Isolating Signals: How & Why?

At some point in ancient technology history, someone realized they needed to isolate a control signal. By “isolate” we mean that no common electrical signal path would exist between the control signal and the controlled device. Another term for this is galvanic isolation.

Before 5V, 3.3V, and 1.8V logic, this isolation was usually used in industrial control applications. “Ground” or circuit common in one piece of equipment could easily be at a significantly different potential than ground in an adjacent piece of equipment (and this is still an issue). Control signals could be passed from subsystem to subsystem via relays, making the circuit common issue moot. Also in use: an early, simpler form of the optocoupler — a small light-tight module with an incandescent or neon lamp and a cadmium sulfide (CdS) photo cell mounted side-by-side.

(Source: Avago)

(Source: Avago)

Crude, but effective. And, of course, the predecessor to the silicon based opto-isolator IC. The incandescent and neon lamp version are useful in applications where the controlling voltage is in the range of 12V to several hundred volts, AC or DC. These optocouplers were not useful for conveying rapidly changing signal information, but at the time, they worked just fine.

(Source: Fairchild)

(Source: Fairchild)

Once the need for passing fast-changing signals came to the fore, the silicon-based optocoupler really began to shine (pun intended). These use an LED and a phototransistor or a photodiode plus amplifier, so they could be pretty speedy. And it was easy to put multiple devices in one package — handy for moving lots of digital signals to and fro.

Now we are starting to see a new type of signal coupling device. It does not use light — no LEDs and phototransistors. Instead, the coupling is done magnetically. The device shown below is the Analog Devices ADuM144x series.

(Source: ADI)

(Source: ADI)

Again, putting multiple channels in one package is common. Better yet, this one has bidirectional capability on two of the four channels, useful in some data-com applications. And these devices can pass data about as fast as or faster than the opto devices. Isolation looks quite good and current draw is better than the LED devices — plus these can easily operate at lower supply voltages.

These nice features come at a price, however. By which I mean that the price you pay for such features is the increased price compared to a 4N25. Note also that the ADI parts do have some amount of susceptibility to external magnetic fields. A closer look at the data sheet (cited above) will shed some light (so to speak) on the degree of susceptibility.

Which methods of isolated coupling techniques have you used? What works well and what problems have you run into?

Related posts:

— Brad Albing, Editor-in-Chief, Planet Analog and Integration Nation Circle me on Google+

29 comments on “Isolating Signals: How & Why?

  1. etnapowers
    November 26, 2013

    I think that a comparison between the two methods of isolation has to take into account not only costs but also the error in codifying the same reference signals. Are there any data available about such a comparison?

  2. Vishal Prajapati
    November 26, 2013

    I have used Relays, Opto-isolators and magnetic isolators also.


    For slowly changing signals which relays can handle, it has been the most reliable and my favourite solution. As it provides much greater voltage isolation compared to opto-isolators and magnetic isolators due to its contact saperation distance. And much higher current capability. But it has limitation of frequency of operation, lifetime and current requirement.


    The opto-isolators are much cheaper than relays and a bit cheaper than magnetic isolators. They have good amount of voltage isolation capability and have less current consumption. I have successfully used it in RS232 with baud rate upto 9600 or more.


    I have used magnetic isolators from Analog devices for RS485 comm. and They have much higher baud rates compared to opto-isolators. They have lowest power consumption but having less isolation voltage.

  3. eafpres
    November 26, 2013

    @Brad–I have seen some extreme cases of isolation in transformer yards.  In a distribution transformer yards, you have 3 phase power coming in over high voltage lines.  At the point the lines enter the yard, measurements of the current are done using current transformers which allow the measurement of the line current but isolated from the line per-se.  Since the lines are from 100kV up to 500kV or so, you don't just walk up with a clamp type current meter and clamp it around the line!

  4. studleylee
    November 26, 2013

    I designed a controller for a CNC plasma table cutter. Many of the Plasma torches use a high voltage spark to create the initial plasma streamer channel, then as the current increases the voltage drops to lower DC values. As expected, this HV start coupled into the lines going to the computer and would often cause the PC to reset. I was able to use NVE GMR isolators to fix this issue and still have fast control signal rates for the stepper motors. The rise and fall times are amazing:

    •100 ps typicalpulse jitter
    •2 ns channel-to-channel skew
    •10 ns typical propagation delay
    Several comapanies make compatible parts. TI, AD, LT
    NVE was great in sending me parts to test and a mug. I've used them now in many designs.
    -Lee Studley
  5. Rcurl
    November 26, 2013

    I work on equiment that produces upwards of 200KV. I found that one easy way to get data back and forth between the high voltage side and the control side is to use inexpensive optical fiber transmitters and receivers for plastic optical fiber cable. These devices are only a few dollars each and the cable is under a dollar a foot. Just cut it with wire cutters and poke the ends into the transmitter and receiver. No connectors required. Here's a link to the manufacturer:


  6. jkvasan
    November 28, 2013


    We normally use 4N35s and 6N137s. 6N137 can communicate upto 10mbps and is a good chip but slightly expensive. We AD convert the signal and transmit throught the 6N137s.

