June 22, 2008
The FETish® comes together.
I wanted a superhet receiver to transceive with my Milliwatt CW transmitter. The goals of this project were:
- Learn about FETs.
- Make a usable receiver.
Superhet receivers aren't trivial, but they use the same types of circuits as transmitters: oscillators and amplifiers. The main difference is in signal levels. A receiver must make a minuscule signal from your antenna strong enough to vibrate your headphones. According to Solid State Design, a gain of about 130 dB will do it.
The first question, then, is how to distribute that gain. Direct-conversion receivers get most of their gain in the audio amplifier, leading to problems with hum pickup and microphonics, where the circuitry acts like a microphone. Superhet receivers usually develop most of their gain in the IF stages. ICs used in typical superhets have a gain of about 50 dB; some receivers use two, some only one. An active mixer IC like the SA612 can add 15 dB. If you use one for the mixer and another for the product detector, you've got close to a quarter of your needed gain right there. Because I wanted to use FETs in as many circuits as possible, I first had to determine the gain possible from a singly-balanced FET mixer.
The Front End
Most of the articles I've read suggest first-time receiver builders start with the audio amplifier. But without knowing how much gain I'd need at that point, it seemed wiser to begin at the beginning: the antenna port. I knew I needed a band-pass filter up front, and they introduce some loss. I used a double-tuned circuit, which I designed (using AADE Filter software) for a bandwidth of about 500 kHz. I had previously decided on an IF of 11.850 MHz, with the local oscillator operating above the IF, at 25.850 MHz for 14-MHz reception. That puts the image frequency at nearly 40 MHz. You want to restrict the image response to attenuate noise at that frequency, even if you don't expect image signal interference. Of more concern was signal breakthrough at 11.850 MHz. Using my oscilloscope as an rf voltmeter and sweeping the filter, I determined the loss was about 5 dB. Nothing like starting a receiver project in the hole!
Next up was the preamplifier. Preamps generally aren't needed on 20 M, and can cause more problems than they solve. The preamplifier sets the noise figure of the receiver. Below about 25 MHz, though, external noise usually determines the weakest audible signal. While a preamplifier also adds gain to the receiver, the lack of selectivity ahead of it can lead to overloading and distortion on strong signals. Still, I was itching to start using FETs, so I built one.
My first FET preamp was a flop. Even without using tuned circuits, with a single FET in common-source configuration, I couldn't keep it from oscillating. I think that, installed in a chassis, it might have settled down, but I didn't want to risk it. The preamp thus set the tone for the rest of this project, in which nothing worked right on the first atttempt!
I tried to treat this project like a job of work. That is, I put in a full shift at the bench, but set aside time to study, and almost always stopped work at dinnertime. But these things will get into your blood. One night I was browsing through W1FB's QRP Notebook, and spotted an interesting preamp circuit, on page 96. Doug said he borrowed it from a commercial Japanese transceiver. The next day I built one, using J310 FETs, and it was completely stable. With the J310s I even got a little more gain than Doug, 15 dB instead of 10. Good! Now I'd compensated for the filter loss and had a few dB in the bank.
An earlier version of the rf preamp, and the mixer, under test.
Mixer
Things really started getting interesting with the mixer. I planned to use the same K6ESE DDS that serves as a transmitter VFO, for the local oscillator. I'll have more to say about this DDS later. For now, it wasn't developing the 2.5 V p-p W1FB suggested for a singly-balanced FET mixer (QRP Notebook, page 93).
After fooling around with FETs for a few hours, I broke down and built a little amplifier to bring the 100-mV output of the DDS up to 2.5 V, using a pair of 2N3904 bipolar transistors. I used the W1FB mixer circuit, with 2N3819s this time. No particular reason, except that their noise figure is comparable to the J310s and I had a bunch of them on hand. In place of the 100-ohm source resistors, I used a 250-ohm trim pot, with the wiper connected to ground and the ends connected to the FET sources. I connected my DVM across the ends and adjusted the pot for 0 V, meaning each FET had the same source voltage. It was possible to balance out the rf feedthrough, but hard to measure it. The LO cannot be balanced out in a singly-balanced mixer.
IF Amplifier and Filter
I think I could write a book about this section, the one that gave me the most trouble. The filter was simple. Having decided to use four crystals in the main filter, and a fifth one after the last IF amp to reduce noise, I dumped the bag of 11.850-MHz crystals (obtained without specs on eBay) on the bench, and fired up a crystal oscillator-buffer connected to my counter. I didn't bother measuring anything except their frequencies, which I matched to within 10 Hz. Some crystals stayed on frequency when plugged into the oscillator, while some drifted slightly. I rejected the latter.
