EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: BennVenn on July 14, 2015, 10:46:39 pm
-
I watched Dave's xantex 300v 4A power supply teardown last night and had a look at the datasheet for the quasi-resonant controller. Reading through the application note it had a few requirements for successful quasi-resonant full bridge drive. One of these was the requirement for the resonant frequency to be at least 4 times the switching frequency.
Why is this?
If the resonant freq was 4x the switching freq, wouldn’t this give only 4 options for zero voltage crossing? Or is this how it works and power is switched via phase imbalance between the two half bridges?
0-300volts is pretty impressive in a single stage converter. Also noticed no power factor correction?
Looking at prototyping a small 200w converter using this topology.
-
The switching transition time would be presumably one quarter cycle of the resonant frequency and if this resonant frequency is too low it would intrude a lot into the total cycle time instead of holding the switched node high or low and delivering output power.
-
Doesnt the LC circuit need to swing back to the rail voltage before the fet switches to effectively switch at zero voltage? Or am I way off
-
https://youtu.be/27c4RTntAPw?t=15m21s (https://youtu.be/27c4RTntAPw?t=15m21s)
http://www.ti.com/lit/ds/symlink/uc1875.pdf (http://www.ti.com/lit/ds/symlink/uc1875.pdf)
http://www.ti.com/lit/an/slua107a/slua107a.pdf (http://www.ti.com/lit/an/slua107a/slua107a.pdf)
for reference.
At a glance, it's not obvious where the control is. They've got a whack of beefy output diodes, but seems to be more than just the controller and full bridge at work. There might be additional stages.
Note that the delayed commutation is simply using an extra-leaky transformer to commutate each side of the bridge; this assumes there is load current present to achieve commutation. Otherwise, it does hard switching. Which really sucks with those old ass IRFP460s in there. Piles of capacitance.
There's no trick to it, it's just setting dead time between high and low side switches.
They sure threw a whole hell of a lot of EMI filtering at that sucker. Geez. Resonant is supposed to save you some trouble, but either they went overboard just because, or they didn't get any help from it whatsoever!
Tim
-
Reading through those references now and what I am taking from it is that ZVS can only occur at specific output conditions based on duty cycle, load current etc otherwise it is hard switched?
Also that the phase shifting actually shorts the LC circuit and holds it until the next transition where it releases and the voltage swing enables ZVS?
Intersting topology, not really what I was expecting. Thought it would give a wider window than the app note suggest.
-
The "active short" condition is advantageous for PWM. A half bridge converter, or full bridge with simplified control, leaves the output alternately high and low impedance (i.e., open circuit vs. constant voltage). Which means, if the load has some ability to send energy back, say from reactive current, then it's going to have a duty cycle and average level different from what the logic signals going into the gate drivers. A phase shift modulated circuit always delivers constant voltage, so that the output is always shorted to itself (at +V or -V), or to +V, or to -V (i.e., backwards).
Note that it's also advantageous that the shorting occur alternately at +V and -V: this way, the power dissipated due to holding that circulating current is shared between high side and low side switches.
This works best for highly inductive loads, like radio transmitters (somewhat akin to class E, but in a push-pull configuration), induction heaters and resonant switching converters.
The reason it's called "quasi resonant" is because, due to the wide range of operating conditions necessary in the switching converter environment, leakage inductance must be limited (otherwise it absorbs a lot of the flux you're trying to deliver via PWM in the first place), and so the energy available for commutation will be small as well. So you're only going to get a limited swing (load current insufficient to achieve full commutation -- partial hard switching occurs), or a full swing but only for a short time before it decides to reverse and swing back.
A conventional switcher actively avoids this by keeping leakage minimal, damping it out completely (there's nowhere to hold that energy -- no shorting state -- in a half bridge), and the rectifier being choke-input means the transformer's terminal voltage is also forced to zero on the secondary side (in CCM).
The rebound looks something like:
(http://seventransistorlabs.com/tmoranwms/Elec_SG3524_Osc1.jpg)
From left:
- (previous pulse cut off on display)
- Everything open circuit (half bridge configuration)
- Low side switches on, then off
- Flyback clamped by +V rail (with ringing because breadboarded circuit)
- Inductor current falls back to zero, voltage returns to midpoint
- Opposite sequence occurs (high side switches on then off)
There was some RC damping on this, so the rebound after flyback is small. A less reactive load (like a conventional half bridge switcher) will tend to look like this, without the flyback pulses (the transformer's inductance is held shorted by the output diodes, so only the little restoring squiggle occurs). A design, like this "quasi resonant" thing in question, will tend to exhibit relatively more flyback (though not as severely as this example), which facilitates commutation if timed properly.
The "on" and "off" durations are not resonant at all, hence "quasi" -- only some small part of the waveform acts resonant. Conventional PWM timings apply.
Tim
-
Ahh well that makes a lot more sense, It's just using the (engineered) ringing for ZVS. I read a report on the efficiency of conventional full bridge vs quasi at various power levels. Primary side losses were much the same with the exception of a narrow load region with the quasi, during full power and that was only a percent or two higher.
Also looked at a fully resonant PWM driven induction cooktop schematic that utilised two op-amps, a gate driver and a pair of IGBT's. This is the standard 1800w model shipped from China. So very simple! Going to pick one up this week to see if I can replace the work coil with a bunch of stacked ferrite toroids and make the secondary resonant instead.
-
They've got a whack of beefy output diodes, but seems to be more than just the controller and full bridge at work.
Two output chokes too, so probably a current doubler rectifier.