150W current source, with a current-mode switcher

But kindly give me some feedback. Also, your time will be well spent studying the LM25117 datasheet.

Lawdy, lawdy, there's even a PSpice model that's NOT encrypted >:-} ...Jim Thompson

That's almost like an old-style Japanese schematic, every square inch is occupied.

Just a quick comment: I'd add 0.1uF caps at input and output. Otherwise you might run into issues at the EMC test lab. Some high-current ferrite beads from Laird or another manufacturer can't hurt either.

I just had a client design review where that was a problem (at the EMC lab).

Why a 100ohms gate resistor for Q2? That sounds high.

I like that as well as adding inductance.

Yes. The gate resistor is R15, 3.3 ohms. It's certainly an overly-crowded drawing.

You comments on EMC testing are good.

SPICE models are often encrypted nowadays? The same *stupid* mindset that leads to encrypted datasheets?

Is it one-upsmanship if I say mine's 2kW?

Granted I didn't test it at full power, but it sounds very similar.

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

Here's something for y'all to knaw on, a 150W current-source SMPS, made with a current-mode controller. It has external monitoring and programming with

Most power supplies have low-Z voltage-source outputs, and when operated in current-limit mode they simply servo the voltage to the achieve the desired output current. A true current source has a low effective output capacitance and its output voltage can change rapidly to maintain a constant current into a changing load.

Current-mode controllers have by their nature constant-current outputs, from the buck inductor anyway, but this is morphed into a voltage source with a big output capacitor and the controller's feedback. So I figured, why not implement one with a small capacitor and constant-current feedback? This turned out to be awkward with most controllers I considered, but NSC's (TI's) LM25117 was an exception. Among its many features is a fast current-monitor output using a cycle-by-cycle S/H circuit, which was helpful and inspiring.

My project name is PWR-758. My PCB layout uses mostly SMT parts, but I avoided the temptation to place any parts on the bottom. I got a bit carried away with bells and whistles, like programmable compliance ranges, external meter, external disable, etc., plus configuration modifications with a 10-mode jumper-header, labeled A,B,C to I, extra connectors, etc. I was forced to stop adding features when I ran out of space on my size-A schematic drawing and my

When operating in constant-current mode, one option uses the controller's S/H circuit, and another option uses an LTC6102 output current-sense IC, which is fast and has very low offset voltages. The PWR-758 also has ordinary voltage-source mode options. Please take the compensation part values with a grain of salt, because I have not yet properly evaluated them, nor made a working copy of the circuit. In fact, take the entire circuit with big grains of salt!

-- Thanks, - Win

Sounds interesting, is the schematic available?

No, I don't have it handy...

The basic design is pretty standard. Oscillator and PWM comparator; two error amps, one for voltage, one for current. Average current mode feedback (via ground referenced shunt in series with power input).

Made the mistake of using all LM358s and not filtering often, which got noisy when I added the next part (a modest frequency inverter for induction heating). :^)

Probably, the most novel part of the circuit is the limiter. It has independent max(V) and max(I) inputs; it regulates to whichever is largest. That means the V error amp runs all the time, but its output (the current setpoint) may be clamped, by the current max signal. (The current error amp always has total control. Its output is scaled so PWM goes 0-90% or something like that, as necessary for gate drive operation.)

That little block is:

This doesn't have a "V limit/I limit" detect output, but that should be easy and/or obvious enough to add. The design goal was preventing integrator windup: this wraps around the volt error amp itself, keeping its output below Max.

You're all welcome to critique and improve it; lower parts count and higher accuracy are always welcome.

Tim

Yessir.

I breadboarded a few-amps current source, intended to be super-wideband, namely high impedance over a huge frequency range. It was homebrew hysteretic.

It simulated beautifully but didn't work very well. The big problem was picking up the small shunt voltage close to the switcher.

I thnk the way to achieve a wideband high-Z out is to simply add an output inductor or two.

Yes, it would have worked with a low-bandwidth constant-current switcher followed by a big composite inductor, like what people use in bias tees; a string of decreasing size inductors and damping bits. But in my application, at 4 amps, the inductor chain would have been impractical. We went linear out of pragmatism.

The final inductor in that pic is the Coilcraft 4310LC, ferrite/epoxy filled, 8 GHz, 4 amps, sort of a spin on the Piconics conical inductor idea. That works nicely in other apps.

Here's a PSPL bias tee:

But, you could have just bypassed the error amplifier, letting it demand full current. Then letting the external current error opamp clamping the COMP pin of the device.

That way you set the peak current on a cycle to cycle basis, and let the opamp do the fine tuning

Cheers

Klaus