|
|
| Elliott Sound Products | Relays - Part III |
 Main Index
Articles Index
|
You could be forgiven for thinking that this topic has been 'done to death', but relays are still one of the most efficient and cost-effective ways to switch power. DC is and always will be a problem, and unless you use a properly designed hybrid relay, contact arcing is an ongoing issue with voltages above 30V. While I have shown a number of solutions in various articles, a hybrid circuit that takes over conduction before there's more than 10V or so across the contacts remains the safest way to prevent contact arcing.
This article is intended to tie up a few 'loose ends' in the first two articles, as well as provide additional information on topics that have been covered but not always in-depth. An example is the 'efficiency circuit', which is primarily intended to reduce the holding current for an electromagnetic relay (EMR). However, it can do a great deal more - in particular speed up both activation and deactivation, something that you won't find much information about elsewhere.
Galvanic isolation is often a critical factor. This simply means that there is no 'galvanic' connection (via any conducting material) between the 'control' and 'controlled' sides. The requirements for isolation vary depending on the usage. Medical applications usually require a very high isolation and test voltage and extremely low leakage, but a horn relay in a car generally requires no isolation at all (many share a terminal for the coil and contacts). Relay (and IC relay controller) datasheets specify the insulation resistance and maximum working voltage, but it's up to the user to ensure that the isolation barrier can't be breached in normal use. Possible breaches can be caused by insufficient creepage/ clearance distances on a PCB, internal debris created by an exploded capacitor, moisture/ dust ingress amongst many other possibilities (often application specific).
An 'ordinary' relay can switch DC easily if the current is low enough. For example, I'd have no hesitation using a standard relay to switch 100V DC provided the current is limited to less than ~250mA. Most small (PCB mounting) relays only have a small contact gap when open, in the order of 0.5mm. This is just sufficient to break 500mA at 100V, but it's right at the upper limit of the capabilities of most small relays and cannot be recommended. You may (or may not) find this information in the datasheet.
Fortunately, there aren't many applications that require high-current DC to be interrupted by a relay. Loudspeaker protection is one, but that has been covered thoroughly already. Most circuits that use relays operate with AC, where the relay provides very effective isolation between low voltage circuits and hazardous voltages - e.g. the AC mains. The range of equipment that can be switched using a relay is almost unlimited, and they are ultra-reliable when correctly specified.
That doesn't mean that there's nothing more to be said on the topic. It's also worth pointing out that there have been patents taken out for SSRs (in particular) that are fatally flawed. A patent document is often a good way to get 'new' ideas, but a design only needs to be sufficiently different from others and be 'novel' - i.e. not obvious to a person experienced in the field. In some cases, it may perform poorly (in some cases not at all!), so perusing ideas is not always fruitful. A verified design means that (hopefully) the author has built and tested it, and can say with some certainty that it works as claimed.
The relay style used for the following explanations and tests is shown above. This is the type of relay that's used with Project 39, and they are readily available from most suppliers. It's rated for 10A at 250V AC or 30V DC (3A with a power factor of 0.4). The coil is 12V, with a resistance of about 270Ω. It is unremarkable in all respects, and the contact separation of 0.5mm is typical of most relays of this type.
With MOSFETs (and IGBTs), there is no static drive current, because the gate circuit is insulated from the rest of the device. However, for fast switching, you may need over 1A to charge and discharge the gate capacitance. A charge is placed on the gate to turn it on, and has to be removed again to turn it off. The current is determined by the effective gate capacitance, switching voltage rise and fall time, and is limited by any series resistance (generally between 4.7 and 22Ω). If the rise/ fall time is 1μs (pretty slow by modern standards), the charge current is determined by the voltage, rise/ fall time, and capacitance. For example ...
Icap = Vpeak × C / Rise/Fall Time
Icap = 12 × 6nF / 1μs = 72mA
Most MOSFETs and IGBTs specify the gate charge in coulombs, so to convert from coulombs to capacitance simply divide the gate voltage (typically 12V) by the charge. A gate charge (Qg) of 71nC (an IRF540N for example) has an effective capacitance of 12V/71nC, or 5.9nF. This is something of an over-simplification though, because the gate charge varies as the drain-source voltage changes. A simulation of an IRF540N switched with a 12V, 1μs rise/ fall time pulse showed a peak gate current of up to 180mA. While important for high-speed switching, this isn't a major consideration for SSRs. Relatively low switching speeds do cause high dissipation, but switching is usually sporadic - generally less than one transition (i.e. 'on' to 'off' or vice versa) in any one-second period.
