I promised this schematic of a two terminal oscillator to some of you last week.
The original of this article was written a few years and I sent it to half a dozen RnT experimenters. The Q multiplier stayed on the bench in front of me and continued to evolve. I have updated this piece with most of the changes. The current choke, which you see in the photos, is homebrew and un-obtainium commercially. You can sub a 27 mh, see text. The view of the bottom of the dip socket is not quite the same as the diagrams. The 1k resistor is not used with the 113 mh choke and the pins which hold it are not not connected.
Here are some photos of the current edition. The originals of these photos are at such high resolution that I can zoom in and see everything at 10x scale. The socket in the photos is a little scruffy since it has been modded a few times. Too bad I didn't clean the flux and dirt off it.
Some of the qualitative observations in the text may not be quite accurate. I can't remember what has changed and what hasn't. I'm happy to answer enquiries thru this thread so everyone can benefit.
SIMPLE CRYSTAL SET Q-MULTIPLIER
The essence of this circuit came from a correction to an article by Bodizar Pasaric, 9A2HL, RIP, which was published in the British QRP Society Club journal call SPRAT about 2003. See the SPRAT index on the web. Bodizer was very inventive and his projects were always clever and entertaining. Thanks to some RnT regulars for their testing and feedback.
The odd coupling of a FET and BJT is called a Lambda device and it exhibits negative differential resistance (NDR) similar to that of a tunnel diode. In this setup, the NDR is hard to measure. A sister circuit using an N-ch and P-ch fet back to back gives a very nice NDR curve but is not controllable as an oscillator / regen setup. I have also experimented with true tunnel diodes. Their high current drain/low impedance makes them much harder to use - impractical for this sort of application.
The construction technique shown here allows you to substitute components easily so you can sweeten it up for your application.
I have tried all the devices mentioned here. They all work, some better than others. The FET has more effect on the circuit than the BJT. Different types as well as different samples within a type will give different results in terms of the point of regen and how smooth it is, quality and strength of beatnote, current drain, capacitance, tolerance for different xtal sets and the diode detector characteristics etc. The ultimate sensitivity is about the same for all.
I like the J310 FET which has a low capacity, a strong beatnote and good control. It will not work well with germanium diodes ( see below). The 2N4416 is the best (of my lot) for that. MPF102 and 2N3819 are in the middle. The J305 and U310 work at low current levels and are best with low Is diodes and high Q coils. If you have to buy FETs, get a 2N4416 and J310. Both are common and one or the other will work well depending on the crystal set.
The PNP BJT is not as critical but there is some dependence on the FET used. The 2N3906 is common, works fine with any FET and has low capacity. Or use a 2N2907a, MJE2907, 2N4403, or 2N5087 if you have one. Bodizar used a BF245 FET and BF509 BJT.
In this circuit, the power flows from the battery through a series resistance and diode to the Lambda device. When the Q-multiplier (QM) is connected to a xtal set (XS) coil, the circuit back to the battery is completed. There must be continuity through the coil for it to work. The QM adds some parallel resistance across the XS coil which reduces circuit Q about 2% when little or no current flows ( off) and it also adds a few pf of parallel capacity which reduces the maximum tunable frequency. Otherwise, the crystal set acts normally. As the regen is increased, the current and voltage at diagram point "X" will increase. The apparent gain of the XS increases and stations normally not heard will become listening volume. There is no added noise. If a weak station is tuned, weak oscillation will begin at some voltage/current point which depends on the QM and XS circuit components but will be between 50 to 300 ua and .5 to .8 volts. Rotate the control more and the signal and oscillation both increase. Eventually, a low frequency oscillation (growl) will begin. I think this is a relaxation oscillation caused by a drop in the XS diode resistance. It is the end of the useful regen range. The most sensitive point is just before the growl begins. You have to zero-beat to hear the modulation. Further advance of the control lowers the growl frequency until it is just clicks.
There is no hysterisis or backlash in the control. However, when the control is advanced so that prolonged ringing occurs on spikes, the set may then slowly, over 10-15 seconds, begin to oscillate continuously. Back the control off a smidgeon and it will slowly stop oscillating.
There is a frequency shift as the control is rotated. It can be small with the right components, almost nothing in the control range of interest. The switched capacitor in the control box will allow you to turn the QM on and off w/o a frequency shift. This is important for Dxing. You want to be able to turn the QM on for spotting and the turn it off to see if the station is audible barefoot. The exact cap value depends on the other components. 100 pf works OK for me but I am not fussy. You may want to use a trimmer so the exact compensation can be dialed in.
