2
$\begingroup$

I am designing a simple transmitter using an SI5351A to generate the carrier, then modulating it with a diode ring modulator to generate a dsb-sc signal, then somehow converting it to ssb modulated signal to send into a class E amp, but I'm not sure how to extract just one sideband from the dsb signal, I know it can be done with filters but with my frequency changing I would need to change the filters dependent on my transmitting frequency, is there another way to achieve this?

$\endgroup$
2
  • 2
    $\begingroup$ Traditionally, with filters at IF, then mix up to where you want it. But first, I don't think you can transmit SSB with a class E amplifier, at least without some careful predistortion. Can you clarify - why class E - what's driving the modulator? - why not just modulate, filter and mix up? $\endgroup$
    – tomnexus
    Mar 1 at 22:09
  • $\begingroup$ i'm was thinking of driving the modulator with a carrier of the frequency i want to transmit from the SI5351A, i'm using class E because i'm a beginner to PA's and class a or ab amps are well out of my knowledge, but i suppose i could use an if, then mix it, then feed it into the amp and maybe it would work $\endgroup$ Mar 2 at 7:06

1 Answer 1

2
$\begingroup$

That makes no sense: you wouldn't convert the already modulated, high-powered signal to SSB. That would require steep filters at the RF frequency, and that filter needs to also be low-loss and high-power capable. That's prohibitively complicated and expensive, and sometimes not even physically possible.

But: you already have a chip that can produce two different tones, and you want to have suppressed carrier: This would suggest you want to go with a superheterodyne transmission scheme!

You'd mix your audio with a relatively low frequency (say, 140 kHz), use a filter to get rid of one sideband and the carrier, then mix with your target frequency minus 140 kHz, and use another filter to get rid of the carrier and the unwanted mirror at the target frequency - 140 kHz. Done! Amplify that.

Of course, that's a very 1940s way of doing this. Realistically, instead of analog mixing to IF and filtering there, you'd just directly produce the single-sideband IF signal, e.g. in a microcontroller. Much easier, much cleaner. Your listeners will prefer that ;)

ssb modulated signal to send into a class E amp,

Class E?? Oh well, then this does make even less sense.

With a class E amplifier, your amplifier is also your mixer.

Figure from original class E amplifier paper
Figure from the original paper introducing the class E amplifier, [1 ]

As you can see in the diagram from [1 ], there's a switch you need to actuate. In fact, you need to actuate it at the intended RF architecture. The switch switches (haha) the input signal and thereby amplifies its amplitude; class E is highly efficient for constant envelope signals, because in that case, the "active switching device" (essentially, a transistor) will be "on" or "off" most of the time, and only very shortly in a transition between these states, where it dissipates power as heat. So, you drive the amplifier with a square wave of the up-conversion frequency.

Now, you're applying this amplifier topology to an AM signal; that means you need to vary the power flow into the amplifier, and in the diagram above, that equates to modulating the "DC POWER SUPPLY"[2 ]. That's fine – but it's not lossless thing (it depends on the efficiency of another amplifier supplying the DC voltage), compromising the efficiency advantage of the class-E amplifier.

But: That's fine! Square waves can easily be used in the generation of SSB signals; use two class E amplifiers in parallel, but drive them with two square waves that are ¼ period out of phase (that will require a PLL if you want to derive these square waves from your SI5351A; you usually wouldn't do that, but get a synthesizer that offers you an inphase and quadrature oscillation to begin with), and add their output (using an RF combiner, typically). You'd then numerically generate a sine and a cosine of half your audio bandwidth, mix your audio (typically, also numerically) with that separately, and use a low pass filter on each branch. Feed one branch each into one of your class-E amplifier's modulation input.

Done, you just built a Weaver modulator; the magic is that adding your sine and cosine-modulated audio cancels out one sideband.

Honestly, though, this is where the complications begin, and you'll notice that gain differences (which we'd call IQ imbalance in a direct conversion mixer) need to be compensated for spectral cleanliness. It's really not clear to me why in a AM waveform application of a beginner, a class E amplifier is the way to go. There is a lot of attractiveness in combining phase-offset class E amplifiers, especially in suppressing the harmonics, but to realize these advantages would take feedback loops, thus modelling of the behaviour of the amplifier pair as control system, and thus a lot of math. And at least in the microwave region, I don't think this is done commercially – certainly not because it's mathematically intractible, but because component (and especially switching speed) imperfection make this an architecture hard to tame. There are experimental application of class D / class F with some matching (which kind of is close to being class E, but is distinctively not class E) for HF that I'm aware of. That's a clever choice: with modern FETs and modern gate drivers, we can directly apply pulse-width modulation, and get something that works over a larger frequency and amplitude range without changing component values.
Compared to the class C, a practical class E would probably only have minor advantages on the efficiency front - and with digital predistortion being a commodity function in silicon-integrated RF chains, I'd assume that you'd much rather find class C amplifiers than class E among those practically employed with a matched load for fixed frequency.

i'm a beginner to PA's and class a or ab amps are well out of my knowledge,

Wow! Honestly, I'm more of a discrete-time, Laplace and Fourier kind of guy than an analog circuit design guy, but even for me, the way class-E works is (much) harder to understand than class-A or class B, which can essentially be understood from DC inputs (and extended to modelling in a couple parasitic capacitors should one need to understand the high-frequency behaviour).


[1 ] Nathan O. and Alan D. Sokal, "Class E – A New Class of High-Efficiency Tuned Single-Ended Switching Power Amplifiers". IEEE Journal of Solid-State Circuits, June 1975
[2 ] Fredreick H. Raab, "Idealized Operation of the Class E Tuned Power Amplifier", IEEE Transactions on Circuits and Systems, December 1977

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .