A Regenerative Receiver with AGC for 80 Meters
March 23, 2006 by Rick Andersen, KE3IJ
or 3-10 MHz Shortwave
[Revised 12/2/06, 2/16/07, 3/10/07]
Here's a Regenerative Receiver project that I designed for the 80 meter Amateur band, but which can also be used in a wider-tuning configuration that covers approximately 3 MHz to 10 MHz, for Shortwave (World Band) listeners.
Built in the same style as my previous "40-Meter Tweeter" which can also be found at my Radio Projects menu at http://www.ke3ij.com/radios.htm, this radio features:
* Colpitts-style regen detector which eliminates the need for any tickler coil
* 3-10 MHz tuning set by pc board trim capacitor, but main panel bandspread via varactor diode tuning which is renamed Main Tuning and covers a band of frequencies variable to 150 KHz from where the pc board trimmer is set
* An audio-derived Automatic Gain Control (AGC) circuit that helps prevent overload and blocking by strong signals
* Low input impedance matches 50 ohm antenna system -- giving relatively little AM overload/bleedthrough (although you may want to add a 3.8 MHz bandpass filter after the antenna jack and input diodes; see text and schematic below).
This is version 1 of this design, so it is to be expected that there will be some modifications, especially as I get feedback from other builders who try the circuit out for themselves. Other features that I will probably include in the next version of this basic 80 M regen are:
* Audio filtering: Variable frequency plus the ability to vary the filter smoothly from a bandpass, to a flat, to a notch response
* I may add a zener diode regulator -- So far, the 80 M prototype drifts in frequency very slightly; not nearly as much as the 40 meter version did, which absolutely required a zener regulator
* A built-in audio amplifier and speaker. Right now there is a headphone jack, which I am running into a Radio Shack amplified speaker (they still have 'em! Get 'em while they're hot and not-yet-discontinued! $12.99 last I checked)
* An S-Meter circuit. I actually designed one with the AGC included in this radio, but had some problems with the "zeroing" of the meter tending to drift off zero after a while -- so I left it out of this first version and plan to include it in the next, when the bugs are worked out
* I will probably build a switched crystal converter for the front end of this design and make the radio into a Regenerodyne multi-band. Have not built a Regenerodyne yet but it is a wonderful idea and could make for a simple, basic station receiver
The Circuit and Description
Though I built the entire radio on a piece of 5x7", double-sided copper-clad pc board (from Radio Shack), with front panel and partial sides and back, (see photos at bottom), I decided to break the schematic up into 3 sections for easier analysis, below.
First, the "front-end". Built around Q1, a common 2N3904 NPN silicon transistor, we have an RF amplifier/buffer to keep the regenerative detector isolated from the antenna. Regen builders are aware that a swaying antenna on a windy day can cause the oscillating detector frequency to sway also, which of course is very annoying. On top of that, the oscillating detector is actually transmitting its signal out the antenna and causing potential interference. This front-end RF stage keeps that from happening. I am able to verify that this stage isolates the detector pretty well in that I have an ATU (Antenna tuner or Transmatch) connected between the external antenna and the receiver.
With no isolation, the regen can be detuned by turning the ATU knobs back and forth around the resonant spot on the ATU. With Q1's circuit in place, hardly any frequency shift occurs.
Now as far as gain or amplification is concerned, we home-builders always assume "the more, the better". I have learned that this is not always the case -- especially with regens. Regenerative circuits are already extremely sensitive, which means that if you shout in their ears, their little eardrums are going to hurt (overload). A problem I've often encountered is what is known as "blocking" -- when a very strong signal overloads and swamps the detector, being stronger than the oscillator "injection level" of the detector itself. So the detector's frequency is "pulled" and it tends to "lock onto" the incoming signal, sometimes stopping oscillation altogether. [We're assuming that we're in a ham band listening to CW or SSB voice here, which requires the Regeneration control to be advanced far enough that the detector is now oscillating.]
When this strong-signal overload occurs, your CW tone will chirp violently, and single-sideband voice sounds completely unintelligible.
That's why I decided to add an AGC circuit to my regen, which I have not seen anywhere else (correct me if I'm wrong).
This same stage that is the RF amp and buffer, doubles as the AGC'd amp. Looking at the first schematic, below, we see a 50-ohm antenna input via a BNC connector (you can use the standard SO-239 ham connector here if you like) that immediately connects to a pair of parallel, oppositely oriented small diodes. This is a clipper that ordinarily doesn't have any effect on the incoming signal, which is in the microvolt region. If a strong voltage spike were to come through, the diode(s) would turn on at .6-.7 volts and conduct the excess to ground, protecting the receiver from front-end zapping.
Next, the incoming signal goes either to an optional LC bandpass filter peaked at about 3.8 MHz, of which I give two variants in the schematic diagram, or straight to a .001uF coupling capacitor and on into the first transistor stage... Let me explain:
NOTE! All of the following, in italics, assumes that you are using the AGC-80 receiver in its narrow-band mode for the 80 Meter Amateur Band (3.5-4.0 MHz). If your intent is to use the receiver as a 3-10 MHz "Shortwave / World Band" radio, you will want to ignore the following, since you don't want a front-end filter to restrict your wide-band tuning.
The original schematic published in March 2006 showed the straight-through connection. When I designed the circuit, I failed to take into account the filtering action provided by my resonant antenna tuner.... and later realized that there is some out-of-band "bleedthrough" [interference from carriers far-off in frequency] if your setup lacks a resonant ATU and you connect your AGC-80 directly to an external antenna. To remedy this, I inserted a 150 pF ceramic capacitor in series with a small T-50-2 toroid with roughly 20-35 turns of #28 enamelled wire on it. Theoretically you need 12 uH with a 150 pF cap to resonate at 3.75 MHz, the middle of the 80 meter band. The low impedance (50 ohms) requires that the L and C be in series, rather than the often-seen parallel LC "tank" type of bandpass filter.
In practice, I wound about 30 turns and then began subracting turns, one by one, until I heard a peak in the loudness of the background static, with the receiver tuned to about 3.85 MHz. I can't recall the exact number of turns needed, but since moving/squeezing the windings on the toroid does have some effect, it is best that you do what I did and find the exact number of turns that peaks your radio at 80 (or 75) meters. Then, drip some hot wax from a birthday candle onto the toroid, to hold the turns rigidly in place. This 50-ohm series bandpass filter will knock out the out-of-band bleedthrough almost entirely.
A NEWER AND BETTER BANDPASS FILTER
Later (February 2007) I came up with a sharper-cutoff 50-ohm bandpass filter, based on a "PI" lowpass filter preceded by a series capacitor of .001 uF; this cap and the original circuit's .001 uF blocking cap (giving a hipass filter response) combine into a narrow bandpass filter with sharp skirts, centered on approximately 3.85 MHz in my prototype-- see the "optional" filter between the "X's" in the schematic.
Notice that I use a 1-inch length of plastic tubing-- cut from a McDonalds straw(!)-- as the coil form for the inductor in the PI-filter. Pierce 2 small holes in the middle of the 1" section of straw, with about 1/4 inch between the holes. Within this 1/4 inch, wind approximately 15 turns of #28 enamelled copper wire, close-wound, bringing the ends of the 15-turn coil through the holes and out each end of the straw. Add or subtract a turn, or squeeze together / separate the turns slightly, until the received signal strength peaks at maximum, then drip some melted candle wax onto the coil to hold the turns in place. This filter will sharply attenuate out-of-band signals.
