Test laboratory of our factory asked me to design an isolated high voltage amplifier. Here are the requirements:

  • The circuit will be built with transistors and passive components in the lab. (No IC usage - because we don't have any and can't buy)
  • The circuit will be supplied from lab power supply (Vs < 40VDC).
  • The input signal will be applied from the function generator (1Vrms sine - frequency may vary between 10kHz and 100kHz).
  • The output will be isolated and have a peak value of 1.5kV.
  • Output power will be 100mW max.

Here's my design on paper:


simulate this circuit – Schematic created using CircuitLab

As you can see, the circuit is a simple Class AB amplifier with inductive load. Since the voltage gain is 10V/V, the voltage across the primary will be 10Vrms, so the output will be 1kVrms = 1.5kVpk.

Is there any problem with this circuit? What can you suggest?

  • 1
    \$\begingroup\$ Nothing wrong with the concept AFAICS, only details. 30V may be a tad low for 10V rms output, depends on your distortion. BC337 seems a bit small for the output device, and a bit big for the pre-output device, perhaps those part numbers are placeholders? I'm mulling on the idea of something almost identical so I can have 20W of 'mains in my pocket' to test the flicker of commercial LED light bulbs in shops before I buy. However, instead of building an AB amplifier, I'm spending $4 on a 50W class D module online. You can afford $4, right? It's worth 100x that in your time. \$\endgroup\$ – Neil_UK Mar 11 '19 at 7:11
  • 2
    \$\begingroup\$ Ah, read your frequency figures, and a commercial class D audio amplifier is out. A 1.5kV transformer, big enough core for 10kHz, with low enough SRF for 100kHz? That will be fun to wind. Make sure you thoroughly study how to wind high frequency transformers. It might also be easier to split it into two seperate 10k/30k and 30k/100kHz ranges. \$\endgroup\$ – Neil_UK Mar 11 '19 at 8:00
  • 1
    \$\begingroup\$ Last time I checked your profile, you were in Turkey. As such, you should have easy access to all the world's IC distributors; something isn't right about "not being able to buy an IC", to be honest. \$\endgroup\$ – Marcus Müller Mar 11 '19 at 8:12
  • 1
    \$\begingroup\$ @MarcusMüller That's what I doubt. The guys in the lab spoke to the tech director first, then the director forwarded to me at 5:50 PM on Friday! Today I designed on paper, then simulated. I have a lot of experience on Class AB audio frequency power amplifiers but I just worry about the concept/idea of getting isolated HV output from it (Neil told that the idea seems OK). Even if the simulation results are acceptable I still couldn't trust on it. That's why I also asked here. Now the layout-guys are drawing the circuit on Altium and the trainees are winding the transformer. \$\endgroup\$ – Rohat Kılıç Mar 11 '19 at 10:30
  • 1
    \$\begingroup\$ this discusses capacitance leakage inductance tradeoffs. Still looking for good disc/pancake references. Turn to turn capacitance is negligible compared to layer to layer. Think about the energy storage as proportional to v^2. To and fro winding doubles peak inter-layer voltage compared to unidirectional layers. Disc winding 'shortens' the layers significantly to further reduce the interlayer voltage. \$\endgroup\$ – Neil_UK Mar 12 '19 at 13:57

Is the diffpair tail current at least 3X the Miller Capacitance charging current, for 28 volts PP at 100,000 Hz? Otherwise you risk SlewRate limiting.

===== Linearize the high-gain (bottom transistor) with 10 ohms in the emitter ===

The load is reactive, or even resonant at an unknown frequency. I'd add 10 ohm base resistors in each of the 4 output transistors. And a Zobel series RC from output to Ground.

There will be lots of cross-over distortion. Why not use 5 diodes, and place 4.7 ohms in output emitters.

Place 100pF cap across the 3-diode (Or 5-diode) network.

To avoid sag in the upward-swing output Vout, I'd increase the VDD to 40 or 45 volts, and increase that 100 ohms above the bootstrap capacitor.

The resistive feedback network adds phaseshift to the feedback waveform. The diffpair input capacity, base-emitter-emitter-base, must be charged.

If the VDD bypassing resonates, life gets tough. And 1uF and 1uH resonate at 160KHz. 10uF and 0.1uH resonate at 160KHz. 1,000uF and 0.1uH resonate at 16KHz. How to dampen? Use Rdampen = sqrt(L / C) = sqrt( 0.1uH / 0.001F) = sqrt(1e-7/1e-3) = sqrt(1e-4) = 0.01 ohm. Which may be provided by ESR of the bypass capacitor. Or not.

If you use small VDD bypass caps, use two caps in parallel, and place 1 ohm in series with one of the caps.

Again, the large caps may self-dampen.

By the way, 4" of wire is 0.1uH.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.