Discretion Is the Better Part: Choosing Transistors

Editor's note: The column “Rarely Asked Questions: Strange but true stories from the call logs of Analog Devices” will be published monthly in Planet Analog for your learning enjoyment. Here is the second installment.

Q: How do I choose a discrete transistor?

A: Discreetly. (Sorry, I couldn't resist that.)

Actually, not too carefully. I have seen engineers agonizing for days about the best choice for a discrete transistor for a slow logic inverter, and I have seen them near break down because they could not obtain exactly the same transistor that was used as an emitter follower in the application note.

In fact, provided that some basic issues are addressed, many very different transistors will work equally well in a large number of circuits.

In many applications, it really doesn't matter whether you choose a bipolar junction transistor (BJT) or a MOSFET. They'll need slightly different circuitry, of course: the transistor will probably need base resistors, while the MOSFET will not; and the MOSFET will have a larger input capacitance, which must be considered when assessing stability. Both work well as solid-state triodes in amplifiers, oscillators, and logic, however. The BJT draws base current but has lower capacitance; the MOSFET has infinitesimal gate current, but quite large gate capacitance. The physical properties and choice of device do matter in some cases, of course, such as where the thermal characteristics of a BJT's base-emitter junction are used for temperature measurement.

Take care is to ensure that whatever device you choose has the correct polarity, (Is it NPN/N-channel or PNP/P-channel?) and make sure that your circuit does not allow whatever transistor you choose to exceed its absolute maximum ratings, both steady state and transient.

For the rest, you must understand what characteristics are important and which ones can vary greatly without having much effect on your circuit's performance. Choosing Transistors considers these issues in some detail, and concludes that the best procedure for choosing a transistor should conclude:

Any device having characteristics better than xxxx, yyyy, and zzzz is likely to work in this circuit; SPICE analysis has shown that the 2Naaaa, 2Nbbbb, and 2Ncccc all work in simulations; and prototypes built with the 2Naaaa definitely work well. However, many other transistors having similar characteristics should be equally acceptable.

Whatever device you choose, do make sure it is discreet as well as discrete. You don't want it to babble to all your colleagues about how little care you seem to take when selecting transistors, even though it's actually good engineering not to over-specify. In the words of Henry Ford, “An engineer can do for a nickel what any damn fool can do for a dollar.”


James Bryant has been a European applications manager with Analog Devices since 1982. He holds a degree in physics and philosophy from the University of Leeds. He is also C.Eng., Eur. Eng., MIEE, and an FBIS. In addition to his passion for engineering, James is a radio ham and holds the call sign G4CLF. He can be contacted at .

21 comments on “Discretion Is the Better Part: Choosing Transistors

  1. Davidled
    April 22, 2014

    In most case, heat is generated surround by MOSFET. The size of heat sink could be considered, not too big, not too small.  All heat should be cool down.  Also, location in the board provides the efficiency of heat divergence. MOSFET package type reduces the heat amount.  

  2. samicksha
    April 23, 2014

    You keep good point @Daej, keeping fact in mind that devices such as power transistors could not dissipate heat and are insufficient to moderate its temperature which makes heatsink even more significant.

  3. Davidled
    April 23, 2014

    Adding Fan and making hole in the package of PCB Board could help heat management. As alternative method, Aluminum package also support heating dissipation compared with plastic.  

  4. Steve Taranovich
    April 23, 2014

    Please also see James' article on EDN 

  5. amrutah
    April 23, 2014

    There are also techniques used where the body is connected to the ground planes which are usually made wide and they are routed to far-off to vent out heat.

  6. samicksha
    April 24, 2014

    Anyone out here who tried using online heat sink calculators for prediction of fluid flows.

  7. Netcrawl
    April 25, 2014

    @Daej I agree with you adding fan could help heat management or reduce heat, if the heatsinking is poor MOSFET can be destoryed by excessive temperature. 

  8. Davidled
    April 25, 2014

    Heatsink would be selected by package cooled, attachment method with length, width and diameter.  Its thermal management might be calculated by material and volume with shape.  Material could be a main factor. Hence, formulation would be changed depending on material.  

