Paul Lukas radio

Photo and Text information

Text chapters


revisited: 2000 01 08

Introduction / Background

I have built my first crystal radio at the ripe age of 7 from a potato! (described elswhere). My second radio followed shortly and it was better one: a weekly radio schedule magazine (no TV yet those days!) occasionally featured simple radio building plans, I made one of those. The crystal set was of the venerated 'single-slider' type, the coil was made of the windings of a defunct transformer over a somewhat wavy cardboard cylinder I glued together. My antenna actually was not my antenna at all – I tapped into a neighbor's large "L" via a wire hook using a coat hanger judiciously applied to the downlead. The owner of the antenna, a nice Air Force radio man, got a laugh out of this when he discovered the theft of his signals, but encouraged me to continue my exploits nevertheless and allowed me to use his antenna forever. My conscience now cleared, I could proceed with further work.

The slightly unsighty (according to my mother) set was exiled to our verandah which necessitated a 60-foot run of wires under the carpets through the house (to my mother's displeasure and horror) to my bed to power the ear phones tucked under my pillow. I went to sleep with radio, I woke-up with radio – I still do. Every morning at 6 o'clock, a 440 Hertz tone (middle 'A') came on serving as a wake-up signal for me which was used by already awakened musicians to adjust their instruments to the correct pitch. The transmitter was 50 miles away, a 120 kW job transmitting on 536 kHz in Budapest, Hungary with the tallest cigar-shaped antenna (to my knowledge) in Europe at that time (100 pi = 314 meters = 1030 feet, that is). Since this station started later and stopped transmitting earlier daily than the Vienna station located in Austria (about 140 miles away) around 445 kHz or so (in Europe the broadcast stations are placed 9 kHz apart due to overcrowded conditions, not 10 as here in the US), thanks to my not-so-selective set happily brought in that station too – my first DX! That did it! I stayed with radio ever since.

Mechanical design

The crystal set is built on a piece of ¼ thick scrap plexi glass plate. The legs are fashioned from four 4" 'L' brackets held together by four 6-32 flat head screws. It is a see-through design, one can look at the guts of the radio while listening to a football game. Since the radio was built as an experimental device, no housing was made for it. I keep a cloth cover over it, however, when not in use (sometimes while it is working too) to protect the the galena, the tuning capacitor and the switch contacts from dust and dirt. The lucite front plate can be edge-lit to give a handsome and mysterious appearance in the dark. The set is shown on Figures 1, 2 and 3.

Circuit description

The radio described here was built in 1996 as an answer to a challenge from a friend of mine and is the result of a quest to make it as efficient as possible with the least amount of components, providing maximum utility at the same time. The heart of the radio is the coil-capacitor parallel-resonant circuit. Due to the anticipated relatively high circulating currents in the resonant circuit, I decided that no switch contacts or other devices shall be connected to that circuit, but direct, short and heavy lead connections only between the two. As a piece of information, I measured some modern "D"- type (the ones used on computers) connector pin resistances – they are in the 20-50 milliohm (20-50 thousandth of an ohm) range. Switches, relays, due to usage and dirt accumulation can become much worse. A few milliohms (thousandths of ohms) in a high-quality circuit can ruin the expected results. I did not want to take chances. In order to find out which coil and capacitor combination is best, a series of experiments were carried out. For about a year, different coil configurations – mechanical and electrical – were tried. I settled for the toroid-on-toroid system with Litz-wire winding. To offer the anticipated relatively high circulating currents the least amount of RF resistance (coil) and loading (capacitor), different strand number and diameter Litz wires and a large variety of variable capacitors and iron cores were tested.

