Variable Output Buck Converter

Go To Last Post
25 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I'm in the early planning phase of a new project: A variable output buck converter. I've done some research and lots of reading, but several questions remain. For now I'm researching the power section (shown below), to drive this I plan for an AVR, running at 5V, the precise chip is not yet determined.

I've tried to find examples on the web, but most of them are geared to a fixed voltage on the output, I'd like it to be variable an controlled by the AVR.

For now I have two applications in mind: One as a variable output lab power supply, the other as a freely programmable battery charger. The basic schematic and layout PCB should go to 50V/10Amps (not necessarily at the same time). I think that ripple may be a problem for the lab supply application, but in many cases it is not a problem.

For now I plan for a synchronous buck converter, I plan for synchronous because I think it gives me some more flexibility. I may replace the lower transistor with a diode in the beginning, to simplify things. If I use a 20Mhz AVR and plan for 10bit resolution my PWM will run at 20kHz, I'd like it higher, but thinks I need the 10 bit too.

Initially I'll supply power from a 20V Laptop power brick (20V, 4Amps), but I'd like to be able to make versions with higher voltage or current. This using the same PCB, just with some variations in components.

I'm interested in discussion and critique, I'd like to start with the least amount of smoke possible :-).

Markus

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

20KHz is pretty darn slow for a buck converter. That'll require some big inductors and some big capacitors. Especially considering the power that you're looking for - things will get expensive.

Take a look at the ATMEGA32M1. It can do much, much faster PWM. Really sweet part IMHO.

I'm working on a similar project - designed for high power battery charging. I'm targeting 250W output.

If you try going fast (I'm aiming for 1MHz switching frequency), you'll want a FET driver.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Call me lazy, but i'd use a controller chip to do the converter stuff and have the avr just provide the required voltage via a dac.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

nleahcim wrote:
20KHz is pretty darn slow for a buck converter. That'll require some big inductors and some big capacitors. Especially considering the power that you're looking for - things will get expensive.

Take a look at the ATMEGA32M1. It can do much, much faster PWM. Really sweet part IMHO.

I'm working on a similar project - designed for high power battery charging. I'm targeting 250W output.

If you try going fast (I'm aiming for 1MHz switching frequency), you'll want a FET driver.


I know that 20Mhz is slow, but I do want to start from some known territory and was not aware of the MegaM1 series. That power switching controller in there looks exactly like what I need. Even has an emergency shutdown for my short circuit protection. Will study the data sheet :-).

Any suggestion for the FET driver, I'm unconvinced by my driver.

Kartman wrote:
Call me lazy, but i'd use a controller chip to do the converter stuff and have the avr just provide the required voltage via a dac.

There seem to be very few dedicated buck controllers designed with variable output in mind, most of them have some restrictions. I'd like to be able to start close to zero (Volt & Amp). But I'm open for suggestions.

Markus

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Had a read of the megaM1 datasheet and the PSC appnote. It looks like the is the best AVR for my job, it can special provisions for the synchronous part of my converter. The PWM-clock can go to 64Mhz, so I can run at 60kHz (10 bit). If I need more resolution it can go to 12 bit.

Looks good, but I'm still unsure of the FET drivers.

Markus

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Happened to construct lab PSU with specs exceeding

Quote:

50V/10Amps

The configuration was as following:
buck converter after rectifier followed by a
linear high current regulator.
The tricky thing was that converter's feedback voltage was red from linear regulator Darlington. Actually that LDO linear was engineered not to exceed around few watts of power dissipation.
Worked pretty well despite was completely discrete. Low ripple, good regulation.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Also be aware of the ATtiny25 and others that can run the PWM clock at 64 MHz, and the controller at 16MHz. It also includes dead-band control.

It all starts with a mental vision.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I have a similar project in my "to do" list but i chose to use isolated push-pull topology. This way i can have multiple and expandable isolated sources derived from a 48V rail. Driving NFETs with grounded sources only seems to me being easier and cheaper.

