Comments for "Micropower FM Broadcast Receiver"

25th February 2014 15:05

fan wrote ...

great site Alan!

i like the detailed explanations of the design process you've given here, discussing the various design considerations and how decisions are made accordingly, plus tips on practical "starting points". very helpful to me as a fan without engineering background but would like to learn, thank you.

some questions if you don't mind. other than an old scope, i don't have any equipment such as a frequency analyser, are there ways of telling whether a BJT is oscilating or not without special equipment? also, when the quenching frequency becomes audible, how does it sound? i'm hoping there are ways to tell with certainty whether oscilation is occuring or not, and whether quenching frequency is too low or too high, with no more than a multimeter and an old scope.

maybe i'm asking for too much, all when you get a chance, thank you Alan!

4th September 2013 20:22

tersio wrote ...

Thanks a lot Yates,

I've truly appreciated your help but i still am in need! please let me have the "SWG" ratings of the wires you used there cos i still haven't got it right yet I'm sure.

Thanks again.

6th August 2013 06:51

Alan Yates wrote...


The 2N7000 is a MOSFET so it would need biasing, also it has a much larger gate capacitance than the J310 JFET, so it would not work at ~100 MHz. I have had a 2N7000 oscillating at about 50 MHz, but that is about the limit.

The FT23-43 is a code number of the ferrite core I used. I think it is a micrometals code system, not sure. The FT means ferrite toroid, the 23 is the OD of the toroid; 23/100 of an inch (about 5.8 mm). The 43 is the material, type-43 ferrite.

The choke itself isn't too critical, if you can't find a small ferrite toroid a few turns on a ferrite bead should be fine, as will a few uH RFC as long as its self-resonant frequency is beyond 120 MHz or so. Even ten or twenty turns on a soda straw should be fine.



29th July 2013 22:17

tersio sparck wrote ...

great job Alan,

but i'm wondering if a 2n7000 TO-92 type FET would equally do the trick at the detecting end, and i don't understand what FT23-43 means?

15th March 2012 07:39

Brian Vaughan wrote ...


Might you have an LT-spice model of a super-regen receiver using a JFET?

I love this website.



10th March 2012 05:27

Brian wrote ...


Thank you for such a fun circuit. I have been building the Albin Ace 27Hz receiver for radio control. I have never tried a JFET equivalent.

In order to get a feel for the super regenerative part of the circuit I am using LT-spice. I am using the stock 2N5484 JFET. However I can not get the circuit to quench. I don't know (yet) if the spice model does not have a pinch off voltage modeled or not. In my experiments, I have put a negative voltage source between the gate and ground to see if I can get quenching at some voltage. I find -1.4 volts to be at the hairy edge but still no quench action (-1.4 oscillation, -1.41 no oscillation).

Might you have some insight as to what is going on?

The next step of course is to pull out the soldering iron and try it out for real.



22nd October 2011 04:53

owen morgan wrote...

Dear Alan, I was so pleased to find your website today. For ages I have been promising myself a super-regen receiver to go in my hat on my dog walks in the morning. I will try to knock up your design with your permission and see how it sounds and how sensitive it is. I did want to make a 648 kc/s receiver for world service on the MW band, but now they have gone QRT. Pity. I was thinking about receiving it on digital and modulating a 648 kc/s carrier with good old AM and putting it back on the air again in my neighbourhood. But I will need it to be stable or eyebrows might be raised. I was quoted £75 for a crystal for that frequency. What a nerve. It might have to be a stable VFO. Many thanks to you. Owen Morgan. UK [X G3UDG]

14th January 2011 15:44

Alan Yates wrote...


Yeah the higher impedance ones should work, not sure how well, try it. Use a DC blocking capacitor, say 10 uH or more at the transistor collector and connect the phones from that to ground.



14th January 2011 03:52

Richard wrote ...

Hi Alan,

I haven't got crystal earphone. I have only high impedance balanced armature headphones. (1 k ohm and 200 ohm).

Will the receiver work with these headphones?

Thanks and Regards,


15th July 2009 13:11

Sergey Bakin wrote ...

Hi Alan,

Thanks very much for your very informative website. I was wondering if you could explain the following point a bit more:

"By tapping down towards the cold-end of the coil the feedback isn't as critical as your usual source-drain capacitor feedback and it tends to be far less difficult to get to work across a broad range of frequencies."

Thanks and Regards,


2nd March 2009 00:11

Alan Yates wrote...


The BC550C is a resonable substitute for the MPSA18. The transistor isn't too critical if you don't mind using a little more current. You could use any generic NPN but lower noise devices are best so the BC550 is quite a good choice. A BC547, 2N2222 or 2N3904 will do fine too. Similarly for the JFET it isn't too critical a BF245 or 2N5484 will be fine.

The simplified two-resistor self-biasing circuit is used for minimalism. If you wanted better quality audio it would probably be a good idea to use a stage with some emitter degeneration to improve its distortion, but the detector doesn't have brilliant WBFM quality anyway, so I didn't feel it worth while. You may find you don't need the audio amp stage at all if your amplified speaker has enough input sensitivity.

It is very hard to precisely calculate the operating point of the simple two-resistor biasing circuit. The device properties make very large differences. It is also hard to control its gain, but it is relatively stable and simple so lots of people use it. A starting point is to pick an emitter current, based on the transistor properties. For the MPSA18 I picked just 50 uA. With a supply voltage of 6 volts I wanted the collector to sit near mid-rail, 3 volts (actually a little higher because of the transistor saturation point, but for a small-signal audio amp like this the exact operating point isn't too critical). So, for 50 uA of collector current (approximately equal to the emitter current because of the large beta), the collector resistor needs to be about 60 k (this is also the roughly the output impedance, ideally we need something closer to 12 k because that is the approximate impedance of the xtal earpiece at 1 kHz, but I wanted less current drain, you could pick a current of about 250 uA to get closer to an impedance match and more power output capability). We know the base will be about 600 mV above the emitter (which is grounded) so the voltage across the bias resistor will be near 3 - 0.65 = 2.35 volts. The particular MPSA18 I picked had a measured DC beta of 785. So the base current needs to be 50 / 785 = 64 nA (tiny!). The bias/feedback resistor is therefore about 37 M. In practice I needed to reduce that value quite a lot, down to 12M - 8M2 depending on the exact transistor, but the estimate calculations are a good starting point. I also checked the input impedance (about 380 k ignoring feedback which tends to reduce it) and small signal gain ~ 112. Simulation with spice is probably a good idea, but the back of the envelope "design" is generally good enough if you then test the practical circuit and tweak its biasing.

If you were to use a "generic" NPN I'd use 1 mA as a good starting point. A 2N3904 from the junkbox has a measured beta of 173 (multimeter says so anyway). At 1 mA current 3k3 will drop near mid rail at the collector. Base current needs to be about 57 uA, so the feedback resistor of 470k is a good starting point. The exact bias point is very sensitive to these gross approximations, so assemble the circuit and measure the collector voltage. You can can then either vary the collector resistor to get it closer to mid rail, or vary the feedback resistor to get the emitter current closer to what you wanted.

For a 9 volt supply (instead of 6 or 3) in practice 4k7/1M is a good combination for ~1 mA emitter current with "average" transistors. The detector part of the circuit will operate right down to below 3 volts, so two AA cells is a practical power supply. A 9 volt is popular for quick projects like this, so that is an option too. For 3 volts something like 1k5 Ohms and 220k should be a starting point for a generic NPN.

Barbie's Musicbox appears to have been designed with more current an a less centred bias point in mind. The circuit's lower collector voltage drives the output emitter follower directly, assuming a 32 ohm DC resistance the quiescent current through the headphones is probably about 10 mA - this was likely a design factor. I would have gone with 390 k for the feedback in the first DC coupled stage which would stand about 4 mA if the collector loads had to be 1k (which isn't a bad idea for audio circuits, I try to keep them < 10k). Burkhard Kainka's circuit should actually work pretty well for driving headphones or a small Mylar speaker from with my detector and might actually have too much gain, using a 1k pot in the pre-amp stage collector would let you control the volume. I do worry about the bias point being fairly low, so it might clip early if over-driven, but otherwise I'd use it as a starting point.

Core wise the material and size matters more than the precise part number, many brands will work. That is why I generally specify the design inductance instead of just x-turns on a y-core. Any small type-43 ferrite bead or toroid will work for the RFC in the source. I often use junkbox beads with a few turns on them as RFCs in the source/emitter of VHF super-regen detectors. Its precise value is not very critical, it could even be air-wound.



1st March 2009 08:19

John wrote ...

I forgot to mention that I can´t get those MPSA18 transistors here in germany !

And to be honest, I am too dumb to calculate new resistors :D

I´ve mainly used MOSFET´s in all of my projects (high voltage in general , solid state teslacoils from 100khz to 4mhz, SMPS, boostconverters , LED drivers , constant current sources for laser diodes)

So I was looking around burkhard kainkas site (he has lots of projects that are related with amplifying AF ) and found this little circuit

So maybe a combination of a 27k resistor and a 1k resistor should do the trick ?

Thank god that my suplier carries the same torodial cores you use .It can be sometimes quite difficult to get the parts people from foreign countries use, because in germany there are few suppliers.

Best regards,


1st March 2009 06:10

Alan Yates wrote...


Thanks mate, I try to document all the "little details" of my projects so people can replicate them if they wish to... I must admit at times I leave out stuff like pin numbers on common ICs, often because I am too lazy to open the datasheet or trace out the finished device. Eventually someone asks and I go update the article with what's missing.

Funny you should mention class-E amps. I've spent most of today in LTSpice simulating them for 30 metre QRSS work. Efficiency is the main goal rather than lots of output power.

Yes an amplified speaker like PC speakers or that one you could once get from Radio Shack will work. Just replace X1 with a 100 nF - 1 uF or so capacitor to block the DC. I can plug my device straight into my amplified speakers despite not having a cap in there as they are AC coupled, but the capacitor can go in even for use with the xtal earpiece so if I built it again I'd probably add one for flexibility.



1st March 2009 05:01

Visitor wrote ...

Hi Alan Yates, your site is really good, I´ve succesfully rebuilt the 500mW pushpull RF amplifier.

I am addicted to class-e amplifiers, (I´ve built a nice 30W AM class-e 4MHZ RF generator with some transistors and irf630 driving a tesla coil which then produces a very clear sound) so many thanks for scripting your design tool.

But now I want somethin to actually recieve rather than transmit.

This Micropower should do the trick, ais it is very easy to built and rather compact.

My major problem is to find a crystal earpiece.

Would active pc speaker do the job (or a simple lm386 based amplifier?)


Many thanks for writing down your inductor specs (core,windings) !

This makes rebuilding much easier.



17th December 2008 00:11

Alan Yates wrote...


The 1/(2.Pi.R.C) formula is for the cut-off frequency of a simple 1st order RC filter. The RC time constant is how long it takes to charge the capacitor through R to (1 - 1/e) times the supply voltage (about 63.2%), or discharge it down to 36.8% of its initial voltage.

In most RC timing circuits (and this particular circuit) the quench frequency is not determined by the cut-off frequency. There is generally a fixed reference voltage at which point the charging (or discharging) will be stopped and the cycle repeated. How long it takes to reach that point and how quickly it is re-estabilished determines the frequency. Generally the full solution involves two differential equations (one for the charge, the other for the discharge part of the cycle) with boundary conditions determined by other circuit parameters which are often constant voltage/resistance or constant current approximations. A completely analytical solution, even for simple models can be truly nasty!

I haven't modelled this circuit extensively so as to give a nice formula for the quench frequency as a function of the emitter RC values. It would not be a simple thing to calculate, and would be an approximation at best as the quench frequency changes with the transistor operating point (emitter current changes, temperature, noise/signal forcing of the oscillator start-up, etc).



