BAT85S Peak Detector: What the Capacitor Loses Between Crests

A diode and a capacitor can hold most of a sine wave’s positive peak. Between peaks, the load draws charge from the capacitor and the held voltage falls. Change one resistor while watching both ends of a Vishay BAT85S Schottky diode. Then use a small part of the trace to test the discharge equation without guessing a constant diode drop.

A peak detector charges a capacitor when the input rises high enough to make the diode conduct toward the capacitor. The diode then blocks most reverse current when the input drops below the held voltage. The repeated variation between the held peak and valley is ripple. The two loads in this lab deliberately trade ripple against how long the circuit remembers a previous peak. All numerical examples are calculated; the worksheet has no bench readings.

Build one grounded low-voltage fixture

• D1: Vishay BAT85S-TAP, the axial DO-35 (DO-204AH) small-signal Schottky diode. The black band marks the cathode. A Schottky diode uses a metal-to-semiconductor junction and can have a lower forward drop than a comparable silicon junction diode. Forward voltage still depends on current and temperature. Do not replace the exact part or infer orientation from the glass color.
• C1: WIMA MKS2C031001A00JSSD, a nonpolar radial 100 nanofarad (nF), ±5%, 63 V direct-current film capacitor. The 2026 WIMA catalogue gives the 5 mm lead pitch and the tolerance/packing suffix. If you measure capacitance, do it with power removed and the part discharged. Enter the measured value and record the instrument; otherwise use and label 100 nF nominal.
• R1 and RL: Vishay MRS25, 1 kΩ series, and either 22 kΩ or 100 kΩ for the load. Use 1% parts and measure the actual loads while disconnected. Kilohm (kΩ) means 1,000 ohms.
• Source and instruments: Tektronix AFG1022 channel 1 (CH1), Rigol DS1054Z, and two Rigol PVP2150 compensated passive probes at 10×. This attenuation mode scales voltage down inside the probe; set the matching 10× factor on the scope so it displays the real tip voltage. Each probe presents nominally 10 MΩ and 10 pF at its tip in that mode.

Keep the protective earth connections intact. This fixture uses only the generator output and one shared ground. Do not connect it to mains, another powered circuit, an amplifier’s floating output, or a charged storage capacitor. Turn the output off before moving the load or diode. Leave a load fitted and verify that H returns near zero before moving leads.

At the proposed ±2.0 V open-circuit signal, R1 and the source’s nominal 50 Ω series impedance limit the positive startup current to about 1.90 mA in the limiting model of a zero-volt output and zero diode drop. Even with 2.2 V positive and R1 at 990 Ω, the conservative ceiling ignoring source resistance is 2.23 mA. The negative half-cycle can reverse-bias D1: if H remains at +2.2 V and A reaches −2.2 V, the difference is 4.4 V. The BAT85S data sheet’s 30 V reverse rating is a limit, not a target for this experiment. Record actual amplitudes and stop if the setup exceeds the proposed low-voltage range.

Capture one full ripple period and a reverse interval

1. With the fixture disconnected, set AFG1022 CH1 to a 1.000 kilohertz (kHz) sine, 0 V offset, 4.0 volts peak-to-peak (Vpp), and the High Z display load. High Z tells the display to assume a high-impedance load. Confirm the open-circuit signal with the scope. Disable the output and connect the fixture with the 22 kΩ resistor.
2. Set both scope channels to direct-current (DC) coupling and the matching 10× factor. Trigger on CH1 with a stable view of several cycles. Use vertical scales that show both traces without clipping; H has a positive offset that must remain visible. Re-enable the source. Let the trace settle over repeated periods, then measure one complete period plus the H peak and valley for that cycle. Save the trace or note the scale and exact cursor positions.
3. Choose a contiguous falling segment of H after the recharge ends. As a starting point, try approximately 0.2 ms and 0.7 ms after the positive source crest. A millisecond (ms) is one thousandth of a second. Keep both cursors in the same gap between recharge pulses. Confirm visually that A stays at least 50 mV below H throughout that interval. Then record the time difference and A/H voltage at each end. The 50 mV margin is a lab selection rule, not a diode threshold.
4. Disable output, let H discharge, fit the 100 kΩ load, and repeat. Keep the source amplitude, frequency, channel factors and capacitor unchanged. Take a second capture with the 22 kΩ load if either trace is unstable or the measurements disagree with your repeatability. Keep the peak and valley from the same stable cycle; avoid single noise spikes.

