Relays are very handy components. Relays allow current to both directions on the controlled load circuit so they can be used for both AC and DC. However, relays have a finite lifetime if connected/disconnected often, produce a loud audible click and consume power in the controlling circuit because of the electromagnet.

Now, when replacing a relay with a solid-state component, you might want to:

  1. Use a BJT. However, BJTs despite their name are unipolar devices, they conduct current in only one direction.
  2. Use a MOSFET. However, MOSFETs have a body diode which means current in the wrong direction is passed always despite the switch state.
  3. Use a gate-turn-off thyristor. However, it conducts current in one direction only.
  4. Use a TRIAC. It conducts current in both directions. However, it requires AC, because the only way to turn it off is to wait for the commutation of the power line signal.

Is there such a thing as a solid-state relay that could replace a relay in practical applications? Requirements would be that it conducts current to both directions and can be turned on and off electronically, and ideally control would be as simple as it is for a relay.

I know that in most cases, you know if you have AC or DC so you can choose between (1)-(3) and (4): if it's DC, use (1)-(3), if it's AC, use (4).

A gate-turn-off TRIAC might be close to meeting the requirements but might require more tricky control than MOSFET/BJT. Is such a component as a gate-turn-off TRIAC even theoretically possible?

Or can a practical solid-state replacement for a relay be constructed as a circuit from simpler components?

The Wikipedia page of solid-state relays does in fact mention TRIAC types that work only for AC (so it doesn't satisfy the generality requirement), failing entirely to mention the existence of a gate-turn-off TRIAC. It also mentions two back-to-back MOSFETs with source pins connected together but I have some trouble understanding whether such a component can actually replace a relay, as the control mechanism might indeed be very tricky and might not work in some applications where relays satisfy an isolation requirement.

None of the top Google search results mention how a useful solid-state relay might be constructed, apart from the Wikipedia page which I have trouble understanding (how on Earth are two back-to-back MOSFETs controlled in practical applications?), and quick search of this Electronics StackExchange forum doesn't give any useful indications on how a useful, practical solid-state relay working for both AC and DC might work.

  • 3
    \$\begingroup\$ BJTs are emphatically not unipolar devices. They are unidirectional, with much more gain in forward modes than reverse, but they are definitely bipolar. That, in fact, is why they are much slower to switch than MOSFETs and why Baker clamps are frequently required. \$\endgroup\$
    – Hearth
    Jan 16 at 16:47
  • \$\begingroup\$ There are many options. Which choices are good will depend on whether you are switching AC or DC, the current and voltage levels, etc. \$\endgroup\$
    – Mattman944
    Jan 16 at 17:03
  • 9
    \$\begingroup\$ I’m voting to close this question because the OP has failed to do the bare minimum of research, such as googling Solid-State Relay. \$\endgroup\$
    – brhans
    Jan 16 at 18:22
  • \$\begingroup\$ @brhans Googling does in fact mention a trick of using two back-to-back MOSFETs with source pins connected together, but produces more questions than it answers, since controlling such a device might be tricky, and the accepted answer to this question provides a very ingenious control mechanism neither googling nor Wikipedia answers well. \$\endgroup\$
    – juhist
    Jan 18 at 17:23
  • 4
    \$\begingroup\$ Your edit has completely changed the question and invalidated existing answers. However, there are now answers that answer the new question and not the old one. I have no idea what to do here--normally you would roll back an edit that invalidates answers, but now that rollback would be an edit that invalidates answers! This question really should not have been reopened, you should have asked a new question. \$\endgroup\$
    – Hearth
    Jan 18 at 23:29

6 Answers 6


Is such a component even theoretically possible?

Of course. Use two MOSFETs back-to-back with sources connected and driven by a photovoltaic opto-coupler like this onethat can switch off a speaker: -

enter image description here

Or can such a component be constructed as a circuit from simpler components?

This is a typical solution for switching AC loads (pin B is not normally needed except if you build it yourself and want to access it): -

enter image description here

Image from here. Choose the MOSFET rating carefully and you should be in business but, take care because high-ish voltages can be very dangerous.

