Isolated Low-Power Capacitively Coupled Supply (Low EMI)

Thread Starter


Joined Oct 30, 2016
Hi Guys,

I'm designing a low-power isolated supply (3.3V, Max ~50mA) and am keen to steer clear of transformers as EMI is critical to the design and I've had some noise issues with my previous transformer-based design.

I have been looking at using a capacitively coupled circuit plus bridge rectifier as per Maxim's AN4553 app note below (or a similar charge-pump type circuit which will allow me to get ~150mW of power across the isolation caps) but their MAX256 chip is far overspec'd from a power (and size/cost) point of view (3W) for my requirements.

Do you have any suggestions for a driver to replace the max256 here, which would differentially drive a square wave into a circuit like this at <2 MHz? Or is there an alternative way to achieve a capacitively coupled low power supply? I've had a look at some cmos drivers, even 555 timers, but nothing seems to provide the combination of (a) required source/sink output power and (b) dual, complimentary square wave output I require to drive into the capacitor/bridge circuit.

Since it is such a low power application, am hoping there is a cheap, small, quiet way to do this easily, but so far I haven't found it.

Thanks in advance,


Thread Starter


Joined Oct 30, 2016
The transformer I am using is quite a large part, and all of that wound wire over ground is acting like a big antenna in the centre of my board. Swapping this for some small passive ceramic caps loaded neatly on the ground plane should reduce my EMI right down, that's the theory anyway.

The actual isolation need only be 50/100V so some 0603 caps would do the job ok.



Joined Nov 6, 2012
You might be surprised to find out that it's not the Transformer,
but the Board Traces, or Trace Layout, that are acting like Antennas.
This can also be a problem of inadequate, or improperly sized, or placed,
Power Supply Bypassing CapacitorS (plural).
If the problem is actually caused by the transformer design, go with a Toroidal Core Transformer.

A Megahertz range, square wave is very tricky to achieve, if not almost impossible.
You'll have huge losses in the diodes, which all have capacitance.
Board Trace and Component Layout are extremely critical, and may require numerous "cut & try" sessions.
Efficiency always drops at some point as you go higher and higher in frequency.

What are you trying to isolate from what, and why.
The Capacitor solution may provide DC Isolation, but you will have AC interactions no matter what you do.

You are trying to find an unusual solution to a problem that is not fully understood.
You need to provide the what, when, were, why, & how of the project to get a useful answer.

Thread Starter


Joined Oct 30, 2016
Many thanks for the detailed reply.

I essentially need to build 2x small isolated 3V3 supplies (~50-100mW each). One to power the LNA in an externally connected GPS antenna, and another to power a bank of ISOLATED digital IOs. Both of these need to be isolated from the main system, to avoid any potential for ground loops and to minimise the risk of an overvoltage event making it's way back to the main product circuitry.

A previous iteration of the design used 78602/9JC, but it was causing substantial EMI issues with a receiver circuit. The XFMR itself was found to be the culprit during test and debug and as much due diligence as possible has been applied to ensuring appropriate decoupling, track width/length/impedance optimisation have been used.

I've looked at other pulse transformers that might be suitable but they all look similar to the MuRata one. I've seen some RS232 transceiver ICs use a charge pump approach for a low-power built-in isolated supply, and it seems like this would be a nice solution, to remove the XFMR completely, and just use a pair of ceramics to realise the isolation barrier. I don't need kV of isolation, so this in theory would work well. But there doesn't appear to be a dedicated driver for this type of approach. Am I right in assuming then, that for a low EMI-solution to this problem, the industry best practise is to try to optimise a XFMR-based solution? I see TI have some integrated transformer packages that look interesting.

Ultimately I only need <100mW from each supply, so a simple, low EMI solution would be ideal. I also don't need the switching frequency to be >1MHz, although this would in theory make filtering out any differential-mode switching noise easier.

I welcome your thoughts.
Many thanks,


Joined Nov 6, 2012
There are Fully Isolated, completely self-contained, Power Supply Modules.
That is to say, I briefly looked at them, but I don't have any need for them.

I'm still trying to figure out why you think you need so much isolation.
Have you had problems with a similar setup in the past ??, or,
do you know for a fact that,
whatever this setup is designed to accomplish,
will not work without everything being Isolated ??

And, as far as I/O goes, there are Chips made specifically for the purpose of isolation,
which can also provide Level Shifting if needed.
Sorry I can't give you details on these parts, but like I said, I have an interest, but no need.

Each one of these systems starts out with either, High Voltage AC, or,
is Battery Powered and recharged by High Voltage AC, so,
I would start at the source, and run 3 separate Transformer Isolated Power Supplies, ("Modules").

After you have the needed DC voltages, plus ~10 to ~15% overhead,
run Foil Shielded, Twisted Pair Cable between the Supplies and Loads, ( Belden "BelFoil" cable ),
then use a Linear Regulator right at the point of use, even a simple Zener Shunt Regulator will work fine.
These Cables can be run well over 100 yards with no unusual considerations or losses
when used for low current Power delivery.

The Shield/Drain Wires should all be attached to a Star Grounding system at the AC Power Entry Point.
Do NOT connect the Drain/Shield Wire at the point of use.

If the Powered Equipment actually needs a ground,
create one, by driving a Copper Rod into the Earth,
AC Mains Grounds are notoriously noisy.

Unless you're running some sort of on board RF Transmitter, you won't have any problems.

All this might be solved with a simple Common Mode RF Filter on each of the Power Supply Pairs.
There are SMD Chips for exactly this purpose, (teeny-tiny ones),
and should be used at both the Power Supply End, and Load End,
that is, if these are not on the same PC Board, or are more than a few inches in length,
use these regardless of anything else that you may decide to do.
Last edited:


Joined Jul 29, 2018
Here's an oddball suggestion: Don't use digital parts. Use a 1-transistor phase shift oscillator to generate a 25 KHz sine wave, and a small audio power chip to drive it into a 150mW rated audio transformer. Rectify the output with fast diodes and away you go. No RF level interference!