Frequency doubler 10 - 20 MHz

[bob91343]

Due to the forward drop of the diode, it will indeed conduct less than 180 degrees.  How much less will depend on amplitude.

[T3sl4co1l]

Out of a square source, you're more likely to get harmonics due to capacitive loading (asymmetrical rise/fall).  Still, something.

Tim

[fourfathom]

Out of a square source, you're more likely to get harmonics due to capacitive loading (asymmetrical rise/fall).  Still, something.

With a single-ended multiplier you will get even and odd harmonics.  The transistor multiplier mentioned previously is essentially a class-C stage, and squarewave drive works as well (or better) than sinewave.  As you know, the beauty of a balanced multiplier is the minimization of the odd harmonics.

I once redesigned a clock tripler in a piece of telecom gear: 51.84 MHz to 155.52 MHz.  We had a nice square wave clock input, and the original design had a single-transistor Class-C style multiplier with fancy filtering and amplification to get rid of all the undesired harmonics (especially the 2nd).  I replaced all that with a simple circuit that sent the squarewave clock into a 155 MHz SAW filter, followed by an ECL buffer to square up the 3X output.  Worked great.  Wouldn't have worked at all as a doubler.  Fourier is your friend.

[T3sl4co1l]

I was referring to the diode idea.

With a single-ended multiplier you will get even and odd harmonics.  The transistor multiplier mentioned previously is essentially a class-C stage, and squarewave drive works as well (or better) than sinewave.  As you know, the beauty of a balanced multiplier is the minimization of the odd harmonics.

Well yeah that works great, for a sine input.  For a square (not rectangular, I mean 50% square), if you prefer the Fourier picture, the harmonics mix, superimpose and cancel out.  You don't get even harmonics, you just get more odd harmonics.  In short, as I said at the start of this thread, you can't square (the math function) a square (the waveform), you just get DC.  The class C amp is just a unipolar class D amp, reproducing the signal with actually unusually good fidelity.  There is no instantaneous (stateless time domain) function, even or odd, which can break the squareness of a square wave.  At best you get skewing of rise/fall times; in which case I suppose we should really be discussing a trapezoidal wave, which is the more practical case after all.

Anyway, the point about a diode was, given various assumptions about the signal path, one could introduce even harmonics by unbalancing the edges.  An ordinary amp with no additional assumptions, won't do that, but given similar assumptions, can do the same thing (i.e., a capacitive load driven by asymmetrical on/off source resistances, to unbalance the edge rates).

I once redesigned a clock tripler in a piece of telecom gear: 51.84 MHz to 155.52 MHz.  We had a nice square wave clock input, and the original design had a single-transistor Class-C style multiplier with fancy filtering and amplification to get rid of all the undesired harmonics (especially the 2nd).  I replaced all that with a simple circuit that sent the squarewave clock into a 155 MHz SAW filter, followed by an ECL buffer to square up the 3X output.  Worked great.  Wouldn't have worked at all as a doubler.  Fourier is your friend.

Indeed, triplers work very nicely from whatever input, and you can basically use a comparator (ECL input stages are basically diff pairs) to not just generate rich 3rd harmonic (as in ~1/3 the fundamental amplitude), but by overdriving the input, the rise/fall time and amplitude becomes much less significant, i.e. it acts as a limiter too.

Consider a square wave into a filter.  The filter will essentially produce repeated step responses, overlapping (superimposed) as they do.  The effect is that, though the 3rd harmonic has been selected as dominant, the 1st and 5th combine as sidebands, producing apparent amplitude modulation at the 2nd harmonic (being (3-1) and (5-3)).  Feed this into another stage, and the amplitude modulation can be overdriven to give a flatter output, even without a follow-up filter; in Fourier terms, the sidebands have been mixed together and canceled out.  Which of course works fantastic for FM radio, where insensitivity to AM (environmental fading, tuning error) is a virtue.