  • 1
    \$\begingroup\$ I think this answer is very useful, as the use of an optoisolator to charge the gate capacitances of two back-to-back MOSFETs is something I would have never invented on my own, and googling solid-state relay also doesn't reveal the fact that an optoisolator can actually be used to generate electricity, just like solar panels generate electricity. \$\endgroup\$
    – juhist
    Jan 18 at 17:22
  • 2
    \$\begingroup\$ @juhist it's a specialist opto-isolator that is used; not run of the mill types. \$\endgroup\$
    – Andy aka
    Jan 18 at 18:53

Use a BJT. However, BJTs despite their name are unipolar devices, they conduct current in only one direction.

BJTs are not switches per se. It is a common theme I see in many EE SE questions, perhaps because textbooks simplify things and don't quite explain the limits of the simplifications.

BJTs, just like any other device you have listed, are building blocks. You use them to make a switch, but as a designer it's on you to make the part you chose do the job you need done. The BJT is not a ready to use switch. Never is, never was.

The devices below all "conduct current in only one direction" (doesn't get much better than that with a diode, right?), and yet they form a rudimentary SSR, shown in the ON state.


simulate this circuit – Schematic created using CircuitLab

quick search of this Electronics StackExchange forum doesn't give any useful indications on how a useful, practical solid-state relay working for both AC and DC might work

Like above, as just one example. Diodes can be Schottkys or ideal diodes based on mosfets. The switching element can be a mosfet. The dissipation can be as low as you can afford it to be, more-or-less.

produces more questions than it answers, since controlling such a device might be tricky

That is a valid concern, but it's not a problem with the approach, but with understanding of the approach. There is no law of Nature that demands that just because you stick a label "The Switch" under a circuit, it's supposed to be simple. I'm not sure where that expectation would come from - perhaps because mechanical switches are supposedly simple to understand? I've been working a lot with relays over the years and the best way to describe mechanical switches to my mind is deceptively simple.

relays have a finite lifetime if connected/disconnected often

It all depends on how much load you put on them. I've tested a couple of rather run-of-the-mill telecom armature relays in a ring oscillator that runs at about 50-60Hz, and I got bored of it after about a billion cycles. The datasheet "ratings" for mechanical and electrical life were 2 orders of magnitude smaller. The contacts were marginally out of spec, but most the armature, coils, etc. were just fine. So it really depends on what you do with relays. They can have stupendously long lifetimes in some applications, even when switching like crazy :) Running them in room temperature ambient air and with loads snubbed so the contacts don't arc much helps a lot.

relays [...] consume power in the controlling circuit because of the electromagnet

Latching relays are used when switching is infrequent. They only consume energy to switch states, just like say a MOSFET device does :)

A MOSFET of course doesn't inherently consume energy to switch states, it's just a common way to use it. In switching power supplies optimized for ultra high efficiency, the gate charge of the power devices can be recuperated. It's an additional complication, but if you're making a supply that will be sold in hundreds of thousands of units, every Watt saved makes a big difference.


Yes. It's called a solid-state relay, as you might have guessed. MOSFET output types can be used much like electromechanical relays (beware: some are unidirectional, though most use back-to-back MOSFETs for bidirectional operation), while TRIAC output types need some care as they do function just like TRIACs.


Bidirectional AC / DC / Audio / Video solid state relays with MOSFETs can be as small as an 8-pin DIP or SO-8 package. Clare, Vishay, IXYS.


You're missing the most important traits of mechanical contacts: on/off ratio, and by extension, surge immunity.

There does not exist -- and most likely cannot -- a solid-state device as we know it, which offers such an equivalent.

Note that contactors, reclosers, automated disconnects, etc. are the same general type of component. They offer much higher voltage and current ratings, but aren't called relays, so I can reduce scope a bit. Though still not by nearly enough to give solid-state elements a chance to compare!

Off State

Mechanical contacts have an off-resistance in the deep GΩ+ range, even TΩ or more. The actual resistance or leakage is difficult to measure, as, well -- the contacts are physically not touching, there's no intended or preferred path between them, only sneak paths over insulating surfaces, and through ambient ionization in the filling medium (usually air or pressurized gas, but oil and vacuum are also used) due to cosmic rays. Generally, the resistance is inversely proportional to dimensional scale, because more air between contacts means more ionization to scavenge. So, smaller signal relays have lower leakage current than large power relays and contactors.

