Voltage feedback in Buck-Boost converter

[dziegiel]

Hallo everyone!

I'm facing big problem with my buck-boost converter and I hope You can help me

Main requirement: Vin 20-400 V (sic!), fixed output voltage -80 V (becouse of topology it's opposite to the ground), Pout = 25 W.
It's works, but I have problems with feedback loop. Let me shortly describe the circuit and the problem:

It's typpical buck-boost topology. The PWM 61,5 kHz is generated by Atmega32 (I've got evaluation board, in final model I'm going to use smaller one) and ampliefed by typpical mosfet driver. 0,47R+33n is a snubber. What is a big problem in Buck-Boost converters is inevitability to connect PWM's generator ground to transistor's source, so 2 "grounds" are in the circuit. In my model I use external extra socket-supply (not sure if it's proper name in English, it's kind of cell phone charger) to supply Atmega and Mosfet Driver.

Problem: At the moment I use a potentiometer at evaluation board to set PWM duty cycle to keep output voltage = -80 V for various input voltage. I want to apply some feedback, sense resisotrs (see 2 resistors, i.e. 800k + 39k at output) to automaticly sense Vout and set duty cycle, but: Atmega's ground is different than supply voltage. So my -5 V on sense output resistor is measured with respect to "big" ground. With respect to "small ground" (Atmega's ground) is around |Vin|+|Vout|, so 100 ... 480 V ! What is more it changes with respect to PWM cycle.

Now is the question: what is the method to apply feedback and sense resistors to such type of converter?

[Neilm]

Why don't you have the micro connected to ground and drive the FET with a proper driver?.

In answer to your question, you are entering the realm of digital control theory. To under stand that you need to know about Z transforms, PID algorithms and reasonably high speed DSP. For example a good digital control loop will have a high speed low resolution ADC so it will respond quickly to transients. 

If regulation is not too much of a concern you could fudge it with a slow ADC but the regulation would be terrible.

Neil

[dziegiel]

Dear Neilm,

at the moment I really don't care if A/D is slow or not, because the problem is how to "feed" this A/D converter the voltage measured by sense resistor placed on the output. Voltage on this resistor is below |5| V (with respect to ground), but with respect to A/D converter's ground it is +100...480 V, in addition changes in PWM pulse... In other words: how to avoid this huge difference between output's ground and Atmega ground?

What is unusual in this project is very wide range of input voltages. Right now 20-400 VDC. In the future also 24-380 VAC + PFC. I don't think if any driver or in general Integrated Circuit will meet such requirements

[Bored@Work]

You try to solve a minor problem, while you want to ignore your major problem. And that major problem is using a dead slow microcontroller in the control loop.

Regarding your minor problem, did you even spend a minute to study existing designs? Have you never seen an opto coupler or any other kind of isolation barrier in an SMPS design? People put them there for a reason. But since you have to ask such questions, do you think you are qualified to work with 400 V DC?

[BravoV]

Where is the control loop for regulation ? I can not spot it. 

[Kremmen]

Please stop while you are ahead and above all, alive.

I am not saying this to be rude but if you ask questions on this level it shows that you need to study the subject a lot more. Off line SMPS circuits can be tricky bastards and definitely unsafe at such voltages. What you are attempting is not a project for a novice and everything in your question spells novice.

Still, to answer your question - isolate the drive using a pulse transformer or level shifting fet driver. This is the done thing.

[Neilm]

at the moment I really don't care if A/D is slow or not, because the problem is how to "feed" this A/D converter the voltage measured by sense resistor placed on the output.

What is unusual in this project is very wide range of input voltages. Right now 20-400 VDC. In the future also 24-380 VAC + PFC. I don't think if any driver or in general Integrated Circuit will meet such requirements

Personally, I would investigate the use of an isolated fly-back topology with an opto-isolator to feedback the output voltage using a referance to switch the opto.

Re ADCs, to get decent regulation ADCs in the order of >1MHz are not uncommon.

I would suggest doing this using some off the shelf chip that will control this.

Neil