A very simple synchronous rectifier?

complicated

gate,

consumes some power,

patents...

Q1 is 'on' when V1 is low, so you're pulling Vout through R1 (20 ohms) to ground. That's a lot of dissipation, possibly more than the sync. rect. saves.

Let's see--12v output, v^2/R1 x 50% duty = 3.6 watts.

I've thought of such things, but the complication doesn't usually pay compared to a fat schottky.

A few extra turns on a flyback winding is a good driver for a sync rect. though.

Cheers, James Arthur

Depends upon the voltage. If you need 3.3V and significant current (e.g. a PC motherboard), the diode drop is a significant inefficiency. Even more so for the ~2V used by the CPU.

OTOH, synchronous rectification can be much worse if you get it wrong, i.e. shoot-through from turning on too early.

Yes and yes.

OTOH, Jan gets a great advantage from using a PIC: he can simply drive the sync. rect. FET from another output, and time it however he wants.

Cheers, James Arthur

On a sunny day (Thu, 26 Feb 2009 16:45:21 GMT) it happened James Arthur wrote in :

complicated

gate,

below

consumes some power,

patents...

R1 is 220 Ohm, the picture is not that clear.

Yes, in real life, with real pulses, things are more complicated. I did test the circuit in my setup tonight, and it promptly blew the fuse. The pulses are quite different, causing the MOSFET to be on when reversed. So next is a small PIC solution using the hardware comparator[s]. Or maybe I can transfer the driving pulse from the other PIC somehow. PIC could be cheaper then an opto coupler... or even inductor.

If Jan decides to go that route, I'd suggest keeping the freewheeling diode in parallel with the FET, to avoid frying the FET due to e.g. bugs in the code or an unexpected reset.

On a sunny day (Thu, 26 Feb 2009 23:29:20 +0000) it happened Nobody wrote in :

I made the frequency settable in kHz via RS232, however that was not so simple, as it needs terminating the current pulse, reloading the timers, setting the maximum PWM width (different as higher frequencies have less bits resolution in the PIC PWM), but it works, but did not first time. If I suspect problems I just put a car headlight in series with the 12V battery that I use as source for the low load tests, it will light up :-) Else a small fuse. Added input protection against user typos or errors too for the minimum and maximum frequency. The MOSFET has an internal diode, if the MOSFET does not conduct when it needs to be on in the synchronous rectifier, then the internal diode will work as diode. For test with low load this is OK. Maybe in the weekend I will have some more time to progam a PIC for the rectifier control.

With a FET you get the freewheeling diode for free.

Timing is another matter--what's that phrase Joerg likes? Phsst...pop... BANG!

Cheers, James Arthur

Try this.

1940s technology to the rescue w00t!

Tim

omplicated

t

he gate,

below

o consumes some power,

patents...

Most of the H-bridge circuits I've seen have a separate freewheeling diode for each FET, in spite of the body diode and the synchronous rectification.

I had assumed that the body diode's maximum current was typically lower than the channel current. But e.g the IRF3205 data sheet lists 110A for both Ids[max] and Is[max] (the body diode current).

OTOH, the voltage drop for the body diode is given as 1.3V (about twice what I'd expect for a Schottky diode). At the TO220's maximum current of

75A that translates to ~100W, compared to 45W for I2R (Rds[on] = 8mOhm), so power dissipation could be a serious problem if you're relying upon the body diode.

On a sunny day (Thu, 26 Feb 2009 20:32:39 -0800 (PST)) it happened Tim Williams wrote in :

Yes, nice, I have worked with technology like that. It is 60 Hz though, and for some the inductor will be too big :-) It can be done even simpler with a UJT:

------------|

Granted, the body diode isn't great; I was responding to:

It's usually good enough for that. YMMV.

Cheers, James Arthur

The body diodes of MOSFETs are usually a lot slower than the diodes you can buy. At low voltages, it is very useful to put a schottky across the MOSFET. It helps a lot at high frequencies because it reduces the power lost in the MOSFET and hence the temperature rise.

I work with someone who was one of the first to do a sync rectifier. He was using a germanium power transistor as the rectifier. He needed the larger EB breakdown voltage.