Measurement Results – the Coil

The optimum configuration was achieved with two toroids on top of each other with the 'figure-8' winding scheme to reduce the winding capacitance – see Figure 4. Figure-8 winding vs. straight winding increased the quality factor Q by ~20%. What I call 'figure-8' is as follows: before winding the coil, fold the intended length of coil wire in half and thread it through the toroid. Temporarily fix the half-point to the core with tape. Start winding one branch until half of the desired number of turns are wound. Affix the end with tape to prevent the turns to come loose. Now start winding the other half in the opposite direction similarly to the previous one. This way you end up with the same number of turns but the 'hot' end of the coil will have less parasitic capacitance to the 'cold' end, increasing the coil quality. (The 'hot' end will be adjacent to ½ of the winding at a ¼ of impedance point). Furthermore, it turned out, that a 'naked-Litz' wire – the insulating threads wound around the strands removed – improved the quality of the coil too! The reason being that this way the wire strands were allowed to hug the iron cores tighter, reducing the escaping magnetic leakage flux – which is a loss. Care must be taken at the winding of the 'naked Litz' not to pull the wire too tight against the sharp edges of the toroid cores to avoid the insulation be damaged or even to break some of the strands, as these will seriously degrade the quality of the coil. I wound adhesive tape over the winding to force the wire turns still closer to the core bodies. I tried different size, material and number of toroid and rod iron cores as well. Although the large diameter toroid cores are easier to wind and require shorter wires (less ohmic resistance) for the coil for the same inductance, the longer flux paths with more leakage flux reduced the quality – bulk effect. On the other hand, the smaller cores are more difficult to wind, but smaller dimensions mean higher unwanted stray capacitance increases. So, it is an optimized compromise (as most things in life and nature) to find the best suitable solution. Also tried to wind the coil on two identical but separate cores to reduce winding capacitances, but the results were unsatisfactory.

The antenna coil is wound with #26 enamel magnet wire over the tuning coil at the 'cold' end, near the end which is connected to ground to minimize undesirable capacitive coupling between the two coils. Also was tried a thin, grounded open-turn copper sheet coil between the windings for decoupling, but this introduced other losses, so this technique was abandoned. I ended up creating a coil with a quality factor (Q) of ~700! – open circuit, unloaded, of course. In comparison with the conventional coils used in older radios with a Q-factor of ~100-200, this is very high. I used my faithful Boonton Radio type 260A Q-meter for reviewuation, which had to be recalibrated because the original meter range went up 'only' to 625! In the old days, apparently it was not envisioned that somebody comes along and makes a really high Q coil! The output of the radio is = transmitter power from the antenna minus the losses encountered within the radio. The smaller the losses, the larger the output power at a given input power. With a low-loss design, more power goes through the radio to the phones, less power stays in the radio heating your room! To reduce the effects of added stray capacitances by metal proximity, the coil is held in place with a plastic cable clamp mounted on a ¼ inch by 2 inch long flathead nylon screw – see Figure 5.

The reasons for the mall diameter magnet wire used for the coupling coil are due to considerations that the antenna radiation and resistive losses are going to be much larger than the combined coil losses at that low impedance level in question, therefore this loss can be neglected. The small diameter wire also presents less capacitive coupling to the tuning coil and less bulk copper intersecting the magnetic field lines.

The 'Q and Tuning Capacitance' curves hows the frequency vs. Q-factor as measured on the Q-meter with the antenna coil installed is depicted on Figure 6. (The lower end of the band is not shown due to the limitation of the tuning capacitor range of the Q-meter). The mere presence of the copper metal and it's lossy capacitance in the form of the antenna coil reduced the Q of the tuning coil to 665 – but I had no choice - I needed the antenna coil! Still smaller diameter wire probably would give better results, but is difficult to work with it, so I left it alone. A smaller inductance tuning-coil (less turns) may have resulted in an improved Q (less stray capacitance, lower ohmic resistance of the winding) but I had to wind the coil to match the highest quality variable capacitor I could find, so the inductance value was given in order to be able to cover the broadcast band. Smaller inductor means larger capacitor on the same frequency(band) which also is desirable from the point that the tuning capacitor/stray capacitances ratio is better. Stray capacitances are considered to be lossy. The final tuning coil has 14+14=28 turns in the figure-8 fashion.

The antenna coupling coil consists of 14-turns with taps at 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 and 12 turns. I use the full winding - #12 switch contact is connected to the last (14th) turn, leaving out tap #12. The taps are connected to both of the 12-position rotary switch contacts, both switches are wired parallel – see schematic, Figure 15 and Figure 7, the switch configuration showing the little 'pigtails' to accommodate the antenna coil taps. One switch matches the antenna, the other the crystal load to the tuning coil through mutual coupling. This arrangement allows a wide range of operating conditions from super-selective to super-wide, also allows different size antennae to be optimally matched to the set at a low impedance level.

