Updated  13 Apr '08

--> home/projects/RTX5x20/


NEW! Go and see the Pic-plot2: our GPIB to USB converter
                Visit our Ham Radio pages to know about our 10GHz activity

"Five by Twenty"
A 5 tubes SSB QRP Transceiver for 20m Ham band

Can you build a complete SSB QRP transceiver with just 5 vacuum tubes? I actually did it, and it works without any transistor or IC.
A prize-winner project at MAdS 2006 by IW9ARO

WARNING: Vacuum tubes work at dangerous or even lethal voltages! DO NOT replicate this project if you are not used to work with tubes.










Many projects of homebuilt ham transceivers obey to some intriguing rules, that often are quite challenging in miniaturization, or refusing to use any Integrated Circuit, or being powered by a lemon. This project has a similar inspiration: only tubes and diodes (then no transistors or ICs at all). Looks easy? Let's try to do it with just 5 tubes, operating in true SSB, 5W transmission, and superhet RX decently driving a loudspeaker. This is not a theory project, but a real QRP transceiver to make real QSOs, and in fact it stands in my shack together with other radios I built or bought.

Nevertheless, it is not an easy design. And even if I'm trying here to give all informations to anyone willing to make a copy of it, skills, experience and instrumentation are needed to get it working. Bottom line, I hope you will like the ideas and solutions I used, and perhaps some of them will be re-used in your own tubes-radio design. Don't forget that the high voltages around vacuum tubes CAN KILL, so take extreme care when working with them.

Because of its original concept and the good execution, this project has won the first prize of the homebuilt equipment competition at MADS 2006 edition in Italy.


To get the maximum possible from 5 tubes and a bunch of diodes, you cannot separate RX from TX. Each tube must play a role in either mode, similarly to what is normally done with the SSB filter in all commercial SSB transceiver. A wise choice of multi-function tubes (I.E. triode+pentode) has been exploited, still remaining in the range of easy-to-find tubes. Passive (diode) mixers are intrinsically reversible, making the two conversions from BaseBand to RF and vice-versa very straightforward.

Another important choice is the LO generation. A really usable SSB transceiver must stay on frequency, and a stable VFO is a tough design by its own with tubes. So I made my life more easy with a 9MHz VXO, using CB crystals oscillating at the fundamental frequency. A "coarse tune" rotary selector chooses one out of 12 CB crystals (can be more ore less), and a variable capacitor makes the fine tuning within 3KHz.

From LO oscillator choice, it comes that the IF frequency is a non-standard (~5MHz), so the SSB IF filter has to be designed, built and aligned. With stock 5.2428MHz crystals I built a quite good 2KHz BW ladder filter, whose characteristics are shown later on these pages.

TX output power is an important parameter that affects design choices. With a simple design and one EL84, 10~15WPEP is an affordable target that I actually missed, mostly because of HT limitations. But I'm happy with my 5W pure QRP, expecting DX contacts as good as my FT817 with an 1/2wave dipole does.

So, in summary, what I request to my 5 tubes is:

  • True USB
  • Covering the SSB portion of 20m ham band
  • VXO with crystal selector and fine tuning +/-3KHz
  • Superhet RX, sensitivity <0.2uV
  • Driving a loudspeaker, >0.5W
  • TX PEP power: >5W
  • S-meter
  • AGC
  • No semiconductor components, with exception (sorry...) of diodes


After many trials and some architectural changes, the final configuration I reached is:





EF80 pentode

RF RX Amplifier

PA driver TX


EBF83 pentode+diode

IF RX Amplifier + AGC detector

IF TX Amplifier


ECF80 triode+pentode

AF preamp + AGC buffer

mic amplifier + modulator buffer


EL84 pentode

Audio PA



6BK7A dual triode

Local VXO + BFO


Homebuilt dual balanced diode mixer

RF to IF downconverter

IF to RF upconverter


Homebuilt dual balanced diode mixer

Product demodulator

Product modulator


Homebuilt 4poles ladder filter

SSB IF filter, 5.2425MHz 1.1KHz (-6dB)


   Back to top



The block diagram shows how the blocks are related and switched between RX and TX. To simplify, the power supply section is omitted. Each block works in both RX and TX, therefore many switches are needed to reverse the signal path from antenna to audio and vice-versa. 3 relays, for a total of 8 switches, are simultaneously driven by the PTT switch on the microphone. Apart of this "reversible" architecture, it's still possible to recognize a quite plain single conversion SSB receiver/transmitter. The big step is done by the tubes: just a 4 stages lineup is capable of about 140dB total gain from antenna to speaker.