As about your PFET driver, i don't think you will find it acceptable in practice. I did it in the same way in the past for a rapid improvement of an MC34063 (for higher voltages and currents) and the high side PFET went pretty hot. I had to improvise an accelerator with resistors, capacitors and diodes to reduce its turnon and turnoff time to keep it fairly warm in full load. All these without having an oscilloscope or many components at hand.

Anyway, IMO R20 and R21 are too high for reasonable turnon/turnoff times. I would reduce them both to 100 ohms and put a small capacitor across R21 if i have to stay with this schematic.

Edit:
I would also use an LM339 instead Q3. This may help you reduce power losses in R25-R27 and may be usefully for programmable current limiting.

Dor

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Kas wrote:
Happened to construct lab PSU with specs exceeding
Quote:

50V/10Amps

The configuration was as following:
buck converter after rectifier followed by a linear high current regulator.
The tricky thing was that converter's feedback voltage was red from linear regulator Darlington. Actually that LDO linear was engineered not to exceed around few watts of power dissipation.
Worked pretty well despite was completely discrete. Low ripple, good regulation.

That's what I have in mind in the back of my head. But this will be an extension for the future, (version 2). MCU control is important to me, so I don't want to stay digital. But, you are right, if you need low ripple then a second, linear stage is probably unavoidable.

Are you willing to share your schematics ?

Kas wrote:
Also be aware of the ATtiny25 and others that can run the PWM clock at 64 MHz, and the controller at 16MHz. It also includes dead-band control.

Yes, now that you mention it I remember having seen that. But, after reading about the mega-M1 types and the built-in PSC I think I have found love :-).

Kas wrote:
As about your PFET driver, i don't think you will find it acceptable in practice.
...
Anyway, IMO R20 and R21 are too high for reasonable turnon/turnoff times. I would reduce them both to 100 ohms and put a small capacitor across R21 if i have to stay with this schematic.

I'm not attached in any way to my driver circuit. I mistrust it exactly where you said: I suspect it is too slow. A good driver chip/or circuit is missing. Any suggestions ?

Markus

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ltdor wrote:
I have a similar project in my "to do" list but i chose to use isolated push-pull topology. This way i can have multiple and expandable isolated sources derived from a 48V rail. Driving NFETs with grounded sources only seems to me being easier and cheaper.

That is pretty much my next project. But instead of a push-pull, i'll be using a forward converter, in a open loop, followed by a linear regulator. Everything will be powered by a 24-48V (still haven't decided) power supply like these ones.

Felipe Maimon

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

markus_b wrote:

Any suggestions ?

I would use a NFET for high side and i drive it through a small impulse transformer made on a ferrite ring as a real freak i am even if buck topology it's not freaky enough as SEPIC/Cuk.
If freakin' doubts burdens me, i would use a low/high side driver as IR2101 and many others. The last option i think is the most handy.

fmaimon wrote:

That is pretty much my next project. But instead of a push-pull, i'll be using a forward converter, in a open loop, followed by a linear regulator.

Time for a contest ?

Dor

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ltdor wrote:
If freakin' doubts burdens me, i would use a low/high side driver as IR2101 and many others. The last option i think is the most handy.

After lots of looking I found interesting looking drivers, like the LM5101, similar to the IR2101, but does not need 15V supplied and is quite a bit faster.

After collecting all this freak wisdom now off to the drawing board to create a complete schematic.

Markus

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

markus_b wrote:
ltdor wrote:
If freakin' doubts burdens me, i would use a low/high side driver as IR2101 and many others. The last option i think is the most handy.

After lots of looking I found interesting looking drivers, like the LM5101, similar to the IR2101, but does not need 15V supplied and is quite a bit faster.

After collecting all this freak wisdom now off to the drawing board to create a complete schematic.


The LM5101 is what I plan on using. That part has fantastic specifications.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

@ Markus
Anytime, sir. Just basic layout, since contemporary elements shall be put in use.