At my previous location in Long Beach, California, with a ~55-foot random length asymmetric "T" antenna strung at a moderate height between house structures, I was able to hear and separate 40+ broadcast-band stations without the use of traps, some in Mexico, ~200 mile range. For this I used an old German-made 4000 ohm magnetic ear phone I picked up years ago at the Dayton HamFest for 3 bucks. I disassembled and cleaned the phones, adjusted the diaphragms with thin spacer rings to a minimum distance/maximum sensitivity and the magnets were re-magnetized. In the condition I bought the phones the magnets were quite weak.

In my present location in Torrance, California with a 25-foot vertical (Sommer T-25) antenna I can receive KNX 50 kW transmitter at 1070 kHz at 1.3 air miles with such power, that a 5 inch car loudspeaker covers my garage and 2 adjacent rooms with enough sound level to clearly hear the news. In the 'open circuit' mode, the S-meter indicates well over 15 volts! The low impedance voice coil of the loudspeaker is matched to the radio via a 117/6.3 volt filament transformer. The transformer is held in place by the speaker's magnet – see picture. The slight magnetizing of the transformer core by the speaker magnet had no detectable ill effect on the operation of the radio since the sound flux going through the windings is minuscule in comparison with the saturation-point of the iron. That small amount of magnetizing force is still far away from the flattening (saturation) of the hysteresis curve of the iron core. The 117 volt winding is connected to the phone jacks and the secondary to the speaker. Trying different transformers, this arrangement gave the maximum power transfer with average settings of the two loading switches as measured with an ac voltmeter at the voice coil.

The Tuning Capacitor

Next issue was the selection of the proper tuning capacitor. Dozens of capacitors and capacitor configurations were reviewuated, with some surprises. Among the types were: small transistor-set tuners, larger air capacitors with 2, 3 and 4 sections, 'floating rotor' – grounded rotor configurations, etc. Even a supposedly high-quality 4-section capacitor with very heavy ceramic support rods - rescued from a Russian military radio. Before testing, the supporting insulators for the stator plates were carefully cleaned and dried and the rotor bearings cleaned and lubricated.

General rules were derived from the tests for quality, reflecting on the configuration and build of the capacitors. The rules I found are listed below:

The Detector

I experimented with several types and configurations of detectors. The conventional galena crystal (lead sulfide, PbS) – see Figure 10 - probably provides the best results with a high impedance sensitive piezoelectric phone. I had difficulty finding a sensitive enough one and could not carry out meaningful experiments. Lower impedance magnetic phones – in the 4000 ohm dc range - work well with the galena but the crystal curve is 'soft' and can not supply higher currents as the case would be with a close-by strong transmitter. The galena crystal curve shows this soft, very gradual 'knee' transition from being an insulator to being a conductor. The curve represents the averaging of 150 measurement points along the curve at varying currents with different spots on the galena crystal. I found that the rectifying capability of the galena detector can be slightly improved if the sharp point of the cat whisker's end is wiped lightly with fine emery cloth or with a fine file before using it. This process most lightly removes any oxidation products from the tip. Most oxides are insulators or lossy conductors. Oxidation can be a problem with copper and copper alloys especially if the detector is exposed to corrosive environments - smoke, smog, etc. Stainless steel whiskers are less prone to these effects.

As substitutes for the galena detector, theoretically the germanium diodes should work better than their silicon brothers due to the lover knee point on the conduction curve. I tried the household 1N34A and 1N91 diodes with good results, but at higher volumes they exhibited the same problem almost to the same extent as with the galena. I connected several of the diodes - up to 10 in parallel – see Figure 11. This gave a marked improvement by stiffening the knee point, but after 3 or 4 diodes in parallel, the improvement was only minuscule. Figure 12 illustrates the different conduction curves for comparison between the galena, the 10 parallel 1N34-s and a 'hot carrier' diode. Note that the Galena crystal has an appreciable amount of reverse current: leakage, which subtracts from it's efficiency. The hot carrier diode had less than 500 picoampère (500-6 microampère) leakage current even at 10 volts reverse bias. I also tried different types of hot-carrier diodes, all gave excellent results without fading out at higher volumes. Could see practically no difference between a single diode and several of them connected parallel, one single diode proved to be stiff enough at higher currents as can clearly be seen by observing the curves

