The Double Tuned Mystery Crystal Set Radio
By, Dan McGillis, WB3KBW
Hi all.
A double tuned Mystery set is my favorite crystal radio. I’ll bet most of the “old timers” on the RadioBoard have built at least one.
By virtue of the Mystery set design, the tank circuit is lightly loaded so sensitivity and selectivity are generally better than most single tuned crystal sets. Double tuning the Mystery set removes even more load from the tank circuit. Even with budget components, the performance of a double tuned Mystery set is outstanding.
Some Mystery set background info can be found at the following web sites.
Ray Creighton’s description & “Proton’s” 3 Mystery set articles from 1932-33:
http://www.clarion.org.au/crystalset/mystery.html
the MRL #39 Crystal Set:
http://www.modernradiolabs.com/
Mike Tuggle’s modified #39 set:
http://www.crystalradio.net/crystalsets/miketuggle/index.shtml
Mike Peebles Telefunken set:
http://www.peeblesoriginals.com/projects/images/telefunken2.jpg
Ben Tongue’s explanation of how the Mystery set works:
http://tinyurl.com/yaqwqu9
Below is the schematic of my latest double tuned Mystery set. The inductors are Amidon 1/2”x 4” ferrite-61 rods wound with 25’ of inexpensive 165/46 litz. The capacitors have phenolic insulators.
Mystery set builders know that you have to play around with a single tuned Mystery set to get the behavior you want - trying the antenna and ground at 5 possible connection points. Connection to points 4 and 5 generally give the best sensitivity AND selectivity as a single tuned set.
As a double tuned crystal set, you can also play around with the various coupling options between the antenna tuner coil and the Mystery’s detector coil, plus the various antenna - ground connections to the tuner.
The two tables under the schematic show the results of fiddling with the tuner-detector coupling and the antenna-ground connections to the tuner. This particular setup “liked” to have the hot ends of the coils coupled to each other. And, it “liked” particular antenna & ground connection points: the antenna to the “black” point, and the ground to the “red” point on the tuner. These conditions increased the signal strength by about 0.6 to 3.4 dB over the average of the other possibilities. Since a 3dB increase in signal strength is quite noticeable - at least for me - the fiddling around was worthwhile. Your mileage may vary.
Overall, double tuning this Mystery set increased the signal strength by about 2.5 to 4.1dB over that of a single tuned Mystery set.
As is expected of double tuning, the selectivity is greatly improved over that of a single tuned set. The selectivity starts to fall off noticeably for freq>1400 kHz, probably because of the phenolic insulators on the capacitor. But the selectivity is really quite good at freq< 1400 kHz.
I was very pleased to be able to easily hear the daytime ground wave signal from 5 kW stations at the low end of the band - 560, 580, 590, and 610 kHz - that are 45, 84, 56, and 63 miles away. Hearing low band stations on a crystal set has always been a problem for me.
It picks up (and clearly separates) virtually all of the daytime stations that I can hear on an inexpensive table radio. With a Benny + Bogen T725 + piezo element headphones in the audio chain - 18 of these stations are “armchair copy”.
This simple little crystal set is on it’s way to being a viable alternative to a powered table radio.
73, Dan
Does the following make any sense.??
A.) IF I have it right, your points are that in the double tuned configuration, the Mystery coil is acting as a step-down transformer with Np = 72t, and Ns = 37t.
So the impedance of the primary coil (Zp) appears to the secondary to be given by:
Zs = Zp*(37/72)^2 ~ Zp*0.26
The diode, on the secondary side, sees a stepped-down tank impedance (resistance at resonance).
B.) The tank’s equivalent parallel resistance at resonance is: Rpt = QL*XL
From Jim Shaffer’s and Todd’s (WD4NGG) Q measurements of Amidon & CWS Bytemark 1/2”x4” ferrite-61 rods wound with 165/46 litz -- as reported in Rap ‘n Tap posts -- the unloaded Q of the tank coil may be in the 700 - 800 range at 1 MHz.
With this capacitor (phenolic insulators), I’ll guess that the tank’s unloaded Q is nearer to 400 at 1 MHz.
I’ll further guess that coupling the tank to the Tuggle tuner drops the tank's Q to a loaded QL of about 200. This would give a bandwidth of 5 kHz at 1 MHz, which seems reasonable from what I hear listening to the set.
So the tank’s Rpt ~ 200*(2*pi*1MHz*368 uH) ~ 462k; pretty high to match a diode to.
C.) BUT, the stepped DOWN equivalent parallel resistance that the secondary sees -- ie., what the diode sees, would be = 0.26*462k = 120k.
