Supplier photo of an assembled infinity mirror light cube illuminated on a dark background.
Supplier photo
Temu affiliate offer · Project idea

Build an infinity light cube

Listing specifies 5-volt direct current. Soldering required; battery and glue not included.

View kit on Temu

Tricky Circuit may earn a commission through these links. Product ideas have not been bench-tested by Tricky Circuit and are not verified lesson parts lists.

LM4040: A Voltage Reference Needs a Current Budget

A small voltage reference can show a beautifully steady number until you connect one more resistor. The missing idea is current allocation. Build a two-terminal shunt-reference experiment in which every microampere has a destination, then deliberately change the source voltage and load while recording both voltage and the inferred remaining current.

The exact integrated circuit (IC), a circuit containing many components in one package, is Texas Instruments LM4040AIZ-2.5/NOPB, A accuracy grade, industrial temperature range, three-lead Transistor Outline (TO-92) package designated LP. The suffix selects the 2.5 V reference and package. A similar LM4040 name, a transistor-shaped body or a different manufacturer’s part does not establish matching pins or specifications.

Verification: the complete supplied calculator and its failure paths were tested on a host computer. The circuit, meter readings, noise, heating and device accuracy have not been physically tested. No microcontroller firmware is required or claimed. All example values are calculations; the observation sheet contains no measurements.

The useful difference between a reference and a power rail

A voltage reference supplies a defined comparison voltage under specified conditions. A shunt connection places the reference across the load, rather than in series with it. The source first passes through a resistor. At the output junction that current divides between the load, the measuring instrument and the reference. The reference draws the portion left over to hold its reverse voltage near the specified value.

This experiment is useful when planning a comparator threshold, a small analog test voltage or a converter reference that draws little current. It does not power a board, motor or radio. Our earlier supply experiments study series-regulator dropout and heating; here the important calculation is the current that must remain in a parallel branch even when the load increases.

The cathode is the positive reference terminal in this circuit, and the anode is the grounded terminal. This is intentional reverse operation, not an ordinary forward-diode drop. Its regulation behavior needs a minimum cathode current. A source resistor is essential: connecting the cathode directly to a stiff 5 V source would not create a safe 2.5 V rail.

Exact parts, pin functions and limits

  • TI LM4040AIZ-2.5/NOPB in TO-92/LP. Match the marking and the manufacturer package drawing before breadboarding.
  • R1 = 1.00 kΩ ±1%, at least 0.25 W, as the mandatory source resistor. Use separate 4.7 kΩ, 2.2 kΩ and 470 Ω ±1% load resistors, each at least 0.25 W.
  • An adjustable low-voltage supply with a 10 mA current limit, starting at 5.0 V; a breadboard and short wires. The normal design envelope is 4.5–5.5 V. A later lower-voltage trial intentionally leaves that envelope.
  • A digital multimeter, an instrument that measures several electrical quantities, set to direct-current voltage with a known input resistance of at least 10 MΩ. Check the actual manual/range; “digital” alone does not establish that resistance.

The current TI datasheet served at review is SNOS633N, revision N, August 2025. Its detailed 2.5 V A/B industrial table gives a maximum required minimum operating current of 60 µA at 25°C and 65 µA across the industrial temperature range. We use 65 µA as the minimum calculation budget, with plenty of extra margin in the normal build. The 45 µA typical value is not a guaranteed design minimum.

The operating upper current is 15 mA. The separate absolute maximum is 20 mA reverse current: an absolute rating describes a stress boundary, not a useful operating point. Power dissipation and temperature limits also still apply. The A grade’s initial ±0.1% condition at 25°C is not a guarantee of that same total error across arbitrary temperature, current, age and meter conditions. We do not claim a calibrated standard from the marking.

Wire the three branches explicitly

A current-limited low-voltage source feeds a 1 kiloohm series resistor and the LM4040AIZ-2.5/NOPB cathode pin 2. Anode pin 3 and unused pin 1 join ground; a selectable load and voltage meter are parallel branches across the reference.
Original exact-part schematic. TO-92 numbers follow the manufacturer bottom view; they are not a generic transistor pinout. Open full-size diagram.
Exact TO-92 pin Connection Purpose
2, cathode R1 output, load top and meter positive Reference/output junction; receives the available shunt current.
3, anode Supply return, load bottom and meter negative Common ground, the circuit’s zero-volt comparison node.
1, marked * Tie to the anode/ground The datasheet permits floating or anode connection; this build chooses the latter.

