Does anyone have experience laying out OptoCouplers for max isolation?

I’m making a board with a section for 6 opto isolators. I’m thinking of using sfh620a in a SMD configuration. Beside 120V mains isolation on some of them I’m hoping to reduce the possibility of voltage spikes passing through the isolation.

One thought is to mill a slot under the isolators.

Another thought would be to run a small ground strip down the middle of the isolation path and connect it to earth ground.

I’m leaning toward the first as it maximizes physical isolation.

Has anyone had any practical experience doing this?


Definitely not an earthed guard track as that would have nasty consequences on proving that the ground is permanent


Hi, something to consider:

  1. use a coupler that is certified for safety isolation (of course)
  2. use the creeping distance that is prescribed in your country
  3. Milling might shorten the distance if this is allowed in your country
  4. Keep in mind that the distance is PAD-to-PAD not Pin-to-Pin
  5. Take care that the PCB Manufacturer DOES NOT use this space for his Copper Logo or Copper Writing (This has happened !)
  6. Optocouplers have an Input-Output capacitance that may transfer fast spikes.
1 Like

Definitely a slot. And spread IC legs. I think I’ve seen IC packets with wide seperated pads.
What are you switching?

concerning the coupler:

at least for Germany there are THT Versions with wide Pins to achieve 8mm creeping distance. They have a VDE sign that is the equivalent to UL.

I have rarely seen 8mm compliant SMD couplers so you might have to use milling anyway. Maybe even with a part of the housing (plastic of course) going through the slot.

1 Like

If you’re worried about creepage, you can consider bigger optocouplers such as the CNY65.


Pin spacing between primary and secondary side is 15.24mm
(Size of DIP14 with pins on each corner)

It also has other hardened features such as a 3mm internal distance between the LED and the detector, with 8kV AC and 14kV DC test voltages.

If you want to go extreme you can use fiber (or plastic) optical cable. The audio versions (S/PDIF) are the cheapest.

But for normal mains voltage related applications the ubiquitous DIP4 sized optocouplers are apparently sufficient, but would you trust a bag of cheap PC817 optocouplers from Aliexpress?

SMPS circuits often have a capacitor (rated for 2kV or more) between the primary and secondary side. I’m not entirely sure why. I think it’s to ensure that the capacitance in the optocoupler is not the dominant factor in picking up noise, which could lead to faulty switching.

“Medical grade” power supplies have extra demands for failing in a safe state if they fail and for lower capacitance and who knows what else.

Bottom line is you first have to know what to protect against before you can build something.

Just remembered. In high voltage generation / distribution optically triggered thyristors are sometimes used.


Thats right, the capacitor keeps GND levels even for high frequencies. We had to use it sometimes for a SPI.

I imagine SMPS (for common mode) as AC generator in serie with its internal capacitance. Generator connected to GND and capacitor to isolated output.
For EMC measurement that output is loaded with 150 ohm resistor to GND and disturbance at that resistor are measured. In such circuit harmonic levels would be semi equal up to high frequency (the decrease in harmonic amplitude with frequency is compensated by the decreasing impedance of the capacitor).
If you add capacitor between output and GND (in fact in parallel to that 150 ohm) you get capacitive divider. If your trafo has capacitance of 20pF and you add 2nF capacitor you reduce disturbance voltage at output 100 times.
When I first time designed isolated RS485 (before I understood that the way I described it here) I used SMPS with 4kV and 4pF trafo and without that 2kV capacitor. It passed EMC tests. But later I read that people more often use SMPS with higher capacitance (like 100pF) and short isolation with capacitor in range 1…4.7nF (forming that capacitive voltage divider).

1 Like

This can get complicated as you have to plan what pollution level your board has to survive in many countries.
Conformal coating may be necessary.

1 Like

Thank you all for your insights. Rest assured I’m not designing anything for production or other countries. I am simply building a hobby data acquisition for my furnace. The isolators are to allow me to monitor the mains voltage (at worst). I simply want to do the best I can.