D.) I’ve measured the FO-215’s “zero-crossing-resistance” (or “cross-over” resistance, or “critical-resistance”) for very weak signals as ~ 200k. It’s resistance for large signals would be less.
So the stepped down resistance of the tank at resonance is a better match to the diode’s resistance than would be the case without the secondary coil.
Or maybe the other way I should look at it is -- the stepped UP diode resistance seen by the primary is not only a better match to the tanks resonant resistance, but is also a lighter load on the tank so it’s loaded Q is higher.
To look at this coil issue some more, I did the following experiment using signals from daytime stations.
I measured the signal voltage across the 500k benny with:
a.) the diode on the hot end of the secondary coil (the 37t coil) as in the figure above - vs -
b.) the diode on the hot end of the primary tank coil (the 72t coil), ie. not using the secondary coil. This makes the set an ordinary double tuned crystal set.
c.) And, I listened to close stations at:
1040 kHz & 1060 kHz,
1370 kHz & 1400 kHz
1600 kHz & 1630 kHz
to try to hear a selectivity advantage between case a.) or case b.).
The results are summarized below.
My conclusions from the above results when the Mystery coil is used in the DOUBLE TUNED configuration are that:
a.) The output voltage is reduced using the Mystery coil - stepped down - as expected if the Mystery coil is acting as a transformer.
b.) The selectivity is improved using the Mystery coil - as expected if the diode’s resistance as seen by the primary (tank) coil is stepped up by the Mystery coil transformer. (Less loading on the tank, with subsequent higher loaded Q and smaller bandwidth.)
c.) The usual crystal set trade-off is operating - more selectivity at the expense of sensitivity. Darn, there’s no free lunch.
I tried 8 different diodes in the double tuned Mystery set to see which gave the most audio output.
diode meas'd Rd (k)
1N34A 20
1N191 36
1N270 55
1N141 141
FO-215 258
BAT85 727
5082-3835 3500
1SS98 4100
The zero crossing resistance's (Rd) for these diodes were measured following a procedure suggested by Mike Tuggle in Ben Tongue’s Article #16:
http://www.bentongue.com/xtalset/16MeaDio/16MeaDio.html
Here’s the experimental setup I used.
If I’ve done something dumb - please point it out so I don’t do it again.
1.) A 26 mVp-p, 995 kHz signal, modulated at 400 Hz, from an Eico-315 signal generator was fed through a 0.1 uF disk capacitor to the ATU. This gave a just barely audible 400 Hz tone in the piezo headset when using the least sensitive diode in the Mystery set.
The headset was made with two Kyocera KBT-44SB-1A piezo ringer elements in series - each shunted by 1 meg. I estimate that the impedance of these elements at 400 Hz is about 6k, about 12k for the headset.
The headset was connected to the 20k tap of a Bogen T725, so I assume the audio impedance presented by the Bogen was about 24 k.
2.) RF bandwidth was measured using an Hitachi V-212 scope and a Yeasu FT-757 transceiver for frequency read-out.
A 10x scope probe was connected across the secondary coil at points 3 & 4 on the Mystery set. After maximizing the RF voltage between points 3 & 4 by tuning the Mystery set’s tank circuit and the ATU to 995 kHz, the frequency was varied to reduce the measured voltage by 3 dB -- which determined the -3 dB bandwidth. The average bandwidth was found by varying the frequency both above and below the resonant frequency.
NOTE THAT I measured the bandwidth of the Mystery sets tank circuit (the primary coil + capacitor) by putting the scope across the SECONDARY coil. I’m not sure if this is correct (??) I did it this way because there was a lot less capacitive loading and detuning by the scope probe. Suggestions appreciated.
3.) The Benny voltage was measured by connecting a DVM across the 500k Benny resistor - which was shunted by 1 uF.
4.) The 400 Hz audio voltage across the piezo element headset was measured by putting the 10x scope probe across the headset and measuring the 400 Hz peak-to-peak voltage.
Before evaluating the different diodes, the first order of business was to choose an appropriate spacing between the Mystery sets’ detector coil and the ATU coil for each diode.
In a double tuned crystal set, as the spacing between the detector coil and the ATU coil is increased - the sets’ sensitivity decreases and the selectivity increases. The behavior of this particular Mystery set is shown below where 3 kinds of “signal” voltages are plotted as the detector coil-ATU coil separation is changed.
The “signal” voltage Vs is the RF 995 kHz peak-to-peak voltage across the sets’ secondary coil - measured between points 3 & 4 with a scope.
The “signal” voltage Vb is the dc voltage measured across the 500k Benny resistor in the audio chain. This is an easy “signal” voltage to measure - requiring only a DVM.
