Electronics > Projects, Designs, and Technical Stuff
Buck converter, ringing issue
(1/3) > >>
dzseki:
I have designed a complex temperature controller system. The output stage (hetar driver) was a result of compromise: the output power had to be rated 10A@30V, but because of other constraits this power supply has to be lowest noise possible. Because of size and practical consern a PWM regulator (Buck converter) topology was chosen, with a rather heavy output filtering.
Since this was  intended as a somewhat universal controller, various loads are expected on the output. Furthermore by the nature of the system the duty cycle can vary between 0-0.95.
It is clear from the above that it is inevitable for the circuit to operate in both continuous and discontinuous modes.

The circuit uses a P-FET as switching element driven by an LM5134 that sits on a „tracking ground” that follows  Vcc by -5V. The output filtering can be seen on the schematic excerpt, it is tuned for ~200Hz cut off with a damping factor of 1.


In discontinuous mode I see some strange behaviour from the circuit. Articles -at least the ones I looked up- did not explained the discontinuous mode in much detail, other than basic equations.

In the given setup, with an output load of 4 Ohm and supply voltage of 18V the operation switches over between continuous and discontinuous operation at about 0.4 duty cycle.
So the problem is in discontinuous mode before the FET would conduct there is some sort of ringing happening on the drain of Q4.
Playing with the PWM repetition frequency: lowering the frequency will make the ringing more severe, higher PWM frequency shifts the switching point towards the lower duty cycles -as expected.

Below is an oscillogram taken in discontinuous mode (with duty cycle: 0.2), lower trace is the drain of Q4, upper trace is the output voltage ripple (AC coupled).
The output is not affected much by this strange behaviour as I see, but I am also concerned about the radiated EMI. Also I am a little bit confused what is happening here exactly. To my understanding in discontinuous mode neither the FET nor the freewheeling diodes carry currents, therefore nor the coil,  but what is oscillating then and why? As I saw adding more capacitance over eg. D9 would definitely detune the frequency of the oscillation, but the energy involved here is clearly larger than it would be practical to dampen with a simple RC snubber, perhaps adding active switch (FET) instead of the freewheeling diode would help to cut the ringing, but I would still like to know why is this happening?
David Hess:
The inductor forms a series resonate tank circuit with the parasitic capacitance of the power MOSFET and diode.  It is completely normal when operating in discontinuous mode.

Check out the MAX1722 datasheet for a little discussion about this.  It seems like if you used a synchronous rectifier instead of the diodes, that it would clamp the ringing.

Benta:
What you are seeing is 100% normal for a buck converter in discontinuous mode.
As no DC current is flowing in the choke at some point (this is where the oscillation starts), both power switch and free wheeling diode are off.
This isolates the choke DC-wise.
The result is, that the choke is a parallel tank circuit comprising inductance and inter-winding capacitance. There is basically nothing you can do about it. Reducing inter-winding capacitance will simply increase ringing frequency.

The upside is, that the ringing has low energy, so it won't radiate a lot. And the ringing is only present at the input side of the choke, the output is kept steady by the output cap.


T3sl4co1l:
C15 / R25-R26 would seem to be the wrong value.  Or they are there for high frequency (loop inductance + switch capacitance) damping.

You can add another R+C to dampen that, same C, different R of course (R ~= sqrt(L / C)), at the expense of more switching loss of course.

I don't usually mind that, because it's lower frequency and not prone to radiating.  The energy stored is either dissipated (when the transistor switches on during a low peak) or saved (during a high peak), but it averages out either way.

Beats me what the hell the resistors are doing there -- the FET is 5mohm yet there's 80mohm in series after it.  And the diodes drop ~0.6V at most load currents.  The efficiency optimization seems really off balance here.

So, the combination of an unusually "stiff" (i.e., low impedance, because Z ~ sqrt(L/C) and C is large) ringing on the switch node, and an excessively low Rds(on) transistor, seems to suggest that the component choice is very well past peak efficiency.  The diode, resistors (and inductor value, to some extent) imply a much lower operating current than the transistor ("70A"!) does.  Or if high current and efficiency are required, then this needs to be a synchronous switcher.

Mind that PMOS have 2.5x worse performance than NMOS, so there is further savings in not just using a modestly sized transistor, but using a bootstrap gate driver and NMOS type as well. :-+

Tim
dzseki:

--- Quote from: T3sl4co1l on December 14, 2018, 06:14:57 pm ---C15 / R25-R26 would seem to be the wrong value.  Or they are there for high frequency (loop inductance + switch capacitance) damping.

You can add another R+C to dampen that, same C, different R of course (R ~= sqrt(L / C)), at the expense of more switching loss of course.

I don't usually mind that, because it's lower frequency and not prone to radiating.  The energy stored is either dissipated (when the transistor switches on during a low peak) or saved (during a high peak), but it averages out either way.

Beats me what the hell the resistors are doing there -- the FET is 5mohm yet there's 80mohm in series after it.  And the diodes drop ~0.6V at most load currents.  The efficiency optimization seems really off balance here.

So, the combination of an unusually "stiff" (i.e., low impedance, because Z ~ sqrt(L/C) and C is large) ringing on the switch node, and an excessively low Rds(on) transistor, seems to suggest that the component choice is very well past peak efficiency.  The diode, resistors (and inductor value, to some extent) imply a much lower operating current than the transistor ("70A"!) does.  Or if high current and efficiency are required, then this needs to be a synchronous switcher.

Mind that PMOS have 2.5x worse performance than NMOS, so there is further savings in not just using a modestly sized transistor, but using a bootstrap gate driver and NMOS type as well. :-+

Tim

--- End quote ---

This circuit is the second generation of its kind. R25-6 and C15 is for high frequency damping, their value was determined with the help of a Richtek application note, and I am satisfied with those results.

The original circuit was based on the LT1158 synchronous NFET driver, which had other (more serious) issues, and after all I was only able to use it reliably when I gave up with the synchronous mode, so I was thinking to try something new with the second generation.

The resistors are there to form a critically damped filter on the output, without them the filter network would have a Q of 4.

Again the first generation of this circuit used even heavier output filtering: 100uH coil with 100 000uF capacitance, with that setup the cutoff frequency was 50Hz, and much lower resistance was needed for critical dampening, but a 100uH inductor that is rated for 10A and the 10pcs 10000uF capacitors took some PCB space, it was overkill anyway.

Efficiency was not in the first place but low noise level was, it is clear that these power levels are not suiting for linear regulators (even when combined with buck preregulator). The circuit might not be super efficient, but I think it is still decent for what it is.

In my philosophy the FET in the circuit is for switching an not for its series resistance, If I want something to drop voltage on I put there a resistor, so I don't have to care about the cooling of the FEt to maintain performance, while the resistor have no problem with heat :)
Navigation
Message Index
Next page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod