Practical Crystal Set Bandwidth Measuring
Radio Under Test
Hi Gang. There is a lot of hobby interest in making the best crystal radio possible. In particular the large litz radios designed for dx reception squeeze every micro watt out of the air and deliver it to the headphones. There is much discussion on which is the best type of coil, the best wire and form to be used. Much of this boils down to the measurement of the coil or circuit Q.
A lot of good work has been done by several master crystal set designers, including Dick Kleijer's article on some LC Experiments he recently did. His 3 part article shows the effects of different coils, and tuning capacitors. He also outlined how difficult it is to measure Q on high performance litz coils. Ben Tongue has done some excellent writing concerning circuit Q, diode data and bandwidth calculation. His series of articles are the basis of my dx circuits.
I have decided to take a somewhat less scientific approach. I did a study on the bandwidth of my #50 radio, which is pictured above. A lot of you understand what a 20 or 40 db change means in radio or even audio. Think of that guy driving down the street with his kilowatt music blaster thumping you to death. Imagine how nice it would sound at 40 db less. People know what I mean, but how about the relationship of doubling the circuit Q? That number has to be translated into selectivity before it can be understood by most.
My #50 radio is the best one for doing bandwidth tests. The coils use very large 660/46 litz wire. You can't get better variable capacitors than found in this set. The coils are on separate wooden bases, so any separation is possible. The hobbydyne circuit with variable selectivity and multi diode selection has everything that is needed to test bandwidth.
My test gear is a little on the light side. The Jackson 640 signal generator is one my dad bought over 50 years ago when he was in the radio repair business. There are much better generators out now, including some really nice digital types. Those are on my shopping list, but for now I will use what I have. Dick Kleijer provided me with a good circuit to place between the signal generator and the radio under test. This approximates how a long wire antenna would work.
The multi meter that I use is a Fluke model 87. This has .0001 volt resolution on the lowest scale. This will work fine connected to the detector output of the radio. Do not attempt to use a meter with less than a 1 meg ohm input impedance. Most analog meters do not have this high impedance input and will load the crystal set circuit.
I peaked the detector tuning each time I changed a diode or turned the differential capacitor. Then I varied the frequency, first up 10 khz and then down 10 khz. At each setting I took a voltmeter reading. Then I figured the db drop off.
The FO-215 and schottky diodes (2 wired in parallel) are the "premium"
type diodes. The 1N34A is the Volkswagen of germanium diodes.
The chart below is first divided by frequency, 550, 1000 and 1600 khz center frequencies. After the two capacitors are peaked, then I moved the generator up and down 10 khz and measured the level. From these numbers, the db difference was calculated.
EXPERIMENT 1 Maximum Sensitivity Setting Maximum Selectivity Setting - 10 khz 550 khz + 10 khz - 10 khz 550 khz + 10 khz mv. db. mv. mv. db. mv. db. mv. mv. db. FO-215 0.1 -59.7 96.7 1.7 -35.1 0.1 -47.3 23.2 0.1 -47.3 1N34A 0.3 -50.0 94.9 2.5 -31.6 0.1 -49.6 30.2 0.3 -40.1 2 Schottky 0.1 -59.4 93.9 1.1 -38.6 0.05 -51.6 19.0 0.05 -51.6 Maximum Sensitivity Setting Maximum Selectivity Setting - 10 khz 1000 khz + 10 khz - 10 khz 1000 khz + 10 khz mv. db. mv. mv. db. mv. db. mv. mv. db. FO-215 8.8 -27.3 204 55.6 -11.3 0.6 -42.9 84.0 8.2 -20.2 1N34A 11.9 -24.0 188 58.9 -10.1 1.1 -38.1 88.0 11.5 -17.7 2 Schottky 5.4 -31.7 208 47.0 -12.9 0.3 -48.3 78.5 5.3 -23.4 Maximum Sensitivity Setting Maximum Selectivity Setting - 10 khz 1600 khz + 10 khz - 10 khz 1600 khz + 10 khz mv. db. mv. mv. db. mv. db. mv. mv. db. FO-215 40.5 -13.6 194 38.3 -14.1 7.6 -21.7 93.0 4.4 -26.5 1N34A 33.0 -13.1 150 45.0 -10.5 10.3 -19.8 100.4 8.7 -21.2 2 Schottky 27.0 -16.7 184 31.0 -15.5 4.5 -25.7 87.0 2.4 -31.2 Radio under test: Dave's Homemade Crystal Radio #50, Coils spaced 10 inches (25cm) Signal Generator: Jackson 640. Output connected to the antenna and ground terminals via dummy antenna network pictured below Modulation : Unmodulated carrier. Frequency readout: Zero beat to Icom 706. Generator output : Adjusted for best scale on meter on each frequency test. Meter : Fluke Model 87 on DCV range. Connected to O-X and G on radio. Frequency shift : + and - 10 khz. Millivolt readings taken and db calculated.
