You can’t work ’em if you can’t hear ’em

You can have a modern transceiver with a good sensitive receiver that won’t turn turtle under the pressure of strong signals connected to a good antenna, but the signal still has to make it from the rig to your ears (unless you’re running some digital mode).


Some of today’s transceivers throw in a small downward (or upward) firing speaker.  These mostly sound rather tinny with some buzzing (how bad depends on where you’ve set the radio down on your operating position).  The built in speaker is better than nothing, and will probably even be just fine for a rag chew with a strong local on a quiet band.  However you won’t want to use it for serious DXing, or in a contest.

Yeasu, Kenwood, Icom, and Elecraft offer matching speakers for their transceivers, which solves the problem of having a front facing speaker.  Hopefully, these speakers are designed acoustically correct to avoid unwanted internal resonances, and to fully cover the frequency range of the human voice necessary for communication (about 300hz to 3000hz).  Expect to pay between $100-$300 for such a speaker.

You can, of course, get a very good communications quality speaker for a lot less by buying something originally intended for another purpose.  I have a small Optimus (Ratshack) speaker that nearly matches my IC-746, and it does a good job.  The shack made a few different small two way book shelf hifi speakers that make decent communication speakers.

You can also build your own.   Let me tell you more than what you probably wanted to hear (unless you are into building speakers).   What you need to find is a suitable driver (raw speaker), and then build a correct sized box to house it in.  Since the frequency response required is one octave from 300hz to 3000hz, a mid-woofer or mid range driver of about 3.5″ to 5.5″ in diameter will suit our needs perfectly.  It has to be a fairly efficient driver because most transceiver will have an audio output power range somewhere around 1.5 to 3 watts of power.  This isn’t as limiting as it sounds, since we aren’t going to fill an auditorium or even a living room with sound, just a semi-circle of space, about an arm’s length in diameter.  Several sources of such drivers come to mind, ranging from the auto sound after market, and drivers made for musical instrument practice amps.

One thing to look out for in selecting a driver, is the cone surround.  This is the flexible band that connects the edge of the cone to the driver frame.  Foam, rubber, treated cloth, and pleated paper are the most common materials used here.
Avoid foam based surrounds like the plague.  They have a half-life of about two years or so before they start to fall apart in a pile of dust.  Your mileage will vary on this, but if you live in a warm-humid climate you are in prime foam rot land.  (I threw out two different sets of expensive tower speakers that had their woofers and mid-range drivers develop this cancer).
Rubber surrounds are vastly superior to foam, and will probably last forever.  I have a pair of home made speakers in my media room that are over 25 years old, their rubber surround drivers are still going strong.
Treated cloth surrounds were once very popular in wide range speakers (woofers with a ‘wizzer’ cone for mid-tweeter range).  They will also last forever.  Finally pleated paper.  This is just the edge of the cone.  During manufacture, the outer inch of so of the cone is soaked in water (maybe with some other solvent) and is compressed between two halves of a mold that form the pleats.  Some heat is applied and when the paper dries the pleating is permanent.  Drivers for PA systems and musical instrument amps are built this way.

In order to know how large a box you need to build for a speaker, you need to know the driver’s Thiele-Small parameters.  These are the acoustic specifications for the driver.  We need be concerned with only three of these.   Rs (or Fs) is the drivers free air resonant frequency.  This must be lower than the lowest note we must reproduce.  So for our case, it must be below 300hz.   Qts is the total Q of the driver.  It will determine the shape of the high frequency cutoff of the driver.  Different types of speaker enclosures require different ranges of driver Qts.  Finally there is Vas.  This parameter gives us the effective volume of the driver cone.  The speaker enclosure must have a larger internal volume than the Vas, how much larger depends on the enclosure type, and desired response.

The common types of speaker boxes are sealed, ported, and infinite.   The sealed box enclosure is the easiest to design and build.  I won’t discuss the math here, there are plenty of websites that have on line calculators for working out suitable designs.  Drivers for sealed box enclosures usually have a Qts from 0.33 to 0.55.  The box itself is usually designed for a Q of 0.707 (Butterworth  LP response).
Ported boxes have an opening in either the front or the back via a tube (PVC pipe) inside the box.  The box is ‘tuned’ by adjusting the length and diameter of this pipe.  Ported boxes do not have rational cutoff response curve (I’m referring to the math here), but there are several design equation sets that are commonly used.  At the low end the speaker impedance drops suddenly like a rock, and a high pass filter is usually employed to protect the driver from excessive movement.  Drivers with lower Qts values are suitable for ported enclosures.
At the high end of driver Qts ranges (0.707 and higher) we have the infinite baffle design.  The ideal infinite baffle is an infinitely sized board with the driver mounted in the middle.   Open back cabinets, car doors and rear decks approximate what is required here.  Those old time, tube era transceivers had such open backed matching speakers.


Even the best speaker isn’t going to cut the mustard during weak signal conditions in the middle of a DX pileup during a contest.  You know, when the guy you REALLY need to hear is being stepped on by the QROO crowd.  What you need are a pair of CANS!

Communications headphones used to be rather uncomfortable things.  That’s probably where the radio slang term ‘cans’ came from, they felt like a pair of tin cans clamped to your ears.   High impedance, magnetic headphones go back to the days of Marconi.  They have a thin flexible iron or steel diaphragm that is pulled by a magnet.  A coil wound with many turns of very fine wire surrounds the permanent magnet.  The weak radio signals drive this coil, which adds to, or subtracts from the magnetic field, causing the diaphragm to flex and move the air next to the ear, thus producing sound.  There is a variation in the construction of this type of headphone where a lever arm is attached to a smaller diaphragm inside of the earpiece.  This lever arm transmits the movements to a thin mica diaphragm that generates to sound.  This type of headphone (Baldwin) is based on the structure of the human ear.  Magnetic headsets were made with impedance values (with both earpieces wired in series) somewhere between 2000 to 5000 ohms, with 2000 and 3000 ohm values being common.   Early magnetic headphones used individual ‘tip’ plugs and jacks for connection to the circuit, but even the first available commercial communications receivers were set up for the now common phone plug and jack system.
During WWII, most communications headphones were magnetic types, but these had to be ruggedized.  Extreme sensitivity was no longer a requirement as the radio equipment had plenty of internal audio gain.  These headsets usually were built to a 500 ohm impedance, a standard that carries over today in the aviation industry.

Amateur radio equipment from the pre-war era, usually had receivers set up to drive medium to high impedance headphones.  By the 1960’s, during the AM to SSB switch over, lower impedance headsets became common.  I’ve looked back at the specs on old Drake, Collins, and Heathkit rigs.  Only Heathkit continued to require the 2000 ohm headset impedance on their SB line receivers and transceivers, the others had switched to speaker level impedance headphones.  Perhaps the adoption of hifi headsets in the shack was the thing driving the lower impedance of today’s headphone jacks.  BUT, they are still mostly wired as monaural jacks.  BTW, hifi headsets are not magnetic drivers, they are dynamic drivers, having actual pint-sized speakers in each earpiece.

If there is one place where a communication quality headphone has to keep outside noise from interfering with the operator, it is in the cockpit of an airplane.  Technology changes slowly in aviation, and bits of the past are ingrained in newer tech.  Many standards have been inherited from the military, the 3/16″ diameter tip/ring microphone plug for example.  (The only maker of amateur radio gear to use that style of microphone connector was Collins, probably due to their involvement in avionics.  BTW the sleeve is ground, the tip is PTT, and the ring is the microphone).