There are several ways that the gate capacitance can be determined. One datasheet value you'll see is Ciss, which is the sum of the gate-source (Cgs) and gate-drain (Cgd) capacitances. For the IRF540N, that's 1,960pF (1.96nF). However, it doesn't consider the effect of feedback via Cgd, which increases the actual current that will be required from the gate driver. There are so many interdependencies that no simple formula can hope to provide an answer that's accurate, but fortunately we don't care much.
We aren't making high-speed switchmode supplies, but a comparatively simple MOSFET/ IGBT relay. Being able to provide up to ~100mA instantaneous gate current is 'nice', but people also use photovoltaic optocouplers that can only provide a few microamps. Switching is slow, but it may not matter. This is where the designer has to do his/ her homework. It's always nice to know what you can get away with and what will come back and bite you. This article is not intended to cover gate charge in detail, and here the discussion ends.
Hybrid relays have been covered in an article and a project, and the project version has been built and tested to verify that it can break any likely DC fault current up to 20A or so. This would normally cause a fatally destructive arc that will not just burn the contacts, but will probably cause the entire contact structure to be completely destroyed. One major advantage of a hybrid relay is that the semiconductors don't have to carry the load current, other than for a brief period at switch-on and switch-off. For most applications, this means that smaller devices can be used, provided they are rated for the voltage and current of the supply and load. No heatsink is required, because they conduct for such a short time.
A hybrid relay can completely solve any issues with breaking high-current DC, at any voltage up to the rated maximum. EMRs have the advantage that no external cooling is required at anything up to the maximum continuous current rating (DC or RMS). It's not unusual for the contact assembly to operate at an elevated temperature when used at full current. For this reason, many relays have a derating curve, similar to that shown for semiconductors. If the ambient temperature is greater than 25°C, the maximum current falls accordingly. This also applies to the coil, which is generally limited to a maximum of 120°C. Any heating from the contacts is also experienced by the coil, as they are in a sealed enclosure with mechanical interconnections. Not all datasheets show this information.
Another form of a hybrid relay hasn't been covered, and that uses a miniature relay (most commonly a reed relay) to control a semiconductor switching stage. While this is a hybrid in the strict sense of the term, it doesn't solve any of the issues that afflict semiconductor switches (notably SCRs and TRIACs). It's uncommon (and irksome) to use a reed relay to activate MOSFETs, because there's normally no voltage available for the MOSFET gate(s). Providing an isolated voltage source is harder than it sounds, because the DC-DC converter must provide isolation that meets international standards for safety. This generally means that it must be rated for continuous operation with at least 275V AC between input and output, with anything up to 5kV used for testing. An example is shown next.
The control is shown using a switch, but it can also be a transistor, small-signal MOSFET, logic circuit or whatever is available. Cheap and cheerful converters such as the commonly available B1212S-1W (12V in and out, 1W [83mA] rating) are completely unsuitable for mains usage, but is fine for lower voltages. These are readily available for under AU$5.00, but it's not recommended that the voltage differential exceeds ~100V for most. There are exceptions, but you have to look at the datasheets very closely if you need mains voltage isolation.
A reed relay can also control SCRs or TRIACs. Their isolation voltage and current capacity is more than acceptable, but you must select the appropriate base - some have minimal clearance between control and switch pins. Reed relays are particularly rugged devices, and while the contact clearance is small, they are usually capable of at least 200V. High-voltage types have the contacts in a vacuum, and can switch up to 15kV (for a price of course). I tested a reed switch (with minimal contact clearance), and it arced at 1kV DC, but was perfectly able to make and break 500V DC. Used with 230V AC I'd expect it to be just fine, although I have no specifications that claim that's within ratings.
The latest gate driver ICs solve all issues with providing a separate supply - see Project 198 for an example. There's another that will be covered shortly, as I have some samples on order and will run tests when they arrive. I did purchase an evaluation module, and these new ICs are very good.
There is a new type of relay available now, which is very different from anything we're used to. At present, Menlo Micro is the only known manufacturer, and these relays are electrostatically actuated. They are only available as an SMD part, and are designed for RF switching at up to 3GHz. They can be used with lower frequencies (including DC), but there are some particular restrictions if you wanted to switch DC.
These are MEMS (micro electromechanical systems) devices that use IC processing techniques to fabricate sub-miniature mechanical structures. The technique isn't as new as you may think though, as it was patented by NASA in 1974. The first patent I came across dates back to 1933! There are quite a few patents covering this technique, but adoption has been minimal because a comparatively high voltage is needed to create the piezo deflection needed to activate a set of contacts. The Menlo Micro device uses an internal charge-pump to generate the voltage needed. The biggest issue with this technique is getting a fairly rigid piezo element to flex far enough to operate a set of contacts.
