A simple, low parts count, efficient, isolated DC supply.

[Odysseus]

After destroying my cherished but dated mp3 player with a sub par 12V cigarette-to-USB adapter, I started looking into building a simple isolated DC-DC converter. I since stumbled upon the interesting circuit shown below.  It has the following nice features.

1) Synchronous rectification, no diode drops, and thus good efficiency.
2) Symmetrical, so power can be transferred in either direction.
3) Coupling between L3 &L6 strongly reduces the output ripple voltage. (Eliminates it in the theoretical ideal case)
4) Sinusoidal switching waveforms (Low EMI, ZVS and ZCS).

I haven't built it yet, however.  I need to wind a suitable transformer.  It remains to be seen if this circuit performs in practice as well as it does in simulation.

[kizzap]

what happens if you make the coupling between the inductors not ideal?

[LukeW]

Complete galvanic isolation just for a 12V->5V regulator seems like complicated, completely unnecessary overkill.
You can make a quality, robust regulator with no need for isolation.

[mikerj]

Complete galvanic isolation just for a 12V->5V regulator seems like complicated, completely unnecessary overkill.
You can make a quality, robust regulator with no need for isolation.

Without knowing the OPs application you can't really say that.  To power an MP3 player from your cars cigarette lighter socket it may well be, but there are applications where such isolation would be essential.  Besides which, it's a pretty interesting circuit.

[rob77]

Complete galvanic isolation just for a 12V->5V regulator seems like complicated, completely unnecessary overkill.
You can make a quality, robust regulator with no need for isolation.

Without knowing the OPs application you can't really say that.  To power an MP3 player from your cars cigarette lighter socket it may well be, but there are applications where such isolation would be essential.  Besides which, it's a pretty interesting circuit.

+1

even some MP3 players might need isolated supply if you connect the headphone output of the player to the car audio's aux in.

[mariush]

If your mp3 player doesn't need more than 5v @ 100mA then you could just use one of those ADUM5000 isolated dc-dc converters : http://www.analog.com/static/imported-files/data_sheets/ADuM5000.pdf

It takes 5v in, it outputs 5v out.. so just put a lm7805 in front of it and you've solved the problem with two chips.

Only problem is price, it's about 7-10$ if you buy just one.

[nuno]

After destroying my cherished but dated mp3 player with a sub par 12V cigarette-to-USB adapter, I started looking into building a simple isolated DC-DC converter.

Maybe your converter lacked proper supply transient protection, a TVS, inductive filtering, ... you would be amazed at how much garbage a, let's say, xeon lights controller, can put at the 12V rail.

Care to open your dead adapter and check what's in its supply input?

[DanielS]

Complete galvanic isolation just for a 12V->5V regulator seems like complicated, completely unnecessary overkill.
You can make a quality, robust regulator with no need for isolation.

When you have a $200-500 device connected to your DC-DC adapter and it costs less than $1 extra in parts to go from a buck design which can fail and kill its load to an isolated design that makes it nearly impossible, I would pay the $1 extra and go with the design that makes catastrophic failure (as far as the load is concerned) nearly impossible.

One problem with this design though is that it has no input regulation, which can be a problem in automotive applications where input voltage can go from 9V during cranking to 15V during charging. If the turn ratio is 27:10 to accommodate a typical 13.5V car running voltage, the unregulated output will be 3.5-6V. Another potential issue with oscillators like these is that they can become unstable when reactive loads are attached to them.

If I had to design a car USB adapter, I would use a double-ended transformer PWM for the output, a common ground and skip optoisolators in the feedback loop. Another possibility would be one of those ringing-choke regulator ICs normally intended for wall-wart low-power AC-DC converters if they can start on 10V, using common ground and skipping feedback isolation here as well.

[nickm]

Very cool.  I always find simple converters very interesting.  One minor drawback of this design is the 4 winding coil which might be hard to by off the shelf.  Another issue could be the gate voltage.  You're showing it peaking around 16V, and since it is subject to component tolerances and transients you could easily get above 20V which is the Vgsmax of most fets. 

You might also look at the self oscillating flyback.  Low parts count (slightly more than yours),  Less switches, regulated, current mode control.  Drawbacks are now you definately need a custom transformer, average efficiency (which shouldn't really be an issue if powered by car).

http://www.deltartp.com/dpel/dpelconferencepapers/S19P6.pdf

[mzzj]

Interesting little circuit but the need for 2 inductors with multiple windings is kind of turn-off.

[DanielS]

Interesting little circuit but the need for 2 inductors with multiple windings is kind of turn-off.

When you want partial electrical isolation to prevent a primay/high-side failure from blowing up the load(s), the transformer is not really avoidable.

For applications in the neighborhood of 10W, winding your own high-frequency transformers is not too difficult if you do not care about achieving the smallest size or highest efficiency possible.

[mzzj]

When you want partial electrical isolation to prevent a primay/high-side failure from blowing up the load(s), the transformer is not really avoidable.

For applications in the neighborhood of 10W, winding your own high-frequency transformers is not too difficult if you do not care about achieving the smallest size or highest efficiency possible.

True. But in this case you need 2 transformers (or coupled inductors to be more precise) The actual inductance value is not critical but has to be in the right ballbark so both transformers need gapped ferrites or something else like iron powder core with low permeability.