  7. yalanand
    November 30, 2013

    @brad, I used bulk relays in  my embedded real time  projects, at that time I faced  back e.m.f problem. To solve that I used another extra diode in parallel with that which cost extra hardware and circuit is also not reliable every time. So will these semiconductor relays  overcome these type problems  to  replace traditional relays?

  8. Brad_Albing
    November 30, 2013

    @etnapowers – Probably someone has published something that speaks to the issue of errors that occur due to the transfer of data or signals thru an isolator. But I don't know specifically about such a study or an App-Note.

  9. Brad_Albing
    November 30, 2013

    @VP – I like relays for applications where they may be abused due to over-volatge or over-current situations. So industrial control applications, telephone switching, railroad signaling. That last one is a good one. The companies hat designed the early signaling systems used relays with carbon block contacts. They were sufficiently low resistance to pass enough current to light the indicator lamps and operate the motors in the signals. And the contacts would not weld together in the event of a lightning strike to the rails.

  10. Brad_Albing
    November 30, 2013

    @eafpres – well, you better not use a clamp-on meter. Else, you'd only do it once. And of course, a relay for switching the high-voltage is very specialized, motor-driven apparatus.

  11. Vishal Prajapati
    November 30, 2013

    @BA, you can still see large application of relay in the railroad signalling in India till date. They are pretty robust and surprisingly reliable. Even today automotive electronics use relays for turn indicators and for what not.


    Whenever I tried to design something in my collage projects, I would try and integrate relay in the application. As I loved to hear my electronics designs clicking.

  12. Victor Lorenzo
    November 30, 2013

    @etnapowers and @B_Albing, “Probably someone has published something that speaks to the issue of errors that occur due to the transfer of data or signals thru an isolator “.

    Back in late 80's I used a lot the datasheet books from Analog Devices and BurrBrown. Those books used to come with articles and sections providing extensive background about many things and many design hints and tested circuits.

    Most times it's a matter of finding (determining) the combined transfer function for the isolation part: the output value (voltage, current, transconductance, etc.) as a function of the input value (again, current, voltage, etc.).

    If the transfer function (light vs current) for the emitting diode is not linear, and the same happends for the collector current vs light for the photo-transistor, then you'll most likely have a non linear system. If the isolated signal is analog-to-digital converted you can compensate the nonlinealities as part of the processing phase.

    Determining the transfer function is not easy task, and it will also vary part-by-part and depend on temperature, but can be approximated.

    This note (Simple Circuit Modifications Enhance Optocoupler Performance ) provides one example that gives some ideas:


  13. Victor Lorenzo
    November 30, 2013

    @Vishal, “you can still see large application of relay in the railroad signalling in India till date “.

    That holds true also in Spain. We use them a lot (including optoisolation) for controlling high voltage and high current motors from embedded systems.

  14. Victor Lorenzo
    November 30, 2013

    Now we are starting to see a new type of signal coupling device. It does not use light — no LEDs and phototransistors. Instead, the coupling is done magnetically “.

    Perhaps my memory could fail now, but I think the idea is not new, I'm almost certain that I've seem the datasheet for one magnetically isolated amplifier before 1996.

    These two pattents cover two different isolation methods: Isolation amplifier, and Isolation amplifier with precise timing of signals coupled across isolation barrier

  15. dec66
    December 2, 2013

    Digital isolators have been on the market for some time now as an improvement over optocouplers. Analog Devices has been producing them for over a decade, and has sold over 1 billion channels of digital isolation. 

    The air core micro-transformers used in i Coupler digital isolators are actually remarkably resistant to external magnetic fields. Intuitively you might think they are susceptible, but it takes a very large field to disturb the signals passing across the isolation barrier since the transformers use an air core and are extremely small.

    The ADuM144x data sheet discusses the theoretical susceptibility. One example given is a conductor carrying 1.2 kA at 1 MHz laying essentially on top of the part (5 mm away) to cause an upset. It's difficult to imagine a real world example that would produce such large currents at these frequencies.

