14 August 2012
Electromechanical relays still have much to offer
Large scale digital logic began not with the vacuum tube but the electromechanical relay. Although vacuum tubes were available at the time – and were used on competing machines – Konrad Zuse chose to build the first operational programmable computer in Berlin using electromagnetic relays, rather than tubes, because he considered them to be more reliable when used in bulk.
Employed for aeronautical calculations, the Z3 used 2000 relays and operated at a clock speed of less than 10Hz. But just two years after its completion in 1941, an Allied bombing raid destroyed not just the machine, but also practically killed off a strand of computer development based on relays.
Although relays can implement Boolean logic easily, vacuum tubes – and, later, transistors – proved to be far faster and effective. The relay was relegated to applications where considerations other than speed or density were important.
Since then, relays have come to be favoured components where isolation between the control and slave circuit is important and where the design needs high assurance that if the relay itself fails or power is removed from the control circuit, the switch will move to its open, non conducting state. For reasons such as these, the relay has long been a contender for implementing simple, but safety critical, logic functions in industrial control systems. Even today, the technology is fending off competition from semiconductor based solid state alternatives.
At its core, the basic relay comprises a metallic switching armature that is held open either by the magnetic field from the current flowing through a coil close to the switch or which is pushed shut by the field (see fig 1). Usually, the current needed to energise the coil is significantly less than the current that is expected to flow through the part of the circuit controlled by the switch. However, that is not always the case. Small signal relays may be used to pass very low currents. The relay is used in place of a transistor because no current leaks through the switch while it is open and the metal conductor offers almost zero resistance, a factor that may be essential in very sensitive instrumentation.
Latching relays make it possible to have a switch control circuit that consumes energy only while switching, rather than during the period that the relay is not in its 'normal' state. Some designs use a permanent magnet to hold the arm in place, with electrically powered coils used to overcome this force when switching between states. Others are based on a ratchet mechanism that forces the arm to its opposite state when the coil is energised.
Despite its apparent simplicity, there are many subtleties to relay design that make selection harder than simply picking a component known to handle a maximum current or voltage.
One problem with relays is that the most sensitive part of the design – the contact – often relies on materials that are expensive, toxic or both. Not only that, the materials used on the contacts need to be selected carefully to ensure that the relay will remain operational over the working life of the product it is protecting. Material selection is not nearly as simple as just picking something with low resistance. While silver remains the key element for use in relay contacts because of its excellent conductivity, it has to be hardened with other elements to prevent it wearing away too easily. Unalloyed silver also has the unfortunate habit of forming a sulphide film on its surface if left exposed to the atmosphere for long periods – and this sulphide film has a much higher contact resistance than the native metal.
One option is to coat the surface with gold, which is also an excellent conductor and practically inert. Unfortunately, gold boils easily and arcing can quickly burn the metal from the surface, as can excessive current. As a result, gold coated contacts are generally reserved for small signal relays that are expected to lay idle for long periods of time.
One of the best performing materials for making high power relay contacts is silver cadmium oxide, which is made using a sintering process, rather than being a true metallic alloy. It works well with problematic loads, although it is more expensive and more toxic than the silver nickel that has been used for decades on a wide range of industrial relays.
The silver-cadmium combination resists the tendency to contact weld when passing high currents that result from switching inductive loads. The use of more than 100mg/kg of cadmium within a product was banned by the Reduction of Hazardous Substances (RoHS) legislation when first announced, although the European Commission published an exemption shortly before RoHS became effective. The cadmium content is relatively low and does not leach out readily from discarded products. However, a number of suppliers have decided to reduce their dependence on cadmium where they can, turning to other options such as silver tin oxide or silver tin indium oxide, which are made using a similar sintering process.
Manufacturers found it hard to use silver tin oxide until the past decade because sintering did not readily yield an even dispersion of the component elements. As process control has improved, the quality of the compounds has become good enough to guarantee even wear across the contacts.
Materials issues and concerns over reliability have helped drive adoption of solid state relays, which use semiconductors to implement the switches instead of electromechanical assemblies (see fig 2). They have the great advantage of being practically silent. Also, with no mechanical switching involved, concerns over welding go away.
The heart of a typical solid state relay is a field effect transistor, although some are based on thyristor structures. The advantage of the thyristor is that current continues to flow through the device even after the activation signal is removed until the current falls below a threshold. This helps avoid large transients when the relay is turned off as the thyristor only disconnects the circuit when the current falls almost to zero.
Most solid state relays use either a single MOSFET or two transistors arranged back to back if the device needs to pass AC power as diodes within the semiconductors will block the current in one of the directions. An optocoupler guarantees the electrical isolation that the use of a simple transistor in the same circuit could not.
Solid state devices still account for a fraction of the total market for relays, but their share is climbing steadily. It should approach, if not exceed, 20% penetration towards the end of the decade based on current trends. Typically, solid state relays are more expensive than their electromechanical counterparts, so are used in applications where downtime is more important than absolute manufacturing cost. As a result, solid state relays are finding application in motor vehicles and industrial automation, to help reduce servicing costs.
