Comparing measurements

Two people measuring the same pedal on different interfaces at different levels produce charts that do not overlay — not because either measurement is wrong, but because a drive pedal’s behavior is a function of drive. This page is about making a measurement citable beyond your own rig: the standard condition that turns “my chart” into “a PedalScope measurement,” what an honest comparison between two units can and cannot claim, and where the method itself stops.

The standard condition: PSC-1

A stated, checkable condition converts a chart into a shareable fact. PedalScope already captures every ingredient the condition needs — the volts-per-dBFS factor anchors absolute level, the source fixture travels in the measurement’s identity, and every record embeds its calibration — so PSC-1 is a declaration, not a feature. It is also versioned deliberately: any future change would be PSC-2, never a silent edit.

A measurement qualifies as PSC-1 when all of the following hold and are recorded. The app records every one automatically except the volts factor, which takes a one-time meter reading (below).

1. Drive level: 168 mV peak (±0.5 dB) at the pedal input for the standard fingerprint sweep. Stated in volts, not dBFS, because dBFS is rig-relative — the same digital level is a different voltage on every interface — while volts are portable. The level sits in solidly-picked single-coil territory: musically representative, not arbitrary. A record whose rig has no volts calibration cannot claim PSC-1.

2. Source: the interface output driving the pedal directly (line and monitor outputs are low-impedance sources, typically ≤ ~100 Ω, which is what “direct” assumes) — or a declared fixture, recorded in the measurement’s identity. The standard fixture for impedance-sensitive circuits (vintage-style fuzz — see rig setup) is a 10 kΩ series resistor, declared as such — how to build one and how to declare it so the condition string says so is on Bench fixtures. Direct and 10 kΩ measurements are both PSC-1, but they are deliberately different questions and are not comparable to each other.

3. Load: the interface input the pedal drives, ≥ 10 kΩ. Use a Hi-Z/instrument input where available; 10 kΩ is the honest minimum, not the recommendation.

4. Stimulus: the standard fingerprint plan — the 10-second sweep and standard probe notes, at 96 kHz native (48 kHz is permitted and recorded; there is no hidden resampling — the app enforces that).

5. Calibration: fresh, unbadged, and from the same session — flatness and distortion floor within the app’s sanity checks, latency measured per capture.

6. Controls documented — positions recorded in the app as a variant or control layout; photos encouraged.

A shared chart quotes the condition string: PSC-1 · 168 mV pk · direct or PSC-1 · 168 mV pk · 10k fixture.

The app grades the combination for you, at read time. A Harmonic Distortion record whose recorded provenance meets the condition wears a PSC-1 badge in its header, quoting the condition string, and the same string rides the provenance strip of every chart you copy or save from that record — so a shared PNG carries its own citation. The badge asserts recorded provenance, nothing more: what no machine can check (that the cable really ran straight from the interface output, what your interface’s input impedance is) stays your claim to stand behind.

A record that doesn’t wear the badge is not being scolded — most measurements are exploration at whatever drive the question needed, and they show nothing at all. The explanation lives under the record’s calibration chip (hover it): either the named ingredients that would need to change, or — for records whose calibration carries no volts factor, including everything measured before the factor existed — an honest “can’t be evaluated,” because without a metered factor the drive in volts is unknowable, not wrong.

The Measure at PSC-1 button in the measure flow works the other way: it sets the standard 10-second plan and computes the sweep level that lands your rig on the 168 mV pk anchor, from the volts factor in your selected calibration. On the reference rig that computes to the −26 dBFS default the app has always used; on any other rig it is whatever level your metered factor says. If no factor has been entered yet, the button explains how to get one rather than guessing.

Getting the volts factor

The one ingredient the app cannot capture for you is the volts-per-dBFS factor, because only a meter on your interface’s output can measure it. The Play test tone button makes it a thirty-second job — the full procedure, for both obtaining the factor and verifying it later, is on the calibration page under Verifying your level factor with a meter. Two honesty notes: the factor is valid only for the rig state it was measured at (the calibration’s measured loop gain is what reveals whether a knob has moved since), and the how-you-measured-it story belongs in that calibration’s notes, where it travels with every measurement that used it.

What PSC-1 does not promise

• Unit-to-unit identity. Two honest PSC-1 measurements of two units of the same pedal model will differ by the units’ real component spread. That is a finding about the pedals, not an error in either measurement — the next section is about reading it honestly.

• Comparability across fixtures. Direct and 10 kΩ-fixture measurements answer different questions on purpose; the fixture is part of the condition string so nobody mixes them by accident, and the calibration’s own source-impedance stamp makes Compare flag a mixed pair (see Calibration).

• Time-variant devices. A device whose behavior changes during the measurement is outside the method regardless of condition — see Limits of the method below.

Comparing units honestly

The register rule first, because everything else follows from it: differences are quantities, never verdicts. PedalScope reports “H2 differs by 8 dB at this note” and stops there; whether that difference matters — or which side of it you prefer — is not a measurement’s call.

