Controlling IC voltage regulator with a voltage

If you want a 5V linear output, how about designing the switcher's resistor divider to provide 6V or 7V as you want 1-2V of headroom? Or is the problem that you want +1.835V above a variable linear supply over the system output range of 1.25V to 34V?

Maybe consider the *other* way of approaching the problem. Control the switcher output voltage directly, and make the linear follower dump a constant drop.

Yes, this will be a bench power supply. Output range of 0-25V or so.

Good idea. But it kind of comes back to the same problem. The power supply will be completely controlled with a microcontroller, and the voltage will be set using a D/A. In addition, the switcher output is not exactly as stable as I'd like it.

What's wrong with just a simple single op amp differential amplifier monitoring the difference between the linear regulator section input and output, and whose output is referred to ground, set to some appropriate low gain, like about 0.6, so it's output is 1.21V with about 2V differential at the input? The diff amp should have sufficient bandwidth that it doesn't add significant phase shift to the feedback loop...and it's possible that you'll need simply a resistor between the diff amp output and the switcher's feedback terminal. You didn't tell us much about just what you've tried so far, but that transistor in there sounds like an unneeded part and potential trouble.

It's also possible that you'll need to reduce the feedback loop AC gain to maintain stability...

Cheers, Tom

You want to make the output of the switcher 1-2V higher than the output of the linear regulator ... so why not do just that? Use some sort of (high side sensing) diff amp to measure the difference between the output of the linear reg and the output of the switcher, and use that as the feedback voltage to the switcher (which will be setup to try to regulate this feedback voltage at 1-2V or so).

This also means that the switcher will track the output voltage of the linear reg even when the linear reg is doing something unusual, e.g. current limiting. (That's probably a good thing.)

Note: you must make sure that it starts. When you first turn it on, the output of the linear reg will be 0, so the switcher output will only be 1-2V or so. The linear reg must be able to start with an input voltage this low.

Regards, Allan

"Some sort of ... diff amp" could be as simple as...

From switcher R1 output ___ >----------|___|------. | | From | linear | reg >| output PNP |-->|--------< feedback /| voltage | D1 to switcher |

I'm going to try this. It looks like the op-amp circuit I was using, only simpler. I was using a diff amp, but it was really jumpy. The again I didn't spend much time on refining it......

And to the other person who replied about omitting the transistor and just use the op-amp output, I might try this too. The thought had occurred to me, but now that I think about it, my op-amp is capable of going very close to zero so it theoretically should work. It's a LM324. Yeah it's ancient, but they're hard to kill and work very well in this application. And handles up to 32V.

I know you said you wanted the SMPS+linear configuration and not to query that, but I will.

Firstly, get the switcher output stable - anything else you do is gonna be a fudge if you don't.

I'd be inclined to go SMPS only unless you have a critical output noise situation. The linear only acts as a filter to improve noise - possibly at the expense of transient behaviour. With a half-decent switcher and a suitable passive post-filter, the performance can be as good.

I must admit that in an EPROM programmer I designed/built, I use a switcher to step the 12V input up to 28V and then a 317 linear to get the required 25V, because a simple linear (at fixed voltage) IS a simple filter.

I want the linear to improve response and noise, AND to go down to 0 volts. The switcher will only go to minimum 1.2V. Ok, the switcher will drop further if I use some negative voltage. But for a bench supply I want rock solid, very clean output power.

You run the risk of destabilizing the SMPS by putting the linear regulator in the SMPS feedback loop. Imperfect linear regulator PSRR at frequencies within the bandwidth of the SMPS can cause oscillations to feed back to the SMPS with excessive phase lag due to the long path, not accounted for in the factory stabilization of your simple switcher. Try additional stabilization with an RC filter on the feedback voltage to switcher to cure your jumpiness.

Then you don't want to do it the way you propose for several reasons. If you're planning on 5A output, the 1.2V headroom may not be enough, especially in consideration of the fact you plan to greatly attenuate the 260KHz ripple from the '2678- which most regulators do not do very well- usually crapping out at 10KHz ripple rejection of any note. And you're not going to improve the response beyond what the switcher can deliver.

I'm no power supply expert, but I'm going to guess the fact that I'm running this thing at a wide output voltage range will create less than optimum ripple and regulation. From the datasheet it looks like you'd pick an inductor and cap based on output voltage and current. Problem is, it will vary as it's going to be a bench supply. I haven't had much time lately to test some things, so I don't know how well this will work. However, I have to come up with a satisfactory solution.