However, the solid state relay faces issues, which is why it is still far from supplanting its electromechanical cousin. Its much higher switching speed can be problematic: voltage transients on the control side can activate the relay at inappropriate times.
Few circuit elements have the linearity of a metal wire, which is one reason why solid state relays are not entirely displacing their electromechanical counterparts. Radar and communications systems are contributing to a demand for high linearity because the complex waveform processing techniques they now employ cannot tolerate the distortion that semiconductor based switches can impart.
The rise of MEMS
The rise of microelectromechanical systems (MEMS) construction is helping to keep the traditional relay concept going – but on a much smaller scale. The operating principle is the same as that of an electromechanical relay, although the reduction in size means that isolation is greatly reduced as is their tolerance to electrostatic discharge. Isolation is not usually a major problem in the RF systems that can make use of MEMS relays or switches and their much higher density compared with traditional relays makes them attractive to builders of IC testers – which may demand more than 10,000 switches in the test head – as well as switched antenna radar systems and multistandard mobile phones. However, reliability is a concern.
In general, circuit design has to be carried out carefully to avoid voltage differences across the switch when it is moving as this can lead to damaging current arcs that, as with larger relays, can destroy the surface enough to prevent reliable conduction. One of the biggest limitations on endurance is atomic diffusion. The relay contact heats up when it is switched on. Eventually, metal atoms move so far that the contact can weld shut.
Stiction is less dramatic, but can cause a MEMS relay to fail in its on state as the electrostatic attraction needed to lift the switch from its contact is no longer strong enough to overcome the atomic level forces. Capacitive MEMS relays were developed to overcome a number of these problems, although they can still suffer from stiction caused by induced charges forming in the dieletric.
In contrast to a conventional relay, the capacitive MEMS switch does not provide a conductive contact. Instead, electrostatic attractions are used to convey a signal from one side of the switch to the other when the switching element is close enough to allow efficient coupling to occur. Naturally, DC signals cannot pass through the contact, but the type of RF signals needed for mobile telephony or radar systems encounter very low insertion losses.
In the early days of RF MEMS, reliability considerations favoured capacitive switches, but improvements in circuit design and construction have seen the pendulum swing in favour of switches with ohmic contacts that can pass signals from DC into the high end of the gigahertz range.
Cost remains an issue for RF MEMS, which is why, for the moment, they tend to be used in military, rather than commercial, RF systems. One further factor that holds back MEMS switches is the need for hermetic encapsulation to prevent moisture contaminating the contacts and, ultimately, wrecking the device. This sealing requirement increases expense. Flexing in the PCB can also disrupt the operation of a MEMS switch. Changes in construction techniques may overcome the cost problems with sealing and changes in design are likely to reduce the sensitivity to mechanical stress.
But the better power consumption and linearity of MEMS compared with semiconductor based switches should eventually see them adopted more widely across radio systems. PIN diodes, for example, require a constant current to maintain them in the low loss on state. A MEMS switch based on electrostatic forces needs practically no current to hold the switch closed.
As MEMS construction techniques improve, the relay may make a comeback in one of its earliest applications: the computer. The reason for the possible move from solid state to electromechanical switching comes down to the drive to bring down the energy cost of computation.
Traditional CMOS circuits face a problem with power consumption: in general, power consumption drops quadratically with the decline in voltage. Unfortunately, as you wind down the supply voltage to less than 1V, the switching time for CMOS increases dramatically. As a result the contribution to overall power consumption from leakage increases until you reach a point where overall energy consumption starts to increase again (see fig 3). The great thing about a mechanical switch is that it consumes no current when it is switched off. It is, potentially, the ultimate low energy transistor.
One plan is to replace CMOS transistors with MEMS relays. The basic design for a MEMS logic relay consists of a bendable bar connected to the source with a gate underneath the middle of the bar such that, when a charge is applied, electrostatic forces pull the free end of the bar onto the drain (see fig 4).
As with conventional MEMS relays, reliability is a concern. However, studies by groups such as one at the University of California at Berkeley have found that devices made so far can operate for at least a quadrillion cycles – which would support a low duty-cycle system such as a power meter or wireless sensor node that runs at 100MHz with a 1% duty cycle over a 10year period.
Speed is a handicap for relay based logic, but that is not necessarily an obstacle for low speed, low energy sensor nodes. Much of the switching delay is due to the time it takes to pull the mechanical beam into place. With CMOS logic, capacitance is the main enemy, so it's good practice to distribute the logic across a large number of transistors, each of which needs to handle only a small capacitance. With relay logic, the ideal is to pack as much logic as possible into one switch, feeding multiple inputs into a single mechanical switch that implements a complete logic function, rather than acting as half of an inverter. It's even possible to build multiple function gates, such as elements that switch between AND and NAND operation based on the state of an input.
According to the Berkeley researchers, relay based logic can slash power consumption by an order of magnitude and still offer throughput and density comparable to low energy CMOS logic while supporting cycle rates of several hundred megahertz. The core device is larger, but by packing more functions into the space, the density need not be lower than that of CMOS.
Despite being threatened by solid state technology on many fronts, the electromechanical relay still has a long life ahead of it and may even start to recover some ground lost to semiconductors in low energy applications.