I don't know that Fourier analysis is all that useful of a way to reason about nonlinear systems (outside of a harmonic balance analysis, and anyway, have fun doing that by hand?), i.e. as thinking about harmonics and how they mix.  The toy, the cartoon picture really, of multiplying sines, is only valid for what it is -- a translinear multiplier applied to pure tones.  It's more complicated when multiple tones are present, and much, much more complicated when highly nonlinear mixers are added on top of that.  Higher order mixing products will completely destroy your image rejection, in the context of radio design.  (For multipliers, who cares, at least all the products are synchronous!)

Tim

[fourfathom]

... The transistor multiplier mentioned previously is essentially a class-C stage, and squarewave drive works as well (or better) than sinewave. ...

Well yeah that works great, for a sine input.  For a square (not rectangular, I mean 50% square), if you prefer the Fourier picture, the harmonics mix, superimpose and cancel out.  You don't get even harmonics, you just get more odd harmonics.

Tim, we have been in general agreement since your observation that a pure square wave into an ideal push-push diode doubler gives a DC output, not 2x.  But fed a pure square wave a simple transistor (or diode) stage will generate strong 2nd harmonics (and 3rd, 4th, etc) if you  adjust the conduction angle appropriately.  Some input filtering and bias control is probably needed to get the right waveshape.

[TimFox]

With a true square-wave drive, one cannot adjust the conduction angle of an active device, only the output amplitude.
With an approximately sine-wave drive, adjusting the bias works well for class-C amplifiers tuned to the fundamental, or with different bias settings to optimize the output into a circuit tuned for the second or third harmonic.

[T3sl4co1l]

Ah, okay.  Then you're following the second part -- adding some capacitance or whatever to draw out the waveform.

I would rather not use terms such as "conduction angle" as this implies it's a linear (other than cutoff) amplifier, with no extenuating (stateful) conditions.  You can have an ideal class C amp (an instantaneous, zero-capacitance, nonlinear dependent source, say Iout = f(Vin) for Vin > Vthreshold, 0 otherwise), and it will never, ever modify a square wave.  But you can have a nonideal one, say with a BJT which overly saturates and thus incurs storage time at turn-off, with no corresponding delay at turn-on, which therefore acts to increase the conduction angle.  And it really is the conduction angle in the BJT case, as it conducts until t_stg has passed; it's not just the output hanging around because current dropped and there's capacitance on the node or whatever.

But the same is likely not true of a MOSFET amplifier.  Mind, given a highly nonlinear Coss typical of power MOSFETs today, it might look suspiciously like it, anyway (i.e. drain voltage seeming to sit around for a while, before shooting off).

Eh, perhaps I'm being too technical / picky / pedantic with the levels of abstraction that come to mind when I see amplifiers and classes.

Tim

[fourfathom]

Example:  A very simple doubler.  3V 10 MHz square wave input, 20 MHz sine output.  I made absolutely no effort to optimize this, I just threw some "probably close enough" values at LtSpice.  Note the R/C coupling into the diode -- this turns the square wave into pos and neg pulses.  The output tank circuit is tuned to 20 MHz.

[TimFox]

Yes, the R/C coupling changes the square wave into something not square.
Conduction angle was a common parameter in class-C vacuum tube amplifiers, where sine waves got converted into sine tips that produced the desired output when filtered by a tank circuit.
For non-linear amplification with solid-state devices and tuned loads, I recommend the discussion (trigger warning:  Bessel functions) in https://idoc.pub/documents/communication-circuits-analysis-and-design-clarke-hess-6nq8yjekrqnw 

[fourfathom]

Yes, the R/C coupling changes the square wave into something not square.

Which is what I meant when I previously mentioned input filtering and adjusting the conduction angle.  Thanks for the link -- looks interesting!

[TimFox]

I love that book for the topics it covers.  However, I recommended it to someone here earlier, and I think he freaked out when he saw the Bessel functions in the calculation of the harmonic content of the collector current in a BJT driven by a sinusoidal voltage.