NOTE: The "McDonalds straw" bandpass filter just described was not designed because the simpler series LC (12uH and 150pF) was deficient. The simpler LC worked just fine in my prototype. I just decided to try a fancier design, to prove to myself that it would work just as well, but also allowing for the possibility that some Hams attempting to duplicate my receiver may live at a location where there might be stronger out-of-band interference than I have at my QTH, and would need a better input filter than my original design had offered. So here I offer you three options, assuming you're using an external antenna system with a 50 ohm impedance coming in to the radio:
1) No input filter at all, which works, but you do notice a bit of 'busy' background carriers and noise. But if your antenna tuner is of the "resonant peak" type (as opposed to the more common "Pi" or "T" types), you will be able to get away with this option-- let the ATU provide the front-end rejection for you! Still, the 50 ohm front-end impedance, by itself, does a lot to correct the 'AM blanketing' problem that plagues simple receivers.
2) A simple series LC filter centered on the middle of the receiver's tuning range [approx. 3.8 MHz]. I used a 150 pF ceramic capacitor and a 12 uH coil, with the 12 uH being attainable in one of 3 ways:
a) As described above, about 20-30 turns of #28 enamelled copper wire on an Amidon T50-2 (red/gray core) toroid;
b) About 60 turns (2 layers, 30T each, close-wound) of #28 wire wound on a McDonalds straw;
c) About 18 turns #28, close-wound on a plastic prescription pill bottle with a 1" diameter -- these are approximate turns values and you will need to adjust them for peak reception. (The #28 wire gauge is not sacred, either. It's nice and thin, however, and I have a roll of it, so I tend to favor it!)
3) A combination Hipass/Pi-Lopass filter with fairly steep skirts, the circuit details of which are shown between the blue X's in the schematic diagram below.
[I'm also thinking ahead to a possible "Regenerodyne" approach for the future, where a crystal converter stage will precede the AGC-80's input stage, giving multi-band reception. The better the bandpass filter, the better the image rejection [the Image Frequency is an unwanted product of the heterodyne mixing process].
Here's Another Option: Use an inside antenna! Maybe you're an apartment dweller who can't string up an 80 meter dipole antenna, which should be around 130 feet long. No problem-- a thin wire anywhere from 5 to 20 feet long can be run along the perimeter of the room. This is a high-impedance, electrically-short antenna that is transformed down to the required 50 ohms by a tuned LC-tank with low-impedance secondary winding on the "L"... the short wire plus the LC-tuner/transformer guarantee that this indoor antenna is sensitive only at the sharp peak frequency of LC resonance. No out-of-band AM blanketing here, either!
Back to the description of the AGC-80 schematic: The desired signal goes through the [optional filter and] .001uf blocking capacitor and into Q1's circuit, which looks like a standard common-emitter amplifier. It is, except that the lower base bias resistor is quite small compared to what you'd expect-- the impedance seen "looking into" the amp is 50 ohms, which (presumably) matches your antenna system, and has a great effect on eliminating out-of-band AM blanketing that so often plagues simple homebrew receivers. The junction between this 51 ohm resistor and the 680K bias resistor above it, is where the AGC action takes place - telling the base of Q1 to vary the transistor's gain in such a way as to compensate for any changes in incoming signal strength.
Looking at the circuitry below that 51 ohm resistor (47 to 68 ohms will work here, by the way), we notice a 22 uF electrolytic capacitor; this cap integrates and stores the AGC voltage, which voltage is generated by the "charge pump" circuit to its left (the 10 uF electrolytic and the two diodes). This is the heart of the AGC.
The input to the 10 uF cap is taken off the output of the 2-stage audio amp to be described later. The + side of that cap sees a DC bias voltage upon which a large-swing audio signal is riding. The 10 uF cap blocks the +DC but passes the large AC swings to the diodes, which rectify them into a negative varying DC, which then spills into and charges up the 22 uF AGC cap. A negative peak detector, in electronics lingo.
This negative bias goes more negative as the audio gets louder; less negative as the audio gets softer. Applied to the bottom of the 50 ohm resistor, this negative AGC bias has a back-and-forth "tug of war" with the voltage above the 50 ohm resistor -- Q1's base voltage -- which is established by the upper base resistor of 680 K ohms. The overall gain of the Q1 amp is determined by the DC bias at its base. So when incoming signals get loud and strong and try to overload our sensitive regen detector, the DC bias on Q1's base is automatically pulled down, lowering the gain of the front-end, so less signal gets to the detector's sensitive ears. When the SSBer finally shuts up and there's an absence of incoming signal, the negative AGC voltage decreases, the 680K pulls Q1's base back up, and the front end gets more sensitive, letting more signal in for the detector.