Back in the late 70’s, I started taking flying lessons in Cessna 150’s.  Rather than use the flying schools crappy headsets with the plane’s hand held microphone, I bought myself a  David Clark H10-30 headset.  This time proven unit is still available today, though with an improved microphone.  It’s a medium impedance headset (around 300 ohm, with a series volume control).  While I haven’t flown in years now, I’ve kept the headset for use with my ham gear.  Even though there is quite the impedance miss-match, most transceivers will still drive it, and I can always use an output transformer ripped from an old transistor radio wired ‘backwards’ to step up the impedance if necessary.  The microphone is an amplifier dynamic type that requires power supplied by the radio to operate.  Like most aviation microphones, it’s designed to work in a carbon microphone circuit.  These noise canceling microphones have a frequency response that matches the human voice range of 300hz to 3000hz.  They won’t sound hifi, but they will cut through the crap.
Most transceivers today provide a 5 to 12 volt power pin on their microphone jacks to power electret units.  I made an inline adapter to power the microphone through a 470 ohm resistor from this power source,   Audio is then coupled to the microphone input via a 10uf capacitor back to the rig with a 10k pot in between.

There is nothing like a pair of well made (for the purpose) headphones for bringing the weak signals right to your ears.  Your radio may have the latest in DSP audio processing (perhaps done at the IF frequency), but the best DSP filter for separating a weak voice out from a crowd of others is still that lump of gray matter between your ears.  But you have to provide your brain with the entire input, as directly as possible.

David Clark makes a whole line of aviation headsets over a wide price range.  They still sell parts for anything they ever made.  I recently ordered new padding and filters for mine that had been turning to dust over time.
There are other brand names in the market today, but I’m happy that this “made in the USA” supplier is still going strong.

Using a weird surplus LED bar graph

The photo I posted in my previous post of a home made project box for a future QRP transceiver project showed what looked like a bar graph meter above the LCD frequency readout.  Most of the common LED bar graph displays have 10 vertical rectangle segments and can be driven by a the popular LM3914 series chips.  These parts can handle such displays directly and are available in linear, log, and VU processing versions of the signal to display.  They can also be expanded by cascading additional drivers and displays to up to 100 segments of common anode displays.

Buried in my junkbox were several Toshiba GL-112T9 LED displays that I got for a few bucks at a past ham flea market.  These are 12 segment displays with horizontal rectangular segments.  There are four green, four yellow and four red segments.  They are not common anode displays however.  Two of the green LEDs and two of the red LEDs are wired in series with a connection to both ends and the common middle.  The four yellow LEDs are wired in two series connected groups.  The end result is that the display has only 20 pins instead of the 24 that would otherwise be required.  Needless to say that this makes it incompatible with the LM3914 drivers.

The obvious solution to the problem (if I wanted to use these bargain bar graph displays) was to make a custom driver using a micro controller such as the Arduino.  I designed such a driver using an atmega16 chip in a 40 pin dip package.  This part is overkill, the required code will fit in only 4K of flash and only 13 I/O lines are required (one being an analog input), so an atmega48 would work fine.  However, I already had several of the atmega16’s in the junkbox, so that’s what I used.

BargraphHere you can see the result.  There are 12 current limiting resistors, and 4 diodes.  The micro controller lights from 1 to 12 of the LED’s depending on the voltage input to the A/D converter.  When both of a pair of series connected diodes are to be driven, only one of the two lines are pulled low.  The common line to the two LEDs is driven through a diode to prevent back flow from the micro controller which outputs a high level (5v) when a segment is off (the anodes of the display all go to +5 volts).

Here is a link to the code:  Bargraph (Github).
The code is licensed under the GPL v3.0, so you can make free use of it for your own projects under the terms of that license.  If you want to port it to a smaller AVR controller the only thing that will possibly need to be done is to modify the some of the register and/or bit names as Atmel has moved things around from the older atmega16 to the atmega48, atmega88, atmega168, and atmega328 processors (the ones used in the Arduino).

To keep things quiet (from an RF point of view) the processor is put to sleep while the A/D converter is running.  The converter is put into the auto-run mode, so it does not have to be restarted after finishing each conversion.  The A/D converter interrupts the CPU after each conversion, which wakes the CPU up from sleep.  At this point the result of the A/D reading is converted into a bar graph display, and the correct LED’s are driven.  The CPU then goes back to sleep, waiting for the next display cycle.
I used the atmega16’s built in 1 mhz RC clock, it’s more than fast enough for this application, the slower speed clock makes less rf noise, and it eliminates the cost of a crystal.

The nice thing about this approach is that I can modify the processing of the input signal to display in linear voltage, “S” units, Log or Vu scales.  There are unused I/O pins so I can even have the processor switch display modes on the fly by adding code to monitor the extra pins, using them as a function switch.  Right now the display is purely linear.  The LED’s light up from green, to yellow, to red, but you could wire the bar graph up in the other direction if you think that makes more sense for your application.





Project Boxs

Finding a nice enclosure to build your projects into can be a problem.  The cost of those ‘mini boxes’ I used to buy as a teenager have gone up by an order of magnitude or more over the past few decades, perhaps more than the cost of living indexed by inflation.  Not withstanding the fact that those ‘mini boxes’ are ugly things that look nothing like the ‘store bought’ versions of what we’re trying to home brew, there must be a better way.

There ARE some nice professional project boxes out there, but they are rather expensive.  I can’t see spending upwards of $100 for a project box to build a little QRP transceiver or other ham radio project in.  If you are thinking small, there are some nice ‘clamshell’ plastic cases that either snap or screw together.  Some of these have a removable door to a battery compartment.  I built my LC meter into one of those, the cost of that enclosure was reasonable.
The box that I built my Heathkit HD1250 dip meter clone into was one of several LMB royal blue project boxes I got a few years ago at a ham flea market.  I think they are still being made, and their cost new is a bit lower than a kings ransom.

If you are willing to perform a “Macgyverism” it is possible to create some nice custom project enclosures out of cheap materials you might have lying about.  Many hams have built custom project boxes by soldering printed circuit panels together.  This only works if you can obtain the material cheaply in bulk.  I ran across a source of badly tarnished PC board stock measuring about a half a yard by a yard in size.  Some pieces were single sided, others double.  Most were ‘standard’ thickness, but there were a few double thick panels in there, as well as a few that were ‘paper’ thin.  Most were fiberglass, but I also got some phenolic panels.

When soldering single sided panel pieces together to make an enclosure the copper clad sides would normally be on the inside of the box.  This will provide a good ground plane for the circuitry, as well as a Faraday shield.  The outside doesn’t need the metal surface, it’s going to be painted anyway.  You will need to use double sided panels in a few places, so save their use for where it matters.   There are many examples of project box construction on the Pic-a-Star project pages.

I made an experimental project box that was to serve as the enclosure for a proposed QRP rig I wanted to build.  The idea hit me when I found an empty one gallon metal container in the garage.  It was a little dented and rusty, but once treated with a rubber mallet against a two by four, a little steel wool, and some paint, that would not be a problem.  The top and bottom of the can were removed with a can opener (also a file and lots of elbow grease).  I then cut it in half the long way, and then proceeded to make a chassis with front and back panels from PC board stock.  The photos show the result.




The front panel and the paint job were inspired by the Heathkit SB series styling.  I’ve since replaced the microphone connector with a 5 bin DIN type, and added some rubber feet to the bottom half of the enclosure.

I haven’t yet actually built anything into this enclosure, what you see is a mock up so far.  The current project I’m working on will fit into a smaller enclosure that I made from a half of a surplus box thrown out by a former employer.  The paint can box will probably get used in a later radio project once I’ve proven out some of the circuitry I’m currently experimenting with.

The idea here is to make do with what you have on hand.  I bought myself a cheap sheet metal bending brake from Harbor Freight, and I’ve used that to form surplus sheet aluminum into box lids and chassis.


Building a Transceiver .. continued

So far, I’ve described the building blocks of a typical superhetrodyne communications receiver, which is the back bone of a complete transceiver.  I’ve left out the frequency generating portion, that is the oscillators.  A single conversion rig has two oscillators, a variable frequency oscillator that sits behind the tuning knob, and a fixed frequency oscillator that provides the missing carrier during demodulation (except for AM and FM modes).  The second oscillator known as the BFO during receive, and as the carrier oscillator during transmit.  If there are multiple conversions, then additional fixed frequency oscillators (usually crystal controlled) are provided.

