Hello,

I recently designed a buck-boost converter within LTSpice using switching MOSFETs but now need to develop a control system for an averaged model. I think I have successfully ported the model from a switched circuit to an averaged circuit as shown below:

Control 1 is a voltage signal which sets a "duty cycle" for the averaged model. Approximately 0.284 to reduce voltage from 12.6 to 5 (with losses showing up) and 0.676 to increase voltage from 9.6 to 20 (again with losses).

My first attempt at a control system does not work as anticipated - I wanted to use a voltage divider off the output and compare to an altered control signal. So I quickly sketched the circuit below and changed control1 to reflect the desired voltage (0.5 and 2 volts respectively for the voltage divider)

Obviously, there is no compensator network or the like - I just wanted to make sure the circuit was operating as expected before moving on to more intricate design. Unfortunately, this circuit doesn't do much of anything to regulate the performance of the circuit.

I have attached the files I used below (CCM2 averaged switch model included) - having a lot of trouble with starting on this design so I really appreciate any tips you can give me.

For reference, here is a completed circuit which accomplishes (to some degree) what I intend to do...its just that I can't make heads or tails out of the control circuit.

Thanks!

 

Hi,

Not sure if i can help here but what do you mean by Averaged Model?
I ask because your idea of an averaged model does not seem to match the averaged models we usually see which have no switches in them.

Starting with a buck circuit for example, the averaged model looks like a linear voltage regulator. It's a good place to start so have you looked at this?

 

Hi,

Not sure if i can help here but what do you mean by Averaged Model?
I ask because your idea of an averaged model does not seem to match the averaged models we usually see which have no switches in them.

Starting with a buck circuit for example, the averaged model looks like a linear voltage regulator. It's a good place to start so have you looked at this?

Hello MrAl,

Thanks for the reply - the averaged model in this case is looking at the average values of circuit parameters (inductor current, capacitor voltages, etc...) rather than instantaneous values - the control system should work basically the same as for a switched model. So in this case the MOSFET/diode of a standard Buck/Boost converter is simply replaced by the CCM2 assembly in this case which uses averaged values over a switching period. (Hopefully that helps explain the what here)

Basically the first image shows the working circuit with no closed-loop control (again, the CCM2 assembly can be considered equivalent to the standard MOSFETs/Diodes) - so the goal here is to design a closed loop control system around the converter.

Looking into a linear voltage regulator it's similar to what we are going for I think - with the difference being that we are trying to convert a voltage to a set value rather than maintain voltage at a set value along with a closed loop control which seems more active than a linear voltage regulator (though I could certainly be wrong).

 

Hello MrAl,

Thanks for the reply - the averaged model in this case is looking at the average values of circuit parameters (inductor current, capacitor voltages, etc...) rather than instantaneous values - the control system should work basically the same as for a switched model. So in this case the MOSFET/diode of a standard Buck/Boost converter is simply replaced by the CCM2 assembly in this case which uses averaged values over a switching period. (Hopefully that helps explain the what here)

Basically the first image shows the working circuit with no closed-loop control (again, the CCM2 assembly can be considered equivalent to the standard MOSFETs/Diodes) - so the goal here is to design a closed loop control system around the converter.

Looking into a linear voltage regulator it's similar to what we are going for I think - with the difference being that we are trying to convert a voltage to a set value rather than maintain voltage at a set value along with a closed loop control which seems more active than a linear voltage regulator (though I could certainly be wrong).

Hi again,

Actually what i meant was look into a buck converter circuit that is converted into an averaged model buck circuit, not just look at a linear regulator. This transformation will help you see how to convert your model effectively i think.

The usual idea is to take the switcher and transform it into a linear acting circuit where the main analysis is concerned with the overall response and ignores the smaller switching effects like the ripple for example.

There are papers written on this.

 

Hi again,

Actually what i meant was look into a buck converter circuit that is converted into an averaged model buck circuit, not just look at a linear regulator. This transformation will help you see how to convert your model effectively i think.

The usual idea is to take the switcher and transform it into a linear acting circuit where the main analysis is concerned with the overall response and ignores the smaller switching effects like the ripple for example.

There are papers written on this.

Ahh, I see what you are saying. So as far as the linearized circuit goes the CCM block models these effects as shown below in a Buck circuit I put together to try and make this task more modular (I made one for a boost circuit as well)

The ripple is removed from the inductor current and we have just a linearized model of the system which is showing the appropriate behavior up until 2ms (after 2ms the input voltage increases but the circuit still responds appropriately by lowering the voltage).

My current path forward is to design a closed loop control system for the buck and boost sections separately and then duct tape them together for lack of a better term (cascade I suppose would be more accurate). Though this approach seems somewhat dubious as the control loop for the system will be unwieldy in all likelihood and it will not be nearly as elegant as a single control loop for a non-inverting buck-boost design.

 

So I have made a decent amount of progress here and I have a "working" model shown below which can get the voltage outputs I would like (the input voltage is supposed to ramp from 9.6 to 12.6 at the middle of the simulation so that isn't some weird artifact)




But I am seeing an issue with my current sensing device as it becomes very noisy for the 20 V range as shown below:



I can tinker with the gain/current sensing resistor but I cannot find a combination where I get the results I want but do not see this strange behavior from the Vcs source. Any idea what might be happening here?