SMPS design

[alexgross81]

hi, I want to design an adjustable switched mode power supply beginning from zero volts. any hints?

 

[mneary]

Define your objectives. Zero is a difficult starting voltage, as is zero current.

 

[alexgross81]

dear mneary

thank you for your reply. I know also that zero is a difficult voltage, I have designed and also constructed a variable voltage spms with 3 to 72 volts and 0.5 to 12 amperes output but i need a 0-72V output.

I thought to design a linear voltage regulator for under 3 volts output and apply a switching mechanism for connecting the linear and the switching regulators around the 3 volts border.

any further hints?

 

[dougy83]

While I'm not sure of how you've implemented your SMPS, I'm guessing that there's a pot on the output voltage connected to a SMPS IC feedback input; the SMPS IC feedback voltage (when output is regulated) being a nominal 3V.

If this is the case, could you not use an summer or amplifier in the feedback path, adding 3V to the feedback signal? This (in my shallow thinking) should allow the output voltage to be adjusted to 0V I would think.

 

[Hero999]

Is this for a lab power supply?

I would suggest using a hybrid regulator combining a switching regulator and LDO regulator to get the benefits of both.

The idea is you build a switching regulator with a voltage range of say 2 to 17V then connect a LDO linear regulator to the output. The linear regulator tracks about 1V to 2V under the voltage of the switcher and can go down to 0V. The advatage is you get the precision, low noise, fast transient response of a linear regulator with the efficiency and compactness of a switching regulator.

Obviously a hybrid regulator isn't quite as efficient as a switching regulator on its own and it's much more complicated but it's a good design solution for a high quality efficient power supply.

 

[dougy83]

Are there LDOs that track the input voltage? And which LDOs go down to 0V output (in regulation)?

The LDO would be nice though as it also filters out a lot of the switching noise and ripple.

 

[Diver300]

Most adjustable SMPS ICs have a feedback input that looks for 1.25 V and it is expected that there will be a potential divider connecting the output of the power supply to that input.

For instance, if 5 V is required, the potential divider will have a ratio of about 0.25, so that when the output is at 5 V the feedback input will be 1.25 V and it all works fine. If the output voltage dips below 5 V, the feedback input will dip below 1.25 V and the SMPS will increase the output current to bring the output voltage up.

If you want to use one of these power supply ICs to work down to 0 V you need to provide 1.25 V to the feedback input when the output voltage is 0 V. The easy way to do that is to connect the feedback input to 3 resistors, instead of 2. The first two form a potential divider as before, so one is connected from the feedback input to ground and the other from the output to the feedback input.

The third resistor goes from the feedback input to a variable voltage source, say 0 - 5 V, that only has to provide a few mA. When the source is at a maximum, the output voltage will be smallest.

Careful choice of the resistors will give control of the output voltage over the range you want, right down to 0 V and up to the maximum rating of your components. Control will be linear until something limits.

 

[Hero999]

Are there LDOs that track the input voltage? And which LDOs go down to 0V output (in regulation)?

No, you'll have to design your own LDO that does that.

Most adjustable SMPS ICs have a feedback input that looks for 1.25 V and it is expected that there will be a potential divider connecting the output of the power supply to that input.

You could bias the 0V node at -1.25V, here's an idea using the LM317 and I don't see why it couldn't be adapted to a switcher.

0 to 13.8V LM317 Power Supply

 

[bountyhunter]

hi, I want to design an adjustable switched mode power supply beginning from zero volts. any hints?

I spent about 20 years designing power supplies and that is just about typical for the specifications we would get........

 

[marcbarker]

Get a 'Class D' Audio Amplifier and modify it to become a Variable Power Supply. As a bonus, it'll be capable of outputting negative voltage as well.

 

[Mr RB]

Get a 'Class D' Audio Amplifier and modify it to become a Variable Power Supply. As a bonus, it'll be capable of outputting negative voltage as well.

That's pure genius! The output should regulator to match the input reference voltage, provided there is no external connection from output to input (since output is probably bridged and one side needs to be output gnd).