[Jay_Diddy_B]

Hi,

There is no need to couple the inductors L3 and L6.

Cooper Bussmann make a Versa-pac series of transformers that are suitable for this design.

Regards,

Jay_Diddy_B

[mzzj]

what happens if you make the coupling between the inductors not ideal?

I was playing around with the circuit in a spice for a while. If you set the inductor coupling to less than 1 the simulator goes wild. It will oscillate at several MHz frequency.
I managed to tame the oscillation with large resistors in series with every mosfet gate(standard trick) 500ohms to 1k looked like usable values.
500 ohm gate resistors, K 0.99 coupling and 0.1ohm resistance for every winding the simulator was still giving 94% for efficiency. Nice.

[mzzj]

Hi,

There is no need to couple the inductors L3 and L6.

Cooper Bussmann make a Versa-pac series of transformers that are suitable for this design.

Regards,

Jay_Diddy_B

Yeah, you are correct about L3 and L6.

Coilcraft hexa-path came to my mind as a transformer but its expensive, like 5usd/piece 

[David Hess]

Interesting little circuit but the need for 2 inductors with multiple windings is kind of turn-off.

When you want partial electrical isolation to prevent a primay/high-side failure from blowing up the load(s), the transformer is not really avoidable.

SEPIC converters provide protection against switch failure but are not isolated.

[David Hess]

For a simple isolated design, I would use something like a flyback regulator with voltage sensing through the flyback transformer.  The transformer is not critical and can be found as a standard part.

http://www.linear.com/product/LT8302

[Odysseus]

Hi,

There is no need to couple the inductors L3 and L6.

Cooper Bussmann make a Versa-pac series of transformers that are suitable for this design.

Regards,

Jay_Diddy_B

capacitace
No, It isn't necessary, and may cause more problems than it's worth, as seen by Mjjz.  The idea, however, is that the coupling largely counteracts the ripple present at the center tap on the secondary, requiring less filter capacitance to meet a given ripple spec.

For a practical USB charger, this circuit has some serious drawbacks, notably a lack of regulation.  I imagine it would be more useful in a different application.

...

even some MP3 players might need isolated supply if you connect the headphone output of the player to the car audio's aux in.

That's exactly the problem that occured.  Turns out the audio amp used by the Sansa E200 is single supply (i.e. headphone ground is vcc/2 above USB ground). Notably, this is not an issue in more sensibly designed portable products, such as the iPhone, which internally generates a split supply just for the audio amp.  So, I suppose the charger wasn't really to blame.

[richard.cs]

Another situation where an isolated supply would be useful in a car: my car is wired positive-earth but so that I can use modern devices the cigar lighter socket is connected with the shell negative. A non-isolated usb charger would result in a flying lead with the connector shield 12V away from the vehicle chassis (and the same problem for downstream audio cables, etc).

An isolated dc-dc eliminates the problem in any car.

[mikerj]

That's exactly the problem that occured.  Turns out the audio amp used by the Sansa E200 is single supply (i.e. headphone ground is vcc/2 above USB ground). Notably, this is not an issue in more sensibly designed portable products, such as the iPhone, which internally generates a split supply just for the audio amp.  So, I suppose the charger wasn't really to blame.

In that case the simplest solution would be to use an audio transformer to isolate the output from the MP3 player.  These are still used in the high end audio installs in cars to eliminate ground loops, and are sold as "ground loop isolators".  They are available on eBay etc. as complete encapsulated units with phono plugs etc. and they are pretty cheap.

[SeanB]

Funny enough I was just parting out a Sony Umatic recorder, and the headphone output has 2 isolating transformers, so that the headphone audio and the headphone volume control is isolated from the main chassis ground. The audio selector switch that allows either left, right or stereo to go to the headphones is also on the secondary side. That way if the ground has been lifted ( naughty but often done to get rid of ground loops) and the whole chassis is floating at half mains you cannot get a shock off of the headphone output.

[David Hess]

I have had problems with ground loops in cars before.

Amateur radio transmitters usually end up grounded through both their negative power lead and antenna ground connection.  The resulting ground loop can result in alternator whine in the audio which is very difficult to remove.  Solutions I have used for this in the past include adding a large common mode choke inline with the power leads and capacitively isolating the ground of the transceiver's coaxial connection between the radio and antenna.  An isolated power supply would certainly have worked but the 10 to 20 amps required are not trivial.

[Zero999]

It looks like a Mazilli inverter.
http://tonic-lab.com/mazilli-inverter
Make the transformer non-ideal. Try adding some resistors and diodes to see if it's any better, then build the real thing.

[wiss]

I have had problems with ground loops in cars before.

Amateur radio transmitters usually end up grounded through both their negative power lead and antenna ground connection.  The resulting ground loop can result in alternator whine in the audio which is very difficult to remove.  Solutions I have used for this in the past include adding a large common mode choke inline with the power leads and capacitively isolating the ground of the transceiver's coaxial connection between the radio and antenna.  An isolated power supply would certainly have worked but the 10 to 20 amps required are not trivial.

I apply a layer of (thin) thread seal tape to the shield in the antenna connector to fix that problem. Isolated antenna would work as well.