Connect with us

Bootstrap Zener with op-amp follower?

Discussion in 'Electronic Design' started by Aubrey McIntosh, Jan 23, 2004.

Scroll to continue with content
  1. On 23 Jan 2004 06:39:59 -0800, the renowned
    No, nothing fundamentally wrong if you make sure it will start up. I
    don't know why you are doing it that way, it makes the zener current
    strongly beta dependent, which is a strange choice, especially
    considering how truly crummy low-voltage zeners are.
    That's a strange choice of $$$ op-amp for that application. I really
    doubt your simulation results- almost 3V across the 1N5221B with small
    10's of uA current? I'd guess more like <1V in that circuit.

    Best regards,
    Spehro Pefhany
     
  2. | On 23 Jan 2004 06:39:59 -0800, the renowned
    | (Aubrey McIntosh) wrote:
    |
    | >Is there anything fundamentally wrong with bootstrapping a zener with
    | >an op-amp follower?
    |
    | No, nothing fundamentally wrong if you make sure it will start up. I
    | don't know why you are doing it that way, it makes the zener current
    | strongly beta dependent, which is a strange choice, especially
    | considering how truly crummy low-voltage zeners are.
    |

    For a more reliable start up, I put a blocking diode back to the
    op-amp. Some of the simulations reported a strange stable state
    where the op-amp output was about -300V. Also I have room for the
    voltage drop. I also put a 1.3M from V+ to the top of the zener to
    force a positive voltage on In+ to the op-amp, and help kick start.

    Does this bear any resemblance to a well designed fail-safe starter?

    As AoE recommends, I'm trying to avoid beta dependency, so I am
    considering the circuit under light of your comment. I did look at
    the ratio of Ie(Q2) / Ib(Q2) in the simulation, and I see values
    around 200. If I have read the .model correctly the beta for the
    2N4401 is about 4K (Bf=4.292K). My interpretation was that I had a
    lot of beta budget left over, and then I assumed I am not being beta
    dependent. Did I miss the path in that reasoning?

    My impression is that the only really critical component in this
    circuit is R_Sense. With an op-amp, I am able to use current sense
    voltage of only about 200mV for about 2uW loss. I actually felt
    pretty good about that. I am stepping the input voltage over an
    absurd range, and it seems pretty stable to that. Temperature
    stepping shows that the load current grows at cold temperatures. I
    assume that is more tolerable than higher current at high
    temperatures. I don't have any sense of whether that should be
    controlled or not.

    I'm sort of in the frame of mind I was in 20 years ago when I had an
    Argus C3 camera and no money for a flash: I want to learn how to
    tame the beast, then I can move to the better substitute. (hints as
    to what to use instead of a zener will gladly be squirreled away)


    | >Concept circuit and simulation output online:
    | >
    | >http://www.jump.net/~vima/led/Active_04.pdf
    | >http://www.jump.net/~vima/led/Active_04_tran.pdf
    |
    | That's a strange choice of $$$ op-amp for that application. I really
    | doubt your simulation results- almost 3V across the 1N5221B with small
    | 10's of uA current? I'd guess more like <1V in that circuit.
    |
    |

    I made up my own .model for the 1N5221B, and now have decided not to
    trust it. I have replaced this with a 1N750 with SwitcherCAD's
    .model statement, and made tuning adjustments to other passives.

    I don't have the "generic user supplied" op-amp under control, so I
    have a lot more SwitcherCAD homework to do. The op-amp in this
    circuit is only a place holder for later.

    I assume that this op-amp does not need to be fast. Other criteria
    that I assume make sense are 1) low power, 2) stable at DC, 3) if I
    want to actually build something later -- under $2, 4) single supply.
     
  3. John Larkin

    John Larkin Guest

    The little bandgap things, LM4040-series or whatever, are very stiff
    over a huge current range and go down to 1.2 volts, no bootstrap
    needed. Low voltage zeners suck.

    John
     
  4. LMV431 is cheaper. 1.24V and 1% tolerance. Both need 50~100uA to
    operate. If you want the circuit to work down to around a single LED
    voltage, you'll need an op-amp that's functional at around 2.2~2.7V,
    preferably micropower, and Vos would be not be a very important
    consideration (you have 160mV or whatever to work with). Maybe an
    LMC7111 or something like that. BTW, you're wasting ~1mA in that
    precision op-amp.

    It might be more fun to try to use the LED itself as a bootstrapped
    reference.

