Saturday, 26 July 2014

Isolation Options

Isolated PSU

So if you find out you need to isolate your circuit, you will need to power it over the isolation barrier, and there are options out there to do that.

Finding myself in that position, and being a notorious tight-arse I started to look at DIY options.

Commercial Option

I'm a fan of using what works, to a point.  In the past I've used COTS Isolated Modules but at over $7.00 each these are not a price-suitable option.  They are pretty damn compact, deliver 200mA and offer 1kV isolation.

Isolated, Tiny, Powerful.  And unregulated.

But for my pet project, I need 4 isolated channels, and $28.00 is a mad starting point.

So as easy as this approach is, it's out for me.

Building Blocks

In pursuit of DIY Nirvana, I hit up Maxim to see what I coudl nab from there.  I love their free options and when they work out they they are burning 50 bucks each and every time the FedEx me three samples, they might wise up and cut me off.  But until then, FREE STUFF!


Maxim have the excellent MAX253 and with the matching mu-Rata transformer it's a piece of cake to follow the application circuit, and boom, isolated 5V at 200mA is a snap.

With my free samples and transformers added to a separate parts order, I built my own version on Veroboard and was able to source 200mA with ease.   But seriously, 200mA?  You can fly to the moon on 200mA!

SO I knuckled down and had a think about this circuit.  If I could get rid of the bulky transformer.. and I stumbled upon this: Capacitor Isolated Power Supply using the MAX256.  Holy crap, I can ditch the 4 buck transformer.  Sweet.

Then I got to thinking:

  • I don't need 200mA.  Heck, I  might only need 1 or 2 mA
  • The MAX253 is still a couple of bucks, and provides an oscillator and two CMOS push / pull stages
  • CMOS logic also use push - pull outputs, at 20mA each
So then the penny dropped - Hex inverters are dirt cheap, so too are capacitors and 20mA is easily obtained from the outputs.

All of a sudden, an isolated power supply for less than a buck is possible.  Time to test the theory!

Putting It To The Test

To test this theory, I put together the following test circuit:

Test Circuit

Using a simple RC oscillator to drive the other gates, I built a test circuit that would:

  • Act as a voltage doubler
  • Provide positive and negative rails
  • Allow for easy current tests
I used voltage doublers as I knew that diode losses would mean that a single stage would not give a full 5V out, and when loaded the rails may sag, so doubling gives some overhead there.

This was then laid out in Altium, milled and tested!

As designed.

As Built!

By selecting from the shunt resistors (10k, 1k, 100R) I can sink 500uA, 5mA and (optimistically) 50mA from a 5V rail.

With no load, I saw + / - 10V on the output, and with 1k, 5mA was very possible however the 10V rail sagged to about 6V.  Still, with a 5V design goal, all was looking good.

5V Regulation

To ensure a constant 5 rail, I decided to add a voltage regulator.  So of course I harvested a redundant PCI WiFI card:

Good Bits!

Random Regulator Installed

And a nice and stable *isolated* 5V was available.  Yeah Baby!


So, next step was to integrate with my target project, which I'll detail in another entry later.

From top to Bottom:  Isolated PSU, Opto Isolated Serial, Single Channel Meter

The prototype (previously run from a non-isolated supply) was kludged together and success was had.  Yay!

From there, I ordered prototype PCB's from Seeed Studio where my meter board runs 4 isolated channels, and thus 4 PSUs.

Parts were loaded, the board powered up and...

Looking Good!

..immediate failure ensured.  The isolated 5V rail was cranked out at a massive 1.05V.  Something was very, very wrong.

I'd used a generic LM7805 SOT-23 regulator on my supplies, and after some head scratching I took a good look at the datasheet...

5mA Quiescent Current

Holy crap-balls.  The quiescent current of a 7805 is 5mA, which belts my supply.  So that's a really, really bad choice of regulator there.  So the 7805 was pulled, and the regulator I ratted from the WiFi card was installed.

Ratted Wedged In.

That too failed. The isolated 5V rail was non-existant.  Damn!

Bits Out

So the parts were pulled, and SIL modules fitted.  At least that would let me test the rest of my design!

Ugly Hack. 

Son of a ... that too didn't work. What's going on?

The Big Picture

After a lot of fault finding, I traced the fault down to my MCP6L0 op-amp on the 5V rail.  In my haste, I ordered MCP6L0R, and that was the root cause of my problem.

If you look at the footprint of the MCP6L01 and the MCP6L01R the VDD and VSS pins are inverted.  So by using the -R part, I was effectively shorting my isolated 5V rail.

Dammit Dammit Dammit Dammit.

Keep Calm and Carry On

So, new op-amps were ordered, and when fitted the 5V rails came up as expected.  At this stage my options were to remove the isolated modules and re-test my PSU circuit, or carry on a develop the rest of my project.  I went with the latter, based on the fact the I'd not yet identified a replacement low-quiescent current regulator. 

Then, in a word, Aliexpress.  500mA 5V DC-DC converters for about a buck a piece.  With 1kV isolation.  With my DIY capacitor supply, 50V isolation was the best I could hope for.  In addition, the modules take up much less board space than my design.

So, with some thinking, testing and redesign, I've come full circle and will be using off the shelf modules to provide my isolated rails.

Live and learn. 

1 comment:

  1. Awesome project! Two things: to increase the isolation voltage you could use X or Y rated caps and for the regulation you could just use a zener diode. The zener diode works as the output impedance of your power supply is so high. You basically end up with a shunt regulator.