BJT hfe drastically varies.

Discussion in 'General Electronics Chat' started by Austin Clark, Aug 3, 2012.

  1. Austin Clark

    Thread Starter Member

    Dec 28, 2011
    409
    44
    Hello, I've purchased a few hundred 2n3904 and 2n3906 BJTs. They're expected to have a beta of around 100. When I went to measure them, however, I get anywhere from 180 to 380 on my multimeter. When I compare the 180 ones with the 380 ones in NOT gates, I'm noticing a serious difference as well (The 180s can't pull down the output nearly as easily), so the meter seems to be giving me reasonable values (at least, relative to each other). The transistors are all of the same part #, but are from different manufacturers and/or batches (they look different, different leg shapes, different markings, etc;). I'd expect a beta of, say, 80 to 120, but 180 to 380? Not only is that a huge range, but it's not even centered on the value given in the datasheets (and the wikipedia). Is this normal? Are my parts defective/out of spec? Seriously, any information at all on this would be greatly appreciated, I'm trying to design/build Transistor-resistor logic circuits and kinda wanted to standardize some rules for base and collector resistor values (and fan-out), and generally the higher the beta the better (In case it matters at all). Thank you :)
     
  2. Markd77

    Senior Member

    Sep 7, 2009
    2,803
    594
    That's pretty normal. The dataheet normally specifies a minimum beta at a given collector current and even parts from the same batch will vary considerably.
    The beta for a single transistor varies with collector current. If you are having problems using them for switching purposes, decrease the value of the base resistor.
     
  3. takao21203

    Distinguished Member

    Apr 28, 2012
    3,577
    463
    It depends where you buy them.
    Cheap transistors might be untested/unsorted.

    Also depends on the type number.

    If you buy a bulk bag/reel from a recognized vendor, chances are the hFe values are more close towards the given value.

    Wikipedia is in no way officially representative information. It is not a how-to guide, not a business/product directory, among other things.

    If the transistors are from different manufacturer's, don't wonder about different hFe.
     
  4. wmodavis

    Well-Known Member

    Oct 23, 2010
    737
    150
    Welcome to Transistors 101.
    You have experimentally discovered an important truth about the BJT beast and it is a perfect example why one should become familiar with data sheets and do your designs based on manufacturer specifications, NOT on what you expect.

    Expectations not in line with specs can lead you down the proverbial primrose path to disappointment.

    The beta of a BJT depends on many variables and varies with actual conditions such as collector current (at what Ic does your meter test?), temperature (at what temperature were your measurements made?) and possible wind direction. Check the spec sheet!
     
    Austin Clark likes this.
  5. crutschow

    Expert

    Mar 14, 2008
    12,977
    3,221
    As noted, a 3:1 (or more) range in betas is quite normal. But for good design practice you need to use the minimum value for beta when calculating the required base current.

    And when using the transistor as a switch it is recommended that you use a beta of no more then 10 to 20 to insure that the transistor turns fully on (saturated) when it is conducting. The beta value given on the data sheet (and which you measured) is only for use in determining the bias point of AC or linear amplifiers, not the value needed to turn it on as a switch.
     
    Austin Clark likes this.
  6. Austin Clark

    Thread Starter Member

    Dec 28, 2011
    409
    44
    I'm not sure what current my meter uses to test transistor beta, but the current I'm using for switching is very low, I usually use a 10k resistor at the collector and a 100k at the base. Sometimes though I'll mess around with a 100k collector resistor and a 1Mohm resistor at the base. I assume that the beta at this point is lower than what my meter is telling me, is this correct? If so, what would the minimum beta be at this point (and/or how could I check for myself?).

    I'll be posting another thread with a few related questions in a moment, to try and keep the information organized for other members in the future. Keep an eye out for it as well please.

    Oh, and thanks again for the quick, informative, and helpful responses. Lifesavers.
     
  7. #12

    Expert

    Nov 30, 2010
    16,248
    6,745
    I was going to suggest sorting them into batches according to gain but I see that first I have to suggest you decide on a "standard" test. It's a practical thing. Choose a current that is somehow relevant to your circuits, like 100 ua or 1 ma and stick with it. Perhaps make 2 batches. Low gain (180-280) for the inputs and high gain (>280) for the outputs. Mark them with something. Fingernail polish works but it's been hard for me to find green for "good". Maybe that's available now. ;)

    Another approch is to spend the money that quality costs. Your choice. Work harder or pay for a sorting machine to do it before the parts are shipped to you.
     