[bob91343]

Where do you get a 633 pF capacitor?

Well that looks like the best idea so far.  I will try it; I have all the parts.

[fourfathom]

Where do you get a 633 pF capacitor?

Well that looks like the best idea so far.  I will try it; I have all the parts.

You don't, except in a simulation.  Instead, you will probably use a trimmer capacitor or tunable inductor.  BTW, that 633 pF cap should probably be smaller by about 1pF, to compensate for the diode capacitance.  And you would also need to compensate for the load capacitance, and impedance.  My simple unloaded tank needs a very high-Z load to work as shown.

But that circuit was just a demonstration of the concept.  You might want to use a transistor with a collector tank-circuit rather than the diode.  If I were doing this I would probably use a double-tuned tank circuit to improve the filtering and make tuning easier.  I will try to put together a simulation of that and post it here.

[bob91343]

It would be an understatement for me to say that I don't trust simulations.  Too many assumptions, and too little attention to details.

I am more a seat of the pants designer, setting things up and testing, then seeing where I went wrong and correcting it.

[fourfathom]

I am more a seat of the pants designer

Me too, but simulations can be extremely useful.  The trick is to know what factors may differ between simulation and the real world.  Model behavior, component tolerances, poorly-known parameters, etc. 
For example, I don't know what the output characteristics are for your 10 MHz oscillator.  One of the spec sheets posted up-thread may have the data, or at least some of it.  And I have no idea about the input impedance of the circuit you plan to drive with this 20 MHz signal, or what the desired drive level is.

Even then you should only trust the simulation so far, but for passive LCR stuff it's quite good.  Experts (not me) can do much better in all of this.

Just for fun, I simulated a trifilar-wound transformer driving two diodes in the push-push doubler we previously discussed.  By putting a small capacitor in series with the input we get a differentiated pos and neg spike from the 10 MHz square wave.  Adjusting the capacitor value I can get a pretty decent duty-cycle on the 20 MHz output.  Adding a 20 MHz tuned circuit at the output cleans up the waveform nicely.  The available output voltage will depend on the load Z.  I haven't tried resonating the trifilar secondary, but that might be helpful in purifying the output.

[TimFox]

With Spice analysis, the .AC mode is essentially the algebra you would do to solve the usual equations, after any active devices (transistors, etc.) have been "linearized".  The results from accurate values of R, L, and C should be exact.  If you apply 1 V directly to the base of a transistor, the calculation should give you an accurate gain for small signals, (i.e., 100 V at the output, even though you have a 5 V supply, means that 1 mV in should give you 100 mV out) but you need a .TRAN analysis to get a reasonable result when actually applying 1 V to such a device.

[fourfathom]

Here's another approach, this time with a transistor.  I have a somewhat more reasonable loaded "Q" on the tank circuit.  Again, the input is differentiated, this time to avoid 10 MHz half-cycle conduction.  Instead, we want about 25% (90 degrees) for the transistor on-time.  I am assuming a 500 Ohm load.  The output shows some ring-down on the waveform, another filtering stage would help this, or a higher Q on the single filter.

[bob91343]

I can't read values on the schematic, and the waveform in the middle has no legend so I don't know what it is showing.  Some spectrum I guess.

[T3sl4co1l]

Try loading the image in a new tab? (from thumbnail state, middle-click; from enlarged state, right-click view image)

Tim

[bob91343]

Okay that worked but what does pp mean on that capacitor?

[fourfathom]

Okay that worked but what does pp mean on that capacitor?

It's a typo that LTspice ignored.  It should read 150p (picofarads). 

If you want to try that circuit, here are some things you can play with:
I have a fairly low "loaded Q" on that tank circuit.  This lets it work in the case of slight component tolerances and load parameter difference, but some tuning might still be required.  If you have a hand-wound inductor, try squeezing or spreading the turns to get peak output.
You can improve the filtering by increasing the loaded Q.  You can do this by changing the C7/C8 ratio, and probably adjusting the inductor.  Or reduce the inductance and increase the capacitance.
If you change the Q, you may want to adjust the transistor current.  Adjust R2 to obtain the maximum undistorted output waveform.

