# Stereophile amplifier measurements used

Measurements by John Atkinson, read on 2026-10-18 from the Stereophile pages
below. Only numbers are recorded here (the pages' text and graphs are
Stereophile's copyright). `tools/amp-calibrate.js` turns them into the
amplifier presets. The 19+20 kHz intermodulation and THD-against-frequency
figures, read on 2026-10-29, aren't used by the presets; they check the model
(see the last section). The Pass Labs XA25, WiiM Amp and Sansui 7070 (Hi-Fi
News) were added on 2026-10-29, with clipping into lower loads and output
impedance at 20 kHz for every amp (`js/amp-output.js`).

## Cayin A-300B (300B single-ended triode)
https://www.stereophile.com/content/cayin-300b-integrated-amplifier-measurements
- Output impedance, 8 Ω tap, zero feedback: 2 Ω at low and middle frequencies, 2.7 Ω at 20 kHz (16 Ω tap: 3.75 Ω; 4 Ω tap: 1.2 Ω). With more feedback, 1.5 Ω and 1.8 Ω at 20 kHz.
- Clipping into lower loads: not published as figures; it meets 8 W at 2% "as long as the load impedance is at least twice the magnitude of the transformer tap".
- Response variation into the simulated loudspeaker: ±1.3 dB (8 Ω tap, zero feedback).
- Meets 8 W (9 dBW) at 2% THD; clipping is gentle.
- THD+N 0.44% at 1 W into 8 Ω (Fig.11, 4 Ω tap, zero feedback); distortion heavily low-order, mainly 2nd harmonic; THD rises with power.
- Unweighted S/N ref. 2.83 V into 8 Ω: 73.1 dB (R) / 65.1 dB (L), dominated by 60 Hz hum; A-weighted 89 / 87.2 dB.
- 19+20 kHz twin tone (Fig.13, 8 Ω tap, zero feedback): 1 kHz difference product −56 dB (0.15%). The caption gives 4 W peak into 8 Ω; the text says the 8 Ω tap was driving 4 Ω just below visible clipping.
- THD+N against frequency is plotted at 1 V into 16, 8, 4 and 2 Ω (Figs.8–10); THD rises at the top of the audio band. The graphs weren't on the page on 2026-10-29, so there are no treble figures.

## PrimaLuna ProLogue Premium (EL34 push-pull)
https://www.stereophile.com/content/primaluna-prologue-premium-integrated-amplifier-measurements
- Output impedance: about 2 Ω from the 8 Ω tap (no 20 kHz figure: the tap went into ultrasonic oscillation unloaded), 1.7 Ω from the 4 Ω tap across the audio band.
- Response variation into the simulated loudspeaker: ±1.8 dB (8 Ω tap), ±1 dB (4 Ω tap).
- Clipping (1% THD+N): 35 W into 8 Ω from the 8 Ω tap; 21 W into 8 Ω and 32 W into 4 Ω from the 4 Ω tap. Into 2 Ω it shut itself down at 40 W from either tap.
- THD+N 0.104% at 1 W into 4 Ω (4 Ω tap), strongly 2nd harmonic; 0.584% at 2 W into 4 Ω (8 Ω tap); 3rd harmonic develops at higher currents.
- Unweighted S/N ref. 1 W into 8 Ω: 57.3 dB (L) / 63.1 dB (R), hum at the supply frequency and harmonics; A-weighted 76.8 / 77.7 dB.
- 19+20 kHz twin tone at 13 W peak into 8 Ω (Fig.14, 4 Ω tap), about 3 dB below visible clipping: 1 kHz difference product −56 dB (0.15%); 18 and 21 kHz products −60 dB (0.1%).
- THD+N against frequency at 4.8 V into 8 Ω, 8 Ω tap (Fig.9), read from the graph: about 0.9–1.3% at 20 Hz, 0.35–0.4% at 1 kHz, 0.5–0.6% at 10 kHz, 0.9–1% at 20 kHz (the figure includes hum). Distortion rises at both frequency extremes and into lower impedances; the 4 Ω tap is more linear.

