MOSFET or bipolar to drive a solenoid valve?

I have to drive a 12V solenoid valve. In the datasheet it's written: Power: 8 W. So the current drawn by the solenoid is 0.666 A.

Usually, to drive small relay (about 10 or 20 mA) I use a bipolar transistor but with this kind of current is it possible? Or do I have to use a MOSFET because my bipolar will be burning?

• I'm not sure about BJTs, but MOSFETs can handle this type of current easily... without even breaking a sweat. – Toby Lawrence Oct 5 '12 at 15:07

I'm afraid I don't agree with Matt's conclusions.

You can find BJTs with low saturation voltages. A quick search at Digikey turned up this one. At 1 A the saturation voltage is typically 45 mV, that's equivalent to a FET with an $R_{DS(ON)}$ of 45 mΩ. Not bad at all for a BJT.
And even the 200 mV Matt used as a typical value is absolutely no problem: at 0.67 A that's 130 mW, and the "lot of wasted energy" after a year of continuous operation is 1.17 kWh. For the NSS60601MZ4 that's 0.26 kWh. Not what I call "a lot".

The problem lies elsewhere. The minimum $h_{FE}$ of 120 at 1 A is specified for a $V_{CE}$ of 2 V, i.e. the transistor is not in saturation. The 45 mV is for a base current of 100 mA and 1 A collector current. Now that's what I call a waste: the 100 mA doesn't flow through the load at all. If you control the transistor from a microcontroller you'll have the problem that a microcontroller can't source that much current. And, finally, the datasheet notes that the 100 mA is pulsed, so it may not be able to do this continuously (though I couldn't find a maximum value for base current).

That's the problem with BJTs, especially for currents higher than about 0.5 A. And here the MOSFET shines. Drive current: near zero. $R_{DS(ON)}$: what you want to pay for, 1 mΩ is possible. You'll have to look at the minimum gate voltage; for driving by a microcontroller you want a logic level gate FET.

So, if FETs are so great, why do we still use BJTs? Cost. A BJT may be 50 % cheaper than a FET. It's that simple really.

edit
Scott mentions a Darlington to overcome the driving current problem with the BJT. I failed to mention it, and I should have. Not because it's a solution, but because it's not! A Darlington has a (much) higher saturation voltage than a common BJT; I've seen values up to 4 V maximum. But even 1.5 V would be much if you have a low voltage supply like 12 V: the reduced voltage across the solenoid means that only about 75 % of the required 8 W is available, and that may be too little to activate the solenoid. Also, the transistor will dissipate 1 W, and a Darlington is more expensive than a common BJT, so the cost advantage doesn't count either.

• To supplement, MOSFETs require a higher voltage to turn on than a BJT so some applications are just simpler to implement using BJTs, e.g., current limiters. There are also subtleties that are just inherent in how each device is made such as effects due to radiation, speeds, physical size and failure mechanisms (ESD sensitivity, burnout, etc.). To elaborate a little more, the higher power MOSFETs means a larger die which means higher parasitic capacitance which means slower switching speeds and a lot of wasted power. For basic switching I'd use whatever was cheaper, smaller and efficient. – Analog Arsonist Oct 5 '12 at 17:35
• One other thing that can be useful is that the rds(on) of a MOSFET increases with temperature but a PN junction's forward voltage decreases with increasing temperature (negative temperature coefficient). Temperature compensation anyone? – Analog Arsonist Oct 5 '12 at 17:38
• Can't you just jump up to a Darlington pair if you need a higher beta? Not that it's the right solution, but the better question might be "Darlington vs MOSFET". – Scott Seidman Oct 5 '12 at 20:16
• @Scott - For me it's obvious that the Darlington is not a solution, but I should have mentioned it and explained why. I'll add it to my answer. Thanks for the feedback. – stevenvh Oct 6 '12 at 5:56

BJTs make terrible swtiches. Their Vce is typically around 300mV. That is a specification that is available in the datasheet. MOSFETs typically have a Rds on of less than 100 mOhms, making Vds almost negligible.

In this case, how much do you care that the entire 12V is dropped across the solenoid? If that is important, a MOSFET is the better option. A BJT could handle the current, but it will be like you're only getting 11.7V across your solenoid valve.

Another consideration is power dissipation, for the same current. In reality the BJT current will be slightly lower due to Vce.

BJT:

$P_q = I \cdot V = 0.666A \cdot 0.3V = 200mW$

MOSFET

$P_q = I^2 \cdot R_{ds} = (0.666A)^2 \cdot 50m\Omega = 22mW$

The MOSFET is 9 times more efficient than the BJT. That may be important if the device in question is under continuous operation.

• Thank you for your answer. I know about the VCE but I don't know about the 12V or 11.7V needed by the solenoid valve because it is not written in the datasheet. Why people usually use MOSFET whereas a BJT usually is really sufficient? – damien Oct 5 '12 at 15:16
• MOSFETs handle higher currents better, and have a negligible Vds. It's 2012 and better parts are available. – Matt Young Oct 5 '12 at 15:36
• "That may be important if the device in question is under continuous operation." For ten times higher currents, yes. But 200 mW is really nothing, and even a small SOT-23 will have no problem with dissipating it. – stevenvh Oct 5 '12 at 16:09
• Should have clarified, yes it's nothing, but operating 24/7, that is a lot of wasted energy over time. Granted, this is all conjecture. It would be nice if we had more information about what is going on. – Matt Young Oct 5 '12 at 16:14