Sorting crystals for the filters.
Having sorted the crystals, I put together the 4-pole filter. Zack, W1VT, used a 12-MHz IF in a 40-M transceiver published in QEX, and later in QRP Power. With 470-pF ladder capacitors he obtained an impedance of 50 ohms. As my crystals were close to 12 MHz I hoped to get lucky, so I also used 470-pF caps. Terminating the opposite end of the filter with a 47-ohm resistor, I connected a pot in series with the input (either end will do in this filter) and connected my signal generator through an attenuator. Then I adjusted the pot so half the applied voltage (measured on my scope) appeared across the pot. The dc resistance came out to 68 ohms. Close enough!
The IF amplifier went through four versions, and consumed more than half the total time spent so far on this project. Before the ladder filter, I had planned to use a double half-lattice filter, just to try it. For some reason not entirely clear now, I decided to put one stage of IF amplification ahead of the filter, and the rest after it. I think the reason was, I was unsure what the filter would look like dynamically, and didn't want to hang it right off the mixer. That filter worked well, but had a lot more loss than I liked. You need crystals offset by some finite frequency to make a lattice filter, whereas a ladder filter's crystals are all on the same frequency. The separation should be about 66-percent of the desired bandwidth.
(Mass-produced crystals are usually within 100 Hz of their marked frequency, and I had a hard time finding two pairs far enough apart. My lattice filter had a -6 dB bandwidth of 270 Hz, and an impedance of about 300 ohms.)
Eventually, I decided it was too lossy and abandoned the idea. Throughout this project I felt the need for adequate gain hanging over my head. Looking back, that drove me harder than it should. Being my first superhet, I didn't have enough experience to make all the right decisions. You learn more from your mistakes, right?
Version I of the IF amp, with double half-lattice filter.
At this point I had one gain stage ahead of the filter and two following it. Plunking the ladder filter in the gap, I was delighted to find its loss was too low to measure. Whatever the total gain of the IF stage was at that point, it's in my notes but I don't remember. It gnawed at me, though. First, was there enough gain? Well, easy enough to get more. I stuck another cascode stage on the board ahead of the filter. These cascode stages were borrowed from Experimental Methods in RF Design (EMRFD).
Version II of the IF amp, with cascode amplifiers and ladder filter.
Now I had lots more gain, but something else bothered me. I didn't like having some 50 dB of gain before the crystal filter. That just seemed wrong. No problem, it's easy to move things around when you use "ugly" construction. Thus began Phase II of the IF amp.
I swapped the positions of the filter and the first two IF amps, so now the filter came right after the mixer and ahead of the IF amps. Lesson learned: Even with care, you can't jam that much gain into a small area and expect stability. But everyone should have a four-crystal, eight-FET oscillator once in a lifetime!
Version III of the IF amp, with relocated ladder filter and crammed amplifiers.
Cutting to the chase, I ended up making a new board, with the filter detached, long enough to get some air between the stages. Stability at last!
Version IV of the IF amp. Stability at last!
I was expecting about 100-dB of overall gain, but I'm only getting a little more than 80 dB. I attribute most of the loss to the ferrite-core transformers. A little more work might fetch another 10 dB from this stage, but I was mighty sick of working on it, so I left it alone. The green LEDs you might notice in the photos provide regulated bias for the lower FET in each cascode pair. EMRFD shows four signal diodes connected in series as a 2-V regulator; the LEDs provide about 1.7 V, plus cheerful confirmation that the FETs are drawing current. The control voltage for maximum gain varied stage to stage, but 6.8 V is close enough for all of them.
BFO and Product Detector
I built this part last, but it belongs here in the narrative. The BFO taught me an unexpected lesson. A few years ago I ordered 2N5484 FETs from Jameco. That FET has a low Idss (saturation current) and the Idss varies little from device to device. This makes them handy for VFOs, where you don't want the device to generate anymore heat than necessary. Unfortunately, Jameco substituted NTE 312 devices. The datasheet for the NTE 312 states a higher and wider Idss range, so I set them aside. I got the real thing from Mouser. I fished one of the NTE 312s out of the drawer to use in the BFO, and consulted the 2N5484 datasheet for the pinout. FET pinouts vary, so if you substitute, always check. Well, I should have taken my own advice. The pinout for the 2N5484 is Source-Drain-Gate, but the NTE 312 is Gate-Drain-Source. You can get away with swapping drain and source, but the gate better be where it's supposed to be! Jameco has been stricken from my list of semiconductor suppliers.
