Note: Al wrote this very good article that needed to be in the public domain but did not have a place to host it so I offered his article a home. All correspondence on this topic should be directed to tubadoc <AT> cfl.rr.com. Thanks!
I have become a master at building electronics from the junk box. I started building amateur radio equipment when I first got interested in radio at about age 8 when I built a crystal set. But everyone used a wooden base with an oatmeal box in those days. But at about age 11 I passed my novice license and was ready to build transmitters and receivers. But 11 year old kids don’t have money and my parts came from trash cans, like old radios and TVs, and other helpful ham’s junk boxes. I started my own rather complete junk box and to this day, 60 years later, still maintain a well-stocked junk box.
This note describes making a 630m superheterodyne receiver but the message is how to “manipulate the numbers” to use inexpensive and readily-available parts.
I built a receiver using I and Q inputs from WSPR but was unhappy with it primarily because WSJT-X doesn’t support I/Q operation. The huge advantage of designing an I/Q receiver or transmitter is it obviates the need for a very narrow band IF filter for unwanted sideband suppression or image rejection when used for frequency conversion. It accomplishes the filtering at baseband using op-amp filters.
But my junk box had some candidate filters for a superheterodyne. Generally, for WSJT you will need a filter of a few kHz bandwidth and some serious skirt selectivity to reject the unwanted sideband and help in carrier oscillator leakage. That usually means a crystal or mechanical filter or an LC filter with a low center frequency such as 50 kHz. This low IF was used on some receivers to avoid the expense of a crystal or mechanical filter but since the image is only 100 kHz away a 50 kHz IF usually followed a 455 kHz IF in a double or even triple conversion receiver. Since I wanted a simple single conversion receiver the IF filter would be either crystal or mechanical. (Actually they both involve mechanical oscillation but that is a subject for another time.)
My junk box had both. The crystal filters in the box were at 4 MHz and higher and did not have an ideal bandwidth for 630m operation. The mechanical filters were 455 kHz types with one exception. A 455 kHz IF for a 475 kHz receiver is going to pose some serious issues being so close to the RF frequency. The “odd” filter was a 250 kHz mechanical filter. It was a 1960’s filter of huge proportions and was used in some sort of military equipment. The bandwidth was 6.7 kHz which would allow nearly the entire 630m band to be included.
When using any IF filter one has to be careful of how they are described. For example, a “455 kHz” mechanical filter usually doesn’t pass 455 kHz at all. It is made to pass a sideband for a 455 kHz virtual carrier. But this 250 kHz filter had a center frequency of 250 kHz. Before I eventually found the specs on the filter I had swept the filter with my signal generator and knew its passband.
When designing a superheterodyne receiver the characteristics of the channel filter, which in this case is the only selectivity at 250 kHz, sets all the other frequencies in the receiver. For my receiver the carrier frequency must lie outside of the passband of the channel filter in order to get sufficient unwanted sideband rejection. With a 250 kHz IF and a 475 kHz channel frequency high side injection, meaning the local oscillator frequency is above the input frequency, would be necessary; there is no other choice. High side injection has other advantages so it would have been my choice even if I could use low side injection. With high side injection you get a spectrum inversion and thus what the IF would receive would be the opposite sideband of the received signal. I wanted to receive USB. Therefore I wanted a carrier oscillator that was just above the IF filter’s passband allowing the filter to pass the LSB.
Now here is where things get interesting. I have some leeway in the carrier/BFO frequency but not much. I wanted the carrier oscillator to be maybe about 20 dB down on the filter’s upper skirt and no more than about 800 Hz from the 3 dB down point of the passband. So I made a spread sheet where I could take standard oscillators, crystals, and TCXO frequencies and divide by integers and show those that come reasonably close to a useful frequency after division. The inputs were from the Digikey catalog. As it turns out a 16 MHz TCXO divided by 63 is 253.96825 kHz. (the 8 significant figures are deliberate) Perfect! That frequency is down about 20 dB on the upper skirt of the filter and provides over 40 dB of unwanted sideband suppression.
A 16 MHz TCXO is a couple of bucks and I can divide by 63 with two 74HC type counters; from the junk box, of course. 63 also allows the 32 weight output of the counter to be used with the product detector. A 50% duty cycle is desired and the duty cycle of the 32 weight output is 32/63 = 50.79%. The product detector will have no problem with that.
However, the carrier oscillator or BFO sets the local oscillator frequency which is 474.2 kHz plus 253.96825 kHz for WSPR and JT9. This is 728.16825 kHz. It’s back to the spread sheet. The LO multiplied by 27 yields 19.66054275 and 19.66080000 is a standard crystal frequency; (71 cents from Digikey) but not close enough. However crystals can be pulled by using higher or lower than their design load capacitance in parallel mode oscillation. So I made a 19.6608 MHz oscillator and padded it down to the correct frequency with a trimmer cap. The division from 19.6608 MHz is accomplished with a divide by 13/14 so the output is a 48% duty cycle and after a tuned amplifier before driving the balanced mixer is 50%.
