Minima - Design Notes

From HF Signals
Revision as of 04:29, 29 January 2015 by KE0BLL (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search


These notes are in addition to the main article that describes the Minima transceiver. They are not required for building and operating the Minima - that information is (hopefully) all in the original article. These are for the curious who will want to hack the Minima.

A number of circuit blocks have been borrowed from other works. These are mentioned with the source. You might find it interesting to read them up.

A number of references are made to other projects on this web site. A popular transceiver, the BITX is frequently mentioned as the source for some generic building blocks. The BITX details are available at

History in ten parts

This is a long capture of how the Minima came about. It includes references on the web that might interest those treading a similar path.

Part 1 - Sweeperino Woes

In 2012, I had completed a computer controlled Si570 based digital sweeper called the Sweeperino.

Sweeperino Woes

It allowed me to generate a signal between 0 and 60 Mhz and in the VHF range with a 1 Hz accuracy. This signal would be fed to any circuit like a filter, amplifier, etc. and the output measured with great accuracy and range with an AD8307 based W7ZOI’s power meter.

As the video shows, the sweeperino had a challenge : The harmonic content of the signal generator produced ghost responses in filters. To overcome them, I was contemplating replacing the open AD8307 with a superhet receiver front-end that would filter out these harmonic responses. I started looking around for a front-end for the sweeper that would match the 90db logarithmic range of the AD8307.

Part 2 - Mixed thoughts about Mixers

Around that time, Chris Trask N7ZWY performed some experiments on diode mixers and H-mode mixers that promised a very high input intercept. He posted his findings on EMRFD’s Yahoo! group in a paper called Mixer Musings. The elegance and the simplicity of the design was quite enticing and I started digging around for more information on H-mode mixers and other forms of passive FET mixers.

Use of FETs as passive switches in mixers has a long history. A recently uncovered a paper by Rick Campbell on his university website details some of it. Campbell refers to earlier work of Wes Hayward and Steven Mass, unfortunately, the paper’s reference section was missing. Gain (I7SWX) sent in his paper in a private email. Gain’s Mixer was built, tested and documented by Martein, PA3AKE. Martein’s site is an awesome resource for anyone who wants to explore high end receivers.

Part 3 - The Front-end begins

A couple of FSA3157s were ordered online and a mixer made with them. An Si570, under Arduino control was used as local oscillator. With an antenna directly connected to the mixer, and its output was fed to the FT-817, tuning up the bands only got grief. The FT-817 was full of noise. Upon probing around, I discovered substantial local oscillator leak into the the IF and RF ports. The local oscillator was probably desensing the receiver.

It was decided to filter out the local oscillator from the IF port. The QER crystal filter described in teh 2012 ARRL Handbook was quite interesting. A quick calculation in the lab notebook created a filter for 12 MHz SSB bandwidth. With the filter added between the IF port of the KISS mixer and the FT-817 antenna socket, the setup was powered up. The stations were weak but remarkably clear.

A single stage of the hybrid cascode IF amplifier by Kopski and Hayward was added to bring up the gain and the noise figure. This is an excellent IF subsystem. The article is available at Roger Hayward's site. Apart from the nominal front-end filtering provided by the Antenna Tuner, no front-end filtering was used. This was just an experiment to try out the KISS mixer.

A second version of the KISS mixer made with discrete 2N7000s failed to impress. It was quite noisy. Probably due to the reverse protection diodes used between the drain and source of 2N7000.

Part 4 - A simple back-end

here. The BFO design was borrowed from the BITX. A 14 MHz, three section, band pass filter was added in the front-end. The receiver was alive. After an evening of listening to twenty meters, it was clear that this topology was far better than any other receiver tried out at VU2ESE. The reception had the clarity of a high end, phasing type direct conversion receiver.

Part 5 - 20 MHz IF

While visiting a local components store, by chance, I enquired about higher frequency crystals and carried a pack of 20 MHz microprocessor crystals. The hunch proved right, these were indeed 20 MHz fundamental mode crystals. The 40 Mhz crystals proved to be third overtone and they oscillated at 13.333 Mhz when plugged into an RC coupled crystal oscillator.