Because these are highly specialised, it's unlikely that too many hobbyists will be experimenting with piezo relays any time soon. I have no idea of the pricing - this isn't disclosed, so we can probably assume that they are expensive.
Electrostatic relays are widely represented in patents, but are few and far between in real life. The general idea obviously appeals to any number of inventors, but the requirement for a high actuating voltage and minimal contact pressure mean that they are generally impractical. There may well be some specific areas in research where they can be utilised, but don't expect to find any from major suppliers. Since they work by electrostatic attraction/ repulsion, the available force is inversely proportional to the electrode spacing. With 'reasonable' voltages, the electrodes must be very closely spaced (and therefore providing minimal travel), and may require a vacuum to prevent arcing and/or contamination. I don't expect that any readers will ever use one, and MEMS processes are the most likely to produce a usable device. I'm not convinced that there's much merit overall, but the actuating power will be very low, which may be an advantage in some systems.
If you want more information you're limited to looking through patent documents, as I found almost nothing other than patents and 'scholastic' papers that have to be purchased. I expect few people will bother.
Note that the term 'static' relay is sometimes used when referring to solid state relays, as there are no moving parts. This is rather unwelcome terminology, as it only adds confusion without adding useful information to the reader. 'Static' and 'electrostatic' have very different meanings, although we refer to 'static electricity' as the high voltage generated by walking across carpet (for example) and the resulting discharge - often accompanied by the person exclaiming 'rudeword!' with some gusto. 
Something that was covered in the Relays, Part I article is a so-called 'efficiency circuit', used to reduce power once the relay has pulled in. However, the explanations were simplified. Most relays will continue to hold with as little as 1/10th of their rated voltage, but it's safer to not allow the current to fall by more than ~65% from the rated maximum. For example, a 12V relay may have a 'must release' voltage of 1.2V, but it wouldn't be sensible to allow the coil voltage to fall below 4V (33% of the rated voltage). For the purpose of this explanation, a 'small' relay is a 10A SPDT (single-pole, double-throw) type with a 12V coil having a resistance of 270Ω.
Mostly, an efficiency circuit is expected to reduce the coil current, but it can do so much more if you need it. The greatest gain can be in speed - with an efficiency circuit the relay activation and deactivation times can be reduced significantly. This point is rarely made, and I've not seen any analysis performed elsewhere to show how this can be achieved.
There are two things that you can do. The first is to use a higher supply voltage to activate the relay. Pull-in time will be reduced dramatically, and the efficiency circuit will then reduce the voltage to (say) 4V while the relay is activated. The second trick is to use a resistor instead of a diode in parallel with the coil (actually a resistor and diode in series).
The resistor speeds up the deactivation time, but because the coil is only receiving 33% of its normal current, the relay will drop out even faster. The lower current means less stored charge in the magnetic circuit, so it will release in less than 5ms (for a relay of the general type shown in the photo).
If the coil resistance is 270Ω, the normal current would be 44mA. If we reduce that current to 15mA using a 1.2k resistor (operating current will be 16mA), the holding power is 390mW (vs. 530mW). The only other part is a capacitor, selected to ensure that the coil gets the full 24V at power-on, and still has at least 12V after ~10ms. That indicates a 33μF cap (close enough).
Even if you only use a 12V relay with a series resistor (270Ω for this example) and a 24V supply, pull-in time is already reduced. This is because the relay coil has inductance, and that delays the current risetime. With a higher available voltage (and resistance), the DI/Dt (aka ΔI/ Δt, where Δ indicates rate of change) is increased. If we assume ~1.5H coil inductance, the risetime is halved when the 270Ω coil is powered from 24V via a 270Ω series resistor. Adding a 33μF capacitor in parallel with the resistor halves that again!
With this, we can decrease the risetime from 12ms (12V supply) to 6ms (24V supply with a 270Ω series resistor), down to 3ms (33μF in parallel with the resistor). The current risetime is one of the things that affects the on-time, with the rest depending on mechanical inertia and the distance that has to be covered - almost always less than 1mm with small relays.
Release time also depends on the coil current. If the current is maintained by a parallel diode, the current takes a significant time to fall below the 'must-release' value. Without a diode, the current collapses almost immediately, but this creates back-EMF that can destroy the driving transistor. It's not at all unreasonable to expect the back-EMF to exceed -400V with a 12V relay. A diode reduces that to -0.65V, but current is maintained for around 10ms. This delays the magnetic release, and the mechanism still has inertia that delays the release a bit longer. The 'typical' release time for a standard small relay is around 10ms when a parallel diode is used.