(All relays in RX position)

Blocks reversibility requires a certain uniformity of drive and load impedances at their ports. Wideband and narrowband impedance transformers have been used to unify impedances to a relatively small value (300~400ohms, excluding the ports that connect to the antenna). This low impedance works well with the SSB ladder filter and the diode mixers. It also helps to avoid self-oscillations of gain blocks due to parasitic capacitive coupling between relays sections.

Diode mixers are intrinsically reversible, so they don't need to be "reversed", at the price of some conversion losses. Furthermore, they have enough dynamic range to handle the weak signals at reception and the large ones at transmission. Each oscillator output (the VXO and the BFO) is permanently connected to the LO ports of respective mixer and, by design, there will be no shift between RX and TX operating frequency.

The receiver has a simple AGC circuit, that will be better described later. AGC voltage and the S-meter drive current is obtained by rectifying the low-impedance audio at the output of the cathode-follower named AF/AGC buff. During transmission, the same buffer drives the AF side of the balanced modulator, that is a low-impedance port. Similarly, the AF signal section that amplifies the detected audio in RX, works as a microphone amplifier in TX.

Quite uncommon is the configuration that allows to the EL84 pentode to play the dual role of AF PA and RF PA with a limited use of relays. The signal to the grid G1 can be either AF from the audio preamplifier or the RF from the driver, and the plate load is made by two transformers in serie: during TX the output audio transformer is shorted and during RX the RF transformer is a short by itself for the audio signal.

   Back to top


The complete schematic shown here includes the supply section, but omits the low-voltage part needed to drive the relays coils. I've used a voltage doubler from the 6.3VAC of the tube heaters because I chose 12V relays.


(All relays in RX position, click on picture to open in new window)

Special parts list


3uH, airwound
800nH, ferrite core, adj
39uH, ferrite core
25 to 35uH, ferrite core, adj


26/3 turns, 4mm dia, ferrite core, adj
23/5 turns, 5mm dia, ferrite core, adj
13 turns, trifilar, 7mm dia toroid


55/10 turns, 5mm dia, ferrite core, adj
5turns, trifilar, binocular 14x8x8mm
miniat. audio push-pull driver transf.
5turns, bifilar, binocular 14x14x8mm
Output audio transf for 6V6
60VA/ 200V, 200mA + 6.3V, 2A

 For a better understanding, I will split its description in the two possible configurations (RX, TX) using two simplified diagrams.

Circuit Analysis in RX mode  

You can refer to the complete schematic or to the simplified diagram below, that shows only the signal processing during reception. The circuit is quite straightforward, but I will give some details here that will be useful also for transmitting operation.

The signal from the antenna, filtered by the bandpass made by inductive coupled T1 and T2, is fed to V1 grid1. Impedance is raised to about 4Kohm for impedance matching. Amplified signal (~32dB) on the plate is then reduced to the 300~400ohms by the resonant transformer T3 and fed to the downconversion. Below the 3dB bandwidth and full response plots of the RF stage.

The passive mixer is homebuilt, using Germanium diodes and tiny toroids (T4, T5). Its conversion loss is 2dB better than a commercial wide-band one (like EMT-4) at this low frequency, probably because of Ge diodes vs. Schottky.

LO generation, as described before, is a VXO Clapp Oscillator, working at fundamental series-resonant frequency of CB crystals. The triode chosen, after some trials, is 1/2 of a 6BK7A (V5a), that has given the best output amplitude and VXO span comparing to other triodes from my small stock (12AX7, ECC82). Many, but not all, of the 20 CB crystals I own can span 3KHz , so don't rely too much on this value. The worst-case to maintain the oscillation is when the variable capacitor is at the minimum value, thus it should be adjusted to a safe value if necessary. I've experimented that the VXO oscillator should be in some way buffered to avoid residual FM and the pulling effect that offsets the frequency between RX and TX. The cause is the variations of the load impedance offered by the mixer LO port. Having no room for a buffer tube, I managed to remove both problems by loosing the coupling of the oscillator output to the mixer. From 3Vpp on the V5a cathode, the amplitude drops to 1.2Vpp, that is still enough drive for the mixer. As an extra care, plate voltage is stabilized at 160V by a serie of two zener diodes.

