My take on a discrete shunt voltage regulator

[AndrewT]

any time you care to look up the various ones posted in the Forum and elsewhere.

[HiFiNutNut]

Would you place the output cap near the shunt MOSFET output whether remote sensing is used or not? physically it could be placed anywhere from the output to right at the sensing point.

[ikoflexer]

We'll see, it all depends on whether something better than the trivial comes up.

I would say where the sensing points meets with the load and the power lines from the shunt.

[keantoken]

Hello all.

I have become interested in the Shunt scene.

I was wanting to obtain some measurements/figures from the latest reg here so I could set a benchmark for my own design ideas. More specifically, I want to know the output impedance value for the flattest area of the curve.

Have you simulated with typical ESR and ESL values for C1 and C2? I'm curious how much of an effect there would be.

EDIT: In fact, I could have just asked for the LTSpice files. Can someone please upload them?

- keantoken

[ikoflexer]

Post 1 has the plots you want. For all caps I used ESR of 1R. C2 ESL of 10nH and C1 ESL of 15nH. I know, these are exaggerated, but that's exactly what I'm trying to do, simulate a somewhat sloppy build, and still get good performance.

Zout at 100kHz is about -88dB, or about 36uOhms which, is pretty academic of course, because in the real world one has to deal with a number of issues.

I have other versions that simulate even better (one I showed you in the Aspen HP amp thread) but they don't make a good overall regulator, IMHO. This one here is something that has more specs that I think are desirable in a real circuit.

If you'd like to play with it I can send you the file.

[ikoflexer]

There's been discussion all over the place about how desirable low impedance is in a regulator. This design promises low impedance and here we go, we build it, we connect it to the load, super! It all works. But... there's a but, yes. It so happens that there is a distance of 5 cm between the load and the regulator. The guy who built it carefully read through the comments here and decided to follow our advice and use relatively thick, solid copper wire of 1 mm diameter.

After looking at the output impedance plot in post #1 our user knows that 36 microohms output impedance is not of this world, but he still hopes that in reality he'd still get something below 1 milliohm. Will he?

The inductance of 5 cm of 1 mm diameter copper wire is about 45nH. What happens to the output impedance? It goes up to about 404 milliohms.

This is all still theoretical, a simulation. But it illustrates the point and we should expect probably even worse effects in real life.

Proper implementation is really essential. To have any hope of low output impedance, remote sensing is not an option, but becomes necessary.

[keantoken]

Thanks for the advice... I wish it was easy to test PCB layouts by the dozen and see what works and what does not. I was asking about Zout because I couldn't tell the exact number from your graph.

So, wires should be as short as possible, remote sensing done by shielded cable.

My current design has 60uohms at 100KHz, but 11uohms between 5k and 10Hz (the MOSFET used has much influence on Zout, here I use IRF9240). It is reasonably well-behaved, when compared with your stress test (although mine is with a 47u, .01ohm ESR bypass). It is completely un-Salas, and a little more complex. I'm using an incredibly fast implementation of the Rush Cascode (very well suited to regulators, if you're not scared away by tempco and its other weirdness).

Only in simulation though. With proper layout, how close to simulated performance do you think we can get?

Thanks,
- keantoken

[ikoflexer]

kt, there's one big hurdle to cross, before even thinking about zout in reality. And that's stability. As soon as you get low zout numbers past 20kHz, stability becomes a huge issue. Sometimes you can get away with compensation caps, if you're not far from stability. I make it a point to build the prototype to see that at least is stable, or if not, to investigate how to make it stable. You may want to test yours especially now that it simulates with good performance and wide bandwidth. If you already did, then disregard my comments.

If stable, then only reality can give you the actual answer, I wouldn't venture. The way I look at it is, if the design promises good performance, and both the stability analysis (not that I trust it so much) and real circuit stability is good, then the better implementation you have the closer you'll be to the ideal result given by simulation. In my view the corollary holds too: if it simulates huge output impedance, how in the world can one expect it to perform well in real life?

Just my 2c, I'm a beginner. The experts may disagree.

[HiFiNutNut]

Originally Posted by ikoflexer

I would say where the sensing points meets with the load and the power lines from the shunt.

Ikoflexer, my scope does not agree.

I rearranged some tracks and enabled remote sensing. The output cap (1.1uF + 1R) was not at the output but at the sense/load point. Measured a couple of 60mV resonances at the output, reduced to about 15mV at the load.

I then removed the "local" bypass (1.1uF + 1R), and put a 22uF(0.34R ESR) ZL at the output. A thin flat line appears in the scope at both the load and the reg output.

So it seems that the regulator needs an output cap at the output.

Retested the 1.1uF + 1R at the output with remote sensing. 15mV appeared in the scope again. This tells me that the output cap is not large enough.

OK, I am going back to do more soldering. I will gradually increase the output cap until I see a thin flat line in the scope.

[ikoflexer]

Originally Posted by HiFiNutNut

Ikoflexer, my scope does not agree.

I rearranged some tracks and enabled remote sensing. The output cap (1.1uF + 1R) was not at the output but at the sense/load point. Measured a couple of 60mV resonances at the output, reduced to about 15mV at the load.

I then removed the "local" bypass (1.1uF + 1R), and put a 22uF(0.34R ESR) ZL at the output. A thin flat line appears in the scope at both the load and the reg output.

So it seems that the regulator needs an output cap at the output.

Retested the 1.1uF + 1R at the output with remote sensing. 15mV appeared in the scope again. This tells me that the output cap is not large enough.

OK, I am going back to do more soldering. I will gradually increase the output cap until I see a thin flat line in the scope.

When you say resonance, what do you mean? Would you be able to take a picture of it? Does it look like a sine wave, or does it look like a spike that occurs from place to place? If the spike, crest, or wave repeats on the screen, how many divisions are there between the two of them on the screen, and how many mS/div was the setting on the scope? I'm trying to infer the frequency of the blip.

Also it would be good to know if the blip goes down in size with higher current.

Another thing to consider, how is the grounding done?