Friday, 3 April 2015

Regen #5

Greetings — Привет! 

In winter 2015, I built 5 HF regens plus 2 VHF super regenerative receivers. I’ve run out of my better quality air variable capacitors, potentiometers and room — each version seems larger and uglier than the previous. With all these experiments, I’ve advanced about 2 mm up the regen receiver design learning curve — and in standard form factor, more questions arose than answers.

Briefly/frankly: I’m more a science officer Spock type than a green smoothie, quinoa and tofu devouring new ager — thus I prefer to avoid emotional messaging and hyperbole. I share my experiments to kindle interest, invoke dialog and sincerely hope we’ll all improve and enjoy what’s left of the SWL bands + analog radio design. Out of the gate — I think of this radio as OK; enjoyed making it and wish you well with your own experiments.

Photo of Regen #5 Front Panel.

 Above — Photo of Regen #5 Front Panel.

1.  Schematics

Regen #5 RF Board Schematic

Above — Regen #5 RF Board Schematic

Regen #5 AF Board Schematic

Above — Regen #5 AF Board Schematic

2.  RF Preamplifier

A common gate amplifier provides reverse isolation. IMO, batteries are best suited for moors and trail. As a dedicated AC power supply enthusiast , I won’t run a regenerative receiver without this isolation to prevent my leaked signal getting 60 cycle modulated and coming back into the receiver antenna port (the cursed common mode hum thang), or perhaps, making the local Hams irate with my QRM. 

Thanks to stuff like marijuana grow ops, our neighbor’s switch-mode lights, dimmers and other dumb dumbs; low band reception proves vexing in many bigger cities. I don’t want anything I’ve built adding to the radio listener interference burden.

I agree with the conclusions of regen wizard Charles Kitchin — the preamplifier should run at least 2-2.5 mA source (or emitter) current to help it avoid rectifying strong local stations. Lay the JFET flat side down on the copper board and solder the gate lead as close to the JFET plastic body as possible to squash UHF parasitic oscillations from that lead’s inductance. Further, the drain ferrite bead shown could just as easily be a low-value resistor. For example, 22 to 51 Ω.

The primary, 4 loose turns on a T68-6 = 181 nH (XL = 6.82 Ω at 6 MHz) generates enough of a magnetic field to couple the JFET output to the Q-multiplier (Q-M) tank. My circuit gives a wide range of input signal amplitude variation without too much overall gain, plus light coupling to the Q-M tank.

Close up of RF preamp circuitry and Q-multiplier inductor.

Above — Close up of RF preamp circuitry and Q-multiplier tank inductor. I wound the 3.34 µH inductor with 22 gauge enamel coated wire on a T68-6 toroid.

3.  Q-Multiplier

I sought a low distortion, high Q, negative resistance oscillator as the heart of this receiver and came up with this fun Colpitt’s variant. You can spend years learning, building and testing a ton of oscillator topologies and I plan to work towards this over time.

(FFT) The worst case distortion of the Colpitt’s Q-multiplier at 6 MHz : 2nd harmonic =  -46 dBc.

Above — (FFT) The worst case distortion of the Colpitt’s Q-multiplier at 6 MHz : 2nd harmonic =  -46 dBc. With lower amplitude, it may drop as low as –55 dBc across the 4.9- 8.33 MHz tuning span.

The regen control changes the oscillator amplitude.  1 annoyance with a Colpitts — as you adjust the amplitude or “regen” control potentiometer, the oscillator frequency changes since bias changes affects the transistor input capacitance (mainly through collector to base inter-element capacitance at the pn-junction).

A diagram showing the various internal and parasitic capacitances that affect a transistor circuit
Above — A diagram showing the various internal and parasitic capacitances that affect a transistor circuit. 

To reduce the tank tuning effects from bias change, I employed 3 strategies that worked:
  • High fT transistors (low input C)
  • Cascode oscillator topology (reduces Miller effect)
  • High tank C to L ratio [(weak effect) --- also minimizes stray C plus may boost resonator Q and decrease phase noise]
Initially, some friends reacted negatively to my transistor choice — the 2SC3355, a low–noise 6.5 GHz BJT. Yes, that’s a bit crazy fT-wise, but I bought 20 for $1.00 from a dying electronic store’s bargain bin and quite frankly — they're amazing.  I ran ferrite beads and VHF-UHF bypass to prevent parasitic oscillations and sniffed out none with my spectrum analyzer and DSO.

All my other BJTs above the fT of the PN5179 and BF199 are SMT parts. I wanted the RF board to only house leaded parts soldered in 100% Classic Ugly Construction plus — no low-Q, unknown temperature coefficient cut or glued pads anywhere on the RF board. Actually, leadless SMT parts probably offer the better BJT choice with respect to wiring parasitics.
It also might be better to run MPSH10 (PNP), BF199, PN5179 or other transistors with an fT between ~1- 2.5 GHz compared to my mega fT choice to reduce the chance of spurious oscillations while still fronting a low input C — I’ll leave transistor choice up to you.