On a different tact, I was messing some time back with a 2 transistor black regulator but using a rail to rail comparator instead of the sensing transistor so it could work as a variable voltage regulator. It worked fine from 10v output, right down to zero volts output, provided a small resistor was put in the output filter to keep duty cycle above a couple of percent. Stability was also good over a pretty large current range.

I don't have my orig schematic but this is the basic format;

 

[indulis]

Check out the book by Keith Billings... there is even a sample design of a off-line adjustable SMPS.

 

[xtrmi]

Mostly clueless about SMPS, but I think this is the next project I want to do. So far I've read about 20 pages of a reference guide posted on this forum and have a question about the basic idea behind the SMPS. Does the SMPS use the transient characteristics of the capacitors and inductors to maintain a certain voltage and current. For example once the capacitor reaches a desired voltage it sources a bit then its resupplied when the switch comes back on a very short time after. So the fast switching would be quickly propping up the voltage every time it comes back on. Same thing would happen with the inductor. I have noticed that inductors and caps are used and wonder if this is how it works.

 

[Diver300]

https://upload.wikimedia.org/wikipedia/commons/5/52/Buck_operating.svg

That is a schematic for just about the simplest SMPS, the buck regulator.

The capacitor value is not important as long as it is large enough. Bigger is always better.

The energy storage device that actually does the work is the inductor. The important thing to understand about the inductor is that the current in the inductor only changes relatively slowly. When the switch closes, current builds up in the inductor, but it only does so slowly, so the switch can open before the current is too big.

When the switch opens, the inductor current cannot change instantly. Initially, it is still at the maximum. The only place current can come from is through the diode. The current will fall, relatively slowly, but continue to supply the load, until the current reaches zero or the switch closes again.

That's about all there is to a buck regulator. The load is supplied some of the time from the supply, and some of the time from the diode, so the output current is more than the input current. The output voltage is, of course, always lower than the input voltage.

The rest is maths, and physical limits to what currents and voltage the components can stand, and imperfections because real components don't work quite like ideal ones.

 

[xtrmi]

Hmmm Thanks man. Let me digest this and I may have another question or two.

 

[indulis]

What do you call

relatively slowly

?

VRM's, which are buck converters, have VERY HIGH current slew rates.

 

[bountyhunter]

What do you call ?

VRM's, which are buck converters, have VERY HIGH current slew rates.

The slew rate of the current in the inductor is directly proportional to:

1) the amount of inductance

2) the voltage across it when the FET is ON and OFF.

He s right that in most bucks, a slower slew rate is good because it allows continuous operation where the inductor current does not drop to zero which gives maximum energy delivery.

There are switching converters operated in discontinuous modes for various reasons, the traditional buck usually is not run that way at rated output current.

 

[Diver300]

What do you call ?

VRM's, which are buck converters, have VERY HIGH current slew rates.

Well any buck converter has to switch much faster than the current can change in the inductor. That was where the relatively comes from. To understand a buck converter on a basic level, it's good enough to assume that the switch opens or closes instantaneously, while the inductor current changes slower than anything else in the circuit.

When you say that VRMs have high current slew rates, can you put a figure on that?

If the load current changes quickly, the capacitors have to support that current until the converter can get going. That will be on a similar time scale to the rate of change of current in the inductor.

I avoided quoting any numbers in my post, as I was trying to explain general principles, rather than scaring off people with what may look like big numbers.

I work with some buck converters that go from 0 - 2 A in less than 50 µs. The inductor is 10 µH, and the rate of change is up to 2 A/µs with a 24 V supply. However, on a microsecond time scale, a 100 µF capacitor really won't change much so the current from the buck converter can ramp up much more slowly than the inductor current can change without getting a significant voltage dip.

I don't know the switching time as it isn't on the data sheet, but it's less than 50 ns, so far less than the about 1 µs it takes for the current to get to the maximum.

 

[indulis]

How high of a slew rate...

Designing high-current, VRM-compliant CPU power supplies - 10/26/2000 - EDN


At these levels you become "current slew limited" and the inductor is made intentionaly small (but with a high switching frequency). Output capacitors are very low ESR... i.e. Oscon capacitors. The crossover frequency is obviously high and synchronous rectification is a must. The bulk of the energy is stored in the inductor... then the trick becomes how much ripple current you can tolerate.