The “signal” voltage Vp-p is the 400 Hz peak-to-peak audio signal voltage measured across the headset with a scope.
There are 4 things to notice in this plot.
1.) Looking at the Vs curve, as the detector-ATU separation is increased from a 1” separation, the “signal” voltage increases - reaching a maximum at around 2 - 2.5” of separation. This is the maximum sensitivity separation. Audio volume is loudest here. As the separation is increased further, the signal level starts to decrease - but the selectivity starts to increase, ie. the RF bandwidth starts to get smaller.
Making the separation less than that required for maximum signal puts the set in the “tight coupling” region. Interesting effects can take place there -- double peaked frequency response, asymmetric frequency response, etc. I stayed out of that area.
2.) The second thing to notice is that all 3 “signal” voltages behave about the same as the detector-ATU separation is changed. They all peak at about a 2 - 2.5” separation and then decrease.
3.) The third thing to notice is that the plot shows the typical crystal set tradeoff -- you can have high signal level (more volume) OR smaller bandwidth -- but you can’t have both.
A good compromise might be to adjust the detector-ATU spacing such that the signal voltage is reduced to ~ 90% of it’s maximum value. That’s only about a 1 dB drop in volume - hardly noticeable. And for the cost of that -1 dB in volume, the bandwidth decreases from 10.7 kHz to 6.2 kHz.
For this particular setup and diode (FO-215), that “90%*Vbmax separation value” is about 4”.
4.) The fourth thing to notice is that the Benny voltage Vb can be used to characterize the sets behavior as the detector-ATU separation is changed. Vb seems to track the other “signal” voltages fairly well - and it’s an easy voltage to measure.
So, to evaluate each diode in the Mystery set, I first measured the Benny voltage as the detector-ATU separation was changed. Then I adjusted the separation distance to the value that made Vb = 90%*Vbmax. I think this procedure puts all the diodes on an equal sensitivity - selectivity footing.
The diodes all had slightly different Benny voltage (Vb) vs detector-ATU separation curves. A representative sample is shown below for 3 of the diodes.
From the plot, you can see that the 1N34A diode requires (in this particular setup) about a 2” separation for Vb to be at the 90%*Vbmax point. The 5082-2835 Schottky diode however requires about a 4.25” separation to be at the same sensitivity-selectivity tradeoff point.
The measured detector-ATU separation required to make Vb = 90%*Vbmax - for all 8 of the diodes in this particular Mystery set - are shown below.
FINALLY, here’s the data that compares the various diodes in the double tuned Mystery set - for a weak signal.
For each diode, the detector-ATU separation was set at the value which made Vb = 90%*Vbmax. Then the peak-to-peak 400 Hz audio voltage across the piezo element (in series) headset was measured with a scope. The results are shown below.
It looks like the maximum headset volume occurs when the diode’s zero crossing resistance, Rd, is about about 200 k.
The audio output using a 1N34A diode is down about -11 dB from the peak response of the 1N141 and FO-215’ diodes. The Schottky diode responses are down about -4 dB from the peak.
On this particular Mystery coil, the 72 turn (Np) primary coil is wound directly onto a ferrite rod. The 37 turn (Ns) secondary coil is wound directly on top of the primary coil. The coils are not bifiliar wound.
The physical turns ratio (Np/Ns) = 1.95 (assuming I counted the turns correctly). But to see what the magnetically “effective” turns ratio really is -- ie., how well the coils are coupled, -- I put a small RF voltage (at 995 kHz) on the secondary coil and measured the voltage across the primary coil with a scope. The result was:
(Np/ Ns) = (Vp/ Vs) = 0.049v/ 0.024v = 2.04 .
This shows that:
a.) the coils are indeed tightly coupled,
b.) and, since the ratio is slightly > 2, I probably miscounted the turns --- old eyes.
Since the impedance relationship between the 2 coils is:
(Zp/ Zs) = (Np/ Ns)^2,
the impedance on the secondary coil (Zs) -- which is the diode + Benny + Bogen + headset -- is seen by the primary coil as:
Zp = Zs*(2.04)^2 = 4.16*Zs.
IF, for example, Zs = 250 k, (a FO-215 diode’s Rd is about 250 k) the primary coil -- and therefore the Mystery sets’ tank circuit (the primary coil + tuning capacitor) -- sees a load of 1040 k.
That’s pretty light loading.
As a side note -- based on what I’ve read -- here’s what I mean by “light loading”. I hope it’s right.
A tank circuit at resonance can be looked at as an equivalent parallel resistor, Rptank, given by:
Rptank = Q*XL = Q*(2*pi*Fr*L),
where Q is the Q of the tank circuit - not just the coil. In this case the resonant frequency Fr = 995 kHz, and the coil’s inductance is L = 368 uH.