True to form for this project, the first FET crystal oscillator had problems. I tried a VXO with the crystal, inductor and trimmer cap from the gate to ground. Worked fine with only the crystal, but as soon as I added the inductor, it quit. Different inductors and parallel resistors didn't help. The next one I tried had the three connected between drain and gate, and it worked right away. Not only that, but it is very stable with changing supply voltages. This is unlike the bipolar VXOs I've built, which all needed regulated supplies. The BFO got a FET buffer, too, which delivers about 1.5 V p-p to the product detector.
The product detector is just like the mixer, except the output load is an audio transformer All I had that was small was a 1000-ohm isolation transformer from Radio Shack. It doesn't have a center-tapped winding. Connecting one end of each winding together, so it looks like a bifilar transformer, seemed to work fine. With the audio amplifier connected the mixer is very sensitive, hearing well below 1 uV at the IF.
Audio Amplifier
To keep the FETish thing going, the first stage of audio amplification is a FET-input preamp circuit by Terry Ritter. Even that had to be modified! I used a 2N3904 instead of the 2N2222A, a 220-uF bypass cap for the FET, and found I had to swamp the input with 1000 ohms to reduce hum. This stage does provide 25-dB of very quiet gain, though. The two red LEDs regulate the preamp supply at 3.4 V.
FET-input audio preamp (below) and product detector.
After that I used two sections of a TL074 FET-input quad op amp, one for a low-pass filter, the other for some gain. I think the preceding stage is overdriving the filter, because the amp didn't like strong signals. I have that bypassed for now, running the preamp directly to an LM380 IC, connected per the datasheet, except... I hung a 1000-ohm resistor after the output electrolytic, to charge it in case there is no load plugged in when the receiver is turned on. Otherwise, there would be a resounding crash when I plugged in my headphones.
To mute the receiver, I put a 2N7000 MOSFET after the preamp. The transmitter is keyed by applying +12 V to the driver stage. I plan to use a bias tee to send that keyed +12 to the receiver, through the receiver antenna cable (I'm keeping transmitter and receiver separate to make for easier modifications). So the 2N7000 is connected across the audio line. When +8 V (the +12 V signal through a voltage divider) is applied to the gate, the FET shorts the audio signal to ground. This got me thinking about a simple audio AGC circuit. A superhet properly has AGC applied to the IF amplifiers, but all I want is protection from strong signals, so they don't blast my ears. And a simple audio AGC would be nice to have in a direct-conversion receiver. Some experimentw with the 2N7000 show that a dc bias of 2-4 V applied to the gate make it act like an audio attenuator. Once I get the rest of the receiver sorted out, I plan to tap off and rectify some audio, and amplify the dc to the right levels to use the 2N7000 as a volume limiter as well. I've never seen this done before.
So, Does It Work?
At last it was time to string all the pieces together and see if they made a receiver. The short answer is, yes, it works. Of course, there are problems. The BFO is leaking into the IF through the antenna terminal. That's no surprise, considering everything is spread out on the bench. There's still some audio hum, but most of the noise problems I was tearing my hair out over, went away when I turned off my soldering station. The temperature-regulator is a great noise generator.
The biggest problem, and the one I'm going to tackle next, is the excessive number of spurs generated by the DDS. Now I'm really glad I built this receiver. Although it's providing a signal at nearly 26 MHz, there are birdies all across the 20-M band. It desperately needs band-pass filtering. Had I put the transmitter on the air with this DDS, I'd likely have transmitted all over the place. All that might have saved me from hot water was the low output power of the transmitter.
The problems will be tackled one by one, and very soon the receiver will find its way into an enclosure. Although sometimes aggravating (which was its way of telling me I didn't know what I was doing!), this project was lots of fun. You never know what you're going to learn when you start a project, so a surprise waits around every turn. I know where I'll be using FETs in the future and where I won't. With this project under my belt, I'm feeling ready to take on projects of wider scope. But first I'll wrap up this one. Another lesson I've learned is to finish one project before moving on to the next. Otherwise, you end up with a bunch of almost-finished projects, that haunt you every time you look at them.