    Austin Clark likes this.
  8. Markd77

    Senior Member

    Sep 7, 2009
    2,803
    594
    A simple enough test if you have 2 voltmeters, for low current testing have a 1M variable resistor and a 1K 1% resistor in series at the base. Put a 1K 1% resistor at the collector. Set the variable resistor to the midpoint to start and then adjust it while watching the voltage on the collector resistor to set the desired current. The ratio of voltages across the two 1K resistors is the gain. To measure gain at higher currents, reduce all resistors as required.
     
  9. JMac3108

    Active Member

    Aug 16, 2010
    349
    66
    Crutchow has the right idea. When using BJT's you need to design your circuit to work over the full range of beta (don't forget temperature dependence too!).

    For linear circuits use one of the typical " beta independent" biasing methods. For switches make sure to design for minimum beta.
     
  10. JMac3108

    Active Member

    Aug 16, 2010
    349
    66
    I wouldn't try to sort parts unless you absolutely had to. Its a much better idea to design the circuit to work over the full range of beta. What if you sort parts, then someone replaces a part in your circuit later and it doesn't work because they happened to get one with low beta?
     
  11. #12

    Expert

    Nov 30, 2010
    16,248
    6,745
    and what if you bought a couple of hundred unsorted rejects for your own experiments and they were never going to be installed in a product that is for sale? Imagine the consequences!
     
  12. JMac3108

    Active Member

    Aug 16, 2010
    349
    66
    They don't sound like unsorted rejects to me! Just sounds like regular old BJT variation. I would take this as a good opportunity to learn how to design for worst case. But if its your project, have fun and do it however you want :D
     
  13. WBahn

    Moderator

    Mar 31, 2012
    17,715
    4,788
    Looking at the 2n3904 datasheet from Fairchild, you are given a minimum beta at five different collector currents ranging from 100µA to 100mA. The Vce for each is 1.0V and the temperature is 25C. The minimum beta starts at 40, peaks at 100 at Ic=10mA, and drops down to 30. Other than that, there is a maximum beta of 300 specified at Ic=10mA.

    In this datasheet, there is also a plot of typical hFE as a function of collector current with Vce=5V at three different temperatures. For room temp, that shows a typical beta of about 230 up to about Ic=10mA.

    So, unless you are measuring with Ic=10mA and seeing higher than the max 300 beta, your transistors would seem to be in compliance with the specs (keeping in mind that I've only looked at one manufacturer's data sheet and not looked at the PNP data sheet at all).

    Given the data sheet information, your circuit should be designed with the assumption that the beta of the NPN transistors is going to be as low as about 50 and, to provide a good design margin, should design for suitable operation down to a beta of 10 (which, as pointed out by others, is the general rule of thumb when using a small-signal transistor as a switch).
     
    Austin Clark likes this.
  14. #12

    Expert

    Nov 30, 2010
    16,248
    6,745
    Here's the datasheet for 2N3904 Fairchild.
     
  15. ramancini8

    Member

    Jul 18, 2012
    442
    118
    I designed some of the last series of discrete gates for large computers. We used the minimum beta specified for the transistor, and we derated it 30% to account for temperature and life degradation. Do not let the reverse B-E voltage exceed the rated value, even if current limited, because this will cause beta degradation.
    100K is large for a saturated gate collector resistor because the collector resistor has to charge the next stage input capacitance, hence large collector resistances slow down the gate. DTL is much faster than RTL. Check the Vce(sat) at the current level you are using; it may be high. I think the 2N2222 may be a better transistor to use in a saturated gate.
     
    Austin Clark likes this.
  16. WBahn

    Moderator

    Mar 31, 2012
    17,715
    4,788
    What time frame was that? What would typical performance numbers be on large computers of that era?
     
  17. ramancini8

    Member

    Jul 18, 2012
    442
    118
    It was the late 60s and the computer series was the Spectra 70 by RCA. Sorry, but I don't remember specs on 50 year old designs.
     
Loading...