    For more information, here's a link to a comparison of isolation technologies:


  16. Brad_Albing
    December 2, 2013

    @Victor — Yep – those nonlinearities are surely trouble in an analog application. If you're just transferring digital signals, then less of a problem – but still an issue that you must deal with. Esp'y at very high data rates (no surprise there). Thanks for giving us a good pointer to some solid applications info.

  17. Brad_Albing
    December 2, 2013

    @eafpres – Oh, also, regarding high voltage distribution transformer yards, 3 phase power, and (motorized) relays, there's this:

    138kv Elkford, BC switch

  18. Brad_Albing
    December 2, 2013

    @Lee – I concur on the need in such applications – I did a similar design for a resistance spot welder. It was powwered from a 3-phase 480VAC input, rectified and filtered a little, so around 700VDC bus voltage. That was applied to a 4-SCR H-bridge; its output feed a series L-C plus the welding (contact) electrodes.

    It would not have had a chance of working were it not for the liberal use of optocouplers.

    Also, good choice regarding the use of optocouplers that were available from multiple vendors. Makes the design and the parts procurement processes go much more smoothly.

  19. Brad_Albing
    December 2, 2013

    @Rcurl – pretty clever way to create an optocoupler with as much voltage isolation (i.e., distance/spacing) as you need. I'm surprised you could cut the fiber with just wire cutters – that you didn't need some special flush-cutting devices.

  20. Brad_Albing
    December 2, 2013

    @JK – I've done variations on that – built a simple voltage to frequency converter or voltage to pulse width converter (in effect, a PWM) and then passed that signal thru the optocoupler. Going the other way, you can generate a PWM signal with the MCU, pass that thru an optocoupler, and then just make an active low-pass filter to turn it back to analog.

  21. Brad_Albing
    December 2, 2013

    @yalanand – Well, maybe. I will say I've had quite good results using the back EMF diode across the relay coil. It using very little space on the PC board and typically costs perhaps 2 cents, perhaps less. The solid state relays are constantly improving and cost is generally coming down. But relays still are the device of choice for very high voltage switching or very low leakage applications; or where abuse is likely in terms of over-voltage/over-current.

  22. Brad_Albing
    December 2, 2013

    @Victor – well, I guess “new” is a relative term. And I was thinking mostly of the devices that ADI makes. Those have been in widespread use for perhaps 5 to 10 years, with regular refinements of the devices every-so-often.

  23. Brad_Albing
    December 2, 2013

    @dec66 – Well, OK – admitedly, I read over that section of the data sheet rather quickly. You're right, unless you are building some very large power supplies or MRI machines, it's likely not an issue.

    Do I recall an issue with the early generation versions of these parts wherein they might latch-up or act funny when first powered up?

  24. etnapowers
    December 5, 2013

    @Victor, thank you for this comment, it's very exaustive. I've read the article at the link you suggested, it's a study of linearization of a non linear function around a working point. As you know this linearization process introduces some errors , I guess the designers have evaluated this error.

  25. etnapowers
    December 5, 2013

    @Brad: it's difficult to find such data because normally these simulation data are issued by the designers during the designing of the application circuit, as Victor Lorenzo said, these simulations include also the evaluation of linearization process of a nonlinear transfer function.

  26. Victor Lorenzo
    December 5, 2013

    @dec66: “It's difficult to imagine a real world example that would produce such large currents at these frequencies

    Receiving an electromagnetic impulse of large magnitude is not so rare. I remember a few years ago, a customer's installation was less that 1.5m over the power lines of one subway, in the station lobby. Some times the embedded PC's USB simply got crazy. Tracking the problem down took us a while, it was caused by EMI emitted when trains were starting to move (high power consumptio) and the train power line contacts (pantograph) passed for some cathenary joins. The activation of a local high current AC motor also produced large EMI on the circuits.

  27. dec66
    December 5, 2013

    @Brad: some digital isolators won't pass a DC signal, which can be an issue at start-up if no data transitions occur. I was burned by this on an isolated ADC evaluation board once. ADI's i Coupler digital isolators contain an internal refresh circuit that guarantees DC correctness 1 µs after power-up.

  28. dec66
    December 5, 2013

    @Victor: I didn't mean to imply that large current/voltage impulses don't occur in the real world. It's all about the magnitude and edge rate of the current and the distance, since those factors determine the magnetic field strength (in Guass). 

  29. jkvasan
    December 9, 2013


    My variant was implementation of a mixed signal mcu which does all the processing and send the data through serial port. The serial port signals were fed through the optocoupler and later processed further.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.