Coupling a load to the tank -- a diode, an antenna/ ground, or an ATU, etc. -- is like putting a load resistance in parallel with the tank's Rptank resistance. The resulting resistance is:
1/ Rptot = 1/ Rptank + 1/ Rpload.
The larger the parallel load resistance is, the less the tank’s resistance is reduced.
The larger the parallel load resistance is, the “lighter” the tank is loaded.
For some good info, see Phil Anderson’s writings at the Rap ‘n Tap main site:
http://www.midnightscience.com/article1.html
http://www.midnightscience.com/download%20files/mag-coupling%20select-sensitivity.pdf
or, as pointed out by Broesel, Dick Kleijer’s excellent site:
http://www.crystal-radio.eu/engev.htm
To get an idea of what this particular Mystery set's tank equivalent parallel resistance (Rptank) might be, I measured the tank’s Q as the separation between the detector coil and the ATU’s coil was changed. Two cases were of interest:
a.) The Mystery set’s tank circuit ALONE -- with no load on it other than the ATU; ie. no diode, no Benny, no Bogen, no headset.
b.) The Mystery set’s tank circuit FULLY LOADED by the ATU, the diode, the Benny, the Bogen and the headset -- ie. a working crystal set.
The Q measurements -- as the detector coil-ATU coil separation is changed -- are shown in the graph below. And in the graph below that, the Q measurements are shown converted to an equivalent parallel resistance (Rp = Q*XL).
There are at least 2 things to note about the graphs:
1.) The tank's Q and Rp (the blue curve) are dramatically reduced by coupling-in the diode/ Benny/ Bogen/ headset load to the tank circuit (the red curve).
2.) Pushing the detector coil and the ATU coil closer together also dramatically lowers the Q and Rp. High Q and Rp =’s good selectivity, lower Q and Rp =’s good sensitivity.
Keep in mind that bandwidth is related to Q (and therefore Rp) by: BW = Fr/ Q .
The whole point of making these measurements was to try to get an idea of how the transformer action of the Mystery coil might affect the set's double tuned behavior.
To see the effect, I mathematically coupled an assumed 250 k diode/ Benny/ Bogen/ headset load to the Mystery set's tank circuit. (It’s not rocket science.) .
1.) At a given detector coil-ATU separation, get the tank’s Rptank value from the above graph (the blue curve).
2.) Put a 250 k load resistance in parallel with Rptank -- under 2 different conditions:
a. — assume the 250 k load is seen by the tank as being transformed to 1040 k (ie. 4.16*Rd);
b. — OR, assume no transformer action, so that the tank sees just the 250 k load.
3.) Calculate the loaded tank’s new Rp value as either:
1/Rploaded = 1/Rptank + 1/ (4.16*250k)
or
1/ Rploaded = 1/ Rptank + 1/ (250k).
The results are shown below.
The CALCULATED Rp curve (red open circles) for the transformed (4.16*Rd) load on the tank sure looks a lot like the previous MEASURED Rp curve (red closed circles).
The calculated Rp curve for a non-transformed (Rd) load (black open circles) is considerably lower than what has been measured.
To put the results in the perspective of bandwidth, BW = Fr/ Q and Q = Rp/ XL,
if the detector coil-ATU coil is at 4”:
— transformer coupling of an Rd = 250 k results in a bandwidth of ~ 6.5 kHz,
— non transformer coupling of an Rd = 250 k would result in a bandwidth of ~ 13 kHz.
CONCLUSION
It looks like the transformer action of the Mystery set’s coil when double tuned, contributes a lot to it’s excellent performance.
There’s another effect of the transformer action.
I’ve read that for maximum audio output from weak RF input signals, the audio load should be about equal (+/ -) to the diode Rd -- and the diode Rd should be about equal (+/ -) to the tank's Rp.
To match the measured Rp value of about 1100 k at wide detector-ATU separations, a diode Rd of about 1100 k is needed. This could be done by:
1.) — using a 2:1 transformer coil - like the Mystery set's coil - plus a diode with an Rd ~ 250 k.
The audio load on the diode should then also be ~ 250k. This requires an audio transformer that steps-up the headset impedance to ~ 250 k at the audio frequency of interest.
2.) — OR, as Broesel has mentioned, use a Schottky diode (or diodes in parallel) that has an Rd ~ 1100 k connected directly to the tank coil.
This requires an audio transformer that steps-up the headset impedance to ~ 1100 k.
It just amazes me how interconnected all the parameters are in a “simple” crystal set.