If the hold trace goes negative, switch the output off and check diode direction, scope ground clips and channel coupling. If the hold trace never rises, check the band, the actual input sine and continuity through R1. If the valley is nearly the peak, zoom CH2 vertically with DC coupling still active. If it remains unresolved, report the scope limit instead of entering an invented droop. Do not move the cursors into recharge merely to get a larger drop.

Calculate only the discharge you actually isolated

When A remains below H, D1 is reverse-biased, so the intended discharge current flows through RL and the finite CH2 probe input. The probe input resistance and RL are in parallel, meaning they draw separate currents from the same node to 0 V. Add their conductances, the reciprocals of resistance. At this frequency, the probe’s nominal 10 pF is about 0.01% of the 100 nF capacitor, so the simple model leaves it out. Diode leakage, stray coupling and capacitor tolerance still matter.

R_eff = 1 ÷ (1/R_L + 1/10 MΩ),   τ = R_eff × C₁

The time constant τ is the time for an ideal discharge to reach about 37% of its starting voltage. With a 100 nF capacitor, the nominal time constants for 22 kΩ and 100 kΩ are about 2.195 ms and 9.901 ms. Starting from a constructed H reading of 1.700 V, the model predicts approximately 1.354 V and 1.616 V respectively 0.500 ms later. These numbers predict a chosen reverse interval; neither is a prediction of the whole-cycle valley or a measured result.

V_B = V_Ahold × exp(−Δt/τ),   τ_window = −Δt ÷ ln(V_B/V_Ahold)

V_Ahold means H at cursor A, not node A. The voltage ratio estimates an apparent time constant only when the hold actually falls by a resolved amount. Whole-cycle ripple is peak minus valley. It includes recharge timing and must be kept separate from the selected window’s drop. The smaller load resistance should make discharge faster under the model. An observed change can also reflect different recharge intervals, source variation, leakage or measurement limits.

Vishay gives forward-voltage limits for specific currents, such as 240 mV maximum at 0.1 mA and 320 mV maximum at 1 mA under the stated test conditions. During recharge, current changes as the sine and capacitor voltage move. Subtracting one assumed forward voltage from the input crest cannot reproduce that waveform. The code deliberately starts from the measured hold value at cursor A.

Enter both loads and run the complete analyzer

Copy the blank comma-separated values (CSV) worksheet. Give each load one row, in ohms, nanofarads, milliseconds and volts. In anode_a_v and hold_a_v, A means the first cursor position; B in the other two columns means the second cursor position on the same trace. Set reverse_segment_confirmed to yes only after inspecting both traces between the cursors. The program checks the endpoints, but two points cannot establish the full interval.

Save peak_hold.py and run python3 peak_hold.py observations.csv. The result reports model τ, observed ripple, observed window drop and the difference from the ideal prediction. An unresolved apparent time constant appears as null if the interval falls by less than 5 mV. That is an analyzer convention; use the actual scale and repeatability of your scope to decide whether a larger reading is resolved. The host tests use constructed examples.

"""Analyze two manually checked diode-off windows of a peak hold circuit."""

import argparse
import csv
import json
import math
import sys


FIELDS = (
    "load", "resistor_ohm", "cap_nf", "period_ms", "dt_ms", "peak_v",
    "valley_v", "anode_a_v", "hold_a_v", "anode_b_v", "hold_b_v",
    "reverse_segment_confirmed",
)
NOMINAL = {"22k": 22_000, "100k": 100_000}
# Well below the fixture's readable precision; only cancels float rounding.
BOUNDARY_EPS = 1e-12


def finite(row, key):
    try:
        value = float(row[key])
    except (ValueError, TypeError) as exc:
        raise ValueError(f"{key}: enter a number") from exc
    if not math.isfinite(value):
        raise ValueError(f"{key}: use a finite reading")
    return value