16th December 2008 15:08

Mike wrote ...

You mentioned that the optimal quench frequency should not be below 15Khz. According to your design, your quench frequency is defined by R1 and C1. In your circuit, R1 is 10K and C1 is 6.8nF. According to my calculations, your quench frequency is 2.3Khz. Shouldn't C1 be more like 1nF or am I doing my calculations wrong?

The formula I'm using is 1/(2 * pi * C1 * R1)

15th October 2008 08:03

Alan Yates wrote...


To move the receiver up to 151 MHz you'd just reduce the value of C5 (and maybe C6, C3 and L1 too). Try dropping C5 to 5p6, if that doesn't let you tune up that far then remove a turn from L1 and reduce C3 as well.

As smooth easy tuning doesn't matter too much for fixed tuning you might like to remove the C5 C6 network and just use the trimmer C7 directly from the drain to ground.

Is your desired signal wide-band FM like a broadcast station? Narrow-band FM won't work with this kind of receiver, its deviation is too narrow to be slope detected but the fairly wide response curve. AM will work fine, even better than WBFM, this radio is great for monitoring the local control tower in the airband.



15th October 2008 02:00

Steve wrote ...

I would like to use your design however I would like to change the receive freq to MHZ. How can I do this? Basically I just want to listen to one frequency. No need to tune it to others. I can tune it to the required freq and leave it.

I know how to build equipment but I do not know a thing about designing or components and what they do.



15th June 2008 17:19

Alan Yates wrote...

Yes. The emitter of a BJT is the one with arrow which also tells you the type (NPN or PNP) of the device.

15th June 2008 03:17

Zohaib wrote ...

Is the orientation of the legs of BC5xx series transistors is like this:

The Collector does to R4, the Base at C2 C8 & R3 Connection?.

If not, then please tell me about this.


5th June 2008 16:17

Alan Yates wrote...

Look at the datasheet. For the J310, looking at the flat-side towards you, legs pointing down, the legs are right to left, Drain, Source, Gate. The Drain does to L1, the Source L2.

Q2 can be replaced by almost any generic NPN, but you'll probably have to reduce R3 to 1M and R4 to 4k7. I'd recommend a BC549C or BC550 for low noise, but it isn't critical. BC5xx series are typically pinned out as Collector, Base, Emitter in the same orientation as the J310. 2N3904 will be Emitter, Base, Collector.

4th June 2008 03:28

ZOHAIB wrote ...

How can we determine which leg of transistors will connect to other components, as they have emitter, base and collector legs?.

Also Q2 is not available.

2nd June 2008 22:32

Alan Yates wrote...

A "greencap" is a polyester capacitor, so called because common types are often encapsulated in a green epoxy dip. Other colours are common too, including red, orange and blue, however the name has kinda stuck over the years even for non-green ones, at least here in Australia.

They are constructed from Tin or Aluminium foil and polyester film or directly metallised polyester film wrapped tightly to minimise its volume. They are quite reasonable for this application, but it is not especially critical. A ceramic device would also work or just about any capacitor technology commonly used for devices in that magnitude range.

The monolithic is a special kind of ceramic capacitor, they are multi-layer devices, typically using Barium Titanate dielectric sandwiched between metallisation layers composed of Gold or Platinum-group metals to remain conductive after the oxidative sintering firing required to construct the capacitor. The metallisation is often printed on as an ink composed of the finely divided metal in a carrier solvent. Their construction is actually quite extraordinary and it amazes me how relatively cheap they are for the exotic nature of their manufacture.

The important thing is they can be very small for the capacitance they contain, in this case 100 nF, for which a foil or film polyester would be quite sizeable, much larger than the few cubic millimetres of the monolithic device. They have pretty reasonable properties as well because of their small size and relatively non-inductive construction. None of this, except perhaps their small size, is all that important for this coupling application. Larger values would probably be OK, as would using an electrolytic or a film cap.

So in summary, it isn't that critical really. Use whatever you have, but stable caps in the RF part of the circuit will help minimise drift of the circuit. The AF parts can be almost anything.