The cascode configuration boosts the main Colpitts BJT’s output resistance to present a higher QL to the resonator. Better quality oscillators often run higher QL to isolate the tank from transistor variations and/or to reduce phase noise and possibly some temperature (frequency) drift.  

RF board built with 100% Classic Ugly Construction.

Above — RF board built with 100% Classic Ugly Construction.

RF section prior to wiring up the pots plus switch

Above — RF section prior to wiring up the pots plus switch. More bolts were added later.

4.  Detector


While factoring each particular JFET’s characteristics, the physics behind this detector at various signal levels and Q-multiplier amplitude settings looms complex. I’ll just give a hypothesis for some points.

The detector is directly connected to the Q-M tank. To decouple it, I center tapped the main inductor. 

A review of regenerative receiver operation 


For maximum AM sensitivity on a signal, advance the Q-M amplitude pot until you hear some high pitched AF noise or oscillation. Then slowly lower the Q-M amplitude just enough to eliminate this audio buzz. Instead, for weak signal listening, we may choose to autodyne the signal by further increasing the Q-M amplitude while adjusting the fine tuning capacitor to zero beat.


For SSB, advance the Q-M amplitude pot until you hear some hiss and then go look for the familiar duck quack of a SSB signal. Fine tune around the signal and tweak the Q-M pot until you hear your desired audio quality. CW is straight forward — just find a nice beat frequency. If you run the RF gain too high, signals may get phase or frequency modulated.

Of course, for AM, or SSB/CW, a delicate interplay exists between the various controls so that you'll often dial in the best sounding audio by expert knob tweaking — the so-called, “art of regen” stuff that conjures mystique and nostalgia. This takes practice — regen receivers sure engage you!

Regen #5 Particulars

Further muddying the waters, I ran a front panel switch that the places an 8K2 resistor in parallel with the 22K detector source resistor to make a "higher current" setting. In my particular case, that’s ~5974 Ω which sets the JFET source current to 441µA. I’ve probably mislabeled this switch on the receiver’s front panel: it might be better to just say lower current and higher current mode than SSB/AM.

When switched to the lower current setting, the measured source current = 138 µA.



The gate-to-channel diode provides detection and the JFET is working as a square-law detector. Thus demodulation distortion will be a function of signal level and bias on the JFET since the square-law operation is only good over a certain range.  The JFET is biased near to cutoff.

Normally, I run Regen#5 with the switch in high current mode for it offers maximum sensitivity. In certain cases such as when tuning strong AM signals, switching to the low current detector mode offers less AF distortion. Often, the switches' effects on recovered audio distortion sounds subtle.

I determined the 22K source resistor experimentally during listening tests.  The goal = to find a source resistor that gives the best audio fidelity.




Ideally, we want our detector to operate as a mixer — in this case, we’re running a direct conversion receiver.  In SSB/CW reception, you run greater Q-multiplier signal amplitude than while detecting AM signals and we’re probably get some square law detection within the Q multiplier. 

1 theory is that rectification results in DC that may be enough to actually drive the FET toward pinch off which kills the gain. Switching in higher JFET source current will help keep the RF detected from the Q multiplier from pinching the JFET off. 

This winter, I built some different, very low distortion AM detectors. In 1 design, I ran a low current pair of BJTs with heavy feedback. While remarkable for AM, they really sucked for CW/SSB detection. So the direct-coupled hybrid cascode detector shown is a compromise circuit that works OK for both but better for SSB than AM in terms of AF distortion via listening tests.

This detector begs for further experiments. For example, what happens when it is DC coupled to the Q-M tank and a high ohm gate resistor is added? Should the 8K2 shunt resistor be replaced with a pot, or perhaps a switch to allow 3 or more different JFET source currents?

5. Audio Frequency Board

AF circuit breadboard

Above — AF circuit breadboard

Regen#5 features an op-amp plus discrete component AF board. I love analog design and the exquisite control that choosing discrete parts offers. My goal =  low distortion + low noise from AF input to loudspeaker.



Nothing special here. A 0.1 µF input bypass rolls off noise and prevents local AM signals from entering, getting rectified and amplified by the AF chain. I opted to not include a multiple pole low-pass filter chain for dissecting CW/SSB pile ups, or a tone control circuit, but in a keeper-grade radio, I would add 1 or both.

I don’t run band-pass AF filtering in my receivers, but that's another option. With all the free, online software, designing good AF filters has never been easier.
You may boost the gain of the op-amps by increasing the feedback resistor value. Too much gain on strong signals may cause a spasm of feedback and create celestial noises.


