08. November 2019, 05:30 Uhr | Wolf-Dieter Roth, HY-LINE Power Components
Galvanic isolation is best obtained with optocouplers – development engineers have long been learning that. But in the meantime there are new techniques that present interesting alternatives.
The (infrared) light beam of an optocoupler effectively isolates high voltages. But optocouplers are no longer fast enough for today’s data technology. So what are the alternatives for galvanic isolation in electronics?
Galvanic isolation is called for in many electronic circuits, whether in test and measurement, in field bus systems or other extensive wiring in production plant to prevent differences in potential with possibly fatal effects, whether in audio and video engineering so that no loops occur, or in medical engineering for safety reasons.
Digital technology, supposedly so down-to-earth, also suffers from this phenomenon, as anyone will know who has joined a satellite receiver by an electric S-P/DIF connection (instead of the variant by fiberglass cable) to an amplifier, and then found that the digital sound, otherwise free from interference, disappeared every time a light switch in the house was operated.
Various techniques were used in the course of time to produce galvanic isolation, starting with the classic transformer and ending with the TMR/GMR coupler as the latest option. All variants are currently offered by the different producers.
Transformers have always been customary not just for voltage or current transformation but also for galvanic isolation. Whether under power as a line transformer, which has remained indispensable to date even though implemented in the meantime in switch-mode power supplies as a variant of higher frequency, or as a pure signal, data and pulse transformer – neutral transformer. Baluns, other toroidal transformers and the coreless variants of radio technology are also in the category.
Transformers work in different versions at frequencies from a few hertz up into the gigahertz region, but they can only operate on AC voltages. They miss out on slowly varying signals and DC voltage. That can only be got around with an amplitude-modulated AC voltage that has to be demodulated following the transformer. But this cuts off the upper limit frequency of possible transformation according to Nyquist to half the frequency of the AC voltage that is used.
A second weakness not where analog signals are concerned but in digital technology is the lacking pulse accuracy of transformers: They may be able to cover some powers of ten in the frequency range, however the typical squarewave data signals are often distorted. Inductances of the transformer take their toll, leading to pulse tilt, overshoot and phase shifts – not exactly right for time-critical edges.
Other weaknesses are the space need and the high crosstalk between several transformers of the same kind. They are scarcely found on digital SMD boards in a conventionally wound version, just as the core memory is long something of the past because of its size. The sole advantage of the transformer: Like its big brother the line transformer it exhibits little power loss, in classic designs, when transforming from the primary to the secondary side. In this way the secondary parts of the circuitry can often manage without their own source of energy if the transmitted signal is strong enough.
In more modern IC design – as a planar transformer on a silicon chip – such devices come with one to four channels. These offer up to 100 Mbit/s, use edge detection, and initially can transfer no DC voltages and lower-frequency AC voltages, for which reason an auxiliary oscillator of about 500 kHz is implemented in the ICs onto which these signals can be modulated. Nevertheless, unlike conventional transformers, these chips are not able to carry notable amounts of energy. So a device also needs a power supply on the secondary side. Furthermore, the magnetic lines of force do not remain in the component as in the classic transformer, meaning poorer electromagnetic compatibility. Here the auxiliary oscillator can also create EMC problems.
One alternative is a capacitive coupler. In simple form – as a potential-isolating capacitor – it can be found in just about every AF and RF amplifier. As a full galvanically isolating solution true to curve the matter becomes more complicated: DC voltage and slowly varying, low-frequency waveforms can again only be transfered by modulation of the signal onto an AC voltage.
In real circuits two channels directly cover the range from 100 kbit/s through 150 Mbit/s, and then the range from DC voltage to 100 kbit/s by pulse-width modulation. After the capacitive isolating path the two channels are again added: a functional but relatively complex solution (Fig. 1). The outputs produce 3 V or 5 V logic level, like all others except the optocouplers. Here electromagnetic compatibility is again restricted, plus lacking rise and fall times of the signal edges and delay times limit practical data rate.
Another possibility is to continue with the capacitive or inductive modulation principle and transmit with high frequency, at 2.1 GHz and with an inductive transformer, in other words no genuine HF transfer. Thus, theoretically, transmission of digital signals up to 1 GHz is possible. In practice up to 150 Mbit/s is offered.
Still, crosstalk between the individual channels is not negligible. Another problem is the influence of other high-frequency transmitters in the vicinity of the frequency band that is used, like UMTS/LTE cellphones or 2.4 GHz ISM transmitters, which are almost ubiquitous with video transmissions, Bluetooth, ZigBee, WLAN, microwave ovens and no end of other applications.
So there are likely to be problems, despite possible screening, if these devices are used in an assembly together with a »genuine« wireless system. Radiation by the components themselves, despite a transformer, is not at all to be disregarded at these high frequencies. In telecom units of every kind this variant will consequently be less than welcome.