Some years ago I was asked to checkout issues with a piece of production equipment (I was in design engineering). The issue was that the controller would lockup several times a day. Out products consisted mostly of plastic parts which were stored in plastic bags what were in plastic bins. Here ESD lives and thrives. Whomever designed the machine had an Isolation board made using your standard garden variety optoisolator. Well he did not follow any rules of isolation and had input and output traces in close proximity but they were “isolated”. Clearly the ESD when right through this board to the main controller resulting in a lockup. The folks operating the machine would often remark the machine stopped after she just touched the machine,

I appreciate all the help the folks in the forum supply. Rest assured if I were designing a product I would have stated such in my post. It just wouldn’t seem right to me to ask for help with a production project without at least stating so.

Opto isolators for nuclear power plants have been known to be 3 feet long :slight_smile:

I don’t use any parts from Aliexpress or ebay where safety could be an issue. Or even if I plan on using it for a hobby installation and don’t want it to fail. (I also do some home automation devices and I really dislike things to fail) Its not worth the few $ that could be saved. And even if the parts are from a well know mfg, they are likely to be old :skull:

1 Like

Never thought they would go to that extreme. Do you have a link to such an optocoupler? How many digits does such a thing cost, and is it possible for mere mortals to acquire such a device?

It’s got nothing to do with nuclear power plants in particular, just high power applications.

A search like this:

Brings up results like this:

And a fancy picture in case the pdf gets (re) moved in the future:

HVDC (High voltage DC) applications go over a mega volt these days, and probably have stacks of these to get to high enough voltages.

1 Like

If safety can be an issue, look into fail safe circuits. All components can fail, and even if you buy your components from a reputable source, they still can be fake (happened to me), or (semi)damaged by ESD.

I had to make an additional external alarm output to a medical device. I made a circuit. Next I made a list of each component with all it’s failure modes (al combinations of a pin open, a pin shorted to another pin, a pin shorted to the ground plane, short or open in the cables and so on). Next I wrote down what that failure mode did to the working of the rest of the circuit, and if this would result in injury to the user.

After a few rounds of this, I came up with a circuit that could fail, but would then always generate a false alarm, and not miss any alarms. This is inconvenient, but fail-safe.

search for FMEA and single fault analysis for details.

I doubt it. I ran across this by talking to one of the operation engineers at the Yankee Power Plant here in Connecticut.

Actually there is an intrinsically safe circuit protocol. I’ve needed to perform such an analysis a number of times. One of the analysis was for an automotive fuel sensor (in the tank).

search for FMEA and single fault analysis for details.

I’ve probably done a hundred FMEA’s starting with aerospace and then automotive. Some mfg (Toyota) have an even more rigorous analysis.

Neat! Rad hardened as well :slight_smile:
Our limit was about 1500 volts with no radiation considerations.

I did a few designs that had rad hardened and EMP requirements. Not fun. The information provided to perform the analysis was sobering.

Thank you, I did not know this. Basically it’s a device that guaranteed operates at such a low power level ( 1.2V, 0.1A or 25mW) that it’s not possible to ignite a gas (most gasses) with it:

I feel for you. That’s a LOT of tedious work.

Decades ago I remember hearing a “BANG” before the power went out in our neighborhood. Turns out a friend of mine in high school was throwing wire rings from a pile of burned out tires at high voltage power lines seeing if he could short out a pair. He succeeded. The friend with him suffered from some sort of eye burn because he didn’t know to turn away from the giant light bulb that ensued.

Years later I learned that General Electric, not wanting to pay the royalties, wouldn’t use an inert gas to extinguish the arc in their circuit breaker so they used air instead. (at those currents just breaking the circuit wasn’t enough. you still had a ‘lightning bolt’ arcing across from terminal to terminal) The result was a large ‘bang’ when the circuit broke.

1 Like