Optos vs Digital Isolators, the truth?

On a related subject, I have failed to find any lifetime data for the opto dielectric breakdown (not LED degradation)

Anyone got data for a random device?

Cheers

Klaus

The optos can have degradation of the LED over time. Keep the LED current low.

The ADI (magnetic) and TI (capacitive) isolators seem fine. We've had issues with the Silabs parts.

That would be cool, but you'd have to do the driver and receiver. Surface-mount inductors, on the same or opposite sides, would be interesting... better coupling.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

I seem to remember seeing a chip that is bidirectional (maybe separate tx and rx) and uses PCB loops to isolate signals. Or they may have been digital with the magnetics on chip... can't recall. I ended up not doing that design so I didn't need them. I'll take a look and see if I can find the info.

--

Rick

Hi Rick

That would be great

Thanks

Yes, something like an unshielded SMD inductor on both sides of the PCB, tuned frequency

Regards

Klaus

John, can you tell some details to your problems? We are going to use them in a pulsgernerator @ several 100Volts and...

regards

- Michael Wieser

We had some that got into the wrong state... output input.

--

John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

I have a colleague who has used them, I think he talked of them being sensible to current into the GND node of the secondary side (when used as a high side driver)

Cheers

Klaus

I couldn't find my info on the one I had considered some 10 years ago, but a google search turned up a number of devices.

ADUM1401, Si8431 and this company who I think is the one I had found previously.

I hope this helps.

--

Rick

An opto isolator has a continuous times signal path between transmitter and receiver. Systems that use caps or transformers need some AC scheme to indicate the on and off state. The TI data sheet doesn't explain the theory of operation. The SiLabs at least makes a cursory attempt to explain how it works (high frequency carrier and tuned detector).

Seems to me a opto scheme is far simpler circuit wise, but requires hybrid construction. The digital isolator is more complex, but uses monolithic fabrication.

My gut feeling is the opto is more reliable. Plenty of weird stuff can happen when zapping chips.

I once rolled my own transformer-coupled isolator. It relied on comparators with hysterisis to hold the logic state after the changing magnetic field in the rudimentary transformer had ceased to have any effect on the output.

Sufficient noise on the comparator power supply could have over-ridden the hysterisis and persuaded it to forget it's last state, but it would have been set right the next time someone went to the trouble of driving the transformer.

Since LEDs produce less output per unit current over time and transformers don't wear out, it all depends what you mean by "reliable".

--
Bill Sloman, Sydney

I've dealt with companies doing industrial controls. These people get a lot of nasty spikes. Power line of course, and RF. In a glitchy environment, the opto scheme would certainly be more reliable. In fact, I was hoping these digital isolators didn't use a latching scheme.

As you can see, I'm trying to be logical about this. I don't see a scheme that produces a constant frequency being all that inferior to the simple LED. As you say, the LED dims, so it has it's own issues, though for binary data, the drop in light output wouldn't be a show stopper.

When switch mode power supplies hit the market, they sounded like an overly complex solution. And even today, there are really shitty SMPS out there. But once people started to use controller chips rather than home brew, the reliability improved.

e

ime

l r e e

the

e

ers

ot

There tend to be lots of spikes around industrial sites - they can be big e nough that a lightening strike somewhere on the site can put hundreds of am ps through the earth connections, which shows up everywhere. Sensible peopl e tend to isolate the power rails on individual boards the inductor capacit or filters, so that kind of spike shouldn't get through to a comparator's p ower pins.

Metcalf's Law was that every prototype local area network had been taken ou t by a lightning strike. Even after this, the original September 30,1980DEC , Xerox Park and Intel DIX Ethernet standard had to be revised early on (in November 1980) to cope with real industrial voltage spikes. Computer scien tists tend to be short on industrial experience. The telephone business had known better for quite a while.

ry

ly

e

For anybody who was using a significant amount of power, 95% efficient swit ching power supplies were a real step in the right direction. Even if you h ad to wrap them in a Faraday cage and smother them with inductors and capac itors to stop them radiating the rest of the system to death, it was still well worth it.

--
Bill Sloman, Sydney

I once did this,

which had terrible CMRR, resulting in indeterminate states on the IGBT for

10s of us at a time (i.e., until the next full cycle refreshed state).

Does provide an example of my using desat protection since before I ever designed a commercial product, though.

The monolithic parts are, needless to say, a few orders of magnitude better: typically >30kV/us at 1kV step. The ADI parts talk at length how they do it (~1ns pulses, with astable refresh, and reciever time-out fail-safe). Competitors are pretty widely available today, including what appears to be DC correctness (like the apparently ASK SiLabs ones), but it's not always obvious.

Tim

--
Seven Transistor Labs 
Electrical Engineering Consultation 
Website: http://seventransistorlabs.com

For serious isolation, use fiber optics.

--

John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

e_Drive.gif

or

w

at

For even more serious isolation used laser-based data links. You'll need so me tolerably serious optics, and careful alignment, to get the source laser s focused on the optical receivers, but if you want to keep the high voltag e you need isolated inside a tank of pressurised SF6, an optical window may be easier to get than pressure-tight fibre-optic lead-through.

--
Bill Sloman, Sydney

Why do we allow airplanes to crash when safety is involved? Because no one would be able to buy a seat on a flight.