  9. geek
    April 25, 2014


    “An engineer can do for a nickel what any damn fool can do for a dollar”


    @Steve: I think that's a very apt statement which not a lot of engineers pay attention to. The idea of controlling the cost is not just restricted to business people and the procurement guys but the engineers also should see how they can contribute to saving the resources within the company.

  10. geek
    April 25, 2014

    @DaeJ: It may be a very naive question, but I was wondering about the role that the casing of the CPU can play in maintaining the heat inside the box. Do you not think that a surface that can release heat out through the casing because of its color properties should work better than a surface that retains heat within the casing?

  11. Davidled
    April 26, 2014

    There is no doubt that the surface of board releases the heating. Also, as some other components rapidly increase the temperature above which the board faces the damage. In that case, heatsink could help keep the temperature cool down. Among other example, let us think about the simply motor control.

    Regarding on the second question, if the picture is shown, it would be better understandable.

  12. samicksha
    April 27, 2014

    I agree you Daej, Conduction is the method Heatsink follow to dissipate the heat, so material becomes more relevant. Anyone experianced CarbAl  which is made up of aluminum and two different carbon-derived materials with excellent thermal conductivity.

  13. amrutah
    April 28, 2014


    “…if the heatsinking is poor MOSFET can be destoryed by excessive temperature.”

      Are you saying that the temperature rise will cause some reliability issues or will it blown away the MOSFET? Any temperature limits?

  14. geek
    April 29, 2014

    “Regarding on the second question, if the picture is shown, it would be better understandable.”

    @DaeJ: I was just reffering to a kind of CPU casing which is made out of a material and has a surface capable of throwing out the heat rather than retaining it inside. I haven't seen anything like that in the market so I'm wondering whether that's a good item to develop.

  15. samicksha
    April 29, 2014

    I believe temperatre limit here would be, the packaging which limits the maximum junction temperature, due to the molding compound and epoxy characteristics.

  16. amrutah
    April 29, 2014

    @samicksha: I agree that packaging might need a high temperature, but when we are packaging we might not have a heat sink attached to it.  The heat sink will be present/attached on the application board to remove the heat when the device is in operation.

  17. Sachin
    April 29, 2014

    Interesting piece that brings up a new perspective. Unlike most design engineers and EEs, I work in an environment where most of my tasks are repetitive; its basically designing different circuits that do more or less the same thing. So my method when it comes to choosing a discrete transistor is pretty simple and straight forward; repetitive trial and error until I find one that works well for me. Fortunately I never have to go beyond the third transistor since I have a rather limited field of choice since I only try out the ones that I have used before.

  18. SunitaT
    April 29, 2014

    It is true that most of engineers today do not consider the factor of cost during their operation. They think that cost handling should lie with the procurement sector, this is a very wrong idea, and as you are putting it Steve, I think it is high time engineers started thinking on how to help in the process of resource management in their places of work, this will help in cost saving. 

  19. geek
    April 30, 2014

    “Fortunately I never have to go beyond the third transistor since I have a rather limited field of choice since I only try out the ones that I have used before.”

    @Sachin: Is that because you don't have trust in a new brand name or company? Or does it have to do with the fact that the need to try a new brand is not there at all.

  20. RedDerek
    April 30, 2014

    A new engineer does not think about cost. And when involved with commercial production of a consumer product, cost tends to be a key factor.

    When getting the procurement involved in a part purchase, so long as they buy the same part number, such as a 2N2222A, I let them grab the cheapest part. However, if the specs are critical to a design, then my design time does take into account for looking for a reasonable priced transistor across a couple of manufacturers. Then I give that specific list to procurement to work with.

    Here the engineer does start to work with the pricing, but is required to maintain specifications to meet circuit performance.

  21. Davidled
    May 1, 2014

    Product will be depending on total price in the market comparing to development cost.  I think that prototype could be built to see how it works and then it is moving on the next step.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.