The middle plot shows the spectrum at the output port.  You can see that the 10 and 30 MHz signals are at least 25dB below the desired 20 MHz signal.  Increasing the tank circuit Q will improve this further, but at the expense of more finicky tuning (including any temperature coefficient-related mistuning)

[TimFox]

In order to maintain backwards compatability with Spice files dating back to the Fortran-based program and Hollerith cards, Spice parses the component values using only the prefix, case-insensitively, (p, m, k, etc.), obtaining the unit (V, F, ohms, etc.) from the initial letter of the component name (VIN, C12, R3, etc.), requiring MEG for mega, since m or M always means milli.  "pF" is optional, since the F is ignored.

[bob91343]

I have no problem tweaking component values once the circuit appears to work.  Even if it works it still may be some time before I implement it in the radio.  I don't know how many hams read these things but I am sure they know how messy it gets when you want to pull a radio and put it on the bench.  Reinstalling is an excercise in 'what is this wire for anyway?' - at least that's how it works for me.

[TimFox]

This information is not directly related to your original problem, but it may illustrate the importance of filtration with frequency doublers.
For an unrelated project, I needed a OCXO at a frequency near 78 MHz.  There are many OCXOs on eBay that have been removed from equipment;  most have "normal" values like 10 MHz, but there are some oddball frequencies available.  Since these odd frequencies were originally special order, the exact part numbers cannot be found online, and the eBay vendor may have incorrect parameters (except for the frequency, which is marked on the unit), including what type of output (sine or square).  Therefore, I need to check them before proceeding.
The unit I tested (Isotemp OCXO 131-132 at 77.76 MHz) worked fine at 5 V supply, pulling about 0.55 A at turn-on and 0.21 A after an hour.  It is hard to judge that waveform with only a 200 MHz scope bandwidth, but it looked like a distorted sine wave at approximately 3 V pk-pk, connecting through 510 pF to either a properly-terminated 50 ohm line, or after 3 ft of RG-58 into a high-Z load (roughly 120 pF including scope input).  The voltage changed only slightly (1.3 dB) with the termination.
The frequency was correct, to within the calibration of my frequency counter (only 2.5 ppm low), with no attempt to use the electronic tuning (if installed).  So far. so good.
Measuring the spectrum with a 400 MHz spectrum analyzer, from 2 to 202 MHz (50 ohm input impedance), I noticed a strong subharmonic line in the spectrum, at half the nominal frequency.  Since 78 MHz is a bit high for a crystal, the construction must have a 39 MHz (rounded) oscillator with a tuned doubler for 78 MHz.  The spectrum contained all of the harmonics of 39 MHz (up to my scan limit), with 78 MHz the strongest.  Specifically,
39 MHz:  -36 dBc (1st)
78 MHz: 0 dBc (2nd), +13 dBm = 1 V rms at this frequency.
117 MHz: -42 dBc (3rd)
156 MHz: -29 dBc (4th)
195 MHz: -44 dBc (5th)
Note that the suppression of the fundamental and all but the 4th harmonic is quite good, about -40 dBc, but the 4th harmonic (2nd harmonic of the desired output) is stronger than the other undesired frequencies.  I shall therefore apply some basic filtering centered at 78 MHz, since I need to amplify it anyway.

[fourfathom]

Yes, on the undesired output spectra.  To the OP, what exactly is this new 20 MHz signal supposed to feed?  Some circuits are fairly tolerant of impure clock inputs, but with others these spurs (harmonic and subharmonic in this case) can cause problems.  For example, if the clock is driving a mixer, this can result in spurious responses at the output.

And while you're at it, what is the desired level and impedance of the 20 MHz clock?