Variation on Junk Box Bipolar Regenerative Receiver

This version removes the direct coupling of Q2 in the previous version and increases the emitter resistor as well, finally the regeneration control is moved to the emitter circuit. The result is little better performance with less loading of the tank and less detuning caused  by  regeneration adjustment. Probably other optimizations possible also. I tried to make this thing not work…and it was difficult requiring extreme deviation from shown component values. It should be very easy to get this thing up and running with all sorts of tank circuits. It works well with a broad range of supply voltages also the values for R7, R8, R9, and R10 have a wide range of functional values. So one probably wants to start with my values and then tweak for best performance, once the circuit is up and running.

Schematic Diagram(modified TDA7052at circuit 07/01/2015)

2222A regen_rev2

An unusual Regenerative receiver circuit using only bipolar transistors

This design uses junk box 2N2222A or 2n3904 transistors. Its easy to build and offers excellent performance. Here is a preliminary schematic. It works well but I am still tweaking it. Look for revisions to follow.

A few quick notes:

It only uses bipolar transistors, and the antenna signal can be injected such that it is  isolated from the tank or directly to the tank through a small cap. The circuit can use a series resonant Clapp type topology or parallel tank. The clapp version reduces the effect of parasitics and provides very stable tuning. It also works really well with varactor tuning as the voltage  swing across the varactor is low. The series tank requires larger inductors for a given frequency then a traditional parallel tank. I am using 20uH(two crappy tiny molded inductors) and a 300pf variable to get in the 5-8Mhz range. You don’t need any tapped coils or tickler…etc – just about any inductor should work.

The parallel tank version  gives more tuning range for a given tank circuit. It may be a better topology..its hard to tell. You get more tuning range for small tuning caps and the the inductor will be smaller. You also eliminate the collector resistor which surely adds phase noise to the clapp version, when oscillating.

Depending on the tank coil you use and/or if you use a different supply voltage, you may need or want to adjust some of the resistor values. These are noted on the schematic. On the bright side, I have use wide range of values for most of these components and the circuit has still worked. It’s all about maximizing gain while still maintaining smooth regeneration.

The darlington detector is a bipolar version of a “plate detector” where the the darlington is biased just at cusp of being turned on. It acts like a halfwave detector but with high Z and significant gain. You could use a FET but I thought it would be fun to only use junk-box type transistors. I am using 100k and 47k biasing resistors for the detector. The ratio is what matters so 1meg and 470k will work fine also.

In the schematic, I show the detector connected to L1 and C4 which make up the tank circuit for tuning. I also show a Q multiplier circuit separately which connects via the “QOut” signal to the tank circuit also. Its just another way to conceptualize a regenerative radio.

The transition from oscillation  is very smooth. Sensitivity is excellent. The circuit is simple.

Schematic Diagram(modified TDA7052at circuit 07/01/2015)

2222A regen

Video demo showing smooth regeneration transition on a scope


Another Quick Demo of my “No Tuned Circuit” Direct Conversion Radio

This is a video of me listening to radio Cairo in my basement while working on other projects. The slight drift from zero beat can be heard at the end. This is from the transmitter drifting… not the receiver. I know this because I have looked with a frequency counter and the receiver stability is within a fraction of a Hz. This is really the only drawback of this radio- that you have to zero beat AM. Still it works pretty well as is.

My next step is to go back to quadrature output again and build a simple DSP demodulator out of popcorn 12bit ADC’s and a PIC micro. This way I can listen to AM without zero beating the carrier. This is done by generating two outputs 90 deg out phase, squaring them, summing them and then taking the square root. This gives you the magnitude of the envelope of the carrier or the AM modulation.