Fun stuff.
73, Dan
Hi again.
So far, for this particular Mystery set, it looks like:
— The isolated tank circuit has an equivalent parallel resistance of about 1100k.
— A diode connected to the secondary of this Mystery set’s coil - when double tuned - has it’s diode resistance “stepped up” by a factor of X4.16.
— A diode that has an Rd ~ 200k (+/ -) gives the most audio output.
— An FO-215 diode, with an Rd ~ 250k, matches the double tuned tank circuit well because it’s resistance is “stepped up” to about 4.16 X 250k = 1040k.
But what about the audio chain after the diode?
The crystal set literature says that for weak signals, the audio impedance load on the diode should about equal the diode’s Rd value, ie., about 250k.
That’s what I looked at in this next series of measurements.
The 400 Hz audio power delivered to a resistor connected across the FO-215 diode - as a function of the resistor value - is shown below.
The maximum audio power is delivered when the resistance across the FO-215 is about 200k. That is - when the load on the diode is approximately equal to the diode’s zero crossing resistance, which for the FO-215 is Rd ~ 250k.
This isn’t news to most folks here, but it’s kind of neat to measure and see the curve for yourself.
The above graph says that the audio chain after the diode needs to present about a 200k impedance to the FO-215 diode to get the maximum audio power.
The Bogen T725 audio transformer accomplishes this by transforming the headset’s impedance up to ~ 200k. But frankly, the actual numbers behind this has always been a little fuzzy to me - so I decided to make some more measurements.
The first thing to settle on was “what is the impedance of the headset?”
I’m using Kyocera piezo ringer elements in a homemade headset. Mallory and Murata make similar elements. And, in a Murata applications manual, they suggest a way to measure the impedance of these piezo elements. See Fig. 13 in:
http://www.murata.com/catalog/p15e6.pdf
My interpretation of their constant current measurement method is shown below:
I checked my setup by measuring various combinations of resistors, capacitors, and inductors -- comparing the measured impedance with the calculated values -- and got excellent agreement.
The measured impedance of various piezo elements, combinations of piezo elements, and headsets is shown in the table below.
The impedance of the headset configuration that I’m using, 2 Kyocera KBT-44SB-1A elements in series (each paralleled by 1M), is measured as 13k.
How does the Bogen T725 transform that 13k up to ~ 200k ?
More measurements.
I used the impedance measurement setup to measure the impedance of the various Bogen T725 windings -- all referenced to the black winding. The results are shown in the table below.
The table will require a bit of explanation.
column (a) -- these are the commonly used “tap” designations for the Bogen T725 wires. See the excellent work of Ramon Vargas Patron at:
http://makearadio.com/tech/files/Bogen%20T-725%20useful%20calculations_updtd1.pdf
column (b) -- these are the corresponding wire pairs for the “taps”. I measured the impedance for each of these wire pairs.
column (c) -- these are the MEASURED impedance for each wire pair, at 400 Hz.
column (d) -- these are the measured impedance RATIOS, referenced to the main primary (white wire) coil impedance, Zp (63.5k).
There are 2 things to remember about these measured impedance RATIOS:
1.) For a particular “tap”, if you know the measured impedance RATIO (Zp/ Zs), then for a given secondary impedance Zs (like Zs = 13k for a headset), you can calculate the transformed impedance Zp that the primary sees - and presents to the diode.
Zp = 13k*(measured impedance ratio).
2.) The turns ratio (Np/ Ns) is related to the measured impedance ratio by:
(Zp/ Zs) = (Np/ Ns)^2 = (measured impedance ratio).
So, if you know the measured impedance ratio, you can calculate what the turns ratio is.
column (e) -- these are the Bogen’s turns ratios, calculated from the measured impedance ratios.
column (f) -- and FINALLY, for a 13k headset on a secondary winding, this column gives the transformed impedance that the primary sees and presents to the diode as a load.
And FINALLY, again, looking at the items in RED ---
from on-air listening tests, the loudest audio was heard when the Bogen T725 was used on the 2.5k “tap”. This corresponds to presenting the diode with a 236k load (~ 18*13k),
ie., about equal to the FO-215’s Rd of about 250k.
The measurements in these double tuned Mystery set posts seem to present a consistent picture of how this particular set operates. I Hope they’re right.
I know there’s nothing new here. Ben Tongue, Dick Kleijer and many others have elegantly laid out the operational principles of crystal sets as currently understood.
But it makes me smile at the “magic” when measurements made in the basement demonstrate the principles. And I sure learned a lot!
Thanks for your patience.
73, Dan
Original Thread: http://theradioboard.com/rb/viewtopic.php?t=2404