[/img]

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Not quite sure what you intend to do with P-OUT-2. I don't think that it will function as the notes suggest. Here is my logic:

1) You note by Q3 "Short circuit protection"

2) For Q3 to assist in short circuit protection, the collector needs to draw current when the output has been loaded to a voltage below its expected voltage.

3) It looks like R25, R26, and R27 are an attempt to do a real-time peak current measurement. However, you loose a lot of filtering capability by adding those resistors. You typically look for filter caps with ESRs on the order of 10mohm. I think that it would be challenging to get enough base drive.

Jim

 

Until Black Lives Matter, we do not have "All Lives Matter"!

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Jim , the basic idea was as this:

- R25/26/27 are the current measurement shunt. I'll measure the voltage at P-Out-2 (not drawn) to calculate the output current (from P-OUT-1 to P-OUT-2).

- If I size these such that the max current provides a voltage drop of 0.6V then the BJT will start conducting above that current and cut off the PWM to the high-side FET.

In any case, I'm in the process of redesigning that entire part now. The mega-M1 has an analog comparator input providing an shut-off function. I'll use that instead.

Markus

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

nleahcim wrote:
The LM5101 is what I plan on using. That part has fantastic specifications.

Yes, the specs are good and the price is similar to the IR2101. But I find the data sheet is not anwering all questions. For example it shows bypass caps, but omits to recommend a capacity. I would like to see a sample application with all parts fully defined.

Markus

Markus

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I found unter recommanded operation condition
of LM5101 that VDD shoould be between 9 and 15 Volts.
So it seems that 5V operation is not recommanded.

5V Driver would probably mean that you have to use
logik-level mosfets.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

After more study of the datasheet I agree with you. The chip wants a 12V power supply. Well, that will be another LDO. If I use a LET driver the I'll go for non-logic level FETs, the same type for low and high side. I suspect I can get better specs and may be lower prices if I forgo logic level. The driver will drive them for me :-) and is not free, so why not recover cost somewhere.

I've ordered samples and asked NS tech support about the capacity required/recommended of the two bypass capacitors. This information is either omitted or well hidden in the data sheet.

Markus

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

For bootstrap cap see page 7 under "bootstrap capacitor":

http://datasheets.maxim-ic.com/e...

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Ossi: I'm talking about the National Semiconductor LN5100A, the Maxim chip you link to is quite different.

http://www.national.com/ds/LM/LM5100A.pdf

Markus

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

But I think the problem how to dimension
the bootstrap cap is the same.
ossi (born in Sarnen, Switzerland :lol: )

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I have to admit that I stopped reading on page one, after seeing that the Maxim is different. But you are right, the advice they give may be universal:

Quote:
Choose a capacitor value at least 20 times greater than the total gate capacitance of the MOSFET being switched. Use a low-ESR ceramic capacitor (typically a minimum 0.1FF is needed).

At least I have now a fall-back plan, if I get no usable response from National Semi.

I'm born in Burgdorf, Switzerland, grew up for 8 years in Ascona, Switzerland too, but Italian speaking, then back North, got my engineering degree in Burgdorf. Later I moved to Geneva and now I'm married to a French-only speaking wife...

Markus

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

If i dont missing something,a classic buck converter does not need a second transistor like Q6,so can be removed.
Needs also the output a resistor divider network continuously compared voltage ratio with a reference voltage for the voltage control of the output.
Optionally if current control is needed shunt resistor can be inserted,like this in the circuit made from resistors R25-R27 but they can act only as a maximum current limiter.

The schematic of the small picture i think will never work for a number of reasons.No startup.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

@geoelec: Yes, the classic buck converter has a transistor and a diode. I plan for a 'synchronuous' buck converter where you replace the diode with second transistor. A second transistor has less loss than a diode so your efficiency is better. Also, I believe that this buys me some flexibility for a wider range of conversion.

Markus