    Vx o
    |
    |
    |
    +--------------+
    | ---
    .-. \ / LED
    Ra | | ---
    | | |
    '-' |
    | +------ Vb
    Va ------+ |
    .-. .-.
    | | | | Rs
    Rb | | | |18R
    '-' '-'
    | |
    +--------------|
    ===
    k = Ra/(Ra+Rb) GND

    If Va = Vb, then If = (Vf/Rs)* (k/(1-k))

    Best regards,
    Spehro Pefhany
     
  5. John Larkin

    John Larkin Guest


    Dang, Spehro, you beat me to it. I was thinking of a grounded LED with
    a highside resistor, and driving the top of the resistor to 1.1 * Vled
    or something like that, right from a r-r opamp. Opamp plus 4
    resistors, including a startup thingie.

    John
     
  6. Aubrey McIntosh wrote...
    The 2n4401 is not a darlington transistor! Like most BJTs, it
    has a beta of 50 to 200, so ideally for circuit robustness you
    shouldn't require a beta of over say 25. Rather than depend on
    Spice, it's wise to obtain and read data sheets and think about
    their specs. http://www.onsemi.com/pub/Collateral/2N4401-D.PDF
    As you know, in AoE we teach back-of-the envelope calculations,
    rather than spice models, for evaluating circuits. :>)

    Thanks,
    - Win

    whill_at_picovolt-dot-com
     
  7. That circuit tries to stabilize the voltage on the emitter resistor to drive
    the LED with a constant current.
    This means the output voltage from the opamap will mainaly depend on the
    current gain of the transistor. It is quite likely too small to drive the
    zener anyway.

    This is NOT a bootstrap circuit for the zener. That would have to produce a
    fixed voltage at the opamp output, above the zener voltage and the drop over
    D4.

    Regards,
    Arie de Muynck
     
  8. Fred Bloggs

    Fred Bloggs Guest

    Aubrey McIntosh wrote:
    [..snip...]

    Just stop wasting time and use two opamps- one for your voltage
    reference and a second to servo the output- and what purpose does that
    ultra-large base resistor serve except to test the output range limits
    of the opamp. What is the circuit supposed to do anyway?
     
  9. Fred Bloggs

    Fred Bloggs Guest

    LOL- How else are neophytes to be exposed to the harsh realities of 5V
    opamps with 300V outputs and comparators that seek out midscale
    equilibrium solutions?
     
  10. Fred Bloggs wrote:

    | LOL- How else ...

    Glad to be of help. More to follow soon.
     
  11. I quickly scanned this morning's postings, and I'll go digest them
    soon. Meanwhile;

    RESULTS
    I calculated values for the bootstrapped LED with a 10 mA target
    current, using the Yellow LED .model posted recently on s.e.d.

    The LED grounded vs sense resistor grounded are anti-symmetric in the
    bridge. I find it conceptually clearer to have the sense resistor
    grounded, as the target signal is the virtual ground on the op-amp.

    In order to stabilize the circuit as the control supply voltage
    changes, however, I ended up switching geometries.

    The problem is that the R4, the "force to start" resistor gives more
    current into the op-amp at higher supply. Compensation is done with
    R5, a value too big for the Mouser catalog. It seemed simpler, late
    at night, to work with the grounded LED geometry to comprehend the
    compensation tweaks.

    Staying with 5% values, I have a choice of the drift being to higher
    current or lower current. It seems philosophically better to have
    the circuit shut down instead of run away, so I drifted on that side
    of the line.

    This approach also seems to stabilize the zener circuits, but as the
    entire exercise is with my 10 Kingbright white LEDs in mind, I
    thought I would just switch over.

    Files at:

    http://www.jump.net/~vima/led/boot_led_01_asc.pdf
    http://www.jump.net/~vima/led/boot_led_01_plt.pdf
    http://www.jump.net/~vima/led/led_curve_01_asc.pdf
    http://www.jump.net/~vima/led/led_curve_01_plt.pdf

    DISCUSSION
    When I tried to step the temperature, I got the failure to converge
    message. Previously I read and found which parameter to relax in
    SwitcherCAD III so that the simulation can proceed in this case, but
    I can't remember at the moment. If someone knows off the top of
    their head, it would save me another search.

