Connect with us

question about temperature characteristics of transistors

Discussion in 'Electronic Design' started by Tolstoy, Aug 5, 2007.

Scroll to continue with content
  1. Tolstoy

    Tolstoy Guest

    I have a project that requires current
    limiting. The sensing has to be done on the
    high side and I don't want to order
    some special chip, so I dreamed up a circuit
    that just uses jellybean parts.


    Icharge-->

    ------+-----Rs-------+------batt, 12v
    | |
    \ PN2907 /
    |---+----|
    /| | |\
    / 1M \
    | | |
    | gnd |
    | |
    +--------------|----
    | | to comparator
    | +---- or op amp
    | |
    R1 R2
    | |
    '------+-------'
    |
    gnd

    Circuit operation:

    the b/e junctions of the transistors will
    experience a differential voltage Ich*Rs.
    For a particular DeltaVbe=Ich*Rs,
    the base currents will exhibit a certain ratio
    and the collector currents will hew to the
    same ratio, at least for matching transistors.
    Choosing R1 and R2 in a ratio equal to the
    ratio that the collector currents exhibit at
    target charging voltage will cause equal
    voltage across the comparator inputs
    at that moment and trip it.

    I haven't tried prototyping this circuit yet,
    I just came up with the idea this morning,
    so please criticize it and
    point out what I may have missed.

    This circuit will get bolted
    somewhere on an internal combustion engine
    operating outdoors, so temperature will vary
    to extremes. The transistors' 2mV
    tempco *should* cancel; but then the circuit
    operation also relies on the exponential
    voltage/current relationship, and
    I haven't found any information about whether
    the terms of this expression exhibit a
    temperature dependence.
    Does anybody know please?
     
  2. Jim Thompson

    Jim Thompson Guest

    Perhaps something along the lines of...

    http://analog-innovations.com/SED/CurrentSense.pdf

    You need to make-up for the drop in the sense resistor by biasing
    (adding R in emitter of one of the transistors).

    Temperature IS going to be an issue. Perhaps a dual transistor
    package would help.

    ...Jim Thompson
     
  3. See:
    http://www.edn.com/contents/images/30101di.pdf

    as a start.

    Jon
     
  4. Is is also temperature dependent. And since the voltage at some
    current depends on Is, that does change. Something like:

    Is(T) = Is(Tnom)*(T/Tnom)^3*e^[-(Eg/k)*(1/T-1/Tnom)]

    Jon
     
  5. John Larkin

    John Larkin Guest

    Why not add a resistor into both emitters? That will linearize the
    transfer function and mostly swamp the temperature mismatch. That
    would allow, say, 50 or 100 mV of peak drop in the sense resistor.



    A little more base current wouldn't hurt. And maybe some base
    resistors to keep the transistors from screaming.

    Note that Icharge will typically be a very nasty waveform.

    John
     
  6. D from BC

    D from BC Guest


    This is starting to look like..
    http://www.zetex.com/3.0/pdf/ZDS1009.pdf
    D from BC
     
  7. Jim Thompson

    Jim Thompson Guest

    A resistor in both emitters cuts the gain... only needed on the left.
    Run the math, you have to set a transition level somehow ;-)

    The 1Meg should go bye-bye. Diode connect left-hand transistor (or
    right-hand) depending on signal-sex desired.

    ...Jim Thompson
     
  8. Jim Thompson

    Jim Thompson Guest

    Yep. Looks like useful "popcorn" circuits are returning, as people
    discover that uP's have limited capabilities ;-)

    ...Jim Thompson
     
  9. Fred Bloggs

    Fred Bloggs Guest

    The comparator does not trip on a ratio, it trips on a difference. This
    difference voltage will be something like:
    Is*exp(Veb/Vt)*(R1-R2*exp(-Ich*Rs/Vt)), assuming matched Is, and so you
    want R1-R2*exp(-IchRs/Vt)=0 or R2=R1*exp(Ich*Rs/Vt). Then the trip
    current at any other T will be Ich=(Vt/Rs)*LN(R2/R1) so that dIch/dT=
    K/(Q*Rs)*LN(R2/R1). But you selected R2/R1 to be exp(Ich*Rs/Vt) at Tnom,
    making dIch/dT=K/(Q*Rs)*(Ich*Rs/Vt)=Ich/Tnom or dIch/Ich=dT/Tnom. This
    is not what you would call insensitive to temperature.
     