The 5.2425MHz IF is filtered by the homebuilt, 4 poles ladder filter. 5 poles or more would be better for USB, but I had only 5 crystals in my hand, and one was needed for the BFO, so take it or leave it...

Instead of make calculations of fixed capacitors, I've tuned the filter using an HP 8753A VNA to get the best possible from it. As visible in the frequency response plots (and as described in literature...) the two skirts are asymmetrical, demonstrating that ladder design better suits LSB filtering (or whenever the rightmost side is used). Unwanted sideband rejection is -20dB at around 300Hz modulating audio, and -36dB around 2KHz (would be respectively -35 and -50dB if it was used as an LSB filter or with an LO-RF=IF conversion scheme). Ripple is quite large : 5dB peak-to peak, but due to the high first overshoot.

Sure, this is not the best SSB filter for the purpose, but it works and -believe me- nobody could ear what a network analyzer sees. I've seen many DIY transceiver projects with a similar filter, and no one tells how critical is its adjustment and how important is to use it the right way.

Both filter ports have impedance around 400ohms, therefore the wide-band transformer T6 (1:9) is needed to match the 4~5Kohm V2 input impedance. After amplification, the resonant transformer T7 brings the impedance back to a lower value. Both V1 and V2 are variable- Mu pentodes, in order to be driven by the AGC voltage, therefore their plate loads are resonant transformers to eliminate the harmonic distortion that those pentodes introduce. After some tests, the EBF83 has been preferred to the EBF80 because of the slightly higher gain. Overall IF stage gain, including filter and transformers losses, is around 27dB.

Note: All plots are captured using Pic-plot and 7470.exe


Amplified IF signal is then fed to the product demodulator, another dual balanced diode mixer. This one is different from the downconversion mixer, because the baseband-side transformer T9 is an audio transformer (a push-pull driver), taken from an old AM transistor radio. To achieve a good mixer balance, I successfully experimented a simple solution with one resistive trimmer and two capacitors. The BFO, built around the triode V5b, is almost identical to the VXO, but runs at fixed frequency with a 5.2428 crystal. Carrier injection point has to be adjusted by L4 at 5.2410MHz +/- 200Hz.

Well, finally we got the audio baseband, starting from a weak SSB signal at the antenna connector. We have used just 3 tubes so far, so we still have 2 in the budget to complete the project with AGC, S-meter and speaker audio.

Generally, the AGC voltage is generated by rectifying the amplified IF signal, but here the sum of RF+IF gain is not enough to get the -10V we need for full AGC. Another option, more suitable for this project, is to rectify the amplified AF. The pentode V3b, with a voltage gain ~150, amplifies the AF from the demodulator to feed simultaneously the volume potentiometer and the grid of the triode V3a, configured as catode follower. The cathode follower doesn't add any gain to the audio signal, but having a relatively low output impedance (~1/1000 of V3b output), not strictly necessary for the high-resistance AGC line, we need it to drive the low resistance coil of an S-meter. The same audio buffer, during transmission, works well to drive the low-impedance port of the balanced modulator, so we would need it anyway.

AF rectification is done in two ways: a Germanium diode for the S-Meter and a tube diode embedded in V2 for the AGC. Originally, the idea was to use the second one for both purposes, but I learned that its high resistance (tens of Kohm) is unsuitable to drive a 200uA FS microammeter. AGC times I found comfortable are about 40ms (attack) and 500ms (decay) with the component shown. It is possible to modify them by playing with R*C values, but not independently.

Being a simple AGC circuit, without a dedicated amplifier, the AGC curve is quite "soft", without a well defined threshold (somewhere around 5~10uV) and a weak limiting effect: after the threshold, 80dB increment at the antenna translates into 20dB at the speaker. Nevertheless, there are no saturation effects because all blocks in the signal processing chain can handle very large signals (they all work in TX, too), and the only drawback is the frequent use of the volume control.