These internal signal sources must be clean, and we must select our IF and injection frequencies carefully to avoid unwanted mixing products that can result in spurs that fall inside the band of frequencies we want to receive.  Otherwise we will end up with ‘birdies’, extra signals we can hear as we tune across the band that are not really there.  On transmit we could end up with spurious emissions that cause interference with other amateur stations, or even worse, other services outside of the amateur bands.

The time honored method of frequency control was a well built VFO as the local oscillator (LO).  Double bearing tuning capacitors, temperature compensated trimmer capacitors, anti-backlash gearing from a reduction dial to the variable capacitor, and solid, vibration free construction.  In the vacuum tube era the VFO tube might have its heater powered on 24/7 (perhaps at 50% power until the rest of the rig was turned on) to eliminate warm up time.  Sometimes a variable inductor was used instead of a variable capacitor for tuning the VFO (mostly in radios made by Collins).  To provide for linear tuning, that is the same number of dial rotations in degrees per khz of frequency variation across the entire tuning range, the plates of the tuning capacitor needed to be just the right shape, or the turns wound on a slug tuned inductor had to have just the right pitch.  Ham builders would sometimes file the plates of a variable capacitor to get this perfect!  Various VFO circuits were used, with their operating points (bias) adjusted for the lowest possible noise.  I don’t think oscillator phase noise was discussed much in the vacuum tube era, but it’s certainly a design consideration today, and along with IMD and IP3 values has become an important design consideration in today’s radios.

To cover more than one band the VFO would have to be band switched.  Imagine designing a perfect oscillator not once, but five (or more) times!  Swan, Atlas, and a few other companies chose this route, but others band switched their LO’s by providing an additional mixer and crystal oscillator.  By using pre-mixing before injection into the first receiver mixer (in a single conversion rig), the VFO only needed to be designed once.  Of course, this added an additional oscillator and more mixer products to contend with.

Sometime in the 1970’s, radios with digital frequency displays began to appear.  The first of these used TTL based frequency counters to measure the VFO frequency, and after adding a band adjusted offset, would display the radio’s operating frequency.

Collins did something a bit different on one of their high end communications radios designed for the government.  They got rid of the VFO, and the tuning knob spun a slotted wheel that generated pulses which drove up/down counters.  These counters provided the input for a frequency synthesizer.  I saw one of these radios at a hamfest in the 1970’s, but I don’t remember the model number.  I do remember that the price of that radio would buy a luxury automobile, or a previous owned single engine airplane!  Not long after that, Icom showed a rig with a similar tuning method, and the other makers followed suit.

These first synthesizers used digital PLLs.  They consisted of a crystal controlled reference oscillator, several divider chains, a phase detector, and a voltage controlled band switched oscillator.  The divider chains were programmed with a ratio that corresponded to the desired LO frequency divided by the reference frequency.  The phase detector developed an error signal that was proportional to the difference between the desired frequency set by the divisor ratio between the reference oscillator and the VCO.  This error signal steered the VCO onto the right frequency.  Any drift in the VCO output would generate an error voltage that applied to the VCO would correct the drift.

PLL’s could only tune in fixed steps.  If the reference signal was 1khz, then the PLL could only be tuned in steps of 1 khz.  Of course we need much finer tuning steps than that to get a SSB or CW signal ‘right on’.  A lower frequency reference could be used, but at the cost of increased lock up time, which means a slower tuning rate.
Early radios with digital frequency displays still used frequency counters, but had replaced a bank of pre-mix crystals with a PLL.  Here a rather high reference frequency such as 500 khz could be used.
Pure PLL tuned rigs used two PLL’s with their reference frequencies separated by a small value, usually between 10 to 100 hz.  By using the difference between these two circuits in a mixing loop a combined step size of only 10 to 100 hz could be achieved.  Such a synthesizer was described in the ARRL handbook.

Motorola produced a family of PLL chips in digital CMOS.  Some of these had a dual modulus divider.  These chips divided the divider chain into two parts, and provided for control of an external prescaler divider.  This allowed the CMOS PLL to control a VCO at frequencies in the VHF range.  One of these parts is used in the Elecraft K2 transceiver.  The reference oscillator here is a varicap tuned VXO that is tuned along with the digital PLL.  In this way the K2 can achieve a step size smaller than the PLL reference frequency.