Thus, the AGC is a negative feedback circuit which always "bucks the trend" and tries to keep the volume approximately the same.
Now you can hear the background level come back up when the guy stops yapping, and go way down when his +30 dB/S9 signal blasts through on the next transmission, performing just like your "real" radio does. The "blocking" problem that plagues regens is now gone! Probing the 22uF cap with a digital voltmeter shows about 790 millivolts when all is quiet, dropping to about 500 mV when a moderately strong signal is coming through, to as much as 2 volts negative or so when extremely loud foreign broadcast stations are tuned in.
NOTE: I revised Q1's emitter resistor in the AGC circuit on 2/15/07, changing it from its original value of 4.7K ohms down to 1K ohms. This seems to have increased the sensitivity of the radio, although it also brought up the background noise (with antenna disconnected) somewhat. I also added a 10K resistor in series with the 10 pF cap leading from the AGC circuit to the Detector circuit; this was necessary to prevent a tendency toward some "pulling" that cropped up after I lowered Q1's emitter resistor to 1K ohms. If this modification does not seem to work well for you (i.e., if very strong signals "pull" or "block" the detector frequency, which the AGC is supposed to prevent), feel free to go back to the original 4.7K emitter resistor, and remove the 10K that is in series with the 10 pF cap feeding the tank circuit of the detector.
Next, the Regenerative Detector itself. This is the heart of the receiver, and could be built as a stand-alone, minimal Regen radio. [See A Single Transistor, Big Loop AM Radio for details.]
I have decided that the Armstrong 'tickler-coil' type of regenerative detector, seen everywhere, is no longer my favorite. This one is. I call it a Colpitts-style detector because it is essentially a variation on the Colpitts oscillator. It does away with the need for a pesky, fussy tickler coil with its turns ratio, coupling distance, and phasing nuisances. The Colpitts oscillator (which this actually is, but for AM is operated just below the point of oscillation) gets its positive feedback or regeneration through a "capacitive voltage divider" -- the two little 100- and 47 pF caps hanging off the collector, base and ground.
The only inductor you need is "L", shown as a 30-turn toroid coil in my prototype. [12/2/06 update: You may also want the optional front-end LC filter; see text above.] If you can't get your hands on some Amidon T50-2 toroid cores, don't fret! You can wind your own air-core coil on a small plastic pill bottle, a cardboard toilet-paper roll, whatever. The point is that this Colpitts design makes your coil very simple.
This Colpitts circuit is not my own (although the particular parts values are); you can find it on various Regen radio websites as an alternative to the Armstrong variety. My previous design, the "40-Meter Tweeter", used a variation of this circuit (except with PNP transistors, which makes the circuit look "upside-down" in the schematic), as do the "Universal" and the "VHF/Aircraft" designs at my webpages. It's a very handy circuit that has become the configuration of choice for my regen designs.
The detector is tuned by means of the simple "L" coil, mentioned above, and a variable capacitor as a Tuning control (in the simplest, stand-alone version of this circuit). In my first tack-soldered "spider-web" slap-together of this circuit, I had a good old metal plate 365 pF variable capacitor as my Tuning control. With 30 turns of #28 wire on a T50-2 toroid, I could tune from about 3 MHz up to 10 MHz. This is probably how you want to build the receiver if your main interest is in Shortwave listening. If your interest is in the ham bands, of which I chose 80 Meters, that 365 pF variable gets replaced (as per the schematic below) with a pc-board mount trim capacitor, 470pF screw adjust [mine came from Electronix Express in New Jersey, USA].
