I'm trying to build a high voltage optocoupler to implement galvanic isolation between a high voltage supply and its control signals. Since the working voltage is several kilovolts, it's beyond the ratings of all isolation ICs. Thus, the plan is to build a discrete fiber-optics isolator, using off-the-shelf transceivers and cables on both ends. I'm considering using either TOSLINK or SFP transceivers.

However, I cannot find any electrical safety ratings of fiber optics cables. When using an optocoupler, the manufacturers provide all the necessary safety data to designers, including clearance, creepage, comparative tracking index, material group, working voltage, and dielectric withstand voltage. The equivalent doesn't seem to exist for fiber optics cables. Intuituively, low-cost cables are made of plastic, and telecom-grade cables are made of glass, thus there must be a difference.

I heard fiber-based isolation has been successfully used in the industry at very high voltages. How is it possible to use fiber optics cables for galvanic isolation in a manner that conforms to relevant safety standards?

  • \$\begingroup\$ I imagine the limiting factor would be whatever the fibre optic is coated with. Perhaps you could use a black (for the wavelength being used) plastic tube between the emitter and receiver instead. \$\endgroup\$ Commented Jul 27, 2022 at 18:51
  • 1
    \$\begingroup\$ I have no personal experience with light triggering but this app note might be of use: infineon.com/dgdl/… Specifically section 4 \$\endgroup\$
    – vir
    Commented Jul 27, 2022 at 18:53
  • \$\begingroup\$ If I were a glass-fiber manufacturer, I'd hesitate to specify dielectric withstanding voltage, since surface conditions affect the result so much. Read the cautions for high-voltage probes - they usually caution about keeping their outside surfaces very, very clean. \$\endgroup\$
    – glen_geek
    Commented Jul 27, 2022 at 19:07
  • \$\begingroup\$ Afaik fiber optics would be handled like an insulator so you primarily have to consider creepage and clearance between the transceivers given by the applicable standard. \$\endgroup\$ Commented Jul 27, 2022 at 19:23
  • 1
    \$\begingroup\$ @Uwe The problem is that these ratings are usually surge ratings or functional insulation ratings. The surge voltage should not be applied constantly, a functional insulation should not be used to protect personal safety, and a basic insulation can only be used in combination with addition insulation to provide safety. It's not uncommon to see an isolator with surge ratings up to 12,800 volts, but only rated for a 3000-volt working voltage, and a even lower 700 VAC Reinforced Insulation rating. \$\endgroup\$ Commented Jul 27, 2022 at 19:55

1 Answer 1


After some additional thoughts, I think I've found the answer.

Lack of Voltage Ratings is a Non-Issue

The key is to realize that, the regulations "take nobody's word for it." The system-level (rather than component-level) safe working voltage across an insulation barrier does not appear just because a manufacturer says so - it's calculated from first principles specified by the relevant safety standards and regulations. In IEC 60950-1 or IEC 62368-1, it mainly depends on clearance and creepage, which are determined by working voltage, comparative tracking index, material group of the solid insulator. In addition, additional safety margins are added based on the pollution degree of the environment, and the level of insulation required (functional, basic, supplementary, and reinforced insulation).

Thus, it's possible to perform an independent calculation.

Furthermore, it needs to be emphasized again that an insulator doesn't have a single "safe voltage", but instead it's determined by the pollution degree and the level of insulation required. Thus, even when we're using a commercially-available, off-the-shelf isolator, we already need to do an independent calculation to find the safe voltage for our purposes. In fact, the vendor's stated "withstand voltage" or "working voltage" is often useless or even highly misleading: the actual safe voltage per regulations is often significantly lower than the working voltage rating of the isolator itself.

Calculation from Scratch

So how do we perform this calculation?

  1. Pollution Degree and level of insulation are easy to determine since they depend on the application. For in-door home, office, or lab use, select Pollution Degree 2. And to protect humans from electrical shock without grounding, use Reinforced Insulation. We also assume a working voltage of 5000 V.

  2. Then, creepage and clearance are determined by visually inspecting the optical fiber cable.

  3. Finally, comparative tracking index, material group are seemingly the most difficult problem since the material property is unknown. What can we do?

    • In turns out, the standard just allows us to assume the worst case by default.

    If the material group is not known, Material Group IIIb shall be assumed. (IEC 60950-1,

    • Alternatively, if size matters, another option is performing an independent lab test, stated by the standard:

    If a CTI of 175 or greater is needed, and the data is not available, the material group can be established with a test for proof tracking index (PTI) as detailed in IEC 60112. (IEC 60950-1,

Problem Solved!

We now have all the data we need for a calculation:

  1. Comparative Tracking Index: unknown, assume < 175.
  2. Material Group: IIIb.
  3. Pollution Degree: 2
  4. Insulation Type: Reinforced Insulation.
  5. Working Voltage: 5000 V.

Using Table 2N in IEC 60950-1, we can see that the required creepage is 50 mm, and we need to double that to 100 mm for Reinforced Insulation. We also check the clearance requirement from Table 2M, we find the required clearance is significantly lower than creepage, using linear interpolation between 2800 V @ 8.4 mm and 7000 V @ 17.5 mm, we find it's 13.2 mm.

Therefore, we just need to make sure that the high-voltage is separated by a 100 mm optical fiber. We also need to ensure the circuit board itself doesn't violate the creepage and clearance requirements, and we're good to go.


Some possible objections:

  1. IEC 60950-1 has been deprecated and replaced by IEC 62368-1.

    • True. But the creepage and clearance requirements are almost identical (because both requirements are based on IEC 60664-1, the "upstream" standard - insulation coordination for equipment having a rated voltage up to AC 1 000 V or DC 1 500 V). For simplicity I used IEC 60950-1 as an example.
  2. IEC 60950-1 and IEC 62368-1 are designed for IT and office equipment with a rated voltage not exceeding 600 V. Why are you applying this standard to a 5000 V power supply?

    • I'm fairly confident that the "rated voltage" is defined to be the input voltage from the power plug (Definition in IEC 60950-1). It's permissible for the equipment to generate a high voltage for internal use. For example, a scanner may use a high voltage Cold Cathode Fluorescent Lamp (CCFL) and a computer may use a Cathode Ray Tube (CRT). In fact, IEC 60950-1 specifically mentions the issue of high voltage starting pulses for CCFL lighting. So, for a hypothetical example of a high-voltage supply in a large office photocopier, IEC 60950-1 is appropriate.

    • For other equipment, you may need to consult a different standard, such as IEC 61010-1 for lab equipment. But the creepage and clearance calculations in those standards are mostly also derived from IEC 60664-1, and thus will have similar tables and calculations.

  3. Creepage.com says the working voltage must be under 1000 V.

    • It's a useful website for calculating creepage distance per IEC 60950-1. Unfortunately, its calculation is outdated. It's based on an outdated version of IEC 60950-1 - Edition 1 from 2001, with a creepage table that stopped at 1000 V, so it misleadingly states that the working voltage must be under 1000 V. This is incorrect. Since Edition 2 (2005), higher voltages have been added. IEC 60950-1 does not disallow a working voltage above 1000 V. Someone should contact the author of this website.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.