The frequency synthesizer method most commonly used today is Direct Digital Synthesis.  In this circuit a high frequency clock drives a special type of counter called a digital accumulator.  This counter is really an adder circuit that produces the sum of the previous operation with a frequency tuning word.  The output of this counter is a phase angle that is applied to a sine (or cosine) look up table in a read only memory.  This value is then applied to a D/A converter to produce an analog output signal.  The DDS circuit synthesizes a sine wave.  The output must go though a low pass filter whose cutoff  frequency is less than half the DDS clock frequency.   The available tuning step size for the DDS, as well as the output S/N ratio depends on the number of bits used in the digital accumulator.  The purity of the output signal is a function of the number of bits in the D/A converter, and the difference between the output frequency and the DDS reference clock.  Analog Devices makes several families of DDS chips.  The earlier generation AD9850,  AD9852, and AD983x parts are popular with many QRP builders.  Later generation AD995x series DDS parts raised the number of bits in the D/A from 10 to 14 bits, as well as raising the clock frequency from 125-180mhz up to 400mhz.  The spectrum purity, and phase noise content of this latter generation is much improved over the earlier parts (but these parts are much more expensive).  The AD9951 is used in the Pic-A-Star superhetrodyne-software defined radio transceiver project.

Yet another synthesizer technology was introduced by Silicon Labs.  The Si570 programmable VcXO combines a crystal oscillator, PLL, and a multi-synch programmable divider to generate a wide range of frequencies.  An I2C micro-controller interface is provided.  This device was intended as a crystal replacement, not as a variable frequency oscillator.  However, it is capable of rapid frequency hopping over short distances.  The part has good spectral purity, and low phase noise, and is used as the VFO in the Elecraft KX3 transceiver.

A  much lower cost version of the Si570 is the Si5351 family of parts.  These chips require an external crystal (25 or 27 mhz), and can provide 3 to 8 separate outputs on different frequencies.  The three output Si5351A costs less than $1 each!  It can provide both the VFO  and BFO (LO and CO) functions in a single unit.  While its phase noise is not quite as low as the Si570, or AD DDS parts, it is still quite good.  The Si5351A is used in the Elecraft KX2 transceiver.  The Si5351A is available on a breakout board with a 3.3 volt regulator and a 5v-3v logic level converter for about $8 from Adafruit.  Combined with an Arduino micro-controller module and an LCD or OLED display, you have the basis for the necessary frequency generation of a QRP transceiver.


Designing and building a Transceiver

I’ve been accumulating all sorts of parts in my junk box over a span of  nearly a half a century.  Some of this stuff was removed from war surplus electronics obtained from ham flea markets, and mail order surplus houses.   Included are components from ARC-5 command sets, and other famous military radios, things like variable capacitors, IF transformers, vacuum tubes and sockets, transmitting mica capacitors, etc.
There are also commercial surplus inventories and discarded engineering samples from places I’ve worked at, flea market finds, and more recently bargains on eBay.

Vacuum tube oriented components can sometimes be repurposed for use in solid state designs, though they are usually too large to fit in as the current trend in construction is miniaturization.

Many of the solid state parts I’ve collected over the years, some of the transistors and integrated circuits that were once popular with the amateur home brew crowd have become obsolete.  Still, having an ample supply of them, I would probably use them in a project or two.  Some of these are still available today, but in SMT package versions of  the now obsolete through hole variants.

Consider the design of a QRP HF transceiver.  I’ve been gathering ideas from the designs of  W1FB, W7ZOI, N6QW, DK7IH, EI9GQ, G4LFM, VK3HN, and others.  I’ll go into the actual circuitry I’ve decided to use in later posts, for now let’s look at the available parts.

transceiver is a device that functions as both a communications receiver, and a transmitter (obviously that’s where the name comes from).  The assumption is that most of the circuitry can be used for both functions, ie: the transmitter is sorta kinda the receiver run backwards.  (It’s not really that simple though).