This can be reduced to around 4ms simply by just using the diode in series with a resistor having the same value as the coil. The back-EMF will be 24V - it can be determined for any resistor value by using the ratio of the external resistance divided by the coil resistance, plus 1. A 270Ω coil with an external 560Ω resistor will generate a back-EMF of 36.8V (with 12V across the coil). The diode now only serves to prevent the external resistor from dissipating power needlessly.
If an efficiency circuit is used, the coil current is reduced during normal operation, so there's less energy to dissipate when the current is interrupted. This makes the release time faster again. If you really need the fastest possible activate and release times from an EMR, the next circuit employs both a high-speed efficiency circuit and a rapid dropout due to reduced operating current and allowing a higher back-EMF.
The circuit shows both techniques in use, using a higher than normal voltage and a very basic efficiency circuit that drops the coil voltage to just under 4V after C1 has charged. With 2kΩ in parallel with the relay, it releases out in about under 2ms, vs. ~6ms if the resistor is shorted (leaving just the diode). Because the voltage is reduced, the back-EMF is limited to about -17.5V.
Although the 'control' is shown as a switch, it can be a transistor or a small-signal MOSFET, wired either in the +18V or 'ground' connection to the circuit. A typical use may be a 2N7000 wired from the bottom of the relay circuit to ground/ -Ve supply. The efficiency circuit is a single 'block' of circuitry, with the relay, 2 resistors, one diode and one capacitor. It's polarity sensitive, but can otherwise be used with any switching circuit that you may already have wired in the equipment. Note that the back-EMF is higher than normal though - typically about 30V with the values shown (and the 18V supply). The supply voltage can be anywhere between 15V and 24V, with R1 adjusted to suit. R2 will typically be somewhere between 1k and 2.2k - a higher value releases faster but has a higher back-EMF.
This arrangement provides more than enough current to keep the relay activated, but still ensures that the release time is minimised. It's possible to get it better, but 2ms is very respectable, and the arrangement only adds three cheap parts - two resistors and one capacitor. Operating current is reduced from the nominal 44mA to ~22mA, a total power dissipation of 264mW (compared to 528mW for the relay alone). Pull-in time is around 4ms - much faster than if the relay were powered from 12V.
It's also worth examining the overall efficiency improvement. The coil (and added series resistor) power may be reduced to around two-thirds the normal (so from 530mW to 390mW as described above), but the contact dissipation will remain unchanged. It's unlikely that it will be increased, because the relay's armature will remain fully engaged. However, even if the contact resistance is only 10mΩ, the contact dissipation is 1W at a current of 10A. For the test relay, I measured a N/O contact resistance of 10.4mΩ when closed, and this is 'typical' for this style of relay.
The results described above are based on simulations, which are very accurate if all influences are allowed for. However, if I'm to make assertions about the operation of a relay, then a proper bench test has to be performed. Without that it's just supposition because we're dealing with an electromechanical sub-system.
With an 18V supply and a 1k series resistor in parallel with 33μF (not 100μF as shown - I wanted to test 'worst case'), I measured 1.6ms dropout time with a 2k back-EMF resistor. Without the diode and resistor dropout was only 1ms, but with no suppression the back-EMF was far too high (well over 100V). The static coil voltage was 3.8V, so the holding current was just over 14mA. Used 'normally', dropout was 10ms with 12V and an anti-parallel diode. Activation time was only 4ms with the efficiency circuit, vs. ~10ms without. The pull-in measurements are difficult to perform accurately due to contact bounce.
I used a scope set up for single sweep, triggered from the relay supply (after the switch). Contact release was picked up using the second channel of the scope, with a resistor from an external supply to the 'N/O' (normally open) contacts, and with common connected to ground. This lets you see the instant that coil current is disconnected and also the instant that the N/O contacts open.
There is something you need to be aware of when an efficiency circuit is used. This is especially true if the supply voltage is the same as the relay coil's voltage. If the relay is activated/ deactivated repeatedly, there's a maximum operation rate that can be achieved. This is imposed by the feed circuit (R1, C1), and if you don't wait long enough for C1 to discharge, the relay may not reactivate.
With the values shown, C1 can be considered fully discharged after 5 time-constants. One time constant is simply R1 × C1, so ~60ms. After deactivation, you need to wait for at least 300ms before attempting to activate the relay again. This isn't a real limitation, because if you tried to operate an EMR more than 3 times per second, it won't last very long. Even a relay with a claimed life of (say) 1×106 (1,000,000 operations) would only last for about 3.8 days.
If you need that sort of switching frequency, there's only one choice - an SSR. Use MOSFETs for DC, and a TRIAC (or back-to-back SCRs) for AC. There are countless industrial applications that do need fairly high repetition rates, but even there a maximum rate of 7 operations/ second isn't a limitation.
Somewhat predictably, I have no intention of trying to cover every possible application for relays (of any kind), because they are limited only by the imagination of designers. If you are designing a circuit that requires switching, you need to select the best switching device to suit the application. I doubt that anyone would consider a switchmode power supply using an EMR to be sensible, even if so many explanations show 'switches' that look just like the schematic representation of a mechanical switch.
You must also be careful if the assembly housing the relay is subjected to mechanical shock or vibration. Because the coil current is reduced, there's less magnetic force available to keep the relay closed. Mechanical shock might cause the relay to release spontaneously, so if vibration (etc.) is present, you must perform thorough tests to ensure that the relay remains activated under all foreseeable conditions. If not, you'll need to increase the holding current until it's stable.
There are other options for reducing the coil current, but most get complex and expensive fairly quickly. An integrated switch such as the MAX4624 can be used for example, but there are some serious drawbacks. For example, the IC has a maximum voltage rating of 5.5V, and if you used a 12V relay, the maximum voltage you can apply is only 10V using a sensible 5V supply. An example is shown next, but at almost AU$8.00 each, the MAX4624 will cost more than the relay it's controlling. The example comes from Stackexchange (by 'stevenvh'). It's an elegant solution, but is not without its problems.
Cost aside, there's also an inevitable delay as the 100μF cap (C2) has to charge before the switch is allowed to change state. This delay is provided by R1 and C1. When the 'control' switch is closed, voltage is applied to the relay and C2 via D1. A few milliseconds later, the voltage on pin 1 is high enough for the MAX4624 to change state, and the relay voltage is boosted to about 9V, which should be enough for it to energise. C2 discharges, leaving the relay coil powered with about 4.3V, enough to keep it held in. Unfortunately, the control circuit has to provide the power to charge C2 and the relay current, making it a fairly unattractive proposition. It is clever, but somewhat impractical.
Other switching schemes can be used instead, but most will simply add parts for no great advantage. The advantage of the circuit shown is that it lets you use a 12V relay with a 5V supply, with a resulting power saving. The disadvantage is that the relay must have a 'must activate' voltage of no more than 8V, and the IC is expensive. The fact that the 'control' circuit has to provide the current to charge C1 and power the relay is a further disadvantage. Control circuits are normally expected to be activated by minimal current. That can be achieved, but it needs more parts.
One technique that has been used is to power the relay from a PWM (pulse width modulated) supply. This avoids dissipation in resistors, but circuit complexity is increased quite dramatically. I'm a little surprised that no-one has offered an IC solution, as it would be quite useful.
There are several ICs designed for driving relays, one of which is the DRV110 (TI), which is designed to provide a period of full voltage, after which the IC operates with PWM (pulse width modulation) to reduce the power. Everything can be selected with external resistors and capacitors, and an external MOSFET is used to drive the relay coil. This is a good option, but for relays that may only require 500mW or so it's not worth the effort.
The circuit is adapted from the datasheet, and with Rosc grounded the default frequency is 20kHz. This is a simplified circuit, based on the 8-pin version of the IC. Some of the pin names are (IMO) suboptimal, and 'keep' doesn't quite measure up - the capacitor determines the time that full current is applied to the relay coil. If Rpeak is 0Ω, the maximum current is the default of 300mA. The 14-pin version of the DRV110 has a 'hold' pin, so the holding current can be specified. The default is 50mA, so this IC would be pointless with a general-purpose 12V relay with a 270Ω coil, as they only draw 44mA anyway.
This type of device is well suited to relays that have a high coil current, particularly those that require minimal wasted current or where the maximum coil current is designed to be short-term. High voltage relays and small contactors are examples. A simple PWM efficiency circuit can be made using a 555 timer, with a bit of extra circuitry to stop oscillation (with a 'high' output) for a couple of seconds after power is applied.
PWM efficiency circuits usually provide a useful reduction of the relay's release time, because the holding current is lower than normal. This isn't why they are used, but it comes free.
