Apply the same small voltage steps to two diode circuits. In one, the diode and resistor decide how much signal gets through. In the other, an operational amplifier, or op-amp, adjusts its own output to compensate. Measure the amplifier output before the diode and the useful output after it. Those two readings reveal what the feedback changed.
A half-wave rectifier keeps one polarity of a signal and suppresses the other. This circuit accepts inputs both above and below a 2.50 V reference called REF. The useful active output rises above REF for input below REF and stays near REF for input above it. These direct-current (DC) steps show the settled transfer. They do not measure switching time or bandwidth.
Use a Texas Instruments OPA2323IDR on a soldered adapter. It contains two amplifiers in the D-package, an eight-pin small-outline integrated circuit (SOIC-8). The midpoint buffer uses channel B so the two rectifier loads do not pull on an unbuffered resistor divider. REF is the signal’s zero for this experiment. G0 is the power return. Connecting them together would short the buffer output.
| Pins and parts | Connection |
|---|---|
| 8, 4, C1 | Pin 8 to the 5 V rail, pin 4 to G0. C1 goes across those pins. |
| 5, 6, 7, R4, R5 | R4 from 5 V to pin 5, R5 from pin 5 to G0. Join pin 6 to pin 7 directly. Their joint node is REF. Pin 5 stays separate. |
| 1, 2, 3, R1, R2 | Pin 3 to REF, pin 2 to SUM, and pin 1 to DRIVE. R1 goes from VIN to SUM. R2 goes from ACTIVE to SUM. |
| D1, D2, R3 | D1 anode at DRIVE and cathode at ACTIVE. D2 anode at SUM and cathode at DRIVE. R3 goes from ACTIVE to REF. |
| D3, R6, R7 | D3 anode at PASS and cathode at VIN. R6 goes from VIN to G0. R7 goes from PASS to REF. |
With power off, connect supply channel 1 positive to the 5 V rail and channel 2 positive to VIN. Join both negative terminals at G0. Confirm these are independent outputs before sharing the return. Set channel 1 to 5.00 V and channel 2 initially to 2.50 V. Set both current settings to 20 mA. Enable channel 1 first and confirm 4.90–5.10 V and REF within 50 mV of half the measured rail. Then enable channel 2. Disable channel 2 first at shutdown. Before changing wiring, turn both outputs off and check the rails have fallen. The OPA2323’s recommended supply range is 1.7–5.5 V. All input pins in this fixture stay near the midpoint and well inside the rails during normal measurements.
On a low step the rectifier networks can feed a little current back toward VIN. The 2.2 kΩ resistor keeps the adjustable supply sourcing net current at every intended step. For example, at the lowest screened midpoint and most negative step, VIN stays about 2.09 V or above. The preload then draws at least about 0.94 mA with 1% tolerance. Each rectifier branch can return only about 32 µA in the ideal settled model. A supply that cannot sink current can still hold VIN under this load. Confirm the actual signed VIN reading at every step.
Call the signed input x = VIN − REF and the signed active output a = ACTIVE − REF. When x is negative, D1 conducts and closes the main negative-feedback loop. Negative feedback is the path that moves the output so the two amplifier inputs stay nearly equal. That makes SUM close to REF. Current balance at SUM gives a ≈ −x × R2/R1. Equal resistors give about +100, +200 and +300 mV for the three negative steps. Independent 1% resistors put the ideal magnitude ratio between 0.99/1.01 and 1.01/0.99, or about 0.9802 to 1.0202.
DRIVE has to sit above ACTIVE by D1’s actual forward voltage. A −300 mV input gives about +300 mV at ACTIVE, plus whatever D1 needs at DRIVE. Near that step, R2 and R3 each draw about 30 µA from ACTIVE. When x is positive, D1 blocks. D2 then closes the alternate feedback loop and pulls DRIVE below REF. R3 holds ACTIVE near REF, with a smaller path through R2 and SUM. Both loop states are important. Leaving D2 out can let DRIVE run toward a rail on the blocked half.