The very first transceiver was built for what was then called the ‘ultra high frequency’ bands, back  when hams were first trying to make use of the five meter band.  A super-regenerative detector and one or two stages of audio comprised the receiver.  On transmit, the detector circuit functioned as a modulated oscillator, and the audio stages functioned as a microphone pre-amp and modulator.  A multi gang switch, or a bunch of relays handled the task of switching the necessary circuits around from receive to transmit modes.  In the late 1960’s (it was in the 1967 version of the ARRL handbook), a version of this idea was presented for the 450 MHZ band, a 6CW4 nuvistor was used in the detector and transmitter, while two transistors were in the audio stages.   Fully transistorized versions of the same concept were used in cheap, made in Japan, CB walkie talkies.

We’ll take a look at the design of the receiver portion of a transceiver first.

A Modern HF communications receiver is still based on the superhetrodyne circuit.  We won’t consider SDR or software defined radios at this point, although even they do make use of the superhetrodyne principle, though the functions are implemented in software rather than hardware.  Early receivers used a single conversion to an IF frequency that was usually lower than the frequency of the lowest band being covered.  Multiple stages of radio frequency amplification along with several tuned circuits in a preselector attempted to reject image response.  The gain distribution between the RF and IF amplifier circuits wasn’t optimal here, resulting in strong signals overloading the first mixer.  Going to a higher frequency first IF would allow for not needing as many (or any) stages of RF amplification, but double (or even triple) conversion was needed to provide the necessary selectivity.  It wasn’t until good piezoelectric (crystal) filters with steep slopes to provide selectivity at HF IF frequencies were available that receivers could provide the required selectivity before taking gain.

Some receivers still did make use of double or triple conversion.  If a first IF in the VHF range is used (up conversion), a single low pass filter in the front end will allow for general coverage over the entire MF and HF range.  IF frequencies between 45 and 75 mhz are common here.   A moderate bandwidth roofing filter is used at this IF, usually around 5-6 khz wide, this allows for reception of AM and NBFM signals, as well as SSB and CW.  A second conversion, down to a lower IF between 5-10 mhz follows, usually without much gain between conversions.  Several IF filters are now provided at this second IF, usually a choice of 200-500hz for CW, 2100 – 2700 hz for SSB, and 3.5-6 khz for AM or NBFM.   Finally, sometimes a third conversion is done down to a low frequency IF of 455 khz.  The third conversion oscillator and the BFO oscillator are made variable.  This allows for moving the signal between an overlap of the bandwidth of two roofing filters, one in the HF IF, and the other in the LF IF.  By doing so, we can continuously vary the effective receiver bandwidth over a wide range.  Until the advent of IF DSP, this feature was standard in many amateur transceivers.

Today the last IF of 455 khz is replaced by an even lower one between 15-50 khz.  The processing at this frequency is handled in the digital domain by a Digital Signal Processor.  The functions of the roofing filter, IF amplifier, product detector, noise blanker, AGC, beat frequency detector, and notch filter are all performed in software.  If the rig doesn’t have to provide for general frequency coverage the up conversion to a first VHF IF frequency isn’t provided and the first IF is in the 5-10 mhz range.  If general conversion is provided, then up conversion is available, but is usually only performed outside of amateur band receiver coverage, and sometimes for the transmitter chain.  The Kenwood TS-590 uses this dual conversion chain process.

For our amateur designed and built transceiver will first consider only the receiver circuit chain, and then decide which elements will be either reused, or duplicated for the transmitter chain.