After the volume control we have the audio PA, operated by the power pentode V4 (EL84). The AF signals at its plate ignores the RF transformer T10 and goes totally to the audio output transformer T11, and then to the loudspeaker. I also put an headphone jack in the front panel of my prototype, not shown in the schematic.

OK, now the receiver is complete, but we have used all 5 tubes. So the question is: can we transmit with the same tubes?

  Back to top


Circuit Analysis in TX mode  

It can be useful to refer also to the simplified schematic below, where all blocks are shown as they are connected during transmission. Many considerations and details described in receive mode apply also to transmit, so I don't repeat them.

Starting from the microphone, and going counter-clockwise, there is an RF blocking filter, then the V3b that, as told, amplifies the amplitude by 150 and feeds the cathode follower that drives at low impedance the balanced modulator. With the series trimmer I've been able to adjust modulation level with electret, piezo and dynamic microphones. The AGC detector has been disconnected so that the V1 and V2 will work at full gain. The unwanted LSB is filtered out and the SSB signal amplified by V2 is ready for the upconversion at 14MHz. The same RF bandpass filter used for the antenna signals is reused to reject the unwanted mixing products: yes, there is an impedance mismatch here, but I've verified that adding a 1:9 transformer here doesn't add anything.

The same V1 amplifies the modulated 14MHz signal to be sent to the RF PA (V4) and then transmitted. V4 plate now has the audio transformer shorted, so the whole RF power is transferred by T10 to the classical PI filter. T10 works simultaneously as plate RF choke and as 1:4 wideband UnUn, thus operating a first impedance matching step to the antenna. RF PA is an "A-class" amplifier, whose EL84 could be capable of 10~15W with an appropriate anodic supply. In my design I've achieved the 5WPEP minimum target, so I ' quite happy. A simple RF detector monitors the output power using the S-meter microammeter.

  Back to top


As tradition rules, amateur tube equipments are built by point-to-point wiring, with components directly soldered to tube sockets pins and to insulated supports. Mine is no exception:


The chassis is open, with the tubes visible. On the top of the chassis there are mounted the supply transformer, the audio output transformer, one of the relais, one tunable RF transformer and two submodules. The square submodule close to the crystal set is the 1st mixer, while the rectangular one in the center contains the RF filter together with the SSB filter.


Front panel is essential with the few controls needed. On the two sides of the chassis, a touch of style with the wooden finish:

  Back to top






Noise Floor: ~70nV (-130dBm)@3dB S+N/N

Sensitivity: 0.14uV (-124dBm)@10dB S+N/N

Blocking (deltaF=50KHz): -46dBm

Blocking (deltaF=10KHz): -50dBm

3rd order IMD (deltaF=25KHz): -66dBm

3rd order IMD (deltaF=100KHz): -66dBm

IMD dynamic range: 64dB

Calculated IPC (ARRL method): -34dBm

2nd order IMD (8+6=14MHz): -26dBm

2nd order IMD (22-8=14MHz): -16dBm

IF rejection: 92dB

Image rejection: 96dB

S9 reading: 50uV


NOTE: tests as described in RadioKitElettronica 5/99

Power: 5.5W PEP

Carrier suppression: >33dB

2nd harmonic: ~40dB below fundamental

3rd harmonic : >50dB below fundamental



I've been pleased to make some DX with this radio (not many: I prefer to spend my time in my lab), mainly to test its operating capability. It's a QRP, so it's not a breeze: with an 1/2 wave dipole antenna I made contacts from Italy with north Europe, Spain and, my record, with Brazil. So the conclusion is: it works!


  Back to top

The projects presented in these pages are our own design and have been tested and verified by ourselves at the best we can. However, they might be inspired by concepts, ideas, solutions coming from known-art or free resources on the Web. We provide them as  reference designs to skilled hobbyists and technicians  who are willing to reproduce them for non-commercial use. Your results might be different from ours and we cannot be considered responsible for that. Similarly, we are not responsible for any damage or injury you might incur while building, assembling or using the equipments, projects or ideas presented in these pages. The firmware embedded in our projects is our property unless differently stated and, when available in the Download Area, it is license-free only for non-commercial purposes.  

Hotwater 2007-2009