In the passive branch the diode’s forward voltage reduces the output magnitude. For x below zero, PASS lies between VIN and REF in the ideal circuit. Its signed voltage p = PASS − REF is negative. Its magnitude is approximately −p, so compare the active response using a + p. Vishay’s 240 mV maximum forward voltage is specified at 0.1 mA and 25 °C. It is a bound at that test current, not an on/off threshold. The branch here uses lower current and the actual voltage is a measurement. Do not use the data-sheet maximum as a fixed drop in the predictions.
The analyzer uses the Fluke 115 accuracy expression ±(0.5% of the absolute reading + two display counts). On the 600.0 mV range, two counts are 0.2 mV. On the 6.000 V range they are 2 mV. The manufacturer’s accuracy conditions include operation at 18–28 °C, relative humidity at or below 90%, and calibration within one year. For each negative step, the calculator combines the input interval with the resistor-ratio range, then subtracts it from the active output interval. It requires this entire residual to stay within ±12 mV. At zero and on positive steps the active output interval must stay inside ±8 mV of REF.
On a negative step, the lower bounds of a + p and DRIVE − ACTIVE must reach 30 mV and 40 mV respectively, and the upper bound of PASS must stay at or below +8 mV. On the −300 mV step the full PASS interval must also reach at least 20 mV below REF. This check catches a disconnected or reversed passive branch that otherwise stays near REF. On positive steps PASS must also fit inside ±8 mV and the DRIVE interval must be at or below −40 mV. At zero there is no required DRIVE sign. The calculator screens displayed DRIVE readings outside ±650 mV and compares the two −200 mV steps for midpoint, input and output drift. These are experiment choices, not TI or Vishay acceptance limits. A valid worksheet can report an unresolved result.
The diode-only branch is negative because its load returns to REF and the diode pulls it toward the lower VIN. Reflect its sign to compare magnitudes. The active magnitude is 198 mV and the passive magnitude is 61 mV, so the nominal difference is 137 mV. The analyzer then applies meter intervals before drawing a conclusion.
Save the completed worksheet as observations.csv. Run python3 rectifier_steps.py observations.csv. The experiment ZIP includes the analyzer, tests, editable drawing and blank worksheet. A valid worksheet returns JSON and exit 0. Read its status before claiming that the effect was resolved. Invalid input returns exit 2. All included test readings are synthetic. No physical measurements accompany this article.
"""Check signed direct-current readings from the OPA2323 rectifier fixture."""
import argparse
import csv
from decimal import Decimal, InvalidOperation
import json
from pathlib import Path
import re
import sys
D = Decimal
HEADER = ("step", "rail_v", "ref_v", "vin_minus_ref_mv", "active_minus_ref_mv",
"passive_minus_ref_mv", "drive_minus_ref_v")
LEVELS = {"zero": 0, "neg100": -100, "neg200": -200, "neg300": -300,
"pos100": 100, "pos200": 200, "pos300": 300, "neg200_repeat": -200}
GAIN_MIN, GAIN_MAX = D("0.99") / D("1.01"), D("1.01") / D("0.99")
def display(text, resolution, maximum):
"""Read a fixed-range display value with no invented digits."""