This goes for the other components in the circuit too.

2nd June 2008 03:16

zohaib wrote ...

"C1 is a greencap, C8 is a monolithic"

what does it mean? not ceramic?.

1st June 2008 12:22

Alan Yates wrote...

The circuit has such high gain that it doesn't need much signal to produce a reasonable signal to noise ratio. That said, I live in a strong signal area, and an antenna will be needed in fringe areas. You can feed RF into the circuit from an antenna through a small capacitor, say 1p2, tapped down on L1 towards the +ve rail. Alternatively you can couple a signal into the source at L2 (or at the drain too, but stability may become worse with more significant hand capacitance).

Ceramic NP0 capacitors are used in the RF sections, including the decoupling cap for the detector. All caps except C8 and C1 are ceramic. C1 is a greencap, C8 is a monolithic.

8M2 is 8.2 Mega Ohms, similarly 6n8 is 6.8 nano Farads.

1st June 2008 04:04

zohaib wrote ...

how can it receives without antenna?

what type of capacitors used?

what is mean by R3=8M2 & C1=6n8

please tell me about all these things


24th April 2008 00:26

Alan Yates wrote...

JFETs are widely available in most parts of the world, I'll admit the J310 is obsolete, but it could be replaced by a MFP102 or BF245, or almost any modern small signal JFET. I can send you J310s if you are really stuck.

You can use a BJT easily enough, I'd use a BF199, MPSH10 or similar, but BC547 or 2N3094 should work too. Arrange a voltage divider to bias the base such that the collector current is about 100-200 uA. It need only sink a few 10s of microamps - no need to be very stiff. The base should be bypassed at RF, 10 nF to ground should do the trick. Some other values might need to be tweaked to accommodate the change.

I picked ~100 nH so the resonating capacitance for near 100 MHz would be at least 20 pF. (20 pF comes from an estimate of the stray capacitances and something larger than C3, which has some effect even though it is tapped down on the inductor) If you increase the inductance to 200 nH the capacitance halves and circuit's stray capacitances become more significant (and body capacitance, etc). 120 nH is only about +j75 Ohms @ 100MHz, normally I'd start higher often with 200 Ohms, but at VHF it is generally good to keep impedances low. 200 Ohms equates to about 320 nH, matching capacitance is just 7.5 pF, the circuit stray capacitance is estimated at 4.5 pF, more than half! It would likely work, but stability would be more of an issue. Going the other way, I wouldn't go much below 50 nH (about 50 pF to resonate), 50 nH is about the same inductance as 50 mm of wire the same diameter as the component legs, it doesn't take much for the strays to start becoming half of the tank inductance.

The optimum value is hard to calculate. The inductor Q is a factor, as is the capacitance and the impedances in the rest of the circuit. I've never invested the time to model it, so I am not completely sure how to optimise the value for a super-regenerative detector. However in practical terms it can't be too large (self resonance and stray capacitance limited; it *must* be less than about 450 nH or you won't make it to 108 MHz with no additional capacitance at all!) and it can't be too small (insufficient reactive magnitude compared to other circuit elements, and again limited by minimum practical circuit wiring inductance, etc).

Remember also that if you don't add an antenna to this circuit the coil is the only thing picking up RF from space. (Actually a surprising amount of signal appears to come into the circuit from the headphone lead! How it induces a significant signal in the tank isn't completely clear, possibly capacitively, as the headphone lead is near a quarter-wave at 3 metres.) All that said, the value does not appear to be very critical in my experience. Within reasonable limits you can compensate by changing the resonating capacitance, but ideally you want most of the L & C to be something you control - not strays.

23rd April 2008 00:48

C. MOHAN wrote ...

How to arrive at L1 inductance value. instead of 120nH why not 200nH. how to calculate optimum value for inductance.

23rd April 2008 00:46

C. MOHAN wrote ...

Dear Sir,

Now a days JFET is not available if the circuit is given using transistor that will be good. is it possible to replace JFET in your circuit with transistor

Thanking you,


Leave a comment on this article.