    I did find the LTC.lib file with many models in it. Presumably, I
    will read the spice details there, and draw some circuits with a few
    op-amps in them. This should keep me out of mischief for a while.
    Since I have separated the "control" and "power" supplies, I don't
    really need an op amp that will work from 4 -- 32V, so there must
    also be a better (low power) op-amp that springs to mind, and
    available in the SwitcherCAD library.

    Now that I have a clear picture of the concept, I am ready for an
    actual engineering concern or two. With the deisgn goals of :
    stable, as close to 10mA as possible without going over, generally
    well behaved according to your long experience, and available at
    Mouser, what would you choose for the values in the bridge, and why?
    Carbon Film? Metallized film? Do you always choose a certain way
    between KOA, Xicon, others?
     
  12. Thanks. I now have the datasheet and am reading over it.

    I went on a Google expedition for a SPICE model of the 2N4401.

    1. I found a link to a 1988 spice model which uses the same model as
    SwitcherCAD III at
    http://matrix.coe.montana.edu/sshaw/classes/2000/EE216/
    http://www.qsl.net/k7qo/models.html
    http://www.qsl.net/k7qo/all.mod

    2. I found a download site at fairchild. They prohibit additional
    disclosure of their model. Thus far, this has been enforced by
    non-delivery of the promised email.
    http://www.fairchildsemi.com/models/email_model_file.jsp?file=2N4401.mod

    3. I found a very abbreviated, but unique model:
    http://www.xcvr.com/elmer101/spice.htm reports
    ..MODEL 2N4401 NPN (CJE=30pF CJC=6.5pF)

    4. Our own Spehro, in the very recent past. Note his Bf value !!

    5. And, finally, I found something that isn't what I was seeking, but
    which resonates.
    In ECE6416 at GaTech, they roll their own. An academic low noise
    amplifier design with specifications that the students derive certain
    values, IS, BF, RB, VA, CJC, CJE, and TF, for their spice models from
    their own laboratory measurements or mfr datasheets.
    http://users.ece.gatech.edu/~mleach/ece6416/Labs/design.pdf

    I like the idea of measuring the characteristics of a specimen, and
    calculating the SPICE values that it exhibits. A sanity check, kind
    of like measuring the resistance of a resistor when it is marked on
    the outside, plain as the colors in the rainbow. Sounds like a
    PICLIST project.
     
  13. Ben Bradley

    Ben Bradley Guest

    Perhaps this is why Bob Pease doesn't trust simulation programs -
    how else could an op-amp powered by 5V give such a dangerously high
    output voltage?
    Don't touch the schematic on the screen...
     
  14. Fair enough.

    The larger problem that I am working on is this. Design, build, and
    characterize a "microdistillation" probe for mass spectrometry.
    Interpret the data with factor analysis, according to "On the
    discrimination and quantitation of isomeric mixtures of compounds by
    mass spectrometry," Dissertation, McIntosh, 1999. Using known and
    unknown mixtures, report assay values. Do subtle chemistry to enable
    chiral determinations. Build at least one mass spectrometer which
    allows such determinations to be performed quickly.

    In use, the probe will be in a high vacuum chamber, 3E-7 Torr
    pressure. Sample will be on the probe, and the temperature programmed
    specifically to facilitate interpretation by the Arrhenius (or
    Clausius-Clayperon) equation. Multiple compounds will be mixed, and
    the exact amounts are not only unknown, but the object of
    determination. In replicate runs, known amounts of the compounds can
    be added to the sample (Method of Standard Additions) or else run in
    pure form.

    Total amounts of material will be small. Typically a 1 uL deposition
    of solution is made, and the solvent evaporated. The solution
    deposited is typically 25 mg / 25 mL w/w concentration. This is
    general purpose, so the analytes are whatever comes in the door.
    However, you may wish to Google for
    to get an idea of the likely
    candidates.

    This probe is structurally similar to a "hot wire" probe that is in
    common use, with more finesse. That is, it will be a helix with a
    diameter of about 0.5 -- 1.0 mm and a length of 1-2 mm. The wire
    diameter is small. It can be fabricated from Wollaston wire and
    etched. The dimensions should be such that a droplet of solution
    should be held by the wire prior to solvent evaporation. Various
    solvents, such as water, ethanol, and methylene chloride should be
    accomodated.

