Hi,
it is pretty difficult to find something useful about a buck-fed push-pull design, so below is my LTspice simulation of the core circuitry. I hope it will be useful for somebody. Any improvements are highly welcome.
Best regards, Piotr
-----------------------8
Thanks Piotr, that is my favorite isolated topology, very flexible and the overlap conduction can help the rectifiers.
piglet
Is it [difficult]? You may find more info with "current fed inverter".
It's a few transformations away from the oft-seen half bridge forward converter (choke input rectifier, current mode control).
Circa 2012, I built one of these, very much like what you're doing, actually. Twas for a gate driver array, to upgrade an IGBT assembly. Think it was 100W total input, 8 channels, +/-15V outputs. 300kHz sync buck,
200kHz chopper, and a bunch of planar transformers all wired in parallel to the chopper. My first design with planar, in fact; fortunately it wasn't a disaster!* Hmm, can't remember what the buck controller was, if it was IC or discrete. Chopper was just driven from a multivibrator (CD4047??) and drivers. Full bridge.*I do recall going through a revision or two. First proto was 2-layer, which got way too hot. Simple fix, double up the layers. Components ran much cooler too, of course. Relatively big board (think it was something like 3 x 12"), but who cares, PCB is cheap.
Oh, or maybe it wasn't synchronous, also I remember observing that one diode gave higher losses as schottky versus PN type, but I can't remember if it was the buck diode (if applicable) or just a snubber diode. Want to say snubber. Which would've been to clamp peak voltage on the chopper, which is developed by leakage/strays through the transformers.
Note that leakage inductance is exactly equivalent to capacitance in the voltage-fed forward converter; but we almost never have to worry about that, because SMPSs are usually Zsw < Zo, i.e., the transformer looks inductive. Flipping the topology (current-fed) also flips that condition.
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website: https://www.seventransistorlabs.com/
For me it was. There are literally two controllers (LM5041 and UC1827) that can do that, both with unusually crappy documentation. Most buck chips I had some experience with went to great lengths in order to explain the obvious. Here you need some inductor and some transformer. But what the heck is "some"? I am used to designing voltage-mode converters and this one required a significant perspective shift. Hence the sim, now it is much clearer.
UPS has some delivery problems, no prototyping today... :-/
Are you sure we refer to the same topology? The chopper needs to run at a half of the buck frequency, the math doesn't work. Did you have a cap after the buck?
That's exactly what I need. Stealing some RF AC with a relay would significantly simplify another part of the device. Even more windings, this is creepy. :->
[snip, thanks] -- a cumulative appreciation for your continual contribution, Tim, I have learnt a lot that way. :-)Best regards, Piotr
To someone with your experience the Abraham Pressman book is probably old hat but he does have a good analysis of buck fed current p-p.
piglet
Bingo, thank you! Many years ago I had read a great book with a significant part devoted to current-fed topologies, but I was unable to recall the author when badly needed that information. Pressman, indeed! At that time they looked bizare to me, but they no longer are.
My only prior experience with current-fed converters were some experiments with the Baxandall resonant push-pull topology. I was impressed by how clean and powerful the signal was, relatively to its simplicity. A really decent 2W pump for small-signal magamps. But then the IR21531 with its output drivers abused in a way to form a full-bridge converter powered from a 12V rail turned out to be even simpler, so the Baxandall went to the "curiosities" bin. Hence we can safely assume this is my first time "for real".
Best regards, Piotr
Keep the perspective shifting! :) I would look for a,
- current mode controller, with
- sync output,
- forced CCM option,
- transconductance error amp,
- where the amp output is the current setpoint.
Treat it as a mere transconductance amplifier. Voltage at the setpoint node is current at the output. Voila, current-feeding buck!
Don't need to know anything about the chopper really, just as you don't need to know anything about the FWB at the secondary side of the conventional topology. V and I are constrained by the supply, no need for local limiting. The usual commutation dynamics apply -- considerations of rise/fall time, overlap (deadtime in a current-fed inverter must be overlapping), stray L and C.
It can. I went for ~orthogonal frequency to avoid oddness. Beat frequency is ~100kHz, still above the control loop bandwidth, *shrug*. 2nd harmonic is preferable (by symmetry*), I may've avoided that because 400kHz was pushing it for the parts I was using.
Also used relatively a lot of inductance, a Bourns 2200LL I think. So the ripple was small.
*Assuming no subharmonic instability, which can still happen in average current mode controllers. Evaluate carefully.Might as well throw in some of those saturating-current-transformer-commutated synchronous rectifiers. Or do the whole thing in mag amps. Or something. :^)
Cheers!
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website: https://www.seventransistorlabs.com/
It is a useful book - I learnt the boost-fed-push-pull aka flyback-fed-p-p aka weinberg converter from that book - another sometimes handy topology to have under the belt.
piglet
A cap after L1 makes it into a voltage-fed forward converter. That looks nicer.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
What are the advantages of adding that cap?
piglet
One more follow-up question.
(The only) diagram at page 14: what is that current transformer T1 supposed to do, if the buck around L2 most likely works in CCM? How can that arrangement provide any decent level of overcurrent protection, if majority of the choke current is DC?
Best regards, Piotr
Never mind, T1 is before the buck switch. Hard night, sorry.
Best regards, Piotr
The waveform at the CT of the main transformer looks sane with a cap to ground. It's awful if that node is inductor-fed.