Simplified switching mixer DC receiver uses no tuned circuits

This receiver is a simplified version of my quadrature sampling receiver. It  is simplified because it does not require the phasing filter section and also does not divide the LO by 4. This simplifies the local oscillator requirements significantly. The receiver uses a 2 pole low pass filter for selectivity and a TDA7052 audio amplifier. The CS2000 is a SPI controlled clock generator and is used for frequency generation and tuning, but any VFO, VXO  or any other stable frequency source will work. In this case, one of the extra inverter stages could be used as a linear amplifier to boost up the oscillator output if required(need a good squarewave). If this is done, the micro-controller can be eliminated. The input RF amplifier can also be eliminated if you are using an approximately 50 ohm resonant antenna. In this case the antenna would connect through C13 to pin 9 of IC5. If you do this – pin 9 must also be biased to 1/2 supply. This can be done with a couple of  4.7k  resistors connected in series from V+ to Gnd. The center  junction point will go to pin 9.

This receiver can receive AM but must be at zero beat(exactly tuned). The CS2000 has a resolution of 1Hz or so and is very stable – so the receiver work pretty well for AM. Unlike the phasing receiver this circuit cannot eliminate one of the side bands, but to be honest its not worth effort unless you want contest grade ham receiver performance.

In the video you can see I break out the power switch, audio gain, tuning encoder and display to a daughter board.

I designed the display to have very low spurious noise. The info on this is here:

More info about how this circuit works can be found in this post, which is the quadrature version of this receiver.


DC Receiver

Demo Video:

zip file of code: Filedropper is full of dumb ads but look for the “Download This File” button in middle of screen

Quadrature Sampling Detector Phasing Receiver – 2MHz to 15Mhz range – NO TUNED CIRCUITS!

Well, I finally finished my phasing receiver!

I am posting the schematic and design info.  I have a surface mount layout which I will post also. There is  source code  for a micro-controller required. I plan on providing some sample code  and I am happy to provide coding assistance. Since there are so many variables with regard to possible features, tuning method, micro-controller used, etc, I probably will not post my source code in its entirety.

I am an amateur radio operator W4KIA and have always loved shortwave radio. I designed this radio  to experiment with the basic  platform of the Tayloe detector . It is intended  mainly for AM shortwave listening – so there is a lack of sharp filters to optimize CW selectivity, etc. The radio is designed to used a random wire antenna for casual listening. I say this because one would want to change the circuit to optimize for amateur radio applications. This design should considered a reference or starting point  and not a perfected turnkey project ready to build.

The heart of the circuit is a quadrature sampling detector developed by Dan Tayloe. At first glance, this circuit  appears to be to be a switching mixer used as a product detector. This is not quite accurate. It really is a sampling detector which generates a difference beat(when LO is slightly out of tune with the incoming signal), but not a sum product. The very basic idea is to switch on and off  a RF input into an audio range low pass RC filter and detect (integrate) the difference beat frequency or modulation envelope, when exactly at zero beat. A simple diode detector is analagous to some degree – the difference is; it self switches so it is always perfectly synchronized to zero beat. Also, a diode only switches on for a portion of  a half cycle, hence there are losses.

The fact this is a detector(not a mixer) and that it is quadrature sampling leads to two very important  benefits. One, I and Q signals  are easily obtained. This allows sideband rejection via an analog all pass filter or  DSP processing. Second, it provides near unity gain (since you have no sum product, you do not lose half of your energy in  detection, aswell, you detect at 90% of peak not RMS). There is much information on the web describing the Tayloe detector so I will not go into all the  detail here but I will describe some of the advantages/disadvantages of the detector.


The sampling detector has large dynamic range – primarily a function of operating voltage of the analog switches.

Unity gain (well almost but much better than a diode mixer/ detector – which will be worse than – 6dB). Quadrature sampling allow for near peak detection  instead of RMS and there is not a sum product which otherwise is thrown away and wasted

Requires no tuned circuits and is agnostic to RF below the maximum switching rate the analog switch input can handle ( see disadvantages below).

Its simple and not critical to adjust

Built in front end selectivity – If your input is at 10Mhz and filter knee is at 1KHz(need to double this for the calculation) – you have front end selectivity Q of 10Mhz/2KHz ……which is 5000!


