Remote sensing in buck converters

[joniengr081]

I have attached a circuit of un-regulated buck converter in which the PWM from MCU controls the gate of FET which is the switching element. I call it un-regulated because there is no feedback from the output voltage and no output regulation.

I am wondering how do we add remote sensing in switch mode buck converters to make them regulated buck converter.

I understand the concept of remote sensing in general. When the load is far away from power supplies and there is a voltage drop across the long cables due to large amount of current, we need to introduce the sense wires in linear power supplies to regulate the output of the power supply at the point of load. Such linear power supplies are called regulated linear power supplies.

How the regulation is done in buck converter is a question in this post.

I am wondering what changes in the attached circuit we need to do in order to make it regulated buck converter.

[boB]

The feedback would be applied to the control circuit that feeds that FET gate-source and its PWM duty cycle.

So, quite a bit more needed here.

There is a reason these circuits leave the gate-source drive for you to add.  This is just the basic buck converter and will need high side gate drive.

There should be quite a bit of examples online where the gate drive is actually connected to something.  That circuit is where the voltage feedback is connected to.

Some examples may be synchronous buck converters instead of this one which is non-synchronous.

boB

[Smokey]

https://webench.ti.com/power-designer/switching-regulator?powerSupply=0

Pick a topology.  Pick a part.  Pick performance specs.  Fully formed schematics (with real part numbers) and simulation results  for your learning pleasure.

[Someone]

make them regulated buck converter.

It is unlikely your MCU will have sufficient speed (and peripherals) to produce what is usually considered "regulation". All the theory was worked out in analog:
https://ridleyengineering.com/design-center-ridley-engineering/38-control.html
While that can be done entirely in the digital domain, the vast majority of regulators retain key functionality in analog parts.

[uer166]

make them regulated buck converter.

It is unlikely your MCU will have sufficient speed (and peripherals) to produce what is usually considered "regulation". All the theory was worked out in analog:
https://ridleyengineering.com/design-center-ridley-engineering/38-control.html
While that can be done entirely in the digital domain, the vast majority of regulators retain key functionality in analog parts.

That's a bit bs though innit?
You can easily do a current mode contol loops for 100s of kHz from modern MCUs not specifically designed for digital SMPS, and MHz range for those that are.

Last time I took apart a 4.5kW PSU, the PFC stages were controlled by a low-end STM32 clone, and the output LLC with a higher end Cortex-M4 based part.

[Someone]

make them regulated buck converter.

It is unlikely your MCU will have sufficient speed (and peripherals) to produce what is usually considered "regulation". All the theory was worked out in analog:
https://ridleyengineering.com/design-center-ridley-engineering/38-control.html
While that can be done entirely in the digital domain, the vast majority of regulators retain key functionality in analog parts.

That's a bit bs though innit?
You can easily do a current mode contol loops for 100s of kHz from modern MCUs not specifically designed for digital SMPS, and MHz range for those that are.

Last time I took apart a 4.5kW PSU, the PFC stages were controlled by a low-end STM32 clone, and the output LLC with a higher end Cortex-M4 based part.

Its an opinion. "low power" switching regulators are moving to MHz with corner frequencies in the x00KHz. What goes on inside a typical 10c-$2 switching regulator is very high speed compared to a microcontroller and much finer resolution. Analog has its place in these closed loop modulators be that voltage or current mode.

You think the OP is really talking about an industrial kW converter with xx00V and xx0A switching devices? Just the gate drive would be its own work in analog design.

The basic statement stands, people developing digital solutions start with the analog models of what works. It's not possible to confirm stability without including/analysing the power parts the OP has started with as a "given".

[uer166]

OP wants an MCU controlled buck, while that is not optimal unless it's something special, why mislead them into thinking that it's not possible at all?

[Someone]

OP wants an MCU controlled buck, while that is not optimal unless it's something special, why mislead them into thinking that it's not possible at all?

Possible, but likely a wild misunderstanding. Trying to blindly implement "remote sensing" without starting with an understanding of control theory, which is well developed for switching regulators, is foolish/shortsighted.

How do you propose to help the OP develop the skills/knowledge to implement a switching regulator with an MCU? Completely ignore the established theory?

[Siwastaja]

Well, if one wants to understand how a switch mode controller's control works, they surely would learn more from doing that on a microcontroller, than just copying an appnote of a controller IC. Is it going to be easy? Probably not, but that depends more on the general microcontroller programming experience of the OP than anything else. In fact, if one's head starts to spin with control theory and poles and zeroes, it's possible they grok the whole idea much better in time domain as PID controller running on a periodic interrupt.

And I disagree about the claim that the MCU likely can't do it. Almost any MCU can, for example even the 8-bit ATTiny controllers have analog comparator which can be used to implement peak current control running at many hundreds of kHz, and voltage control loop is fine at lower BW (e.g. in timer interrupts) because output capacitance limits dV/dt anyway.

The crucial key piece between a poor-performance, randomly blowing up converter vs. a robust, easy to compensate, fast one is availability of current sensing (and of course, fast reaction to this signal). And this is well possible with any microcontroller with analog comparator peripheral. Many can be internally routed so that this signal terminates PWM cycle, you don't need to go into top-end microcontrollers for this.

And, if you are ready to invest $3 to the microcontroller, you can get stuff like STM32F334 which are specifically tailored into building complex and high performance power converters. GaN Buck operating at say 2MHz would be easy-peasy. Software-controlled switch mode converters are nothing special in 2010's, let alone 2020's. The primary advantage from a dedicated (mostly analog) IC is cost reduction in mass market products, but if OP wants to play around and learn, this is irrelevant.

[joniengr081]

Thanks for replies. I understand that the circuit shown in the first post is un-regulated and non synchronous. I am still wondering how we can make it regulated output buck converter. In case the load is couple of meters away and we need to have extra cables for sense + and sense -, where we can attach these wires in the buck converter coming from the remote load. 

[T3sl4co1l]

I'm not sure which diagram you're referring to -- the attached diagram shows an always-on output and no gate drive at all.

Notice the MOSFET is P-channel and the substrate diode is forward-biased.  No PWM source or MCU is shown at all.

The diagram seems related to the text, but clearly isn't what the text is talking about.  Perhaps you attached the wrong figure?

...I don't say this to be a pedantic prick (...though I sometimes may be), but to make a point -- technical matters require clear and comprehensive communication.

You can talk about something and hope that the other person gets what you're talking about, but, especially when you yourself don't know a subject very well (i.e., as a novice), that assumption doesn't work well going in the opposite direction: we have to assume you know what you're talking about, but a novice, by definition, doesn't.  You might be talking about something very different -- or making otherwise-basic mistakes, that need to be addressed; like the mistakes in the schematic.

Tim

[mtwieg]

The implementation of remote sensing isn't really dependent on whether the supply is switchmode or linear, or whether it's controlled by an MCU or a purely analog circuit.

Typically instead of sampling Vout with a resistor divider, you use a difference amplifier circuit which connects to the remote load with its own wires.

One pitfall with remote sensing is that if one or both remote sense wires become disconnected accidentally, this may cause Vout to go wildly out of control (either way too high or too low). One way to prevent this is to also locally connect the difference amp to Vout via some resistance (see Rs in the attached figure). Rs should be much higher resistance than the remote sensing wires (so that when the sense leads are connected properly you only sense the remote Vout) but also much lower than the input resistance of the difference amp (so the difference amp doesn't grossly underestimate the local Vout).