With the cap, you have a voltage-output buck switcher feeding a conventional forward converter. The control loop seems to still work.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Does the user care about an awful waveform on one node? If the p-p is voltage fed you have to worry about flux imbalance walking the core into saturation. Plus the overlapping conduction required for current fed should make snubbing easier.
piglet
Jim Williams did a lot of great work on the current fed royer converter:
Instead of the OP schematics with flip flops and what not, that is replaced with 2 transistors. Additional benefit is that adding a cap on the winding can let it run in resonance for ZVS (low loss), and operation of the royer has inherent short circuit protection (Pressman has good explanation is hi s book)
Downside is royer needs extra winding,
Royer Jensen vil use the core better for lower loss in core and transistors , but adds a current sense transformer that is used to force saturation on a small transformer to lower losses.
Cheers
Klaus
OK, a prototype controller based on LM5041 works correctly, all the gate driving signals exactly match the datasheet. Both feedback and current sense are shunted to GND for now, the POC is to show the thingie is alive. And it is, trying to drive the buck part at 250kHz.
Tomorrow I'll try to connect some MOSFETs. The buck inductor is 119uH, wound on an MP-092160-2 MPP core. Way too good for this CCM application, but I had some in my stock, so shrug. It is supposed to have 49uH at 6A and 27uH at 10A. I like this wide swing choke.
Best regards, Piotr
/an65fa.pdf
It's just the Baxandall Class-D oscillator
The paper was published in 1960 in an obscure British journal and it has be en said that Jim Williams got the circuit (but not the literature reference ) from the UK.
Jim Williams certainly ran with it.
I've user it, and my own low distortion variation of it.
ophia-electronica.com/Baxandall_parallel-resonant_Class-D_oscillator1.htm
ed with 2 transistors. Additional benefit is that adding a cap on the windi ng can let it run in resonance for ZVS (low loss), and operation of the roy er has inherent short circuit protection (Pressman has good explanation is his book).
But if it is resonant, it really isn't a Royer inverter (which goes back to 1954).
The great advantage of the Baxandall resonant inverter is that the switchin g transistors switch at zero current and zero voltage, which cuts the switc hing loses.
Jim Williams was getting close to 95% efficiency out of his.
With MOSFET switches, at low voltages, you can drive that gates from the ot her ends of winding - cross connect gates and drains.
It's a bit crude and inflexible, but does save the extra winding.
rs, but adds a current sense transformer that is used to force saturation o n a small transformer to lower losses.
Never heard of it.
-- Bill Sloman, Sydney
Never heard of it.
Core loss isn't low if it's saturating, but you can (and preferably should) use an auxiliary drive transformer that saturates earlier, thus commutating the transistors' base rather than forcefully yanking the collectors up.
The current mode version is famous from ATX supplies, e.g.
The feedback winding is a turn or two, while the base windings are several turns, thus the transistors run at a forced hFE(sat) = N(base) / N(fb). Note that the TL494 must short out the drive transformer; normally the transistors latch in this service.
They sometimes also use weak DC base bias, which causes it to self-oscillate for startup, no aux supply needed. More of an AT than ATX thing, I think?
I once made half a one, like so:
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website: https://www.seventransistorlabs.com/
te:
:es/an65fa.pdf
Nope
The original royer, invented in 1954, was a converter which worked on the s aturation of the main transformer
Baxandal was many years later, and added a capacitor for resonant operation
been said that Jim Williams got the circuit (but not the literature referen ce) from the UK.
Jim Williams main effort was to optimize the efficiency for fluorescent lam ps for displays
It is not the same as the Royer, added extra cap
aced with 2 transistors. Additional benefit is that adding a cap on the win ding can let it run in resonance for ZVS (low loss), and operation of the r oyer has inherent short circuit protection (Pressman has good explanation i s his book).
to 1954).
Indeed. The added cap is to boost efficiency performance
ing transistors switch at zero current and zero voltage, which cuts the swi tching loses.
other ends of winding - cross connect gates and drains.
You can't do that for a high voltage design
tors, but adds a current sense transformer that is used to force saturation on a small transformer to lower losses.
Royer Jensen:
Cheers
Klaus
rote:
er:
otes/an65fa.pdf
saturation of the main transformer.
I know - I read the paper back in 1970. The version I was working on a the time didn't - the base drive to the saturating switches was limited so that they came out of saturation before they come did, and more modern equivale nts don't.
operation.
This had a number of consequences, all of which made for a rather better pe rforming circuit.
s been said that Jim Williams got the circuit (but not the literature refer ence) from the UK.
amps for displays.
His motivation for using the circuit was much the same as Peter Baxanadall' s for inventing it - he need high turns-ratio transformers to get the high output voltages he needed, and they tend to have low self-resonant frequenc ies.
Getting a Royer inverter to work near the self-resonant frequency of it's t ransformer is difficult, and the switching spikes (and losses) become remar kably large. The Baxandall approach finesses this.
htm
That is the point.
placed with 2 transistors. Additional benefit is that adding a cap on the w inding can let it run in resonance for ZVS (low loss), and operation of the royer has inherent short circuit protection (Pressman has good explanation is his book).
k to 1954).
ching transistors switch at zero current and zero voltage, which cuts the s witching loses.
e other ends of winding - cross connect gates and drains.
As I said.
istors, but adds a current sense transformer that is used to force saturati on on a small transformer to lower losses.
rmer_6A.gif
A fairly obvious derivative of the original Royer circuit. There are neater ways of getting the transistors to switch.
-- Bill Sloman, Sydney
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.