if not re.fullmatch(r"-?[0-9]{1,3}(?:\.[0-9]{1,3})?", text):
raise ValueError("enter a signed decimal display reading")
value = D(text)
if abs(value) > maximum or value % resolution != 0:
raise ValueError("display reading exceeds the selected range or resolution")
return value
def read_rows(path):
if path.stat().st_size > 12_000:
raise ValueError("worksheet exceeds 12 kB")
readings = {}
with path.open(encoding="utf-8-sig", newline="") as stream:
rows = csv.reader(stream, strict=True)
if tuple(next(rows, ())) != HEADER:
raise ValueError("use the exact worksheet header")
for row in rows:
if len(row) != len(HEADER) or row[0] not in LEVELS or row[0] in readings:
raise ValueError(f"row {rows.line_num}: invalid or repeated worksheet step")
try:
rail, reference = (display(x, D(".001"), D("6.000")) for x in row[1:3])
vin, active, passive = (display(x, D(".1"), D("600.0")) for x in row[3:6])
drive = display(row[6], D(".001"), D("6.000")) * 1000
except (ValueError, InvalidOperation) as error:
raise ValueError(f"{row[0]}: {error}") from None
readings[row[0]] = (rail, reference, vin, active, passive, drive)
if readings.keys() != LEVELS.keys():
raise ValueError("complete all eight worksheet steps")
return readings
def interval(value, counts):
margin = D(".005") * abs(value) + counts
return value - margin, value + margin
def subtract(left, right):
return left[0] - right[1], left[1] - right[0]
def contains(low, high, bounds):
return low <= bounds[0] and bounds[1] <= high
def millivolts(bounds):
return [float(v) for v in bounds]
def observe(name, reading):
rail, ref, vin, active, passive, drive = reading
target = LEVELS[name]
if not D("4.900") <= rail <= D("5.100") or abs(ref - rail / 2) > D(".050"):
raise ValueError(f"{name}: rail or midpoint outside the fixture window")
if abs(vin - target) > (5 if target == 0 else 10):
raise ValueError(f"{name}: measured input outside its step window")
if abs(drive) > 650:
raise ValueError(f"{name}: DRIVE outside the fixture window")
x = interval(vin, D(".2"))
a = interval(active, D(".2"))
p = interval(passive, D(".2"))
amp = interval(drive, D("2"))
expected = ((-x[1] * GAIN_MIN, -x[0] * GAIN_MAX) if target < 0 else (D(0), D(0)))
residual = subtract(a, expected)
active_minus_reflected_passive = (a[0] + p[0], a[1] + p[1])
if target < 0:
feedback_ok = contains(-12, 12, residual)
drive_gap = subtract(amp, a)
passive_response = -20 if name == "neg300" else 8
diode_evidence = (p[1] <= passive_response and active_minus_reflected_passive[0] >= 30
and drive_gap[0] >= 40)
else:
feedback_ok = contains(-8, 8, a)
drive_gap = amp
diode_evidence = (contains(-8, 8, p) and (target == 0 or amp[1] <= -40))
return dict(feedback_ok=feedback_ok,diode_evidence=diode_evidence,
measured_input_interval_mv=millivolts(x),active_interval_mv=millivolts(a),
passive_interval_mv=millivolts(p),
active_minus_reflected_passive_interval_mv=millivolts(active_minus_reflected_passive),
expected_active_interval_mv=millivolts(expected),feedback_residual_interval_mv=millivolts(residual),
drive_signature_interval_mv=millivolts(drive_gap))
def analyze(rows):
measurements = {name: observe(name, rows[name]) for name in LEVELS}
initial, repeat = (rows[name] for name in ("neg200", "neg200_repeat"))
repeat_ok = (abs(initial[1] - repeat[1]) <= D(".010") and abs(initial[2] - repeat[2]) <= 5
and all(abs(initial[i] - repeat[i]) <= 20 for i in (3, 4, 5)))
if not repeat_ok:
status = "repeat_drift"
elif not all(row["feedback_ok"] for row in measurements.values()):
status = "active_transfer_unresolved"
elif not all(row["diode_evidence"] for row in measurements.values()):
status = "diode_contrast_unresolved"
else:
status = "guarded_small_signal_rectification_observed"
return dict(status=status, repeat_ok=repeat_ok, rows=measurements,
scope="Signed steady-state measurements and fixture screens only.")
def main():
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("worksheet", type=Path)
args = parser.parse_args()
try:
result = analyze(read_rows(args.worksheet))
except (OSError, ValueError, csv.Error) as error:
print(f"Worksheet rejected: {error}", file=sys.stderr)
return 2
print(json.dumps(result, indent=2, allow_nan=False))
return 0
if __name__ == "__main__":
raise SystemExit(main())
See the sensor buffer experiment for another use of a voltage follower to keep a load from moving a reference.