Semiconductors have an off-resistance in the MΩ to GΩ, strongly depending on temperature. Bandgap in semiconductors, is analogous to the ionization energy of the fill medium, or the work function of the electrodes, in a mechanical relay. Typical semiconductors have a bandgap of 1-3 eV, compared with ionization or work function in the 10+ eV range. The lower bandgap, plus intentional doping, allows thermal energy to liberate free charges, without incident radiation; this is a necessary part of operation, as the movement of those same charges enables transistance so to speak, whether of BJT or FET nature. The on/off ratio is therefore inherently limited -- though to what maximum value, I don't really know; it depends on bandgap at the very least.

Modern GaN FETs presumably could outperform Si quite nicely, if optimized (and specified) for these parameters. Mind, power devices are mostly rated for a maximum leakage of some ~µA; even if the device might have typical leakage in the nA or below, it literally costs testing time to measure less than this, so they don't bother. Real GaN FETs might perform much better, but you'd have to measure parts yourself.

Incidentally, SiC typically are rated for high leakage, because they're measured at modest VGS(off), and have generally poor transconductance (hence VGS being significant); they would have to be tested or designed differently to obtain leakage current ratings consistent with the higher bandgap.

Note that I'm using resistance loosely here; consider the worst case figure as an average/total equivalent resistance at ratings, Vmax / Ileakage. In general, these are non-ohmic leakages -- current varies with voltage at low voltages as charge carriers become more strongly attracted towards the electrodes, then saturates at mid levels as carriers are swept out and rate-limited; at still higher voltages, avalanche breakdown starts to take over and current goes up, often erratically/suddenly as avalanche discharges grow in scale as well as frequency.

Avalanche discharge, in the case of air gap, can grow through a range, from a modest multiplication factor, to corona filaments, to complete sparking. In semiconductors, usually avalanche is limited at the multiplication stage, but there are instances where breakdown can become more dramatic. BJT avalanche for example, can experience local heating, resulting in punch-through -- effectively the B layer disappears, C and E layers become shorted together, and a conductive filament remains for some 10s of µs, with a turn-on time under 1 ns. This isn't a generally useful switching mechanism, unfortunately.

On State

Mechanical contacts have an on-resistance in the mΩ or below. Even signal relays are rated a maximum of 10s of mΩ, and regularly do less. Compact relays, contactors and "solenoids"* are available with ratings of hundreds of amperes, and typical on-resistance might be ~µΩ.

*"Solenoid" is an automotive term, here; it's really just a contactor, but for historical reasons, has kept the name "solenoid". To be clear, the solenoid is the electromechanical part that pushes the contacts; a solenoid plus a contact is properly a relay or contactor. Another automotive anachronism is "condenser": much more commonly "capacitor" these days, but the name stuck, so it remains in common use in that context.

The low resistance enables not only low losses, but high surge capacity. Relay wear occurs primarily due to contact arcing, and contacts are primarily type-rated to switch some nominal load current under conditions of open-circuit voltage and load inductance, under which arcing occurs, and thus contact wear and eventual failure (with a rating of thousands or millions of cycles, depending on relative capacity, and materials used). When contacts aren't moving, resistance is static, low (if clean and well settled, of course), and extreme currents can flow, essentially limited by material resistance and time. A 3A general-purpose relay, when statically on, might handle hundreds of amperes, thousands even, under surge or fault conditions, and survive just fine. (A downside: the most likely failure condition is welding, i.e. the contacts don't open when the coil is released.)

Semiconductors can't hope to match this rating; even if we ignore off-state leakage, consider how big of a MOSFET would be required to achieve sub-1 mΩ at a 1200V rating. SiC MOSFETs are readily available in TO-247 package down to 30 mΩ or so, maybe 20; 20 or 30 of these in parallel is therefore required, and the overall package size is multiple cubic inches, and we've only handled one polarity and a fraction of the off-state breakdown rating (i.e. 1200V instead of some kV peak). And the resulting assembly will have many nF of capacitance, and who knows how much leakage.

MOSFETs are feasible at low voltages, where sub-mΩ parts are available for battery-switching application; even in very small packages! The peak ratings of these semiconductors, the chip by itself, typically exceeds that of the bond wires, lead frame or circuit board contacting it, so we can more reasonably say they are devices that can handle pretty much whatever surge you might throw at them. The problem is, on-resistance scales inversely (or worse) with voltage rating, and to handle more than a couple ten volts, absolutely massive die volume is required.