These were characterized using the G3UUR method. Then, using the sweeperino, they were measured again for 3 db bandwidth and both the methods agreed with 10%. The lab notes read as follows:

Date : 4/9/2013
Freq Shift ~ 4 Kz @ 20 MHz Fc
Cm = 22p x 2 x 4/20,000 = 0.008pF = 8 fF
LM = 7923 uH = 0.008H
A QER filter was quickly designed with the parameters given by Dr.Gordon in the ARRL Handbook (Sec.11.23, 2011 ed.)

C = (Fo . Cm) / (Bw * K23), 
Order = 8, q = 1.2532, K12 = 0.7394, K23 = 0.5346

It recommended a termination impedance of 70 ohms. The crystal filter’s sweep brought out a beautiful picture.

Part 6 - Going ‘General Coverage’

With a 20 Mhz IF, the image response (at 40 MHz and above) moves away sufficiently for us to strip it out with a 30 MHz low pass filter. An elliptical, four section low pass filter was added. to the front-end.

At times, when some devices or computers were switched in the shack, one could hear strong carrier(s) in the IF pass-band. These would be fixed and untunable. These were clearly coming from 20 MHz oscillators in various equipment.

Recollecting a W7ZOI paper on constructing a deep notch filter to measure an oscillator phase noise, a similar approach was tried by adding crystals in parallel to the capacitors of the low pass filters. Though it created a deep notch, it was not at the crystal filter’s pass band. This was then substituted by three traps that you now see in the circuit.

Part 7 - Transmission

The microphone amplifier was borrowed from BITX is to amplify the voice. Bob Kopski and Wes Hayward have produced better feedback amplifiers for BITX a few years ago. This circuit was used as the IF amplifier amplifying the signal towards from the modulator to the crystal filter. It has strong 50 ohms output impedance that drives the crystal filter properly.

Part 8 - Power Amplifier

The JBOT, described at is a 5 watt QRP amplifier made by paralleling a number of ordinary, non-RF transistors. The output from the Minima was not enough to drive the JBOT to full power. An additional stage was used to boost the output at the LPF to the 0 dbm (1 mW) needed to drive the JBOT.

Part 9 - CW

Morse Code is the mode of choice for QRP operators. It was important that the Minima should support CW. An easy way to produce CW is by injecting audio sine wave into the SSB transmitter.

The Arduino is capable of producing square wave audio tones. This is a harsh sounding tone. Extensive filtering resulted in a better sounding note. But such a filter was getting more complex than a simple single transistor phase shift oscillator.

A single transistor audio phase shift tone generator was built to produce a 700 Hz tone. The bias was provided by Arduino to key the tone generator on and off. The audio tone is routed through the audio muting relay such that during transmission, the audio power amplifier’s input is connected to the tone generator instead of the receiver audio preamplifier.

Part 10 - New Audio

Tom Gayle of fame, said "With so little RF gain in your receiver, it’s a shame to see an LM386". After such an affront (*grin*) the LM386 was replaced with a discrete audio amplifier.

What’s this KISS mixer anyway?

The mixer used in the Minima has been dubbed the “KISS Mixer” by Chris Trask in his paper. The way it works is very simple.

Consider the transformer in the left diagram of the circuit below. When the lower switch is turned on the lower part of the left side winding of the transformer will become the secondary. The signal at the IF port will follow the RF signal.

However, when the lower switch is turned off and the upper switch is turned on, then the upper part of the left side winding will become the secondary, though in opposite phase. The output at the IF port will now be inverted RF signal.

Why is the KISS mixer better than the diode mixers?

The diode mixers work by using diodes as switches. To turn on a diode, current has to flow through the diode. When a diode is acting as a switch, the signal current as well as the oscillator current are flowing through it simultaneously. For best action, the diodes switches have to be quickly driven to saturation.

A diode mixer works well if all the output at the IF port is absorbed by the next stage’s input impedance. For instance a diode mixer works well with a 50 ohms resistor as the IF load. However, when a diode mixer is connected to a crystal filter, only the energy within the crystal’s passband is absorbed and the rest is reflected back into the diode mixer. This reflected energy, coming in at a different phase will further mix with the local oscillator, incoming RF signal etc and create more spurious outputs.

As opposed to a diode mixer, the JFET channel does not carry the oscillator currents nor is the channel non-linear like a diode. It is more suitable if we wish to directly drive a crystal filter.

In our mixer, the JFET is out of the signal path as it only shorts the transformer to the ground. This makes the receiver very resilient to loud and out of channel signals.