## Parasound Halo A 21+ (Class AB)
https://www.stereophile.com/content/parasound-halo-21-power-amplifier-measurements
- Output impedance: 0.077 Ω at 20 Hz and 1 kHz, 0.1 Ω at 20 kHz (including 6 ft of cable); ±0.1 dB into the simulated loudspeaker.
- Clipping (1% THD+N): 400 W into 8 Ω, 620 W into 4 Ω (both channels driven); 900 W into 2 Ω (one channel driven).
- THD+N 0.01% at 25 W into 8 Ω; predominantly 2nd harmonic, at −74 dB (50 Hz, 100 W into 8 Ω); higher harmonics very low.
- Unweighted S/N ref. 1 W into 8 Ω: 78.1 dB; A-weighted 88.7 dB.
- 19+20 kHz twin tone at 50 W peak into 8 Ω (Fig.9): 1 kHz difference product −86 dB (0.005%); higher-order products lower still.
- THD+N against frequency at 20 V into 8 Ω (Fig.6), read from the graph: about 0.014% at 1 kHz, 0.05–0.07% at 10 kHz, 0.10–0.13% at 20 kHz.

## Buckeye Purifi Eigentakt 1ET9040BA (Class D)
https://www.stereophile.com/content/buckeye-purifi-eigentakt-1et9040ba-monoblock-power-amplifier-measurements
https://www.stereophile.com/content/buckeye-purifi-eigentakt-1et9040ba-monoblock-power-amplifier-new-measurements (June 2025, corrected sample)
- Output impedance: 0.006 Ω at 20 Hz and 1 kHz, 0.0077 Ω at 20 kHz.
- Clipping (1% THD+N), corrected sample: 400 W into 8 Ω, 690 W into 4 Ω, 1,200 W into 2 Ω.
- Low-order harmonics at or below −127 dB ref. 100 W into 8 Ω (APx555); 0.00025% THD+N at 140 W.
- S/N ref. 1 W into 8 Ω: 83.6 dB unweighted wideband, 101 dB audio band, 104 dB A-weighted; supply spuriae at or below −120 dB.
- 19+20 kHz twin tone, corrected sample, 100 W peak into 8 Ω: the few products all close to −120 dB (0.0001%). Original sample, 50 W peak: 1 kHz difference product −120 dB, 18 and 21 kHz products −70 dB (0.03%), plus other high-order products.
- THD+N against frequency at 20 V (50 W into 8 Ω): corrected sample extremely low through the mid-treble with only a small rise above 5 kHz (new Fig.6); the original sample rose steeply in the top two octaves. The corrected sample reaches 1% THD+N at 400 W into 8 Ω with both 1 kHz and 20 kHz signals.

## First Watt SIT-3
https://www.stereophile.com/content/first-watt-sit-3-power-amplifier-measurements (January 2019; read via the Wayback Machine, 2024-09-17 snapshot)
- Output impedance: 0.25 Ω at 20 Hz and 1 kHz, 0.265 Ω at 20 kHz; ±0.2 dB into the simulated loudspeaker; −0.8 dB at 20 kHz.
- Clipping: 8.4 W into 8 Ω at 1% THD+N, the specified 18 W at 3%; into 4 Ω, 16.8 W at 1% and 36 W at 3%. Distortion "rises linearly as the power increases"; it increases into lower impedances (figures not used: the graph isn't archived).
- THD+N 0.244% at 1 W into 8 Ω. Predominantly 2nd harmonic, apparently in opposite phase to the fundamental; the 3rd about 15 dB lower, not rising with power or lower load; higher harmonics from 5 W into 8 Ω.
- Fig.10 (50 Hz, 5 W into 8 Ω), read from the graph: 2nd −45 dB, 3rd −58, 4th −72, 5th −88, 6th −97, 7th −103.
- S/N ref. 2.83 V: 87.3 (L) / 95.4 (R) dB wideband unweighted; 94.5 / 98.1 dB audio band; 107 dB A-weighted. Voltage gain 11.3 dB.
- Model: its distortion doesn't clip at the 1% point, so `tools/amp-calibrate.js` fits a full-output reference (15.89 W), the knee and the 3rd harmonic to all the points above. Result: 0.25% at 1 W, 0.73% at 8.4 W (published 1%), 3.01% at 18 W; at 5 W, 2nd −45 and 3rd −58 dB, higher harmonics below −72 dB. The model's 3rd harmonic grows with level, so it can't stay a fixed 15 dB below the 2nd; that is why the 8.4 W point falls short.
- 19+20 kHz twin tone at 1 W peak into 8 Ω (Fig.11): 1 kHz difference product −54 dB (0.15%); higher-order products relatively low. Similar at the same voltage into 4 Ω (2 W peak, Fig.12).
- THD+N against frequency at 2.83 V (Fig.6): relatively constant with frequency, higher into lower impedances.