It needs to be driven by a Local Oscillator  4X the desired reception frequency. So you need to generate 40MHz  and divide into 4 (90 deg phased) clock signals to listen to WWV at 10MHz…OUCH!

It needs to be driven by a square wave – which introduces harmonic detecting and spurious signals. Front end selectivity will mitigate this.

The gain, bandwidth and noise figure of the detector are very sensitive to antenna impedance matching. This requires a well-defined RF input impedance for best performance.

Although much, much, better than beef stew product detectors, it can still experience AM leakage (I have not observed this with my radio but I don’t use a huge resonant antenna either)

Makes a beat note, so AM detection requires careful tuning with a really stable oscillator( This is what I do) or requires tuning above 20KHz,  band pass filtering, and summing to full wave rectified output (absolute value circuit) to create AM detection. Alternatively, you can A/D convert and use DSP signal processing.

Basic of the design:

IC1 is a PIC microcontroller used to drive my LCD display and generate the  SPI config data for the clock generator IC2. Tuning is generated from a 2bit Gray code mechanical encoder and a push button to select tuning steps(10Hz, 50Hz, 1Khz, 10KHz, and 100KHz). I am using a cheapo LCD for display but it creates noise; so I am going to change over to a charlieplexed 7 segment display or something like that, eventually. Ic2 is a cirrus logic FRAC N synthesiser – CS2000 that I have been able to get to generate output frequencies from 2 Mhz up to 70 Mhz . So dividng this by four, my tuning range is 500Khz to about 17 Mhz. The resolution once divided by four is about 10Hz, using a 22 MHz crystal. One can use a crystal reference as low as 8 MHz but the device generates some spurious birdies from the FRAC N modulation and  22Mhz eliminated the vast majority of these. I think there are two or three spurious signals found across my tuning range, none of which fall in brodcast or Ham bands. The part is very simple to use, has a simple command structure and draws only 10 mA, as opposed to a SI570 which draws nearly 100mA and is more expensive. I looked at making my own PLL or using a DDS  chip but neither would give me the desired tuning range and the DDS requires extreme filtering to reduce jitter(even though you are using a square wave). The CS2000 is a great choice for this kind of radio, one just needs to be thoughtful about reference frequency of the crystal.The clk generator drives IC4 which is a 74HC163 counter which counts up to four providing the enable signals to select which analog switch is enabled in IC 5 the FT3252. You can also create a quadrature clk with a couple flip-flops and use other switches such as the 4066. I chose these for  two simple reasons, I had them and these IC’s lend themselves to a much easier PCB layout.

IC8 is a 100Mhz bandwidth op amp used to set the input  impedance to the detector IC5 at 50 ohms, this along with the resistance across the switches provides the “R” of the RC filter. It also provides some input gain and becomes the dominant driver of noise figure for the receiver. The front end has no pre-selection filtering and there are no tuned circuits anywhere in the design. There is just resistive termination and capacitive coupling. I used  a LMH6642 because I had some in a junk box and it draws very little current. It works pretty well up to the 20 meter band and is stable. I can use any old piece of wire  as an antenna and still provide stable termination for the detector IC5. This particular op amp does not have the best noise figure but works okay. A better choice would be the LT1818 but it draws more current. It has 400Mhz bandwidth and a low noise figure. One can obviously use a discrete JFET, etc, but I figured I would give this a try and it works fine.

After the detector, there is a very low noise dual op amp to sum the 0/180 deg signals and the 90/270 deg signals to generate I and Q (Inphase, Quadrature) having I and Q is the magic sauce of communication because with I and Q you can determine both phase and magnitude of a sine wave. The signals then go into a three stage all pass filter. This filter has critical 1% tolerance values. For best results you should really grade the components a with a decent multi-meter as close as you can get them. The absolute value of the components is not as important as the relative values. For example,  if you chose all the capacitors to be .01uF (as I did) they all can be .011 or all be .095. They just all have to be the same. To design my filter, I used the JTEK filter calculator . I chose one capacitor value and a bandwidth range such that my resistor values could be easily created by paralleling and or series connecting standard 1% values I had. I think my filter bandwidth was 200Hz to 2500Hz or so. I do not specify the majority of the filter components in the schematic, leaving it to the designer chose a bandwidth and component values that are convenient for the builder. Just download the JTEK designer and you can easily create your own parameters