Observations, Experiment #1: The overall readings were pretty much as I expected. What was unexpected was the difference in levels between 10 khz above and 10 khz below the test frequency. I carefully readjusted the tuning and the readings were the same. Therefore, both readings are shown instead of averaging the above and below center frequency numbers.
There were some hand capacitance effects while tuning the antenna tuning capacitor. This wouldn't be at all noticeable while the radio would be in normal use, as the amount of change amounted to a fraction of a db. But it shows how sharp a set with a large litz coil will tune.
On an overall basis, the FO-215 diode is the best for sensitivity, and the 2 schottky diodes had the best selectivity. The selectivity is really not a problem at the low end of the band, so I would recommend using the FO-215 and keep the sensitivity at a maximum.
At the high end of the band, having good selectivity is most important, so I would recommend using the schottky diodes. The selectivity can be adjusted to maximum with the differential capacitor. You do lose a few db signal, but it might be worth the loss to hear a station that is next to another.
EXPERIMENT #2 DATA Coil Spacing 990 khz 1000 khz 1010 khz average db +/- 10 khz 18 inches (46 cm) 0.2 mv. 15.5 mv. 0.3 mv. -36.02 15 inches (38 cm) 0.9 mv. 35.8 mv. 0.8 mv. -32.50 12 inches (30 cm) 2.6 mv. 70.7 mv. 2.5 mv. -28.85 9 inches (23 cm) 10.4 mv. 145.0 mv. 10.0 mv. -23.05 6 inches (15 cm) 69.0 mv. 224.0 mv. 66.0 mv. -10.41 Radio under test : #50 Diode : FO-215 Sensitivity : Maximum setting with differential capacitor, Selectivity widest Meter + Generator: Fluke Model 87, Jackson 640, unmodulated carrier Generator level reduced from experiment 1 Frequency shift : + and - 10 khz. Millivolt readings taken and db calculated. Then an average was taken for final db value.
Observations, #2: This experiment demonstrates the effects of the coupling distance between the two coils. The results are as I expected. The selectivity deteriorates rapidly as the coils are moved the closest. Based on these readings, I believe that 12 to 15 inches (30 to 38 cm) is the optimum distance.
The further down the band, the closer the coils can be positioned. I would recommend below 700 khz that the coils be moved 6 to 9 inches (15 to 23 cm) for maximum sensitivity. Since experiment #1 has demonstrated that this receiver has more than enough selectivity at the bottom of the band, some of that can be given up for sensitivity.
Dick also pointed out that when the sets are over coupled, there will be two peaks on the band. The second peak can as far as 100 khz away from the main peak. To test this, I put the coils 4 inches (10 cm) apart and peaked the tuning at 1000 khz. Then I adjusted the generator frequency over a wide range and found a second peak at about 940 khz! The 940 khz peak was only about 2 db below the 1000 khz peak. I started moving the coils apart, a little at a time, re-peaking the tuning each time. I found that until the coils were ten inches apart, there were two distinctive peaks. As I increased the coil separation, the off frequency peak kept decreasing in voltage, compared to the on frequency value. At ten inches, there was only one peak.
My conclusion is that there is always a trade off between sensitivity and selectivity. Pick the best for your needs.
Your comments, questions and disagreements are welcomed. My aim was to measure on a practical basis, the selectivity of the entire crystal set, rather than only parts of the set. This article has become an international undertaking. I want to thank Dick Kleijer in Holland and Ramon Vargas located in Peru for their time and input. Their comments and suggestions have made this a better article.