def analyze_row(row):
    label = row["load"]
    if label not in NOMINAL:
        raise ValueError("load must be 22k or 100k")
    r, c, period, dt, peak, valley, aa, va, ab, vb = (
        finite(row, k) for k in FIELDS[1:-1]
    )
    if not 0.95 * NOMINAL[label] <= r <= 1.05 * NOMINAL[label]:
        raise ValueError("resistor value or label does not match the fixture")
    if not 90 <= c <= 110:
        raise ValueError("enter the 100 nF capacitor value in nanofarads")
    if not 0.95 <= period <= 1.05 or not 0.20 <= dt <= 0.75 * period + BOUNDARY_EPS:
        raise ValueError("check the 1 kHz period and the cursor interval in milliseconds")
    if row["reverse_segment_confirmed"] != "yes":
        raise ValueError("verify the entire interval on both traces before entering yes")
    if not 0 < valley < peak <= 2.3:
        raise ValueError("check for a resolved positive hold peak and valley")
    if not valley - 0.02 <= vb <= va <= peak + 0.02 or vb <= 0:
        raise ValueError("hold cursors must follow a non-rising part of the waveform")
    if not all(-2.5 <= v <= 2.5 for v in (aa, ab)):
        raise ValueError("anode cursor outside the fixture voltage range")
    if min(va - aa, vb - ab) < 0.050 - BOUNDARY_EPS:
        raise ValueError("anode must be at least 50 mV below hold at both cursors")
    effective_r = 1.0 / (1.0 / r + 1.0 / 10_000_000)
    tau_ms = effective_r * c * 1e-6
    predicted_b = va * math.exp(-dt / tau_ms)
    tau_observed = None
    drop = va - vb
    if drop >= 0.005 - BOUNDARY_EPS:
        tau_observed = dt / -math.log1p((vb - va) / va)
    return {
        "load": label,
        "effective_resistance_ohm": effective_r,
        "nominal_model_tau_ms": tau_ms,
        "observed_window_tau_ms": tau_observed,
        "observed_cycle_ripple_v": peak - valley,
        "observed_window_drop_v": drop,
        "predicted_hold_b_v": predicted_b,
        "hold_b_minus_prediction_v": vb - predicted_b,
    }


def analyze(stream):
    reader = csv.DictReader(stream, strict=True)
    if reader.fieldnames != list(FIELDS):
        raise ValueError("use the header from the blank worksheet")
    seen = {}
    for row in reader:
        try:
            if None in row:
                raise ValueError("too many columns")
            result = analyze_row(row)
            if result["load"] in seen:
                raise ValueError("duplicate load row")
            seen[result["load"]] = result
        except ValueError as exc:
            raise ValueError(f"line {reader.line_num}: {exc}") from exc
    if set(seen) != set(NOMINAL):
        raise ValueError("enter one complete row for each load")
    return {
        "assumptions": "Measured reverse interval, resistor and 10 MΩ probe only; no diode model or bench assertion.",
        "rows": [seen[key] for key in NOMINAL],
    }


def main():
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument("worksheet")
    args = parser.parse_args()
    try:
        with open(args.worksheet, encoding="utf-8-sig", newline="") as stream:
            result = analyze(stream)
    except (OSError, UnicodeError, csv.Error, ValueError) as exc:
        parser.exit(2, f"{exc}\n")
    print(json.dumps(result, indent=2, allow_nan=False))


if __name__ == "__main__":
    main()

Challenge

Why might the 100 kΩ load make the circuit worse at following a falling input amplitude, even when it gives less ripple on a steady sine?

Sources and related reading

Download the experiment files: full program, host tests, blank worksheet and original circuit drawing.

• Vishay BAT85S data sheet, document 85513: exact TAP orderable, package, band, limits and test conditions.
• WIMA MKS 2 March 2026 data sheet · WIMA MKS 2 manufacturer ordering table: capacitance and tolerance/packing selection.
• Vishay MRS16/MRS25 data sheet: exact resistor family and tolerance options.
• Tektronix AFG1000 series data sheet: AFG1022 output impedance, ground and High Z load display setting.
• Rigol PVP2150 probe guide: compensation and 10× input resistance/capacitance.
• Rigol DS1000Z data sheet: DS1054Z channel and coupling specifications.

See the voltmeter loading experiment for a circuit where the instrument’s input is a much bigger part of the discharge path. That setup begins with one stored charge; this detector periodically recharges through a diode.