Derwyn Williams. ZL4SAE.
Published in QST Feb 2017.
Ver 1.2 Corrected some spelling mistakes. Added better image of receiver. Updated schematic.
Having developed an interest in WSPR last year, I purchased a very neat transmitter kit from QRP-LABS. The Ultimate3S. Mentioned elsewhere on this site.
( WWW.qrp-labs.com )
The popular kit worked as soon as it was powered up. After a couple of days of successful ‘beaconing’ on 30m WSPR, and despite owning a couple Elecraft transceivers, the urge to construct a suitable WSPR receiver overcame me and I began researching the project in earnest. Most designs seem to be direct conversion architecture mainly based on the ubiquitous LM386 and NE602 et al. I built a couple of these designs and wasn’t happy with the performance. The 602 is not known for its signal handling, and the LM386 is another problem area as it is usually pushed to its limit in an effort to simplify the circuit. I would happily trade simplicity for proportionally better performance.
The 602 is an elegant part even with its limited performance. This got me thinking about a really narrow Band Pass Filter in the antenna circuit, thus limiting the energy reaching the mixer input. I decided that a narrow bandpass crystal filter would be ideal. A superhet design, although more complex, would give single signal reception, avoiding the usual image problems associated with direct conversion, apart from other known issues. I was surprised to find that some Elecraft QRP rigs use the NE602 as a receive mixer, and even the very well respected K2 uses it as a product detector. Suitably encouraged I pressed on.
I already had a number of 10.140MHz crystals and hundreds of others. While checking through my collection of crystals, in order to select a suitable I/F to work with, I found some 10.7 MHz parts. A quick calculation determined that the VFO or local oscillator would need to be 560 kHz. Oscillator stability shouldn’t be a problem at that low frequency. Something about the frequency jogged my memory about remote controls for TV, Hi-Fi and such devices. A quick search of the web revealed some 560kHz ceramic resonators for sale on Ebay. The seller was in Germany. A pack of 5 of the 560kHz parts were ordered that same day. Shipping cost was reasonable, and the parts arrived within a few weeks. In the meantime a number of circuit designs were sketched out.
I reworked the circuit topology on paper and decided to use the resonators in the I/F and also in the Carrier Insertion Oscillator driving the product detector. The 10.7 crystal would then be used as the local oscillator. This was beginning to look interesting.As soon as the parts arrived I quickly determined that the NE602 would oscillate well with the resonator,using a small inductor and a trimmer capacitor.
The first part of the circuit that was built is shown here. The product detector and Carrier Insertion Oscillator based on a SA612A. Applying a 560kHz signal from my home-brew synthesised signal generator, to the mixer, produced an audio tone in the high Z earpiece that I used for monitoring the very low level output from pins 4 and 5.
The inductor is 470uH and the capacitor is 15pF fixed, in parallel with a 60pF variable. These are parts that were available at the time and a better balance of values should be selected. The oscillator could easily be ‘pulled’ high or low as witnessed by the changes in audio tone.
I/F Phase splitter.
The novel part of this circuit is the phase splitter feeding the input pins of the product detector. I have not seen this done before and I’m not sure if it is more effective than a transformer. However it appears to work fine. This takes the single ended I/F signal and drives the mixer in push-pull. The source and drain resistors are 1k5 in an effort to match the input impedance of the mixer. I removed the internal capacitor from a 455 kHz I/F transformer ( yellow core ), then measured the inductance and from that calculated the required capacitor value. Despite my efforts I could not peak the signal. I then accidentally discovered that the transformer was self-resonant at 560kHz, so no capacitor was fitted. I then easily found a peak in the signal, which was audible even off tune. The FET gate resistor is 1M, making the input very high impedance and less likely to affect the transformer circuit preceding it. An inductor and capacitor were selected as a 850kHz LPF, and fitted between the transformer and the gate of the FET. This removed a 1MHz approx, spurious signal that I could see on the output pins when monitoring with an oscilloscope.
The first mixer.
This has the 10.7 local oscillator used to convert the antenna signal to the 560kHz I/F. The output of this stage is DC coupled into the base of the emitter follower I/F stage. The emitter resistor is 1k5, in an effort to match the impedance of the mixer. The 560kHz resonator is used as an I/F filter, driven from the emitter circuit, and feeds the low impedance winding of the I/F transformer. I later added another resonator. See the alignment notes.