I add in a "bandspread circuit" (as you can too, if you're a SWL and you want bandspread in addition to your panel-mounted 365pF Main Tuning cap) which is a varactor diode circuit tuned by a 10K ohm potentiometer. But because my radio focuses on the narrow band of frequencies making up the 80 Meter band, I move the "bandspread" pot into the main position on my front panel, and rename it the Main Tuning control.
The Shortwave listener's 365pF variable cap moves (as I mentioned earlier) onto the circuit board as a 470pF trimmer called the Band-Set cap.
So the pcb 470pF trimmer sets you in the 80 Meter band initially, when you first calibrate the receiver; then you leave it alone and do all your tuning from the front-panel varactor pot. That's why the schematic below shows the 'bandspread' pot as the Main Tuning pot. Be sure you understand this. You can configure the receiver either way; I chose to limit the tuning to the upper half of 80 meters (SSB/AM voice part of the band), so I can listen to all the nasty old codgers and weirdos that inhabit the SSB portion of 75 Meters late at night.
The Main Tuning (the SWLer's "Bandspread") pot is used to vary the + voltage (reverse bias) on 2 parallel 1N4001 rectifier diodes. These are much more common (and cheaper) than a real varactor diode; they work just fine [although they are nonlinear in their tuning response; you will find that signals are compressed on one end of the dial more than on the other]. I found that 2 of these diodes were needed to give a frequency spread of about 150 KHz across the tuning dial. This isn't quite enough to cover 80/75 Meters but if I were to make it wider, it becomes that much more difficult to tune SSB signals in and "clarify" them easily (tuning becomes too touchy).
The capacitance of any diode varies by a few picoFarads when you vary the reverse bias across it; this small variation is effectively in parallel with the Bandset cap (the 470pF trim cap or the SWLer's 365pF panel cap) through a .001uf DC blocking cap.
The radio signal itself enters the regen detector through a 10pF capacitor [and series 10k reistor, added 2/15/07] connected between the output of the RF/AGC amp (Q1) and the "top" (the Q2 collector point) of the LC tank circuit formed by "L" and the Bandset tuning cap.
The receiver is calibrated initally by listening, with a "real" radio set to SSB/CW mode nearby, as the Bandset trimmer cap is turned back and forth. At some point, you should hear a quick "WHOOP!" as the carrier passes through the frequency your commercial radio is tuned to. I calibrated my prototype for 75 meters, so I tuned the store-bought receiver to 4 MHz (the top of the band) and then turned the screw on the 470 pF pcb trim cap until I heard the "whoop". Then I very carefully fine-tuned the screw adjustment to get the regen to be oscillating as close to 4 MHz as possible (taking into account the frequency-pulling effects of a metal screwdriver!). The regen's Main Tuning pot was set to maximum (full clockwise) for this. The regen receiver was then able to tune from 4 MHz down to about 3.85 on the counter-clockwise end of the tuning pot. Since then I've lowered the tuning by 50 KHz, cuz I wanted to. You can set it anywhere you want. In fact, you can set it for 40 meters if you desire. Or, one of the Shortwave Broadcast bands. (Of course, if you use it for SW listening, you will be tuning to frequencies higher than the 80 meter band, which means you must disconnect any 80-meter filter you may have installed, as an option, after the two parallel diodes hanging off the antenna jack to ground. This, in turn, will increase the likelihood of AM broadcast overload from some of the powerful stations in the band; you may need to install a filter for the particular band you're interested in, and/or use a resonant antenna tuner.)
As mentioned in the AGC section above, varying the DC bias on the base of a transistor in the common emitter configuration, varies its amplification or gain. So we use another pot as our Regeneration Control whose wiper voltage varies the DC bias at regen detector Q2's base. A 10uF electrolytic bypasses potentiometer noise (which can be quite noticeable with this sensitive, oscillating circuit). Resistors above (33K) and below (approx. 3.9K) the Regen pot act to scale the "active region" of the regeneration control so that it occupies at least half of the travel of the pot. In other words, the transistor Q2 only trips the threshold into regeneration at some + base voltage, but then fades out if that base voltage is adjusted too high. The result is that there's only a small spot on the dial where regeneration occurs. We want to spread that out (in the same way that the Bandspread control works with tuning) so we "pad" the Regen pot with extra resistors above and below it, which effectively reduces the percentage of voltage-variation that the pot affects, thereby "spreading out" the "active regen region" across the pot knob's travel. (Clear as mud?)