The package drawing is a bottom view, looking toward the leads from below. Do not substitute the familiar emitter/base/collector order of another TO-92 device. Draw the lead numbers on your worksheet before inserting it. Identify breadboard rail breaks and confirm that R1 is actually in series, not bypassed by a jumper. All connections are made with power removed.

No output capacitor is needed for this static experiment; the reference is specified without an external stabilizing capacitor. Omitting it also avoids confusing a transient stored charge with sustained regulation. Never drive the output from another supply. This is a low-voltage, nonisolated bench circuit; ground is a common return, not protective-earth isolation.

Compute the missing current before trusting the voltage

For an assumed reference voltage VREF, source voltage VIN and source resistance RS, ISOURCE=(VIN−VREF)/RS. The load current is ILOAD=VREF/RLOAD, and the meter adds IMETER=VREF/RMETER. Kirchhoff’s current law, conservation of charge at the junction, gives ISHUNT=ISOURCE−ILOAD−IMETER. The sum includes the measuring instrument because the act of measuring is also an electrical load.

At nominal 5.0 V and 1.00 kΩ, the source branch supplies 2.500 mA if the output is 2.5 V. A 4.7 kΩ load uses 0.5319 mA and a 10 MΩ meter uses 0.00025 mA. Approximately 1.9678 mA remains for the shunt, far above 0.065 mA and below 15 mA. A 2.2 kΩ load leaves approximately 1.3634 mA. These are conditional calculations, not readings from a built circuit.

Now try the arithmetic for a 1 kΩ load. It alone needs 2.500 mA at 2.5 V, consuming the entire nominal source current before the reference and meter receive anything. A negative candidate shunt current means the assumed regulated voltage is inconsistent with that current budget. The program does not report negative device dissipation or pretend to predict the actual voltage below regulation.

With a 470 Ω load, even the idealized case in which the reference draws nothing produces only 5×470/(1000+470)≈1.599 V, ignoring meter loading. This divider calculation explains why the output must collapse; it is not an accurate model of the real reference’s low-current knee. Near the knee, a voltage close to 2.5 V can still have inadequate current and poor specified behavior.

Headroom and tolerance corners

Calculated LM4040 shunt-current margin versus load current at an assumed 2.5 V output. The curve marks open, 4.7 kiloohm, 2.2 kiloohm and 1 kiloohm loads; the dashed underbiased continuation is not a physical current prediction. A separate scale distinguishes 65 microampere minimum, 15 milliampere operating maximum and 20 milliampere absolute boundary.
Calculated current-versus-load curve and separate device-current boundaries. Below the minimum budget the assumed regulation is unsupported; this is not a measured device characteristic. Open full-size diagram.

Headroom is the extra source voltage needed to sustain the intended branch currents. Rearranging the equation gives VIN,min=VREF+RS×(ILOAD+IMETER+ISHUNT,reserved). For the nominal 4.7 kΩ case, reserving 65 µA needs approximately 3.097 V. Reserving 100 µA needs approximately 3.132 V. The latter is a chosen design margin, not a changed datasheet minimum.

The calculator also checks the corners, combinations of worst endpoint assumptions. It uses VIN=4.5–5.5 V, R1 ±1%, load ±1%, a 10 MΩ meter and an explicitly chosen 2.45–2.55 V planning window. That wide voltage window is a declared current-budget allowance; it is neither a predicted distribution nor a replacement for the device’s complete accuracy specification. Include any other real load before accepting a design.

At those endpoints with 4.7 kΩ, the smallest remaining current is about 1.382 mA and the largest about 2.564 mA. All corners retain ample current. The no-load case must also be checked: removing the load gives the reference the largest share. Assuming that an increasing load is the only danger misses the opposite failure, excessive reference current with a low source resistor.

At 5.5 V with the output accidentally shorted, a 990 Ω source resistor dissipates about 30.6 mW. Its 0.25 W rating has substantial margin for this specified low-voltage setup, but a rating alone does not validate every ambient temperature or mounting. Nominal no-load reference dissipation at 5 V is about 6.25 mW. These calculations do not establish a measured case temperature.

Complete calculator and observed-reading helper

Save the file and run it with CPython 3.11 or newer, the ordinary desktop Python implementation. No external package or connected hardware is required. The command prints JSON, JavaScript Object Notation, a structured text format. Its top-level label is CALCULATION_ONLY, and every corner describes an imposed reference voltage rather than a predicted output.

Download complete calculator (ZIP; file: tricky-circuit-batch-006/code/lm4040-shunt-budget.py). The same complete file is below.