The drift of the 19 MHz oscillator was more than I would accept. I then phase locked the 19.66008 MHz crystal to the 16.0000 MHz TCXO to improve the stability. The trick is to find the solution to following equation: 16.00000MHz/M = 19.66008/N. The “padded” crystal oscillator could be set to the precise frequency desired but it was the drift that was problematic. The 19.6608 MHz would provide an RF frequency that is a little over 9 Hz higher than the nominal 474.200 kHz. But this error can be accounted for with the setup menu in WSJT-X. To solve the equation using the precise frequency would result in huge numbers for N and M.
The frequencies of standard crystals and TCXO’x are not random numbers. For example, if you divide 19.6608 by 3 you get 6.5536 MHz which is 100 times 2 to the 16th power. 16 MHz is a million times 2 to the 4th power. After dividing the 19.6608 frequency by three, there are two common factors between the two frequencies; powers of 2 and powers of 10. So, let’s rewrite the equation assuming we divide the 19 MHz frequency by 3. 16.000/M = 6.5536/N = Fref. I have indicated that the resulting frequency is the reference frequency for the PLL. N and M have to be integers and have to produce a reasonable reference frequency. I had an old synthesizer chip in the junk box and one of the divisions for the reference was 256. I decided to set N to 256 and that would set M to 625 however this would reverse the roles of reference and VCO in the synthesizer chip as the 16.000 MHz TCXO is the stable reference. No big deal. Reverse the connections on the phase/frequency detector output. Note that 625 is 10 to the 4th power divided by 2 to the 4th power and 256 is 2 to the 8th power showing the common factors of powers of 2 and 10.
Because the crystal oscillator that is being stabilized has an inherently narrow spectrum, the PLL is not required to reduce any phase noise and therefore almost any reference frequency can be used. The rather high 25.6 kHz reference frequency is advantageous as only a simple RC filter and no additional loop gain is needed to stabilize the loop and eliminate any reference sidebands. But there is a danger to a high reference frequency. That is the possibility of harmonics of the reference frequency falling on the IF or RF frequencies. The tenth harmonic of 25.6 kHz is 256 kHz which is well above the filter’s passband. If anything should “sneak around” the mechanical filter it would appear at 2.03 kHz on the WSJT display. But the input to the product detector is the highest level signal in the receiver and it would take a very strong harmonic to be seen there. The 18th and 19th harmonics appear at 460.08 kHz and 486.4 kHz and are way out of the band.
I did not use a voltage controlled TCXO, VCTCXO, for the 16 MHz oscillator because none was available. The operating frequency is directly proportional to the 16 MHz TCXO frequency and therefore the frequency stability of the receiver is exactly equal to the TCXO’s stability. (It is left as an exercise for the reader to prove this.) The pull range of a VCTCXO is only about 8 to 10 PPM. The error from a precise 16 MHz is about 20 PPM high which could be reduced to 10 PPM or less with a VCTCXO. Lowering the 16.000 MHz master oscillator by, say, 10 PPM would lower the carrier oscillator for the product detector by the same 10 PPM and lower the received signal in the mechanical filter passband. However, a 10 PPM reduction of the 254 kHz carrier oscillator is 2.54 Hz which would have no effect. Therefore, the 16.000 MHz TCXO remains at 16.000 MHz and the frequency error is handled using the WSJT-X setup menu.
Did I get lucky and find just the right combination of frequencies for my receiver? No. I have done this many times. My original I/Q receiver used a TCXO. My 630m transverter uses the same technique and I simply program my Flex 5000 to account for the “odd” frequencies. With the Flex, the transverter input/output only works on 10 meters and I did not want to use a lower band such as 80 meters as I would have to use the RF output with its possible 100 watts. The transverter output provides up to +5 dBm which won’t destroy anything if I happen to forget to lower the transmitter power or warm up the shack with a big attenuator. In my transverter I used a 14.4 MHz TCXO, a couple of bucks again, and doubled it to 28.8 MHz and used low side injection to translate 474.2 kHz to 29.2742 MHz. I could have used high side injection but this would invert the sideband and I chose not to do that although it would work just fine.
Many years ago I replaced the 116 MHz crystal in my 2 meter transverter with a TCXO and some frequency multiplication for EME which still serves quite well.
I hope this note has given some insight in to receiver design and methods of “juggling” the numbers to use inexpensive and junk box parts.
Al Helfrick was first licensed in 1957. He was and remains K2BLA and, more recently WI2 XBV. He is mostly an experimenter and builder but did some DXing along the way. He has 10 band DXCC which includes DXCC certificate number 75 on 2 meters. He is currently on 1296 EME and 630m. He recently “retired” from his position at Embry-Riddle Aeronautical University in Daytona Beach as Professor Emeritus to spend more time with radio and Dixieland jazz bands. “Professor Emeritus means you can still work for the school without getting paid”. He is still active with the university and continues to play with the students in his university’s band that he founded 25 years ago. Retired is a bit misleading as he still teaches week-long short courses for the University of Kansas on avionics 8 to 10 times a year at locations throughout the world.