A constant work-in-progress... my approach to using the marvelous single-signal direct-conversion R2 receiver design in a multiband, modular amateur station.

In the top photo, you see how it was laid out, and used, for several years. Yes, a little haywire, but quite functional. The boxes are made from double-sided PC board material. The BNC jacks are a special PC board type that I found cheap at my local surplus house. They mount with 4 ground pins and a center conductor pin, saving space.

Interconnections were a combination of Molex connectors, BNC's, clip leads, etc. 

The modification of the surplus Collins T-368 exciter is a separately documented project in itself! This unit was available from Fair Radio Sales in the early to mid 90's for about $45; many correspondents tell me this source has dried up. The basic PTO is 1.5-3 MHz, with multipliers to 3-6, 6-12, and 12-24 MHz with a mechanical digital dial. Once transistorized, it is the most stable analog VFO I have ever used, mostly because it is built so heavily. Others may wish to use various combinations of homebrew VFO's, multipliers, or pre-mixers, especially for portable use!

I started out with boxes that combined input filtering and phase shifting.  BNC connectors on both the input module and the R2 were on 3/4 inch center to center spacing, with a set of male-to-male BNC adapters remaining "resident" on the R2. I switched filters and phasing networks with one 4PDT slide switch. Rick's suggested "pi" phasing network in the R2 article is easy to deal with, tuned up predictably, and worked well. I think, though, with the haywired approach, there was more RF radiation than I might have wanted. I got by, but hum started to show up on the high bands (LO leaks out to an AC power line, gets modulated, and re-enters the radio). Microphonics and not-perfectly-noiseless connections also showed up, but usually was tolerable. The photo below gives you an idea. <>

LATER: I turned to the twisted-pair quadrature hybrid, with greater bandwidth, and needing one less piece of iron to wind on. In progress now is the "Quadrature Expressway", a surface-mount approach to making neat networks bandswitched with PIN diodes. Details here.

The input filters are standard 2-resonator bandpass filters found in various homebrew books. They're adapted to tune a range, using dual 250 pF plastic variables (Mouser Electronics).

The "front end" and audio amp of this board date back to my first effort at the "iron-on toner transfer" method of PC board etching - barely enough copper left, but it works! The original audio quadrature section was cut out and replaced with Glen Leinweber's circuit from the R2a.  The circuit board on top contains 2 74HC240 CMOS buffers to generate square wave drive - a good idea in theory, but in practice a bit unstable, so they're now bypassed. Note the cool construction, though - semi-rigid .141" hardline soldered to the board, and SMA connectors!

Rick's R1 article has some precautions for power connections. Because of the high gain on one board, any circulating ground currents generated by the output stage must be minimized. Therefore you don't just connect the DC negative lead to any old point on the chassis, but right to the point where the output transistors are grounded. If you go him one better, and return the SPEAKER ground lead to this point also, you get even more immunity to feedback. This means to insulate your output jack from chassis ground (easy with PC board boxes: simply scrape some copper away).

In my new cabinet setup, I chose to keep the R2 board completely insulated from the chassis, making it possible to experiment with various grounding options to see if it makes any difference.

Call it overkill, but I found some high-level mixers at a hamfest, the SRA-1H (+17 dBm LO drive, I think), and have used them for some time.

I don't have any preamps or front-end filters in there right now, trusting in my transmatch and transmitter lowpass filters to reject birdies. Eventually I'll get them all switched in somehow. You usually can get by without the preamp up to 20 meters, though 20 meters can use it sometimes. I find no need for it on 40 meters and below.

When I've used preamps, these are what I've tried:

1) Mini-Circuits MAR-6 MMIC amp -- 3dB noise figure, 20 dB gain, 1 MHz-2 GHz bandwidth. It works well, but that's a lot of gain, and its output 1 dB compression point is only 2.5 dBm. But if you really want to listen down to the noise on a quiet band, or need something that at least gets you a start on VHF premplification, it's great. A little kit is available: the WBA-6 ($19.95) from Electronic Rainbow in Indianapolis.

2) A 2N4416 grounded-gate stage (fig. 3). I can't remember where I read these specs, but I believe it has about 4 dB noise figure, higher overload margin, and a more manageable 10 dB gain. A better choice for HF, though I found the gain at the high end rolled off (maybe I need a better-selected transistor...)

My amplitude balance pot is put on the front panel, because readjustment is required for each band. You will most often get adequate rejection of the opposite sideband with a reset from memory. Best adjustment requires a steady carrier or VERY strong signal. Your phasing networks should only need to be adjusted once, unless you change your LO drive in some way.

See this modification list for information on small tweaks in the filter designs specified. Another important mod: The capacitor in the mute circuit should be reduced to .03 uF if you want fast recovery for full break in CW. The 0.1 uF originally specified means a good half-second recovery time - far from adequate for a CW man!