Signals whose tank RF voltage is less than the self oscillation voltage will have a nice, clean beatnote. These are primairly stations which cannot be heard or are weakly heard in the phones without the QM. Stronger stations lose the beatnote around zerobeat first and really strong ones have no beatnote at all. This is a function of the very low operating voltage.
The audio level increases dramatically for small signals. Signals which cannot be heard with 10,000x audio amplification and no QM are readable with the QM and no audio amplification. The audio level of moderate signals is improved by maybe 20X and strong signals are almost unchanged.
The circuit works very well for weak signals where it is needed most !!
It might work OK for weak CW and SSB signals but not for anything stronger.
XTAL SET DIODES
A high impedance detector gives the best results since the effective set Q is raised by the QM and that is hard work with a low Z diode. A 1N34, 1N60 etc. is barely usable if tapped down the coil and a transformer is used. Even then, only the 2N4416, which draws around 290 ua, works reasonably well. The FO215 works quite well on the top of the coil with the 2N4416, marginally with the MPF102 and the rest exhibit some gain but no clean oscillation before the growl sets in. Use a schottky such as the -2835 or 1N5711 for best results. A HobbyDyne setup for high impedance is also good. If the QM is used for spotting weak signals then the xtal set diode must be the best for barefoot weak signal performance and its effect on the QM performance will have to be tolerated.
The small coupling caps also affect the performance with different diodes, different frequencies etc. I normally use 5pf and 1 pf with the 113 mh choke - see below - but if you use a 27 mh choke or something else, or you have trouble getting good regen control, measuring the SRF of chokes etc then increase the caps to 5 and 5 or even 10 and 5 pf. If you have to order small caps, get a selection. The only significant drawback to large caps is a greater frequency shift.
The regen control circuit just varies the voltage applied to the probe. This works quite well even although the series resistance approaches the load resistance. In theory, this gives poor regulation but it seems to be better than a "stiff" voltage divider supply in accomodating a wide range of conditions. These pot values are a good start but 50k and 1k would work fine too.
The earlier versions had a fixed resistor instead of the coarse pot. The coarse pot is not changed during normal operation once the best setting is found. It is convenient for trying different components, frequencies, sets, antennas etc., and allows you to oscillate random coils to determine their SRF. (I don't remember why the FO215 is there but it must do something good!! Probably something to do with the diode voltage drop.)
A good way to connect a frequency counter for SRF measurements is to ground and the collector of the 2N3906. A small series cap might be needed. I have only tried my Wavetek DMM. Other counters may need some other arrangement. The output is low so you can have problems with local fields, even 60 hz, which can confuse the counter and give unstable or false readings.
The choke in the photos is one of Dave's 27 mh that I modified by gluing a FT50- 61 toroid core to it- same number of turns etc. but 113 mh with the added ferrite. It works better for measuring the low SRF of chokes because the regen controls can be fully advanced with no danger of spurious oscillation. If you make and use a 100 mh choke, just jumper the 1 k. resistor pins.
The choke value is not critical for Q multiplier operation as long as it does not resonate in the band. Pretty much anything above 10 mh will work The 1 k resistor lowers its Q so it does not oscillate at its own SRF. Using Dave's 27 mh with 1 k in series, I could measure the SRF of another 27 mh choke reliably by setting the regen control so that it was just oscillating.
Keep the hot lead to the alligator short. Putting the active components right at the hot coil or cap terminal and "remoting " the regen control, divider resistors, battery and switch cuts down stray capacity. As long as the "cold" lead is grounded there is no hand capacity at the regen control. This topology is the only way to go for Q multiplier operation. For SRF testing it is not critical.
I have only played with it on single tuned sets. A couple of other experimenters have used it with double tuned with good results.
You can plug the RF components into a machined style of DIP socket for easy swapping. It makes a compact package which probably cannot be bettered. Wiring the jumpers on a socket is a bit of a trial but not bad. You have to be quick with the iron or the plastic socket will melt. This is the layout. It is not the very best but a good compromise for the job.
There is no RF in the control box when the set is grounded so layout is not critical. The battery holder seen here is homebrew. You will probably be wise to use a slightly bigger box or a smaller battery. The wire is from a set of Walkman earbuds. It is good for the probe leads as it is two conductors of litz in a small, flexible and strong cable. The major brands all seem to use the same wire.
Experiment at will. I haven't lost a BJT or FET even though I have reversed everything that can be reversed at one time or another. This has gone through a few versions but there is always room for improvement.
Regards Don Peters