There is  low pass filtering from the detector, also some on the summing amps and finally high pass roll off on the final audio amplifier. In total, this does an okay job of restricting the bandwidth for the phasing filter but it is not optimal by any stretch.  Normally, one would have some higher order filtering to restrict the range of frequencies going through the phasing filter. If you do not restrict the range for the phase shift filter, the sideband rejection rapidly degrades outside the filter’s design bandwidth. My phasing filter works well enough, providing 30dB (measured)  lower sideband rejection. I matched the .01uF capacitors, but did not bother matching the resistors. If you want to change the sideband that is rejected – you have to flip I and Q or take the difference at the output instead of the sum as I have. The audio output goes to IC12, a TPA301 amplifier chip. There is nothing special about this I just had one available, lots choices here.

For AM listening, one could just eliminate the phasing filter and just have some sharper low pass/high pass filtering instead.  You could use 2 switches clocked at 1X frequency with one switch getting an inverted clock. You would then use one differential amp to combine the push-pull output.This would simplify the design even further. The main penalty would be for CW reception.

The receiver works great for SSB and CW, where the sideband rejection filter really cleans up the clutter. For AM the, CS 2000 is  stable and has enough resolution to zero beat AM station and once tuned is pleasant to listen to.

Note: If you are going to do amateur radio work skip the RF front end and use a well matched (50 ohm) antenna. If you you just want make a nice all band receiver – the RF amp is worth adding. You have to bias those switches to half supply. Currently the RF amp provides this- if you remove it you will need to create a half supply bias with a couple of resistors(10K). 

JTEK Link:

Link to Schematic: This schematic has been updated 11/10/2014 to reflect a minor change on the Cs2000 chip. The clk out pin4 should be tied to pin 5 and the cs2000 configured to use this clk as the reference source

PC Board PIC: This is a homebrew board, created using laser printer toner transfer and sponge/ferric chloride etching.


Layout: (expresspcb file format)

Video Demo:

Code Example for CS2000:

Click to access cs2000_demo_code.pdf

Coming Soon..Some New Radio Circuits!

Working on a high performance 2 MHz -70 MHz clock generator for I/Q radio design. The goal is a simple single signal radio with no tuned circuits and good performance. The approach will be using the 4066 analog switch as switching mixer in weaver topology phasing receiver.

The Clk generator utilizes a Cirrus logic CS2000 FRAC N PLL IC. I am completing the software and board layout for the CLK generator this week.


update: decided to abandon the wearver method and use a traditional phasing filter

The Very Cool MRF49XA ISM band Transciever

A few years ago, I remember seeing a highly integrated radio transceiver for 433 MHz and 915 MHz. It was produced by Integrated Associates, which then was bought by Silicon Labs. Recently, I found the same part sold by Microchip as the MRF49XA. It is a really cool part, requiring only a crystal and some decoupling caps, if you use the balanced loop antenna as I have in my design. If you use a standard 50 ohm antenna – you will need a few passive components to provide impedance matching and balanced to unbalanced conversion at the output. The radio uses a DC conversion topology and requires no IF filter circuits. It also provide antenna tuning, and internal TX/RX switching etc. Microchip has a very good data sheet which explains how to use the device.