I’ve discussed this circuit before on this blog post.
Except now, I’ve solved a distortion problem caused by closing the loop in the op-amp connected to the power followers. In high gain feedback amplifiers, it does not take much time delay or phase lag to cause oscillation at high frequencies near the upper end of the bandwidth.  The small 22 pF feedback capacitor lowers the closed loop bandwidth so that there is insufficient gain at high frequencies for oscillation to occur.

FFT of the audio amplifier board.

Above — FFT of the audio amplifier board. Swinging 10 Vpp with the strongest harmonic at - 57 dBc between a 12.2 to 0 VDC single-supply rail provided 1 of the happiest accomplishments in my fledgling amateur radio designer career. I hope to better this 1 day, but that will prove difficult.

AF and RF circuit boards mounted and wired in Regen #5

AF and RF circuit boards mounted and wired in Regen #5

Above — AF and RF circuit boards mounted and wired in Regen #5

5. Thoughts

I’ve seen countless regen schematics spawned by the good and the great. Wow! — how can such a relatively simple concept garner so much attention and appear in so many forms?  I hardly feel I’ve contributed to the regen knowledge base, but gained valuable knowledge of what I want or hope to achieve in the future.

Regen#5 behaves like any regenerative radio should. The Q-M amplitude control feels precise — almost to the point of being too touchy. But does the regen amplitude control work any better or worse than other designs? I don’t know. The frequency shift associated with changing the Q-multiplier bias is minimal when compared to my other builds.

I applied standard LC VFO temperature stability techniques and felt surprised when I didn't have to add negative or positive tempco caps to stabilize it. This receiver stays on frequency for hours even when tuning CW and SSB signals.

Regen#5 suffers from microphonics as a SSB/CW receiver — and reminds me of an unbalanced or single-balanced DC receiver without the broadcast band nor in-band AM detection. The scratchiness of my dust-laden, ancient, air variable caps gets multiplied greatly by the sensitive RF chain. I’ll get some Caig Lab’s contact cleaner and have a go on those caps.

Still, too, the AM and SSB detector poses compromise — maybe it’s better to run 2 separate  optimized detectors with a switch to pass the signal through the best detector for the desired demodulation mode? I’ve got lots of experiments ahead, and more regen circuit ideas to share but plan to stop all regen work until next fall.

6. Out Takes and Sound Bytes

My prototype PA board

Above — My prototype PA board

FFT of the prototype PA board

Above — FFT of the prototype PA board.

Hartley VFO ideas from Regen #3

Above — Hartley VFO ideas from Regen #3

FFT of the signal from the above Hartley

Above — FFT of the signal from the above Hartley.

Sound Bytes


I heavily compressed some sound bytes to show the receiver in real world conditions. You'll hear me tweaking the Q-multiplier control on these recordings:

Mostly SSB during a contest March 28 with lots of QRN   Click

Noisy conditions and weak signal CW listening  Click

SSB during noisy conditions. I probably ran the RF gain too high  Click

Thanks for reading.


  1. Hi Todd,
    ungly but very nice project,step by step assembly....
    You use the J310 for RF preamp that gives you stability also on detector stage can you explain me the fuctionality when it's NC or Hight Current mode?
    After the RF Preamp you drive the Hartley VFO thru 3,34μh toroid to the detector circuit includes J310 and 2N4401?
    Thanks for you time Todd.

  2. Hi Kostas:
    I adopted the separate preamp, Q-multiplier, detector scheme: The Colpitts tank inductor directly connects to the detector which has 2 settings for current. Higher current works better for SSB/CW as discussed. Higher current also = greater sensitivity.

  3. Hello,
    Would you recommend more primary turns for use with a small antenna?
    Cheers, Jeff

  4. Hi Jeff
    Yes, that can work. Still, I fear loading down the main LC tank with a low impedance if you push the primary winding number up to high. You gotta be careful. These preamplifier-laden regens carry so much gain, you might be surprised what you can do with an electrically short antenna… Run the lightest possible antenna coupling consistent with good sensitivity dogma . Best!

  5. You wrote, "If you run the RF gain too high, signals may get phase or frequency modulated." What exactly is the cause of this?

  6. Does not happen during AM reception ---- it's a Ham band thing.. Only on strong stations with too much RF gain. The AF output resembles some sort of phase/frequency modulated phenom. Basically, I think, in this case, detected audio modulates the oscillation frequency. The unbuffered Q-multiplier gets "stim-modulated".

    1. Right, good point. I think David Newkirk has said the same thing here: Quote from Newkirk: "the detector carrier is amplitude-modulated by the audio the detector is producing. (There's probably some PM and maybe even FM as well, depending on multiple factors.)".