The on/off ratio extends to AC. When on, a mechanical contact's lead inductance dominates (perhaps 10s of nH, unless the design is pathological). Off, electrode capacitance dominates (some ~pF). The on-reactance might be some Ω even at high frequencies (MHz), and off-reactance some MΩ.

Controlled-impedance RF types are available as well, where inductance and capacitance are balanced to a characteristic impedance. The capacitance in the off-state is very small, so that isolation even at high frequencies is very good. The main remaining issue is stub length, which limits performance at the highest frequencies; specialized compact devices are needed for optimal performance in the microwave range (GHz).

Semiconductors are pitiful in comparison: a typical photo-MOS SSR has nonlinear capacitance, very large near zero volts, decreasing sharply above ±30V or so. A 3A 600V device might have some ~nF near zero bias, and still 10s or 100s of pF when measured at bias equal to the rating. The let-through even at audio frequencies can therefore be problematic; they become useless at even low radio frequencies (100s kHz).

Photo-MOS SSRs are best suited to transient-limited DC and AC control applications, where the application of large peak voltages, and large inrush, surge or fault currents, can be limited, or avoided entirely.

Photo-MOS SSRs also suffer from very slow switching, typically some ~ms turn-on and off. This limits the switching capacity, as power dissipation is massive during the switching period. This bug can also be a feature, however: whereas mechanical contacts and sparking can occur in fractional nanoseconds, emitting huge pulses of EMI, the slow switching of a MOS SSR causes nearly no emission. This can be advantageous in low-noise applications.

AC-only SSRs are a bit less limited, as they use a thyristor device; this restricts operation to AC only (they latch on while current flows, so can only turn off at a zero crossing), but the on-state voltage drop is low enough that they can bear fault currents, and generally offer a fusing rating -- protecting the device with a fuse of lower let-through (I2t rating), will most likely see the device survive. They are not useful at high frequencies, not so much due to capacitance (which is still nonlinear, though much more modest compared to MOS types), but because high rate-of-change (dV/dt) results in incidental turn-on. They are therefore most often used for mains-voltage switching, industrial (24/120/etc. VAC) controls, and so on.

In short, mechanical contacts vastly outperform semiconductors, in terms of on/off ratings, size and cost.

The main downsides to mechanical contacts are, just that, mechanical: the switching time is inconsistent, contact bounce is typical, the arcing (and even just mechanical rubbing without any current flow at all) causes material displacement, wear and eventual failure; and they are susceptible to contamination, as any mote of dust in the contact gap can cause malfunction. And due to all these factors, dust production is an inherent part of the device (i.e. the metal surfaces are rubbed off or vaporized with use), so the contact resistance can be unreliable.

  • 1
    \$\begingroup\$ Another parameter where solid-state beats mechanical -- the amount of control energy needed. FETs can switch with low voltage and minuscule current, while an electronic relay (solenoid relay) needs much more. Reducing the energy requirement on a mechanical contact is ill-advised because that makes it susceptible to external acceleration / vibration. \$\endgroup\$
    – Ben Voigt
    Jan 18 at 22:58
  • \$\begingroup\$ That's another good point -- and conversely, the power gain of a latching relay is arbitrarily high -- if you don't need it to switch often, that is! (Whereas the FET is limited by gate leakage, a small, but decidedly nonzero amount.) \$\endgroup\$ Jan 18 at 23:13

MOSFET-output SSRs, even relatively high resistance ones, are very pokey because the photovoltaic method produces very little current, and the MOSFETs have gate charge to overcome. They typically have a circuit that snaps them 'off' but the response time is more like a mechanical relay than a MOSFET.

There's nothing stopping you from designing a circuit with back-to-back MOSFETs, a suitable isolated DC-DC converter and a fast isolated gate driver. Such an 'SSR' could (once the DC-DC fires up, which could take some time) switch in well under a microsecond rather than a millisecond or so. You could also use really beefy MOSFETs.

The downsides are that it would need three pins for control and would necessarily draw some power continuously in the 'ready' state.

But even with huge and expensive MOSFETs it still would not be as electrically robust as a mechanical relay in some ways.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.