The front end
The antenna input circuit uses a ready wound transformer of 2.6uH, available from ( WWW.GQRP.COM ) as 2u6L. This needed 100pF to resonate at 10.140 MHz. It’s fairly broadband, but the tuned circuit’s main function is to match the low impedance of the antenna (hopefully 50R) into the high impedance of the crystal. It will of course provide some filtering thereby protecting the crystal from out of band high energy signals. The wanted signal filtering will be done by the crystal.
An alternative antenna circuit could be built around a T-37-2 toroid holding 26 turns, 50pF fixed capacitor and 60pF variable in parallel. The primary winding could be 4 turns. The toroid is best mounted clear of the PCB. I moved the voltage regulator to enable short routes to all parts of the circuit.
Finally the audio stage, which is also driven in push pull. The NE5534 is fairly low noise and at this low signal level is unlikely to overload, despite the high gain of the circuit. A preset resistor is fitted at the output in order to facilitate level setting of the audio into the PC or laptop. In many years of experimenting with computers and radio interfaces, I have never damaged a PC. However I suggest you fit an isolating transformer if you have any misgivings.
As an alternative to an audio isolating transformer I have used a USB sound-card with great success. These devices are very cost effective ($3NZD) considering their utility, and certainly cheaper than a transformer. Windows 10 recognised them immediately and automatically loaded the requisite driver. No further software installation is required. You will need to select ‘PNP USB Device’ or something similar in the WSPR-x audio setup. On the PC or laptop you will need to set the USB microphone gain to at least 50, maybe more.
The NE602 maximum rated working voltage is 8v. On no account exceed this. A low noise linear power supply, rather than switch mode, is highly recommended. I used a SLA 12v battery.
You will need an accurate signal source of 10.140MHz. I recommend using the filter crystal as a temporary test oscillator. Fit a 4p7 capacitor in its place in the receiver circuit. Replace it with the crystal once alignment is done. If you have a transmitter, such as the QRP Labs Ultimate3S already running WSPR things will be a lot easier
First, it always pays to check that there are no low resistance readings from supply pins to ground. Ensure your supply polarity is correct. ( ask me how I learned this ! ) Check that the CIO variable capacitor is set to half mesh. Connect a pair of headphones. Apply 12v DC. If you can power it up with a milliammeter in circuit, the reading should be less than 30mA. My measurement was 23mA.
I have used this oscillator circuit many times with no problems. Just place the oscillator close enough to the receiver so that it can be heard in your headphones and can therefore be detected by the monitoring software later. Set the audio output pot to about 3/4. Tweak the antenna coil for loudest signal. Use the correct tool or you WILL break the core. Similarly tweak the I/F transformer. Connect the audio output from the receiver to your sound-card. I monitored the audio in Spectran.
My transmitter was sending the WSPR messages every 2 minutes. By adjusting the variable capacitor in the product detector oscillator circuit, it was possible to line up the waterfall signal in Spectran exactly on the 1500Hz marker. The waterfall in WSPR-x is unsuitable for this. Switching from Spectran to WSPR-x and waiting a few minutes confirmed reception of my own signal. At this point I strongly suggest you put the receiver in a screened enclosure and surround it with some polystyrene foam in an effort to stabilise the temperature. The device that I found to be most temperature sensitive is the 470uH inductor in the CIO. I might replace this with a different type of coil. Maybe a toroid. It may be necessary to re-tune the CIO after boxing up and allowing the temperature to stabilise.
Later testing using a signal generator and headphones revealed that the opposite sideband was not very well suppressed. So I added a further resonator, in series with the first, and a 180pF capacitor to ground at their junction. This greatly improved things. Best dx heard on the first day using a 35ft vertical antenna, was EA, at over 19,000 km from New Zealand.
Travelling around in a motorhome I don’t have the luxury of a workshop. I am sure that those with access to more/better test equipment could come up with something with even better performance, or at least optimise some components that I have only guessed at. Have fun.
Best 73 Derwyn.
ZL4SAE ex GW4SAE