## Pass Labs XA25 (Class A)
https://www.stereophile.com/content/pass-laboratories-xa25-power-amplifier-measurements (read 2026-10-29)
- Output impedance (including 6 ft of cable): 0.08 Ω at 20 Hz and 1 kHz, 0.1 Ω at 20 kHz; less than ±0.1 dB into the simulated loudspeaker.
- Clipping (1% THD+N): 80 W into 8 Ω, 130 W into 4 Ω. Rated 25 W, where THD+N is close to 0.01%. The protection circuit tripped on the twin-tone test into 4 Ω.
- Fig.4, THD+N vs 1 kHz power into 8 Ω, read from the graph: 0.0067% at 0.1 W, 0.0022% at 1 W, 0.0018% at 5 W (noise dominates below about 5 W), 0.0026% at 10 W, 0.0053% at 20 W, 0.0091% at 30 W, 0.015% at 40 W, 0.024% at 50 W, 0.045% at 60 W, 0.22% at 70 W. Fig.7: 0.0027% at 10 W, predominantly 3rd harmonic.
- Fig.8 (50 Hz, 20 W into 4 Ω), read from the graph: 2nd −105 dB, 3rd −106 dB ("the second harmonic was equal in level to the third"); the other visible spuriae are at multiples of 60 Hz.
- Fig.6, THD+N vs frequency at 8.9 V, read from the graph: into 8 Ω 0.0026% at 1 kHz, 0.012% at 5 kHz, 0.024% at 10 kHz, 0.04% at 20 kHz; into 4 Ω 0.003% at 1 kHz, 0.1% at 5 kHz, 0.27% at 10 kHz; into 2 Ω 0.016% at 1 kHz, 0.4% at 10 kHz. Stereophile suggests not using it with speakers whose impedance drops below 4 Ω in the treble.
- 19+20 kHz twin tone at 10 W peak into 8 Ω: the products all close to −90 dB ref. the peak signal level.
- S/N ref. 1 W into 8 Ω: 91.2 (L) / 88.4 (R) dB unweighted wideband, 94.6 dB audio band, 97.3 dB A-weighted; mainly 60 Hz and its odd harmonics. Gain 20.4 dB.
- Model: free 2nd and 3rd harmonics fitted to fig.4 from 10 W: full-output reference 112.47 W, knee 0.66; 0.0026% at 10 W, 0.0052% at 20 W, 0.0078% at 30 W, 0.010% at 40 W, 0.013% at 50 W, 0.041% at 60 W, 0.27% at 70 W, 0.72% at 80 W; 3rd harmonic about 30 dB above the 2nd at 10 W. Load exponent 0.2, from 8 to 4 Ω at 1 kHz; with speakers whose impedance falls below 4 Ω, Build notes that the model understates the distortion.

## WiiM Amp (budget Class D)
https://www.stereophile.com/content/wiim-amp-streaming-da-integrated-amplifier-measurements-page-1 (July 2024, read 2026-10-29)
- Output impedance (including 6 ft of cable): 0.1 Ω at 20 Hz and 1 kHz, 1.1 Ω at 20 kHz. ±0.3 dB into the simulated loudspeaker, in the mid-treble.
- Fig.1, response at 2.83 V re. 1 kHz, read from the graph, into 8 / 4 / 2 Ω: 5 kHz +0.07 / −0.13 / −0.34 dB; 10 kHz +0.25 / −0.17 / −1.11; 12 kHz +0.42 / −0.2 / −1.57; 15 kHz +0.57 / −0.25 / −2.08. The line input is digitized at 48 kHz, so the output drops sharply above 22 kHz.
- Clipping (1% THD+N, both channels driven): 66 W into 8 Ω, 129 W into 4 Ω.
- Fig.7, THD+N vs 1 kHz power into 8 Ω, read from the graph: 0.014% at 1 W, 0.006% at 10 W (noise dominates below about 15 W), 0.007% at 20 W, 0.0094% at 30 W, 0.013% at 40 W, 0.015% at 50 and 55 W, 0.086% at 60 W, 0.30% at 63 W. Figs.9 and 10: 0.01% at 20 W (12.67 V), "predominantly the third harmonic"; into 4 Ω the same in the midrange; rising to about 0.06% at 20 Hz (the 2nd harmonic) and about 0.1% at 20 kHz.
- 19+20 kHz twin tone at 20 W peak into 8 Ω: 1 kHz difference product close to −110 dB (0.0003%); higher-order products higher, still very low.
- S/N ref. 1 W into 8 Ω (volume at maximum): 70 dB unweighted wideband, 86 dB 22 Hz–22 kHz, 90 dB A-weighted. Gain 20.5 dB.
- Model: output filter fitted to fig.1 as a series inductor into a shunt capacitor: 20.1 µH, 0.51 µF, with the 0.1 Ω resistance (rms error 0.07 dB; at 15 kHz it gives +0.52 / −0.20 / −2.17 dB). Its open-circuit method gives 0.68 Ω at 20 kHz against the published 1.1 Ω. Distortion: free 2nd and 3rd harmonics fitted to fig.7 from 20 W: full-output reference 66.27 W, knee 0.91; 0.0070% at 20 W, 0.012% at 40 W, 0.015% at 50 W, 0.087% at 60 W, 0.74% at 66 W; the 2nd and 3rd about equal at 20 W.

