avoid regeneration voltage varying with AF amplifier current draw.
  • The RC time constant of the regeneration control potentiometer and the bypass capacitors on the potentiometer wiper are important and may be a cause of fringe howl.

The current circuit


Addendum 2015-08-26


The value of R14 (1k) is fairly low, and therefore might cause unwanted interaction between the detector Q1's emitter and C6 (10 nF), the RF bypass capacitor at the input of the AF amp. Because R14 is so low, RF charge buildup on C6 might be able to influence the voltage at the Q1 emitter, which could lead to squegging or fringe howl behavior. Better isolation might be needed between the Q1 emitter and C6. This could be done by increasing R14, but this unfortunately decreases AF output. An alternative isolating approach, with no or negligible signal loss, would be to use an emitter follower stage between Q1 and C6 (e.g. http://qrp-gaijin.blogspot.jp/2015/08/a-12-volt-vackar-style-minimalist.html).

Revision history

  • 2014.10.13: added section on solving fringe howl and new circuit
  • 2014.10.08: added notes on Vackar vs. Colpitts feedback and open loop gain simulations
  • 2014.10.09: added video of fringe howl in operation. Verified in hardware that C1 and C2 are unnecessary.