I have switching for a CW and SSB lowpass filter -- and can switch them both out! It's an interesting hi-fi sound to hear signals above 3 kHz. My ears hear to 10 kHz or greater in my hi-fi headphones, serving as a mini spectrum analyzer. Strong shortwave broadcasters sound stunning this way. At high gain settings, though, you may get some oscillation, sometimes supersonic.

Switching out the highpass filter is another possible option. At first I did it with some trace cutting on the original board. With the changes made when the "R2a" circuit was added, it made me think about playing around with some highpass filtering as well. So I now have a 3-pole highpass filter , the inductors being wound on the ferrite potcores Glen refers to in his writings. It's set to roll off about 500 Hz to narrow the CW bandpass, and for voice modes it's just bypassed. Audio switching, pots, and filtering are precariously hung onto the front panel as displayed in the photo!

Any VFO you use needs a varicap for CW and/or RIT offset. See the Lewallen Optimized QRP Transceiver (1994 ARRL Handbook, QRP Classics) for a good basic circuit. I implemented it this way in my T-368 exciter:

Pick the capacitor for the right RIT range for your needs. Mine had to be big because the PTO operates at 1.5-3 MHz. Also be creative on your varicap diode. The NTE612 is designed as a varicap. But many diodes will work as varicaps. Lewallen suggested a zener diode. Try some old rectifier diodes in your junk box -- something that is likely to have a larger PN junction than a fast switching diode.

Lewallen's RIT switching is also OK as far as it goes. However, you can get offset above and below your frequency just like the big rigs with this circuit.

Pick the capacitors for the right delay for your needs. Remember that when you unkey, the VFO needs to stay at the transmit frequency long enough for the PA output to decay.

The RIT pot is on your front panel, of course. Pamper yourself and use a 10-turn pot -- you won't regret it. The transmit centering can be an internal trimpot, or a front panel control. If you wish, add resistors on either end of the pot(s) to limit the tuning to the most linear part of the varicap's range. The SPDT switch removes RIT by keying the circuit. On CW, you would turn off RIT (switch to ground) to spot an incoming signal to zero beat, then turn on the RIT and adjust the RIT pot for your preferred beat note. (With experience and an ear for pitch, you don't need to do this every time). On SSB, of course, you normally tune with RIT off, and turn it on only as needed. Remember also that the diodes at the output will be part of your VFO thermal stability concerns. As temperature increases, the forward voltage drop goes down, thus the voltage increases, and your VFO frequency will creep up. To some degree this compensates for the way the varactor diode behaves: it creates a frequency drop with increasing temperature (increasing capacitance).

A very interesting document on the fine points on tuning diode temperature compensation, is the old Motorola bulletin AN847 - with today's fancy synthesizers it's marked as an "archive" document, but us conservative hams still using varicap diodes can still benefit from it.

In my first iteration, blocked out above, I started with some driver gain blocks, and the final based on the "10-watt linear amplifier" (2 NTE236's) described by W1FB in the QRP books, with a more elaborate bias regulator.

Funny, that PA never worked quite right. But eventually, I found my way to the circuit's roots: Motorola's classic old application note AN779.  Turns out Doug's circuit was the second half of this. I built it up in my own haywire fashion, and it gave out a good 15 watts... but was never quite stable. Then after a few years of using this, I hit on the ingenious and unique idea of using the board layout provided in the app note - and it worked flawlessly! Lesson:  good layout does the trick, and if somebody's done the work for you, don't re-invent the wheel. So here it is:

It's a nice little 20 watt amp design - 2 MRF476 drivers, and 2 MRF475 output devices, both in push-pull. (These devices are no longer made, so I used the NTE235 and NTE236.) I found the biasing a bit problematic, wasting a ton of current to heat a resistor through a diode, and even letting the transistors go into thermal runaway! So I provided some adjustable bias via pass transistors for both stages, based on the Ten Tec Argosy once again. (All that stuff is slapped onto the heat sinks, which are MUCH bigger than the small black metallic items specified in the original.)

This is driven by a single 2N5109 linear stage (right). The board at the right includes another little 2N5179 driver, used earlier when more gain was needed - it's bypassed now.

The DC switching part is cobbled from some circuit I read somewhere, and modified. The IRF510 MOSFET is used for its high gate impedance and sharp switching characteristic, otherwise the capacitor (.47 uF) that holds the switch closed on key-up has to be outlandishly large. There are probably smaller MOSFETs available for such low-current switching, but this one you can get at your local Radio Shack.

The RF switch itself is drawn from the Ten Tec Argosy, and handles that rig's 50 watts with ease. 1N4007's are 1 kV rectifier diodes that have a PIN structure; they operate satisfactorily, though they are not rated for RF PIN service. The 1N914's rectify the transmitted RF to back bias the diodes on transmit. At 5 watts, I have alsof had satisfactory results with 1N914's in the switching path; in this case you do not need the back-biasing circuit, because the fast switching diodes rectify their own back bias! I don't use the output transistor now, but I may try it to cure some receiver audio pops at high volume levels. The chokes are small molded ones at this power level; actually, the one on the transmit end could be eliminated if your DC return path is through an output transformer winding.