Please note: You may have to adjust that lower 3.9K ohm resistor to some other value in your version of this radio. It may be more convenient to put a small 10K ohm trim pot in place of the 3.9K, which I did in my protoype. Once the trim pot was set to where my Regen control acted the way I wanted it to act, I removed the trim pot, measured it with an ohmmeter, and came up with 3.9K. So then I put a 3.9K in its place, and that's why the schematic shows 3.9K. You may need to modify it.
You definitely need to calibrate that lower resistor so that you have at least the first quarter-turn or so, up from full counter-clockwise, as an area where the detector is NOT OSCILLATING, so that you can bring the detector right up to the point just before oscillation, for good sharp AM reception. Advancing the knob further will cause the detector to start oscillating (you'll hear the background noise come up) and allow you to hear sideband voice and CW (Morse code) transmissions.
NOTE: All regens suffer, more or less, from the following: Once you pop the detector into oscillation, continuing to advance the knob causes a substantial de-tuning of the oscillator frequency. You can use this defect as an advantage -- a "clarifier" for SSB that may be less touchy than the tuning knob -- but I wanted to make sure you were aware of this. We normally don't go "too far" into oscillation when listening to SSB or CW, so as not to de-tune the receiver "too far off" from where we were. Annoying, but a fact of life in the Regen world. By the way, I have read that other Regen designs, such as (possibly) the Armstrong "tickler-coil" variety which uses a variable "throttle" cap instead of a potentiometer for regeneration control, do not suffer as much from this problem of detuning via the regen knob. Someone else has suggested that my fondness for using the Amidon T-50-2 red-gray toroid cores might contribute to the phenomenon-- they may be saturating magnetically when the circuit is set into oscillation, whereas air-core coils would not. But this detuning happens even without toroids in the circuit, which I verified recently when I built the big loop AM radio. So I'm still pondering these claims. Meanwhile, these circuits work, and as minimalists we're expected to put up with some inconveniences in the interest of simplicity of design.
By the way, because the ham-band configuration only allows a range of 150 KHz, you'll have to re-tune the bandset trim cap with a screwdriver to listen to the other parts of the band. E.g., mine tunes from about 3.8-3.95 MHz, which is most of the phone (voice) band (75 meters). To hear the CW (code) portion, I need to tighten up the screw on the Bandset cap. Now I'll be tuning 150 KHz from wherever I re-tuned the Bandset (e.g.,
3.5-3.65 Mhz). You could solve this problem by installing a small-value capacitor (something in the low pF range) that can be switched in- or out- of parallel with the Bandset cap, giving you CW-only versus VOICE-only sub-bands within the 80 Meter band. There are lots of possibilities you could come up with. (Remember, we're minimalists, building minimalist receivers. Add or subtract bells and whistles to your liking, but don't weep if I don't have something that you want in yours! Find out how to add it and DO it!)
Finally, the Audio Preamp.This is the final section of my 3-part schematic. First, go back to the 2nd schematic (the detector) and notice the 10K resistor above the inductor "L". That's our audio pick-off point. The audio is actually detected in Q2 as the "beat" between the incoming signal from the 10pF cap and the internally-oscillating dectector. For AM, the base-emitter junction of Q2 acts like a diode detector.
I like to direct-couple audio stages whenever I can. It saves me having to find the right coupling caps. So the schematic for the 2-stage audio preamplifier, above, shows the base of a 2N3906 PNP directly coupled to the junction between "L" and the 10K resistor. (If you want to use the audio amp in another application, insert a .22uF cap to block the direct DC at the input, then add a 220 K resistor from Q3's base to collector. This gives DC blocking and internal biasing for Q3. I chose to go direct.)
