Optical fibre links

Using optical fibre interfaces and cables, the common mode current is almost totally suppressed and the electromagnetic coupling to or from the link between between transmitter and receiver is eliminated, thereby giving the circuits immunity to RF interference, transients caused by switching high current loads or surges induced by lightning strikes and ESD.

If you have the opportunity to use a fibre link for any interface, take it: it will remove EMC issues for that interface. But of course, this comes at a cost, particularly for the optical transceivers at each end of the link, both in parts cost and in size and weight. So it is normally only justified when other factors, or an extreme interference environment, mandate its use.

Galvanic isolation on board

Signals may be galvanically isolated locally at an input or output with either an opto-coupler or a transformer.

Isolation breaks the electrical ground connection and therefore substantially removes common mode noise injection, as well as allowing a DC or low frequency AC potential difference to exist. The opto-coupler consists of an LED intensity modulated by the driving signal (in differential mode) and the light is transmitted over an electrically isolated link to a receiver or a photo-transistor. The opto-coupler is commonly available integrated into a single package for use on PCB interfaces.

This is a convenient component but its small size does mean only a small separation between transmitter and receiver, hence greater parasitic coupling. Coupling capacitance will compromise the common mode isolation at high frequencies or high rates of dV/dt. This capacitance is around 1 - 2pF per device; where several channels are isolated the overall coupling capacitance (from one ground to the other) can rise to several tens of pF. Electrostatic screening reduces the signal coupling capacitance but not the common-mode capacitance. It is best to minimize the number of channels by using serial rather than parallel data transmission. Do not compromise the isolation by running tracks from one circuit near to tracks from the other: in other words, respect the isolation barrier. This includes ensuring that no ground planes pass across this barrier - this is one place where a break in the ground plane is not only desirable but essential.

A particular problem with some types of opto-coupler is saturation of the photo-detector due to high dV/dt across the barrier. This is most noticeable when the interface is tested for fast transient burst immunity. Bursts of 1kV peak voltage and a 5ns risetime imply dV/dt = 200V/µs, which can “blind” the phototransistor in the receiving section for possibly several microseconds, losing signal for this period. Photodiodes are less susceptible to this effect, and if a phototransistor is used its base should be resistively coupled to its emitter (at the cost of some signal sensitivity). Electrically screened opto-couplers are another possible solution, as are devices specified for high common mode dV/dt: up to 15kV/µs is possible.

Opto-isolation is usually cost effective for digital transmissions, but more costly for analogue transmissions since it requires some extra signal processing. Also, both sides of the link will normally require a power supply. If this is provided by an isolating DC-DC converter, this will bring with it its own issues of RF emissions and immunity, as well as adding cost, weight and space.

You need to be sure that implementation of link isolation is considered at the system level to prevent crosstalk coupling within a cable from bypassing the isolator(s). That is, the isolation barrier must not be split between two ends of a wired cable. All wires going into an interface cable need to be isolated, or none.

Transformer and integrated magnetic isolation

Because a transformer is a passive device it doesn't suffer from photo-saturation effects. Its principal disadvantage is that it can't pass DC, as an opto-coupler can. For low frequencies, its size and weight can be an issue. Set against this, for AC signals the isolation barrier needs no extra power supplies and it becomes very efficient for high frequency data signals - its most widespread application being in the Ethernet interface. As with the optical isolator, parasitic capacitance across the barrier is a limitation.

A modern variant of simple transformer isolation is with the integration of a high-frequency transformer structure onto a single IC with signal processing on both sides, to allow digital data transfer using edge encoding, and also to allow bidirectional transfer (not possible with simple optocouplers). This makes the isolation interface largely transparent to the designer. High levels of isolation (up to 100kV/µs) can be achieved using polyimide film separation between the two halves of the device. It's also possible to integrate the DC-DC supply conversion across the barrier in the same device. Of course, such advantages aren't very cheap; and they come with the related problem that very high frequency (hundreds of MHz) power conversion, while being efficient in size terms, also creates the threat of radiated emissions which wouldn't be caused by other methods. To deal with this, manufacturers advise interleaving ground planes on different layers of the PCB to deliberately increase cross-barrier capacitance (see for instance Analog Devices AN-0971). This will of course reduce the high frequency common mode rejection of the barrier, thus negating one of its principal advantages in EMC terms.

Power supply isolation

A common issue is the requirement to isolate the internal circuit(s) from the external power supply. For an AC mains-powered product this is naturally achieved by the isolating switched mode converter (or in earlier times, the 50Hz transformer) which is needed for safety reasons, but it may also be needed in many DC-powered applications; usually because the DC return conductor cannot be guaranteed by the system builder to be at any particular voltage (for example, 0V) with respect to other system components. So if other system interfaces are not isolated, the DC supply must be.

This is not difficult to do, using either a designed-in or bought-in DC-DC converter, but as with any other switched mode supply it solves one problem by introducing another, that of conducted and radiated switching emissions. The techniques to control these emissions are covered in the section on switching converters.