The Field Strength Indicator

The radio is equipped with a field strength indicator (or 'S-meter' as radiomen call it) instrument which comes handy when precision tuning and by performing other experiments. 'S' stands for signal Strength. The indicator is a 30 microampère meter at full scale and it is controlled by the 3-position bat-handle switch located under it – see picture. Any low-microampère instrument will work, but the higher current meters will perceptively load down the detector – you would be feeding a fair part of the power into the meter, not into the phones. Although any combination of voltage ranges can be had by selecting the appropriate resistor values, for my purposes I have chosen a 0.3, 3 and 15 volt full scale ranges. The first two are predominantly for tuning, the 15 volt range is for no-load, 'open circuit' work. With a long and high antenna you may select higher voltage range resistor(s) to prevent the mater going off-scale. This feature allows one to track the received field strength of weaker transmitters too. Fluctuations in time can and will occur due to different reasons: transmitter power changes (some transmitters have to reduce night time power to avoid interference to other transmissions by other transmitters on the same frequency at another locations), atmospheric condition changes, local phenomena – rain, fog, snow, etc. By connecting a recorder to the phone jacks, a field strength history can be documented of any transmitter putting sufficient power into your location to make the measurements reliable. In this case a fixed rectifier – a diode - other than a cat whisker type is better because it will not drastically change as is the case with the cat whisker if different spots on the crystal are used or if the crystal is bumped.

Since all rectifiers are non-linear by nature, the rectified RF voltage displayed on the meter scale is also non-linear. The beginning of the scale is compressed with respect to the rest. However, the meter can be calibrated in terms of voltages with the aid of a variable voltage dc supply and a dc voltmeter – now you have an RF voltmeter!.

If using electrostatic/piezoelectric phones, the meter switch in the highest voltage range can serve as a dc return path for the rectifier, in this case 500 kohms.

The 12-position Switches

Although ceramic-wafer type switches are the best, I used what I had in the junk box- Figure 7. They are made of plastic but, because they are used a low impedance level, one can get away with these. I have not tried any other types, so there is no data available as to their influence in this service. 4 or 6 position switches can be used in place the 12 position ones, but the utility aspect will suffer. It can be demonstrated that one turn difference can make a large improvement on the reception. You will not have the wide range of settings optimizing the antenna and rectifier parameters. With other words, you may not get the best out of the radio. If a small size antenna is used, you may start the first tap at 2 or 3 turns, and so on. If 'open circuit' operation is planned using a small antenna, the antenna coil turns may be increased to15-20 or so. Just a guess. In some cases a variable capacitor connected in series with the antenna coil may help.

The Phone Bypass Capacitor

For reliable results, the residual RF voltage finding its way through the detector itself and the lead/socket capacitances should be bypassed, shunted to the ground. If too small, ineffective, if too large value, it will cut off the higher audio frequencies. I found a capacitor with a value of 2000 picofarad (2 nanofarad, 2nF) a good compromise. If no capacitor is used, it may causes instability in that your hand (and head) capacitance to the set may effect the rectifier's proper operation. This effect would be more pronounced in the narrow-band service.

It is possible to make the set operate on the short-wave bands with proper coils and capacitances. In this case the bypass capacitor is highly recommended. Also, as short as possible connection leads of the components should be employed. This includes the little spring coil of the cat whisker - it should be made as short and small a diameter as practicable

Some Tips

When placing the radio in service, put both rotary switches in the starting position – positions 1. This results in the loosest coupling to the set. Put the field strength meter switch to the most sensitive position. Rotate the tuning control and see if you can hear/see a station. Regardless, turn both switches in position 2. Tune. Unless you have a very small antenna or leave very far from a station, you should hear something. Keep moving up the antenna switch – tune – try the detector switch, etc. Remember that after changing a position on either of the loading switches, the tuning control has to be readjusted - all three controls are interactive. When you reached a point where you get a deflection on the meter, you will appreciate, how useful it is! Tuning is made much easier with the help of the meter. You will be surprised, how sharp tuning can be achieved with loose couplings.

If station field strength monitoring is your ticket, after disconnecting the phones, you can connect a reasonably sensitive recording device to the phones/speaker jacks. Or you can run 'open circuit', just with the meter. The modulation of the carrier frequency will show up on the meter, trying to follow the modulation waveform. If this bothers you, a 100 microfarad (value not critical) low-leakage elctrolytic capacitor can be plugged into the phone jacks. You have to retune after you applied the capacitor. Attention must be paid to the polarity – depending on the detector orientation in its jacks. Most signal diodes have a ring near the cathode (+) end. This polarity should go to the phones and meter. For the galena crystal, the crystal side is the anode (-). So, the cat whisker (+) side goes to the phones. With the wrong polarity, the meter needle tries to go backward! With the large capacitor in the phone jacks, you must tune very slowly since it takes time to charge the capacitor. Figures 13 and 14 show the car speaker I have used with good results. Figure 15 is the schematic diagram.