Also hanging off that L/10K junction is a .022uF cap to ground. This cap shunts the remaining RF frequencies to ground, leaving just the audio, which we want to amplify without amplifying any RF.
Q3 is a 2N3906 PNP common-emitter amplifier, whose output is at the collector of Q3 / top of the 2.2K load resistor. That resistor has another .022uF across it, for further filtering and softening of the audio signal. There is enough audio output at this point to comfortably drive an external Radio Shack Amplified Speaker, but not quite enough signal for loud crystal earphone or loud 2000 ohm magnetic headphone reception. Also, we need more amplification to generate a good AGC signal..... so we add one more stage (Q4), which is a 2N3904 NPN, also direct-coupled to the previous stage. This is another common-emitter stage, with a 33 ohm "swamping" resistor added to the emitter to knock out a mild tendency toward "motorboating" that suddenly appeared when I built the circuit into the copper box that you see in the photos.
[And before you ask, yes, you can substitute a 2N2222A for the 2N3904 NPN, and a 2N2907A for the 2N3906 PNP. You may have to adjust some resistor values but I doubt it.]
Now the signal (at Q4's collector) is DAMN loud--- it overloads the Radio Shack speaker amp, and drives the headphones to a pretty loud level; it's just what we need, however, for our AGC signal. So we send it directly back up to our first schematic, where it enters the + side of the 10uF cap in the AGC charge pump circuit. Meanwhile, we split it off and send the remaining audio to a 50K Volume Control pot on the front panel, which also doubles as an ON/OFF switch for the radio. (The 50K ohm pot could also be 10K-100K; not a critical value. My prototype uses a 50K because that's all I had lying around the bench; the schematic shows a 10K for uniformity with the other pots in the radio.)
NOTE: This extra stage of gain (which is necessary for the AGC) does add a little "fuzzy" distortion to the audio. If you prefer a cleaner audio, tap it off of the previous stage, Q3's collector. The Radio Shack amp prefers this, because it has an internal preamp of its own. But if your headphones are crappy and not too sensitive, tap your audio off of Q4's collector -- much louder. I leave it up to you to decide. The schematic shows the coupling cap and audio output jack tapping off of Q3.
[2/15/07 modification: As shown in the updated schematic of the audio section, I added a passive audio filter to kill some more of the high frequencies (whistling heterodynes, etc.), which you may leave out if you don't want that option. I happened to have a 100 millihenry inductor from Electronix Express (the blue cylindrical inductors) whose internal DC resistance measures about 106 Ohms (good to know since higher winding resistance lowers filter Q). After trying various filter configurations, I settled on the passive lowpass filter you see in the schematic, with 2 capacitors in series (for voice reception) and one of them shorted (to lower the cutoff frequency) for CW reception. Of course, this requires another switch on the front panel. Again, feel free to modify as you see fit, or leave the filter out altogether.]
Just for good building practice, a 470uF electrolytic cap sits across the +9v lead and ground (on the radio side of the power switch, not the battery side!) which keeps noise and unwanted oscillation away.
Here are a few pictures of my prototype (copper-clad, "Ugly-Construction", no tuning dial yet, mismatched knobs, Oh ain't she a beauty!). The lower left knob on the front panel is the On/Off switch and Volume control. The middle knob is the Main Tuning (varactor pot). The lower right knob is the Regeneration control. The Radio Shack ("Archer") amplified speaker plugs into a miniature audio jack at the right side of the AGC-80. A hi-fidelity speaker from a pair of Optimus computer speakers can be seen behind the AGC-80 and connects to the auxiliary output jack on the Radio Shack amp. Audio quality off Q3 is good; audio off Q4 is a bit fuzzy, but nice and loud. (My Kenwood TS-520-S "real rig" graces the background, but you're not supposed to be looking at that.)
And that's it! That's my AGC-80 Regen Receiver with Automatic Gain Control which I feel gives a substantial improvement in regenerative radio performance. I hope you will consider putting one of these together and letting me know how yours performs. Feel free to email me at email@example.com