In the picture below is a little USB transceiver I designed to demo the device. I will post a demonstration video showing a couple of these in operation. You can see from the picture, there are not many parts at all (there are two decoupling caps on the back)! The challenge was writing all the code to configure and operate the radio circuit and create the USB link. I created a generic HID USB device which requires no special drivers in windows. I ended up building a complete library of routines to operate the radio and configure its frequency, power, network ID, baud rate, etc…. The device works very well and with this antenna has a range of about 200 -300 feet. Much better performance can be had with a better antenna but this requires more board space and or the matching components discussed above.


more to follow…

A High Performance Regenerative Radio

I have built countless regenerative radio circuits throughout the years and some have worked well – some haven’t. I was inspired by the circuit design of the TEN TEC regenerative radio kit. I used some of the same ideas, but changed the design to better match my design criteria. In this  design, I had eight important design objectives:

Simplicity – this type of radio should not be complicated and I have seen designs on the web that may perform well, but seem unnecessarily complicated.

Tuning/fine tuning – I use a cheap poly variable for main tuning and a junk box rectifier as varactor for fine tuning.

No special inductor required – I have tried all sorts of junk box inductors and they all work great. With this design, no tapped coils or tickler windings are required. This design could easily be made into a multi-band radio

Extremely smooth and stable Regeneration control – I adjust a DC bias point condition instead of RF Feedback to control regeneration and the performance is excellent. There is no hysteresis or abrupt transition from regeneration to oscillation.

Ample Audio Gain with no motorboating or instability – I stayed away from the LM386(which could be used) and chose a TPA301 amplifier IC – which give excellent results.

Antenna Isolation – This is achieved with a simple grounded gate input stage which shares the LC tank with the oscillator.

Excellent sensitivity – This design is the best performing Regen I have ever built

No critical adjustments and easily repeatable results – I have built this circuit now three times with different inductors, for different bands and with different JFET device types on bread-boards, etc. The results have all been the same and I have only had to make minor tweaks to optimize performance for different JFET types and significantly higher or lower frequency bands. The radio currently tunes 7-11MHz.

The basic paradigm of this design is to break up the traditional oscillating detector into a separated regenerative amplifier and detector circuit.

The detector is  a “plate detector”, where RF is fed back to the Amplifier via a partially RF decoupled source(normally bypassed all the way for RF when used as a detector).


Version 2: (07/30/2015)


Link to PCB for version 2:(expresspcb format)

Version 1 (shown in video):


picture of prototype:


Video Demo:

Tesla Coil – A Powerful Radio Transmitter with A Lousy Antenna!

I have tried to figure out what exactly a Tesla Coil was for years. I finally occurred to me that the best way to think of a Tesla Coil was as a very high power rf transmitter simply with a terrible antenna. The secondary is really just an ultra shortened 1/4 wave antenna – so short it can barely radiate RF at all. Ultimately the energy boils off the top of this antenna as the discharge us mad scientists know so well.

To prove my conception of a Tesla Coil: I built a class E amplifier that is powered from a simple half wave rectifier and 120VAC straight from the wall. A class E amplifier is simply a power mosfet switched on and off by a square wave drive (highly non linear all on or all off). The trick is to load the output with a slightly out of tune resonant circuit with respect to the frequency of the pulsing. By doing this, the voltage and current waveforms become 90 degrees out of phase. Since power is current times voltage, if they do not overlap (or very little) as in the case of a class E amplifier; no power is dissipated across the semiconductor switch and efficiency can approach 90+ percent. The draw back of such an amplifier is that it is narrow band and must be critically tuned.

A Tesla Coil is ideal as a load for such an amplifier. If you check out the schematic below you will see it is shockingly simple (pun intended). The schematic is incomplete in that I do not show the driver circuit. The driver is nothing special – just a mosfet driver IC (there are a million of them out there) and a couple of 555 timers to create a modulated (30Hz 20% duty cycle) 300KHz pulse source. I can also just use my  bench sine/square generator. The reason I modulate it at 30 Hz  it is twofold, one, to give the crackly long streamers and two, if I run in continuous mode the resonant caps will explode from over heating, also the secondary overheats and starts melting, catching on fire ( both of these events happened to me).

While the design is simple, tuning is challenging – if the current and voltage are ever in phase… forget it.  Get ready to blow some FETS!



Video Link: