Thanks to Rachel Pervin for her help and patience with the above image. And I apologize for the poor quality of the drawing itself; but then, I'm not a draftsman...
The signal passes from the antenna through the variable-coupling antenna coil into the grid coil. Tuning is provided by the combination of the 140/270 and 20 pF variables. S 1 is a switch which allows the 270 pF section to be switched in and out of circuit. The 140 pF is the bandset, the 20 pF the bandspread. The 270 pF variable functions in effect as a range extender, allowing more coverage from each coil. With it engaged, the total capacitance in the tuning circuit is 430 pF, which is a bit much above about 7 MHz, so the range extender primarily works on the 49m broadcast band and below. I used the 140/270 variable because I happened to have one handy. Any value from 100 to 250 pF should work. Just adjust the number of turns on the grid coil to fit your cap.
The grid lead from the coil goes to the grid leak. Many builders like to experiment with the values used in the grid leak. In theory, the larger the resistance, the more sensitivity, and the worse selectivity, and vice versa (this is because the increased R across the grid leak cap lowers the Q of the circuit, and thus the selectivity of the grid leak).
The coil in the upper left hand corner of the drawing is the tickler coil. It is imperative to connect the tickler and grid coils exactly as shown; if they are connected backwards, feedback will be negative and the rig will not oscillate! If you build a regen, and it will not oscillate, check these connections; this is the most common wiring error for the regen homebrewer.
The 500 pF cap that goes to ground just past the tickler is important; again, the radio won't oscillate without it. In fact, radios employing throttle capacitor control put a variable cap here to control regeneration. The 2.5 mH choke is to keep the RF from the detector out of the 90V B+. Regeneration is controlled by the 100k variable resistor the other side of the 47k plate resistor. This is the pot we vary the value of when maximizing the radio, as discussed below.
47k is a very typical value for a plate resistor; it functions as a load, and the rectified audio is dropped across it, and then shunted from the detector to the audio amp side of the dual triode by means of the .1 mFd capacitor. This use of R/C coupling is one of the distinguishing features of this Doerle design; it all but eliminates the chance of fringe howl and avoids the use of a relatively expensive (and these days, hard to find) 3:1 audio transformer, but sacrifices the gain resulting from the step-up of the transformer. The latter is not a problem with the type 19 Doerle; audio levels are good on DX signals, and painfully loud on single-hop domestics (especially Radio Marti!!!!). So, loud, in fact, I added the 500k volume control pot, which is an audio taper model. This control was not used in the original design.
The amplified audio voltage is dropped across the audio transformer, and coupled to the headphones via the .01 mFd capacitor. This is a safety feature I added. The original design (and indeed, many a design of this era) sends the B+ voltage straight through the 'phones. Very few people were ever injured by this; but I have a toddler running around and the use of the coupling cap keeps the B+ exposed in only a few hard-to-reach places above the chassis. (Putting the radio in a cabinet would, of course, be even safer.) I used the old voice coil plate transformer out of a Ward's Airline radio-phonograph that was unrestorable (I had no cabinet for it). This has the additional advantage of providing a source of 8 ohm audio, so I can hook this radio up to the stereo or other common audio amp to demonstrate it! Voice coil plate transformers are hard to find, though, and any choke of several henries would probably work as just as well.
When I was finished wiring the radio, I checked the functioning of the audio section by touching the wiper of the volume control with a screwdriver. No problems. The I moved on to checking the RF section.
When discussing the regen control above, I made allusion to maximizing the radio. This is done in two steps. First, the regeneration control is checked for correct operation. We want the greatest variation in plate current to occur from about the twelve o'clock position of the regen pot to the full clockwise position.
As we start off with no plate voltage, no current flows from the cathode to the plate (naturally). As we increase the voltage, the current flow first starts and then increases; but, of course, only to a point. When further increases in plate voltage bring little or no increase in current we say the plate current has reached saturation. Ideally, this point where plate voltage increases no longer cause increased current flow is also where we want the radio to break into oscillation. In practice, most people set up a rig so the two points are very close, if not exactly the same. It is convenient if this point occurs between twelve and three o'clock of the regen pot's arc of rotation; the actual location of these points will vary quite a bit on different bands or with different antennas, as well as the condition of your batteries, so don't forget to hedge.
You can check this with a voltmeter and ammeter; what I did was to wind the grid coil, insert the coil in the plug-in base, and use my MFJ antenna analyzer to generate a loud RF signal. The detector should function as a simple grid leak detector on loud signals even without regeneration. I made sure I could hear the antenna analyzer and ascertained that the signal had more or less stopped increasing in volume by the time the regeneration control was fully clockwise. The value of the control can be varied to put the point of current saturation where you want it, but be careful; to small a pot allows too much current flow on the B+ and runs the battery down quickly. You should have no trouble if you use the values shown.
The second step in maximization is to wind the tickler coil so that the radio breaks into oscillation at around the same place maximum gain takes place (again, twelve o'clock to fully clockwise on the regen control). This is done by adding turns to the tickler if oscillation does not occur or removing them if it occurs too soon. This takes time but really makes a difference in the sensitivity of the radio. If you want to hunt DX, you will have to maximize the tickler coils (although again, the pot value should be ok as shown).
As implied above, the B+ and A+ can be independently switched. This is valuable for changing coils, as the B+ for the plate of the detector runs through the tickler. So, the B+ must be switched off when changing coils to avoid shock; but repeated switching on and off of the filament is bad for the tube (much as it would be for a light bulb).
See C. F. Rockey's book Secrets of Homebrew Regenerative Receivers and especially T. J. Lindsay's Build the Twinplex Regenerative Receiver for more useful information (available from Lindsay Books). In addition, a quick search of the web will reveal several sites with more on homebrewing regens.