The crystal set described here is performing well, even surprisingly well considering some of the less efficient antennae I have used. I listen to the set regularly while working in the garage where the set is located.

Possible areas of further improvements may be achieved by using a combination of different toroid core materials, Litz wire having more and thinner conductors and thicker wire bundle and by using super-capacitors of one-piece stator and rotor segments. I tried to carry further the 'figure-8' concept by arranging the winding in two figure-8 patterns 90 degrees to each other, but this act gave inferior results. Also tried 1 and 3 toroid configurations. The 2-toroid coil gave me the best results by using the available cores and Litz wire types. Carrying the improvements even further can probably be only an academic interest, since if the radio to be useful, has to be working at a much lower Q-factor (loaded) than the open circuit condition, to be able to extract power. The efficiency differential between a Q of - let us say 700 and 800 – reduced to Q~300 in the operational mode is small indeed, considering the low power levels present in these types of circuits.

A trap (or several traps tuned to different frequencies) may be used in the antenna circuit to help filtering out some undesired strong station(s). Keep in mind that anything extra you connect to the radio will consume/absorb some power. Since all the power come from the transmitters, the losses encountered in the traps will reflect in the overall efficiency. That means that the useful energy will be less at the phones. You can use either parallel or series coil-capacitor combinations for traps. In any case, the traps should have reasonably high Q-factors (sharp tuning curve) to effect the undesired station(s) only, not the adjacent ones you may want to hear.


One can not only hear, but see a station too. If the regular galena or a fixed diode is substituted by LED diode as a detector, it will light up when tuned to a stronger station. Cathode of the LED should go to the phones jack – if you want to see the meter go upscale, not in reverse. I had a heap of old, different color LEDs – old enough before the modern high efficiency diodes have been invented. I chose the red LED for it’s advertised higher efficiency. Tests of different color diodes confirmed this. With the help of the field strength meter, you can easily tune the light to the maximum output. Although not as efficiently as a Schottky diode, my LED drives the inefficient car speaker nicely without audible distortion. Increasing the loading switch settings will not increase the volume over the maximum reading on the meter, but the high sounds will be coming in better due to the wider bandwidth. Even though the LED glows weakly with the 2000 pF bypass capacitor alone without a load, one should use small/medium dc resistance phones – like the speaker transformer primary I use. If you are not interested of the sound at this point, the maximum light can be obtained by shorting the phone jacks. At loose antenna/diode coupling with the load switches, one can not only hear, but see too the narrow bandwidth of the 700 Q [unloaded] coil. I tried the old trick we used to check transmitter output with a small bulb inductively coupled to the tank coil. I found that 2-4 turns of wire wound through the toroid core makes the LED glow nicely. In both cases, if you tune the radio slightly off the carrier center, the LED will flicker according to the modulation. Flashlight bulb will not glow unless you are very close to a large transmitter using a long and high antenna. Next time I will try to build a crystal-controlled [!] oscillator using a tunnel diode – leading to further experimentation! P.S.: My laser diode did not work at all.
Please see the last 2 images at the bottom.

Parts list

2 – toroid cores, Palomar Engineering T-130
broadcast band frequency material Litz wire 52 strands - length sufficient to wind the coil about 2 feet #26 or 28 enamel insulated magnet wire 1 – high quality variable air dielectric
capacitor with ceramic stator support(s) 2 – multi-position rotary switches – suit your taste 1 – 30 microampère or more sensitive meter 1 – 3-position switch 8 – banana jacks 8 – solder lugs for banana jacks 1 – galena cat whisker or diode rectifier 3 – knobs 1 – tuning indicator arm 1 – face plate support feet, cabinet if desired

List of Pictures

1* Front view

2* Back view

3* Side view

4* Toroid showing figure-8 winding+antena coil

5* Toroid mounting

6* Q-curve of toroid coil

7* Parallel wired rotary switches

8* Capacitor rotor connections 

9* Stator connections on variable capacitor 

10* Galena detector closeup

11* 10 1N34-s parallel

12* Galena + diodes compaison curves

13* Speaker with transformer

14* Speaker front view

15* Schematic diagram