250mW @ 5v across galvanic isolation with low EMI?

[max_torque]

I need to push up to 50mA at 5v across galvanic isolation to power a device.  But i need low EMI  eg DefStan Land Class A equivalent:




at ideally low cost, which rules out most off-the-shelf high frequency switch mode convertors.

  it struck me, that as size isn't much of an issue, and my power requirement is low, could i engineer a low frequency sine wave transformer based system from discrete components?  Say a 400Hz to 1Khz fundamental frequency would be easy to filter and it's unwanted harmonics are still going to be a a relatively low frequency, and the magnetic components won't be excessively big.  Of course i'd have to construct a suitable driver to create and push the sine wave into the those elements, and to rectify it back to dc on the other side etc

Anyone done this sort of thing before?

[T3sl4co1l]

Photovoltaic isolator.

Not sure if they come quite that big, but it's a slam dunk otherwise.

How much isolation capacitance can you tolerate?

Tim

[fcb]

I'd roll my own fixed frequency - 50% duty-cycle push-pull transformer design, and then filter it to get your conducted emissions down.  Probably 200-400KHz.

We've just done a design using a CD4047B driving a couple of FDN337's. The high frequency makes the filtering much smaller.  For the transformer we wound our own on a coated ferrite ring to get the isolation (we want a couple of KV).  But if you don't fancy it then you might be able to find a LAN or pulse transformer, Wurth do these natty little 1:1:1:1:1:1 transformers.

After the secondary rectification and filtering, we use an LDO.

If you want super spurious, then a sine driven transformer will be best, if you go low frequency then you'll have some pretty large filters though.

[floobydust]

You can use a low frequency DC-DC with slow controlled edges or sine, but the required iron-core transformer is bigger. I don't know nanocrystaline transformers are out there yet, compared to ferrites. I've used little audio transformers, but not sure of your isolation voltage requirements - galvanic or mains?

Higher freq. say 100-500kHz is compact off-the-shelf modules but as you say much harder to pass stringent EMC. The little 1W converters are convenient but very noisy and you lose any benefits from all the extra filtering you have to add, and then they still radiate enough to warrant a metal shield.

For conducted emissions, the transformer's primary-secondary winding capacitance couples your two isolation sides at RF. So EMI on the primary side can walk out the secondary side. You must use a CM choke and Y-caps on the output of the DC-DC. It depends on your product's noise sources, i.e. a GHz CPU making the primary side power dirty can surprise you radiating GHz emissions out on your galvanic circuits and associated field wiring.

So I'd use a soft/resonant/Royter push-pull DC-DC, with say 6VDC or so unregulated output and CM filtering, then LDO to make quiet 5V.

[mikeselectricstuff]

Photovoltaic isolator.

These only put out microamps for driving MOSFET gates.

[T3sl4co1l]

Photovoltaic isolator.

These only put out microamps for driving MOSFET gates.

The small ones yes, but do they make bigger ones?  I haven't shopped for them.  The efficiency does tend to be crap; but then, so is the extremely restrictive EMI requirement here!

Tim

[max_torque]

The isolation is for functional reasons not HV or safety ones.  Max diff between side of convertor would be around 50v continuous.

Ive looked at the ADuM5211 isolated data bridges that have built in isolated power transfer channel, but the datasheet suggests there are significant EMI/EMC issues with these devices, and i can't afford a long, drawn out emc optimisation exercise to fix those sorts of problems. Hence the idea to look at low frequency, slow edge type power transfer etc

[mikeselectricstuff]

Photovoltaic isolator.

These only put out microamps for driving MOSFET gates.

The small ones yes, but do they make bigger ones?  I haven't shopped for them.  The efficiency does tend to be crap; but then, so is the extremely restrictive EMI requirement here!

Tim

I suspect you'd need enough LED power to require  a buck converter for the LED

[mikeselectricstuff]

i can't afford a long, drawn out emc optimisation exercise to fix those sorts of problems. Hence the idea to look at low frequency, slow edge type power transfer etc

In which case I'd go for a fixed, high-frequency off-the-shelf converter, with LC filters, and provision on the PCB for a shielding can with  feedthroughs.

[T3sl4co1l]

I suspect you'd need enough LED power to require  a buck converter for the LED

As for low frequency oscillators, sure, audio transformers are still around.  They'll have more capacitance than ferrite or whatever, though... how much can you tolerate?  Doesn't matter, just galvanic isolation?

Yes, the ADuM parts are monolithic transformers operating at something like 200MHz plus harmonics. You're never going to shut that up.

There's an LT part that's designed for low EMI: a switching regulator with smooth, linearly controlled switching edges.  That doesn't save you ripple, but it saves you from the most pervasive noise >20MHz.

Tim

[GreggD]

Use a 85x53mm 5-6V Aliexpress solar panel and a bunch of IR leds. The best wavelength should be 940nm. Red will work but not as well.

[David Hess]

A resonant or non-resonate sine wave inverter will have the lowest noise.  If higher efficiency is required, then I would try an LT1533 or similar which has programmable voltage and current slew rates.  Linear Technology application note 70 has a detailed discussion about how to use this part in both switching regulators and inverters.

If an inverter is used, secondary side regulation may be done from the primary side by monitoring the primary side voltage through compensation diodes so an isolated output is provided without an auxiliary isolated feedback loop.