  10. John Larkin

    John Larkin Guest

    Gosh, really?


    If you don't care about DC errors.

    Transition level? What's that?

    His original circuit, fully symmetric, is fine.

    John
     
  11. Tolstoy

    Tolstoy Guest

    [Snip ugly circuit]
    Wow, talk about simple.
    I was going to glue the flat sides of a couple of TO-92 transistors
    together using Krazy Glue. It's amazing how well that stuff works on
    plastic transistors.

    I'm going to study your circuit. Just two transistors. What's not to
    like?

    In the meantime, here's something that looks better than the first
    circuit I posted. You guys can pick it apart if you have a mind to.



    Icharge-->


    ------+-----Rs-------+--batt, 12v
    | |
    Re |
    | |
    | |
    \ PN2907 /
    |---+----|
    /| | |\
    / | \
    | | |
    | '-------+
    | |
    | |
    | ,------------,
    | | | 317-based |
    | | | floating |
    out<--+ | | current |
    | | | source |
    | | |I=Ilim*Rs/Re|
    | v '------------'
    | |
    | ,-------+
    | | |
    \ | /
    \| | |/
    |---+----|
    /| |\
    < PN2222x2 >
    | |
    '------+-------'
    |
    gnd

    Ilim = design limit current

    Thanks everybody for the replies.
     
  12. Robert Baer

    Robert Baer Guest

    Well...Krazy glue is a good thermal insulator, does not hold up to
    temperature, and would make a "thick" interface between the transistors.
    Better yet to use a surface mount dual and layout the PCB for a
    thermal pad for the body to sit upon.
     
  13. Tolstoy

    Tolstoy Guest

    I need something in the largest surface mount package I can find so
    that I can hand solder it, that I can buy in onesies and twosies, and
    totally generic so that it won't get hard to find.
    I found a couple of npn and pnp current mirrors that would have been
    perfect, the BCV61 and BCV72; unfortunately, digikey has a large
    minimum purchase for the npn version. Also, Mouser doesn't list
    either of these parts, which isn't a good sign. So I'm looking for a
    dual npn chip and a dual pnp chip, or maybe a quad with two of each.
    I've never used surface mount, so I have to sort through all the
    different surfmount packages according to size and find out the names
    of the ones that are big enough. Then I can use the digikey
    parametric search to see what's available.
     
  14. Fred Bloggs

    Fred Bloggs Guest

    That ZDS1009 has Camenzind written all over it. It wasn't too long ago
    when you were bitching about it having start-up problems.
     
  15. Fred Bloggs

    Fred Bloggs Guest

    Here is a much better circuit with linear output, better control over
    sense transistor operating conditions, low current and low voltage
    operation too, patented by Analog Devices:
    View in a fixed-width font such as
    Courier.
     
  16. Tolstoy

    Tolstoy Guest

    Nice. I'll file that one. But..
    I want comparator action, not linear output (sorry, I didn't mention
    that before). If my circuit were to have linear output I would just
    end up having to add a load resistor, a voltage reference and a
    comparator to get what I want, an output active high when charging
    current exceeds Ilim, otherwise open or low output.
    Just for grins, I jiggered the circuit around again:


    Icharge-->

    ------+-----Rs-------+--batt, 12v
    | |
    Re |
    | |
    | |
    \ PN2907 /
    |---+----|
    /| | |\
    / | \
    | | |
    | | |
    out-+ Rb |
    | | |
    | | |
    \ | /
    \| | |/
    |---+----|
    /| |\
    < PN2222x2 >
    | |
    '------+-------'
    |
    current sink
    I = 2 Ilim Re/Rs
    |
    gnd
     
  17. Tolstoy

    Tolstoy Guest

    Sorry, that's I = 2 Ilim Rs/Re
     
  18. Jim Thompson

    Jim Thompson Guest

    Good point. If it doesn't leak, how does it start?

    ...Jim Thompson
     
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day

-