Complete Python program
"""LM4040AIZ-2.5/NOPB current-budget calculator; CPython 3.11+, stdlib only.
Numbers are circuit calculations, never voltage measurements or a knee model.
"""
import itertools
import json
import math

MIN_SHUNT_A = 65e-6   # Full industrial-temperature electrical table, not typ45uA.
MAX_SHUNT_A = 15e-3   # Operating ceiling; 20mA absolute is NOT an operating target.


def finite(value, name, positive=False):
    if isinstance(value, bool):
        raise ValueError(name + " must be a finite number")
    value = float(value)
    if not math.isfinite(value) or (positive and value <= 0):
        raise ValueError(name + " is invalid")
    return value


def currents(vin, vout, series_ohm, load_ohm=None, meter_ohm=10e6):
    vin, vout = finite(vin, "vin"), finite(vout, "vout", True)
    series_ohm = finite(series_ohm, "series resistance", True)
    meter_ohm = finite(meter_ohm, "meter resistance", True)
    if vin < 0 or vin < vout:
        raise ValueError("requires nonnegative source at or above output")
    load = 0.0 if load_ohm is None else vout / finite(load_ohm, "load resistance", True)
    source = (vin - vout) / series_ohm
    meter = vout / meter_ohm
    shunt = source - load - meter
    state = "UNDERBIASED" if shunt < MIN_SHUNT_A else (
        "OVER_OPERATING_CURRENT" if shunt > MAX_SHUNT_A else "CURRENT_BUDGET_OK")
    return dict(source_A=source, load_A=load, meter_A=meter, shunt_A=shunt,
                series_W=(vin-vout)**2/series_ohm, load_W=vout*load,
                shunt_W=vout*shunt if shunt >= 0 else None, state=state)


def headroom(vout, series_ohm, load_ohm=None, meter_ohm=10e6,
             reserved_shunt_A=MIN_SHUNT_A):
    vout = finite(vout, "vout", True)
    series_ohm = finite(series_ohm, "series resistance", True)
    meter_ohm = finite(meter_ohm, "meter resistance", True)
    reserve = finite(reserved_shunt_A, "reserve", True)
    if reserve < MIN_SHUNT_A or reserve > MAX_SHUNT_A:
        raise ValueError("reserve is outside the operating current budget")
    load = 0 if load_ohm is None else vout/finite(load_ohm, "load resistance", True)
    return vout + series_ohm*(load + vout/meter_ohm + reserve)


def corners(vin_bounds, vref_bounds, series_ohm=1000, series_tolerance=.01,
            load_ohm=None, load_tolerance=.01, meter_ohm=10e6):
    def bounds(pair, name):
        if len(pair) != 2:
            raise ValueError(name + " needs two endpoints")
        lo, hi = (finite(x, name, True) for x in pair)
        if lo > hi:
            raise ValueError(name + " endpoints reversed")
        return lo, hi
    vins, refs = bounds(vin_bounds, "source"), bounds(vref_bounds, "reference")
    series_ohm = finite(series_ohm, "series resistance", True)
    for tol in (series_tolerance, load_tolerance):
        if not 0 <= finite(tol, "tolerance") < 1:
            raise ValueError("tolerance must be in [0,1)")
    rs = (series_ohm*(1-series_tolerance), series_ohm*(1+series_tolerance))
    loads = (None,) if load_ohm is None else tuple(
        finite(load_ohm, "load resistance", True)*(1+s*load_tolerance) for s in (-1,1))
    rows = []
    for vi, vr, resistance, load in itertools.product(vins, refs, rs, loads):
        row = currents(vi, vr, resistance, load, meter_ohm)
        rows.append(dict(vin_V=vi, assumed_reference_V=vr, series_ohm=resistance,
                         load_ohm=load, **row))
    return dict(assumption="Reference voltage imposed for budget check, not predicted",
                all_corners_pass=all(r["state"] == "CURRENT_BUDGET_OK" for r in rows),
                minimum_shunt_A=min(r["shunt_A"] for r in rows),
                maximum_shunt_A=max(r["shunt_A"] for r in rows), rows=rows)


def observed(vin_V, vout_V, series_ohm=1000, load_ohm=None, meter_ohm=10e6):
    result = currents(vin_V, vout_V, series_ohm, load_ohm, meter_ohm)
    result["interpretation"] = "Inferred from entered readings; not proof of regulation or accuracy"
    return result


def main():
    report = dict(kind="CALCULATION_ONLY", hardware_tested=False,
        design=corners((4.5,5.5),(2.45,2.55),load_ohm=4700),
        nominal_5V={str(load):currents(5,2.5,1000,load) for load in (None,4700,2200,1000,470)},
        minimum_source_4700_V=headroom(2.5,1000,4700))
    print(json.dumps(report, indent=2, allow_nan=False))


if __name__ == "__main__":
    main()

For real observations, enter your measured VIN and VOUT with the actual resistor values into observed(vin, vout, series_ohm, load_ohm, meter_ohm). The returned currents are inferred from entered readings and Ohm’s law; the program has no connection to your meter. A source value below the output is refused because this experiment’s assumed source direction no longer applies. Nonfinite numbers, zero resistance and reversed tolerance endpoints are also refused.