This is Dave, N2DS wishing you good crystal set dx.
Schematic of the radio under test.
Don Peters from Calgary, Alberta sends me the following:
The asymmetrical response is due to a combination of a signal generator output impedance which varies with the generator amplitude setting, the variation in the detector load impedance at the sensitive and selective settings and a further detector load impedance variation with signal amplitude.
When these result in different loaded Q's for the primary and secondary circuits, and the pri and sec inductances are different, in this case by design, the asymmetrical response curve is present. Although the two circuits may be tuned to the same center frequency, there is a small difference in the tuning when related to the +-10 khz offset. This difference is magnified by the Q difference.
In the measurements where the response curve is symmetrical, there is either a close match between the pri and sec or the individual factors compensate each other. These or similar factors are generally present in most practical double tuned crystal sets - part of the design compromise.
Some of the measurements have a +-10 khz amplitude which is less than a millivolt or two and the apparent selectivity is very high. If the offset signal was modulated and monitored with headphones, it would be greatly attenuated at 10 khz offset, exactly as the numbers indicate. However, if a second modulated signal is introduced at the resonant frequency then the offset signal attenuation will be much less.
This is because the diode becomes less sensitive in the weak signal, square law, region and also loads the secondary less, raising its Q. The second signal at resonance biases the diode, increasing its sensitivity, and lowers its resistance which lowers the secondary Q and increases its bandwidth. As a result, a set may not separate closely spaced stations as well as its measured selectivity would seem to predict. This effect is responsible for the situation where a weak station will have silence on either side of it and background stations underneath it.
Additional Measurements, April 2006
I made some additional measurements, having recently purchased a SFG2110 DDS signal generator. I used my scope as an indicator. I knew that a X10 probe wasn't suitable for a direct connection to the top of the tank, so I used a gimmick capacitor of a few picofarads to couple from the tank to the scope.
The coil under test is the detector coil in my #63 crystal set. The first capacitor I used was the one on the detector board. This is a holy grail type. It has the silver plates, high end wipers and ceramic insulators. Then I selected a dual 270 pf that I used on my #60 crystal radio. The 2 wipers look like they are high quality but the insulators are phenolic. The last capacitor is a regular 365 type sold by several vendors. They are shown below:
The signal generator was coupled to the main coil with another coil placed a short distance away. The object of this test is to compare the bandwidth (expressed as a percentage ratio between the center frequency and the -6 and -12 db frequencies.) This was a first attempt and kind of a quickie measurement. I only went one way with the generator frequency and multiplied by two for the bandwidth. This is not quite as accurate as measuring the -6 and -12 db points on both side of the center. But this is close enough to compare the three capacitors. Below is the data. Comments are welcomed. I assumed that 6 db down was half height on the scope display, and -12 db was again half.
Holy Grail Capacitor Frequency -6db BW % -12db BW % 600 khz 1.2 khz .20% 2.8 khz .47% 1000 khz 3.0 khz .30% 7.4 khz .74% 1600 khz 8.6 khz .54% 19.0 khz 1.19% Dual 270 Capacitor Frequency -6db BW % -12db BW % 600 khz 1.4 khz .23% 3.2 khz .53% 1000 khz 3.8 khz .38% 8.0 khz .80% 1600 khz 10.0 khz .62% 18.0 khz 1.12% Standard 365 Capacitor Frequency -6db BW % -12db BW % 600 khz 1.6 khz .26% 3.6 khz .60% 1000 khz 4.4 khz .44% 9.2 khz .92% 1600 khz 14.0 khz .88% 28.0 khz 1.75%
It appears that the bandwidth is a little narrower at 1600 khz using the dual 270 than the holy grail capacitor. Perhaps I made an error in the measurement. I will be doing more of these measurements in the future after I improve my testing conditions.
I also suspect that the coil is being loaded heavier at 1600 khz
due to the constant capacitance of the gimmick capacitor. But the comparison
between capacitors should be good.
What does this show? I think it shows that even a slight improvement in the capacitor will provide better selectivity in a crystal radio.