    Preliminary constraints are:

    1. Go from -50C to 300C as quickly as 30 seconds.
    2. Report the actual temperature of the probe.
    3. Manage the temperature profile so that the value (1 / Kelvin)
    changes at a linear rate.
    4. Deliver sufficient power so that the analyte evaporates without
    deviation from the temperature profile.
    5. Log the power delivered to the probe. Log enough information so
    that the power given to the sample, as distinct from heating the wire,
    is recorded. For reference, read up on Differential Scanning
    Calorimetry. Adapt so that the reference scan is performed on the
    same hot wire, at a different time. Be able to calculate Enthalpy of
    Sublimation.
    6. Oxygen sensitivity contra-indicates tungsten construction
    material. Rhodium and platinum are commonly used and easily
    available. Nickel is also a platinum group material, and may be
    preferred for the early prototypes, depending on availability and cost
    of well characterized material. Platinum thermometers are likely to
    be the cheapest solution.
    7. In the (predominant) operation mode where cooling is not
    energized, record temperature with a minumum of self heating and
    regain control when servo requests heating.
    8. Allow a cleaning cycle where the wire reaches incandesence. While
    not necessary chemically, this is an important human interface design
    aspect.
    9. Respect two emergency input signals: cut power, and maintain
    current power.

    The current area of effort is to move from linear-temperature-in-time
    control without power logging, to linear in "reciprical Kelvins"
    control with power logging and energy integration.

    My choices are to learn enough electronics, or to hire a EE postdoc
    and teach him vacuum systems and chemistry. So I am working through
    various stable voltage controlled power supply concepts. You may
    choose to view the LEDs as a red herring, a double-dipping project, or
    a way to make the project seem interesting to the readership of s.e.d.
     
  15. I read in sci.electronics.design that Aubrey McIntosh <
    a.austin.tx.us> wrote (in <
    om>) about 'Bootstrap Zener with op-amp follower?', on Tue, 27 Jan 2004:
    You could make Bill an offer he can't refuse. To me, this sounds like a
    tailor-made project for him, but I could be wildly wrong.
     
  16. Fred Bloggs

    Fred Bloggs Guest

    The basic hot-wire anemometer circuit looks like so. Step the rheostat
    in accordance with your 1/Kelvin time profile and digitize Vout:

    Please view in a fixed-width font such as Courier.


    Vout
    +-----------+-----------e c--Vreg R1 x Rpot
    | | \ / Rprobe= ---------
    | | --- R3
    / / |
    R1 R3 |
    / / | Vout
    \ \ | Iprobe= -----------
    | | | R1 + Rprobe
    | | |\ |
    +-----------|--------|-\ |
    | | | >-+ Rpot
    | +--------|+/ Vprobe= Vout x ---------
    | | |/ Rpot + R3
    / /
    Rprobe \<+
    / / | Pprobe= Vprobe x Iprobe
    \ \ |
    | +-+ Rpot<-digital pot
    | | rheostat mode
    | |
    --- ---
    /// ///
     
  17. John Larkin

    John Larkin Guest


    Why a postdoc? He'd probably be a researcher in some specific
    discipline. A real circuit designer would probably be better at, well,
    designing circuits.

    The real problem sounds more interesting; there are way too many LED
    questions here already.

    John
     
  18. James Arthur

    James Arthur Guest

    A 1.2V TL431 -- neat!
    If the LED's getting 10mA then a) this is not a micropower application & b)
    1mA lost is merely 10%. Is there a low-voltage requirement? I couldn't find
    it.


    Vf(LED), however, is a rather soft reference, not to mention
    its drift with temp.

    The object appears to be:
    o to drive an LED
    o with a constant 10mA
    o from a single supply
    o with "low" power consumption
    o for less than $2

    Why not go straight at it?

    +5V
    -+-
    |
    |
    .-.
    | | R1
    | | 180 | i=11mA
    '-' V
    |
    .---------+ <--- Va=Vf+1.24v=3.2v
    | |
    | --- LED
    | \ / ~> Vf = 2.0v
    LMV | ---
    431 | |
    --- |
    / \<-------+
    --- |
    | |
    | .-.
    | | | Rs
    | | | 124
    | '-'
    | |
    | |
    | |
    --- ---


    A white LED with Vf=3.6 could be accommodated with
    a higher supply voltage and appropriate R1, or at +5v
    by using an LM10 for the shunt regulator.

    OTOH, if the aim is to get constant *luminous flux* from
    the LED, temperature compensation is essential; closed loop
    control of temp. (or flux) preferred. (An LED heater + servo loop
    stabilizing Vf(LED) should suffice.)


    James Arthur
     
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

-