First of all, we need the input pre-selector.  Once upon a time, this was a band switched set of coils and a multi-gang capacitor.  The operator would have to peak these circuits (with a single knob) when changing bands, or making a large change in frequency within a band.  Thanks to computer aided design we can now design band pass filters that can cover an entire band with reasonably steep walls at the band edges.  No more manual re-peaking required!  The coils can be slug tuned inductors in miniature cans, or they can be wound on toroid cores.  In the latter case, we will use ceramic trimmer caps to get the filters dialed in just right, or we could use precision fixed capacitors (1% tolerance or better) and fine adjust the turn spacing on the toroids to dial things in.  Two popular circuit topologies for such filters are top coupled parallel tuned circuits with link input and outputs, and series tuned sections with shunt capacitors between each section.  The latter may have better circuit Q, but the former are easier to align with a grid dip meter.  You can buy band pass filter kits from QRP Labs, and raw toroid cores can be obtained cheaply in bulk from

An RF preamp isn’t necessary on the lower of our ham bands (160-30 meters) as atmospheric noise will overpower the internal noise of our first mixer.  On twenty meters and higher a preamplifier may prove desirable as it will provide an improvement in the front end noise figure.  In any case, being able to switch this preamp in and out as band conditions demand should be provided.  The preamp should be able to handle large input signals, so the use of an RF power type transistor in deep class A operation (large resting current) is common.  Parallel JFET configurations are also popular.  2N5109’s and J310’s are common here.  The J310’s are becoming hard to find and expensive in through hole TO-92 packages, but are only $0.20 cents each in SMT packages when bought in 100’s.  Another good choice for a preamp are the SMT ERA series amplifiers by Mini Circuit Labs.
Finally, dual gate MOSFETs make good RF preamps.  Gain can be controlled via the upper gate, with signal on the lower one.  Old time parts such as the RCA 40673 are becoming hard to find, but low voltage SMT variants are still available at this time.

There are lots of choices for the mixer circuits.  Diode ring mixers, either the packages ones from MCL, or home made using toroid wound transformers and discrete diodes will give good performance.  FET switches make good mixers.  These parts are made in DIL SMT packages, and builders either use home wound toroids or MCL transformers.  Single, double balanced, or H-mode configurations are used (with a 1, 2 or 3 trifilar wound transformers).
IC mixers are also used.  The NE/SA 602/612 8 pin circuits combine a Gilbert Cell mixer and an oscillator circuit in a single package.  The oscillator circuit supports VFO, VXO, or crystal oscillator configurations, or an external oscillator can be used.  These devices can’t handle strong signals and having some form of input gain control ahead of them is a good idea.  They do work well as the receiver product detector, transmitting mixer or balanced modulator stages where the signal levels can be controlled.
Another IC device that was once popular was the CA3028.  This early device consisted of three transistors and resistors in a differential circuit configuration.  The input signal was provided in a balanced mode between the two bases of the pair, and the output was taken between the two collectors.  The oscillator signal was fed into the base of the lower transistor.  While this device is no longer available (except from Chinese parts jobbers on eBay), you can make your own from discrete transistors and resistors.
Finally, there is the MC1496.  This is also a Gilbert Cell mixer, but without the built in oscillator.  It makes use of external bias resistors so the internal gain and resting current can be configured.  It’s a stronger mixer than the NE602, and it makes a good balanced modulator, product detector, or mixer.  It’s still being made, mostly in an SMT package.

The last circuit block I’ll describe now is the IF amplifier.  The CA3028 was commonly used here in both cascode and differential circuits.  Two AGC methods were used.  Reverse AGC (increasing AGC voltage decreases circuit gain), and Forward AGC (increasing AGC voltage increases circuit gain).  Forward AGC was applied to the lower transistor to control the current though the pair in both cascode and differential circuits.  Reverse AGC was applied to one of the upper transistors which when in conduction mode would steal current from the other, bypassing the signal to ground.
A circuit that made better use of the Reverse AGC control in a differential circuit was the Motorola MC1350.  The MC1349 was a higher gain version.  These were very popular in ham receiver designs (it is used in the Elecraft K2).  Motorola also made the MC1550 and MC1590 which were similar.  All of these are discontinued today, though the MC1350 can still be found from surplus outlets.

Discrete IF amplifier circuits using cascode FETS, or a single FET and a BJT are popular.  Forward AGC is applied to the upper transistor.  With the above mentioned IC amplifiers becoming unavailable, the discrete transistor route is the recommended one.

More in a later post …….