He ended up with a rectifier across the transistor. The circuit started and stopped conducting with only the diode. In the middle of the timing, the transistor was biased on. This way he didn't need super accurate timing.

Before he added the diode, it seems that the only other option was a mechanical transistor changer that installed new ones at a 10KHz rate.

I simplified it some more, here the setup: ftp://panteltje.com/pub/power_pic/power_pic_discrete_driver_synchronous_rectifier_img_0964.jpg

Note the extra turns with flatcable wire on the ringcore to drive the gate of the rectifier MOSFET.

The little blurb of transistors bottom left is the discrete gate driver I added so I could drive the IRFZ44A, that is not a logic level MOSFET, and I use the same MOSFET in the synchronous rectifier here.

This picture shows the same setup working with about 55W output into a car headlight bulb: ftp://panteltje.com/pub/power_pic/power_pic_synchonous_rectifier_working_img_0965.jpg

And this is the diagram of the even simpler synchronous detector: ftp://panteltje.com/pub/power_pic/power_pic_synchronous_rectifier_diagram_img_0968.jpg

The zener is needed, I killed one IRF, spikes happen, the 22 Ohm + zener limits those, the defective one has now gate leakage (spike pierced silicon isolation).

The diagram of the gate driver I forgot to take a picture off, hint: from the picture, all transistors are NPN, should not be too difficult.

The asm source of the PIC driver (but not the very latest) is available here released under the GPL:

..

e of the rectifier MOSFET.

Eww, yellow toroids? Have you measured them? Every single yellow- glazed (one white face) toroid I've tested has huge loading. They suck! Try a black (unglazed usually) toroid for transformer action; gapped black ferrite (C / EI / pot cores, etc.) for storing energy (flyback / filter choke stuff).

Tim

It is not quite so simple because the current is still shared between the external shottky and the body diode. A good fat shottky of the same current class as FET is required so the body diode will not have any influence.

The reverse recovery of the body diode creates nasty spike of current, and it could be the main limiting factor for dV/dt. Once I was considering complete decoupling of the DirectFET body diode using a pair of the external ultrafasts, but it is bulky, lossy and expensive.

We tried the BJT as the SMPS rectifier in the 80x. The recovery from saturation was way too slow. The best variant for that time appeared to be a bunch of several dozens of small germanium diodes in parallel.

What was the operating frequency?

The IRF rep claimed that they have the 100A GaN based SMPS running at

60MHz. Any comments?

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

On a sunny day (Fri, 27 Feb 2009 08:30:09 -0800 (PST)) it happened Tim Williams wrote in :

The color is part of a color code, and relates tot the Al:

\\ Select a -X number from the left column, and see the Al change.

The Al of the black - is 3.2 against 4.7 for the yellow T68, are you sure????

Gapped E core if there is DC flowing in it

I have zero problems with these yellow cores, over a very wide frequency range, from 19 kHz to 150 kHz. The size limits the power.

These came from some old computer supply.

The Fairchild FQP33N10's body diode has a trr of 80nS. That's plenty good enough for an emergency diode that just has to take the one inductor dump in Nobody's scenario above.

80nS -- pretty impressive for what's basically a big zener.

I've done the same. I settled on the extra turns + MOSFET thing that Jan's turned to, but despite being the best, the cost and complexity wasn't worth the improvement in my app.

Cheers, James Arthur

That thing must be a meshugganah of an RFI/EMI emmissions and conduction problem.

complicated

gate,

consumes some power,

patents...

There are a number of problems here. Most would be apparent if you looked at the currents in your circuit.

If the mosfet is on, there will be no voltage developed across it to turn Q1 on, to discharge the gate, when you want the fet to turn off. Probing your model shows that 500A of 'reverse' current through the rectifier is required in order to induce the mosfet gate's driver to turn it off.

A unipolar (or an AC one, for that matter) source PWM'd into a capacitive load is not a good application for any power rectifier, synchronous or not. Call up your references to switched capacitor converters. This is a low power topology.

Simple is relative - "simple discrete" may even be considered somewhat of an oxymoron, these days.

The concept of the normally-off synchronous rectifier is fairly simple. The few integrated versions out there generally do not improve performance by complicating things.

RL