A 555 timer makes a fairly nice PWM efficiency circuit, although it uses more parts than a dedicated IC. The example shown will reduce relay coil dissipation from ~520mW to ~130mW by halving the current after the timeout set by R1, C1 and Q1 (the oscillator starts after about 150ms). The circuit is presented as an example, but there's not much room for simplification. Ideally, you'd use a 7555 (the CMOS version) as that draws less current, and you don't need C2. If this were made using SMD parts it would be tiny, and the cost would be minimal. It is fully programmable if the oscillator is changed from the 'minimum component count' version to a standard astable. As it stands it's pretty good, but unless you are really worried about current drain there's probably little point.
One type of relay hasn't been covered at all in the other articles, mainly because they are fairly uncommon and many people will never of heard of a 'step relay'. I don't have any, so the photo was 'borrowed' from the Net, but they are unique. The actuator operates a small wheel that either forces the contacts to open, or allows them to close. Momentary power will activate the relay, advancing the stepped wheel to the alternate position. The contacts are shown in the closed position in the photo. Note the distance the armature has to move. This indicates that the voltage/ current needed to change the contacts from open to closed (or vice versa) will be significantly greater than a normal relay, but of course it's only momentary.
Some allow a preset sequence, with up to four different sets of connections from a pair of contacts. These are generally fairly expensive, and aren't particularly readily available. The one shown has a single contact, but has provision for a second set that's not fitted. One issue with these (and bipolar latching relays) is that there's no simple option to provide feedback to a controller so it knows the current state of the contacts. This is most unfortunate, because if the controller and the relay are out of sync, there's a chance that 'bad things' can happen. Just how bad depends on the application.
Without a feedback mechanism, one must go to some trouble to find out if the relay is open or closed. This adds complexity, and partly negates the gains obtained by using the step relay in the first place. With a plastic mechanism, I wouldn't expect the unit shown to have a long life, certainly not when compared to a conventional EMR. Is it useful? That depends on your application, and how much trouble you're willing to go to to provide feedback. Without that, a power failure could easily see a controller and the load(s) at indeterminate positions in their normal cycle. For example, the lights could easily be on when the controller thinks they're off, and vice versa.
Without a separate set of isolated contacts to indicate the current state, you'd need to add a circuit to detect the presence of voltage/ current in the controlled circuit, and send that data via an approved isolation device to the controller. It's not a major issue - a couple of resistors and diodes plus an optoisolator will do it, but it's more circuitry that can fail over time. Using parts that aren't readily (and consistently) available leaves you open to a system that can't be repaired once the parts can't be obtained any more.
A step relay that used to be used in the millions was the uniselector, used in old 'rotary' telephone exchanges. These were a work of art, beautifully made, precision stepping switches that were designed to be operated countless times a day. This was known as the Strowger 'step-by-step' system, named after the man who invented it. These exchanges ('central offices' in US parlance) were driven by telephones with rotary dials, but electronic phones were developed that could emulate 'decadic' dialling - a string of pulses corresponding to the digit on the rotary dial. Uniselectors had 10 active connections, corresponding to the digits '1' to '0' (1 to 10 pulses respectively). Later versions used both vertical and rotary positioning, providing greater flexibility. Predictably, a complete discussion is outside the scope of this article, but there's a lot of info on-line if you find this interesting.
Many years ago, we only had BJTs (bipolar junction transistors) for 'solid state' switching. They have been supplanted in almost every application by MOSFETs or IGBTs (insulated gate bipolar transistors). One of the main problems is simply base current - this must be provided to turn on a BJT, but the power is wasted. A BJT can have a very low saturation voltage (fully on), but that requires a significant base current.
If a BJT has an hFE of 100, you need to supply a base current of at least 1mA to switch 100mA efficiently. Normally, you'd add a safety margin and supply 2mA minimum base current. This becomes a real problem when you need to switch 10A or more, as most BJTs have reduced gain at high current, so even more base current is required. On the positive side, a saturated (fully on) BJT can have a low collector to emitter voltage, which may be around 550mV with a collector current of 10A (1A base current, MJL21194 transistor). The power lost at the collector is 5.5W, with another 1.3W dissipated by the base-emitter junction.
Compare that to a MOSFET (even a lowly IRF540N) - static gate current is zero, and the saturation voltage is ~440mV, due to the RDS (on) of 44mΩ at 25°C. It's not hard to see why MOSFETs have taken over for switching, and they are much faster as well.
There are a few applications where BJTs are commonly used, but they are almost all low power, low speed circuits where the limitations are not a concern. Usage in SSRs is close to zero, as they are generally unsuited to this application.
IGBTs (insulated gate bipolar transistors) share the low gate drive benefits of MOSFETs with the high voltage capabilities of bipolar transistors. These are generally faster than standard BJTs but slower than MOSFETs, and are most often used where high voltages and/or high current must be switched. For example, the Toshiba GT40WR21 has a rated voltage of 1,800V, with a 40A current rating. They are available as modules (3-phase, half-bridge [totem-pole] connections) with voltage ratings up to 4,500V and current up to 1,800A (same device!). However, if you have to ask the price, you can't afford one.
An IGBT that can handle 600V at 280A can be obtained for less than AU$12.00 if you ever need to handle that much power. An N-Channel IGBT (by far the most common) essentially combines a low-power MOSFET driving a high-power PNP transistor. They are particularly rugged, and are used in high-power SMPS, UPS systems and inverters (induction cooktops, microwave ovens, motor speed controls, etc.). Their total dissipation might be greater than a BJT, but with no gate current to speak of the overall efficiency is very high. The market appears to remain strong, as new devices are released fairly regularly.
The internal structure of an IGBT is not two separate devices - everything is formed on one die. However, the 'equivalent circuit' is fairly accurate. IGBTs are thought by some to be 'old-hat' due to the availability of SiC (silicon carbide) and GaN (gallium nitride) MOSFETs, but they are still very common, especially where cost is at a premium. It's rare to see them used in SSRs, although it is possible. If used with AC, an 'anti-parallel' diode may be required, because there is often no intrinsic diode as found with MOSFETs it's included in some, but not in others).
Also unlike MOSFETs, IGBTs do not conduct bi-directionally - current can only pass between collector and emitter when the gate voltage is above the threshold. I haven't described any IGBT relays in detail in any of the articles covering SSRs and hybrid relays, simply because they are not suitable for use with audio, and MOSFETs are usually a better choice for medium-power AC control. However, there's no reason that an IGBT cannot be substituted where the designer thinks it's appropriate.
An example of an IGBT SSR is shown above. It uses the same scheme as shown in Fig. 1.1, with added anti-parallel diodes (D1, D2). These must be rated for the same current as the IGBT, because they have to carry the full current when the IGBT is reverse polarised. As noted above, this is not a common arrangement, but it will work well. Each IGBT will dissipate a peak power of 23W (at 10A, and based on a 2.3V saturation voltage). The average dissipation will be around 6.5W for each IGBT plus about 2.5W for each diode (device dependent of course). That's not wonderful, and a heatsink is essential for both IGBTs and diodes. This is one reason that you don't see IGBTs used for SS relays - their dissipation is too high.
In general, IGBTs are used where high voltages and currents are required, along with moderate switching speed. Around 60kHz is usually considered the upper limit, but it depends on the specific device. MOSFETs can operate very much faster and it's not uncommon to see switching speeds of 500kHz or more. An IGBT also has an intrinsic forward voltage, nominally 0.65V but that's only at low current. It's common to see a forward voltage quoted as around 1.5 to 3V or so at maximum current. A MOSFET has an intrinsic resistance, RDS (on), so the voltage across the device can be calculated with Ohm's law. Losses exist in all switching devices (including EMRs). For semiconductors there are two types - forward conduction loss and switching loss, with the latter determined by the transition time between 'on' and 'off', and the switching frequency.
The loss in an EMR is due to the contact resistance along with the resistance of the contact arms. The latter is minimised in contactors (very large relays) by improved construction techniques that don't rely on thin phosphor-bronze (or similar) spring materials to carry the contacts. These are described in Relays, Part I.
So, you can use IGBTs for relays, but unless you have a voltage that's out of range for MOSFETs (too high), they aren't a good choice. On the positive side, there are no issues with holding current, spontaneous re-triggering due to ΔV/ Δt constraints or other undesirable effects (including high electrical noise) that you get with TRIACs or SCRs. You pay for it with higher dissipation though - a TRIAC at 10A will dissipate about 10W, vs. almost double that with IGBTs and diodes (a total of about 18W based on the figures shown above). SiC and GaN MOSFETs are eroding the advantages of IGBTs to some extent, but if you happen to need a 1MW inverter, it will still use IGBTs [1].
The rapid increase in the uptake of electric vehicles has seen an increase in the number and variety of relay solutions offered by major manufacturers. While you might expect that these would all be 'solid state', that's not the case at all. For safety isolation, no-one will rely on MOSFETs or IGBTs because they fail short-circuit. High voltage relays were once more of a curiosity for most designs (power distribution systems excluded), but as automotive battery voltages increase, relays that can safely and reliably break 800V or more have become a requirement. In many cases these devices will be classified as a contactor, but that's simply a word that means "big relay".
One example is the KILOVAC EV200 Series Contactor from TE Connectivity (aka Tyco), which are designed to handle up to 1,800V or 1,000A (but not both at once!). Like many similar high-power relays, these incorporate either internal or external efficiency circuits (economisers) to minimise coil dissipation (see Section 3.4). The contacts are specially designed, and many are polarity-sensitive. Operation with reversed polarity requires derating to minimise contact erosion. Most have hermetically sealed contacts, and the contact enclosure may be evacuated (a vacuum) or filled with gas (hydrogen, nitrogen, or 'exotic' gas mixtures).
Not long ago, a distributor search for 'high voltage relay' would get few results, but that has changed. The EV200 series mentioned above is popular, but as expected, a 900V, 500A contactor won't come cheaply. However, a couple of hundred dollars isn't much in the greater scheme of things, and a great number of those you'll find are designed specifically for electric vehicles and charging stations. The available range can only grow, as electric vehicles become more popular.
Mechanical contacts are the only option when 100% reliability is essential. Even if the contacts arc continuously, the arc will stop when the contacts have been eroded to nothing, so there may be a very nasty fault current, but it won't last long if it has an 800V battery system behind it. Semiconductors will just melt and become a short-circuit, so unless there's a suitable fuse there's no safety mechanism. A fault condition will result in serious damage, but a suitable mechanical contact system may be able to clear a fault before major damage is done.
As you can see from this and the other two relay articles, there's so much that you can do if you really need to. It's not often that you need very fast activation from an EMR, but reducing the release time can be very beneficial. However, like all things it must be taken in context. A DC detector such as Project 33 must have a delay to accommodate low frequencies, and that can't easily be reduced. It's certainly possible to use more sophisticated circuitry to detect DC faster than the 50-60ms detection time of P33, but that simply leads to far greater complexity and more opportunities for things to go wrong.
If a DC detector takes 50ms to deactivate a relay, it really doesn't matter much if that's extended by a few milliseconds as the relay drops out. This reality notwithstanding, there's no good reason to delay the deactivation any more than is dictated by the laws of physics. The efficiency circuit is such a simple concept, but it's not used very often which is a shame. One reason that I didn't suggest it for Project 33 is that it requires some calculations and (possibly) a bench test to make sure that it works reliably - particularly for the hold-in current.
There are so many possibilities that it's simply not feasible to cover them all. Some devices are better suited for use as relays than others, so trying to use BJTs (for example) is not recommended. It's up to the designer to work out the best technology for any given application. Many applications just need galvanic isolation between 'safe' low-voltage circuitry and the mains, and this is something that EMRs have been providing for over a century, and they remain one of the most popular switching devices of all time.
Rarely considered is the gain of a relay. If it takes (e.g.) 44mA to control a 10A load, the gain can be said to be 227 (10A / 44mA). The nice thing is that this requires almost no support circuitry, no heatsink, and it's just an inexpensive part that is soldered into a PCB. Nothing else comes close.
Of course, the end result depends on whether the drive (controlling) and controlled circuits need to be isolated or not. Non-isolated circuits are very common, although they are generally considered to be 'simple' switching circuits. With SSRs, everything gets harder when isolation is required, unlike EMRs where it comes free. DC remains a problem though. 
New ICs have made this easier, especially when you need high isolation voltage (controlling mains voltage for example). This is always a particularly difficult undertaking if you build your own circuit, as it must be safe under all conditions. This is one reason that EMRs have remained so successful - they provide the required isolation easily, and the constructor doesn't have to do anything special.
Active arc quenching, hybrid relays or just an SSR by itself can all prevent contact damage with DC, and the techniques shown here all work ... albeit with caveats in some cases. Any design has to be optimised for the task, and this is done during the development of the project. You also have to get your priorities right, as saving a couple of dollars and ending up with an unreliable product isn't a good trade-off. Compromise is always a part of design, simply because building something that can never fail will cost too much (and it will use commercial products so it may fail anyway!). Engineering is (at least in part) the art of compromise.
There is only one reference in this section, as the others are covered in Part I and Part II in this series. The reader should also read Hybrid Relays, as this discusses more options. The article Solid State Relays has more information on the options available, and covers both advantages and disadvantages of each type.
1 Wide-bandgap semiconductors: Performance and benefits of GaN versus SiC - (SLYT801, TI)
| Main Index
Articles Index
|