If you use one meter sequentially, the output branch exists only when that meter is attached there; the source and output readings are not simultaneous. Keep the supply stable, repeat the readings and record the sequence. Two meters alter both branches differently; draw and include their input resistances rather than silently reusing a one-meter calculation.

A repeatable bench sequence

  1. With power removed, inspect lead numbering, resistor values and common return. Fit R1 but no external load. Set the supply to 5.0 V with a 10 mA limit before connection. Record the instrument, selected range and actual input-resistance specification.
  2. Power the circuit and record VIN and VOUT. If the supply is in current limit or the voltage is unexpected, remove power and inspect; do not change wiring while probing a mystery current path. Compute the inferred no-load shunt current.
  3. Remove power, connect 4.7 kΩ, restore power and repeat. Then use 2.2 kΩ. Record measured voltages and resistor values even if they differ from the nominal examples. A voltage change can contain meter, contact, source, device and temperature effects.
  4. At 4.7 kΩ, lower the source slowly from 5.0 V toward 3.0 V while observing both input and output. This intentionally crosses the calculated headroom region; stop if a connection is uncertain. Restore the baseline before the next comparison. Do not claim an exact knee threshold from one coarse reading.
  5. Remove power and fit 470 Ω for the deliberately heavy-load case. At 5.0 V the current stays bounded by R1, but the 2.5 V assumption must fail. Record what actually occurs, then remove power and restore the light load.
  6. Run the host tests for invalid inputs and impossible current assumptions on the computer. Do not physically remove R1 or force excessive current to reproduce the calculator’s unsafe-design warning.

Blank observation sheet (ZIP; file: tricky-circuit-batch-006/code/lm4040-shunt-budget-observations-blank.csv) separates nominal values, actual instrument readings and inferred currents. There are no prefilled successful results. Record room conditions and a settling interval; warm fingers, supply changes and contact resistance can influence an ambitious millivolt comparison.

Challenge: the output still reads 2.499 V with a 1 kΩ load. Is the reference qualified?

No conclusion follows from that voltage alone. With VIN=5.000 V, RS=1.000 kΩ and RLOAD=1.000 kΩ, the inferred source current is 2.501 mA and load current is 2.499 mA, leaving only about 1.75 µA after a 10 MΩ meter. That is below the required operating-current budget. Check actual resistor values and measurement uncertainty, and choose a lighter load or redesign the feed. A plausible voltage is not evidence of sufficient bias current.

What this experiment does and does not establish

The host checks exercise the delivered current equations, corner enumeration, input refusals, power handling and real command entry point. They do not model noise, temperature drift, transient response, a damaged device or the curve below minimum current. No meter calibration or physical accuracy has been verified. Consult the actual instrument and full device tables before calling any result a reference standard.

Continue with the electronics lessons on current, resistance and measurement, then use this budget when designing a comparator input. Any converter reference with pulsed acquisition current needs a separate dynamic analysis; this static resistor experiment does not establish its settling performance.

Primary sources and version scope

📡Broadcast the signal — amplify the connection.

Leave a Reply

Temu affiliate offers

Find your next electronics project

Explore these do-it-yourself project kits. Check the selected option, included parts and instructions on each offer.

Tricky Circuit may earn a commission if you buy through these Temu affiliate links. These are product ideas; they have not been bench-tested by Tricky Circuit and are not verified lesson parts lists.

Supplier photo of a breadboard, jumper wires and assorted electronic components in a plastic storage case.

Supplier product photo

Try breadboard experiments

Breadboard, wires and assorted parts. Arduino controller board not included.

View kit on Temu

Supplier photo of an assembled blue surface-mount soldering practice board with a circular ring of small lights.

Supplier product photo

Solder a chasing-light board

Surface-mount soldering practice. Battery and battery holder not included.

View kit on Temu

Supplier photo of an assembled two-wheeled line-following car with a blue circuit board, two motors and an empty battery holder.

Supplier product photo

Build a line-following car

Soldering and assembly required. Check battery and track requirements before buying.

View kit on Temu