A worked example from the project’s own bench — its Reference Box, a test box of six switchable stages built from known parts — dividers, clippers, a rectifier, an op-amp crossover — measured 2026-09-04: position P3, a matched antiparallel diode pair (a symmetric clipper), against position P4, the same resistor with one extra diode in series on one side (an asymmetric clipper). Same box, same rig, same session, one part different — and the box’s own arithmetic says where the difference has to land: in the even harmonics. The fundamental agreed to 0.0 dB over the whole band (at 288 Hz, −2.58 against −2.57 dB re input). The even orders separated: H2 read −90.8 dB on P3 and −69.0 on P4 at that note, 20.8 dB apart pooled over 96 notes; H4 read at least 17.2 dB apart, a lower bound because P3’s H4 sat at the floor on 53 of the 83 compared points. The odd orders moved too, and less: H3 5.8 dB (−68.1 against −74.0 at 288 Hz) and H5 5.1 dB — the extra diode raises the clipping threshold on one half-cycle, which reshapes the odd series as well as creating the even one. Overall distance 13.0 dB. Every figure is a quantity: which side of that difference you prefer is not the measurement’s call.

And the honest limit of that example: it is one measured pair, and a pair whose difference was built in. A production unit against another of the same model differs by component spread instead — resistors drift, diode forward voltages scatter — and nobody has yet published measured population spreads for classic circuits. Putting numbers on typical unit spread is planned future work, and exactly the kind of study a shared standard condition exists to make possible. Until then, treat any single reference unit as one sample, not ground truth, and a small measured difference as a quantity — not a defect.

Every difference claim also stands on an instrument floor: how much the same device moves when you measure it again. The Reference Box puts numbers on that too, at two scales. The same P3, measured 2026-08-10 and again 2026-09-04 through a re-patched rig, three weeks and a re-patch apart (one sweep the first time, four averaged the second): at 288 Hz the fundamental read −2.57 then −2.58, H3 −67.8 then −68.1 (0.3 dB), H2 −87.7 then −90.8 (3.1 dB, on a harmonic the symmetric clipper barely produces); the distance between the two reads 0.0 dB overall and 0.0 on every harmonic, because the pooled difference is no larger than the two captures’ own read noise, which the distance subtracts (H2’s six unresolved points carry a bound of at most 6.5 dB, and possibly zero). Read against that floor, the P3-vs-P4 even-order separation of 20.8 dB is the load-bearing evidence, and the 5.8 dB odd-order difference stands nearly twenty times the 0.3 dB the same position repeated to. A same-session bare-loop null run (the loop with no device in it) bounds the other end: against the closing null, P3’s H3 stood 33 dB above the loop’s own read (−68.1 against −101.4 at 288 Hz), while two nulls from the same afternoon read 0.0 dB apart with every harmonic above the fundamental unresolved on both sides — at most 7 to 8 dB, and possibly zero, which is what “nothing there” looks like on this instrument at this averaging. Below the null’s curve, neither position’s behavior is resolvable at all.

Where guitar levels live on the axis

A fair question about any bench measurement: “you test at line level — does that tell me anything about guitar levels?” The premise is mostly false — the standard sweep drives 168 mV peak at the pedal input (the PSC-1 anchor above), solidly-picked single-coil territory — and the stronger answer is that PedalScope doesn’t test at one level at all: the Gain Map and Distortion vs. Level sweep the level axis.

To make that visible at a glance, charts with an input-level axis (Gain Map, Distortion vs. Level, Touch Response, and the Transfer Curve’s input axis) shade the ranges where typical passive pickup outputs live, in volts:

• Single-coil (typical): about 0.05–0.5 V peak.

• Humbucker (typical): about 0.1–1 V peak.

Each band is tagged with a small pickup glyph — a bobbin with one row of six pole pieces for single-coil, two rows for humbucker — so the identity rides shape, not color, and survives into exported charts. The glyph always means “typical range”, never a norm.

These bands are deliberately broad and labeled “typical”, because pickup output varies enormously with the pickup model, coil type, height, and how hard you play. The ranges summarize published pickup-output measurements — Helmuth Lemme’s pick-attack measurements in Electric Guitar — Sound Secrets and Technology and manufacturer output specifications put vintage-style passive single-coils on the order of 100–200 mV RMS in normal playing with attacks reaching a few hundred millivolts peak, and passive humbuckers at roughly double that, with hard attacks on hot units reaching about 1 V peak. Active pickups and piezo systems sit outside these bands. Treat the shading as orientation, never as a norm.

The bands render only when the record’s calibration carries the measured volts factor — on a dBFS-only rig the same shading would be a guess, so nothing is drawn. They can be switched off in Settings → Visualization.

Limits of the method

The measurement assumes the pedal behaves the same during second five of a sweep as during second one. That holds for the classic overdrive/distortion/fuzz families; it does not hold for devices with heavy sag, envelope-dependent gating, slow-recovering blocking distortion, compressor-like behavior, or modulation in the path. Those produce plausible-looking charts that are actually snapshots of a moving target — see What a measurement can’t tell you for the full taxonomy.

The instrument’s defenses are measurements, not disclaimers:

• The Transfer Curve’s memory (nonlinear) metric flags behavior a static curve cannot explain, after the ordinary filter phase that opens every loop has been accounted for.

• Repeatability is checkable directly, and for plugins it has teeth: in one real case, a complex multi-model plugin’s default preset measured twice in a row disagreed with itself by more than 20 dB — a time-variant block, not a broken measurement — and the same pattern is what separates a real aliasing finding from a moving target. The worked example, and the fix, live on Measuring plugins under Default presets can be time-variant.

When the instrument says a device is outside the method, that is the finding. “This pedal’s character is time-dependent” is a true and useful thing to know about it — it’s just not a fingerprint, and no condition string makes it one.