## First Watt F2 (Nelson Pass's manual, not Stereophile)
https://www.firstwatt.com/wp-content/uploads/2023/08/prod_f2_man.pdf (the designer's own figures; a scanned PDF, read from the page images)
- Output impedance 15 Ω (current source with a shunt); 5 W into 8 Ω at 4% THD (1 kHz), 10 W into 4 Ω at 9%; maximum output ±9 V, ±2.4 A; gain 15.6 dB; noise 20 µV unweighted, 20 Hz–20 kHz; measured with a 25 Ω source into 8 Ω.
- Distortion against power at 1 kHz (graph): about 0.09% at 10 mW, 0.8% at 1 W, about 1.8% just below the knee near 4.5 W, then a sharp rise. It rises in step with the signal, as a pure 2nd harmonic does, so the model's 3rd harmonic is 0 below clipping.
- Model: full-output reference 4.2 W, knee 0.99: 0.80% at 1 W, 1.91% at 4.5 W, 3.61% at 5 W. Noise: the 20 µV is treated as hiss; no hum figure is published. Its ±2.4 A current limit sets the clipping voltage below 3.75 Ω (2 Ω: 5.76 W).
- No twin-tone intermodulation or THD-against-frequency figures.

## Sansui 7070 receiver (Hi-Fi News lab report, not Stereophile)
https://www.hifinews.com/content/sansui-7070-stereo-receiver-lab-report (Paul Miller, October 2025; read 2026-10-29). A mid-1970s receiver (1976–78), rated 60 W.
- Power (<1% THD): 76 W into 8 Ω, 88 W into 4 Ω. Dynamic power (10 ms bursts, <1% THD): 75 / 87 / 35 / 18 W into 8 / 4 / 2 / 1 Ω: no dynamic headroom (tightly regulated supply) and current-limited into low loads. Hi-Fi News: "really only suited to life with 8ohm nominal-rated loudspeakers".
- Graph 1, dynamic distortion vs power, read from the graph. Into 8 Ω: 0.010% at 1–3 W, 0.013% at 5 W, 0.017% at 10 W, 0.020% at 20 W, 0.026% at 30 W, 0.041% at 40 W, 0.073% at 50 W, 0.13% at 60 W, 0.19% at 70 W. Into 4 Ω: 0.012% at 2 W, 0.015% at 5 W, 0.019% at 10 W, 0.028% at 20 W, 0.055% at 40 W. Into 2 Ω: 0.019% at 4 W, 0.024% at 10 W, 0.042% at 20 W.
- Distortion 0.0096–0.038% from 20 Hz to 20 kHz at 10 W into 8 Ω ("fairly flat with frequency", Graph 2); 0.01–0.04% from 1 to 60 W at 1 kHz.
- Output impedance 0.205–0.39 Ω from 20 Hz to 20 kHz, 1.57 Ω at 100 kHz. Response into 8 Ω: −0.3 dB at 20 Hz, 0 dB at 1 kHz, −0.7 dB at 20 kHz, −8.2 dB at 100 kHz.
- A-weighted S/N 73.5 dB ref. 0 dBW (1 W; the text says 73.3), 91.3 dB ref. 60 W, "about 10dB below par". No unweighted figure. Gain +44.5 dB.
- No harmonic spectrum, twin-tone or SMPTE figure.
- Model: free 2nd and 3rd harmonics fitted to the 8 Ω points of Graph 1 with the level set at 10 W (its distortion holds at 0.01% from 1 to 3 W, which a model whose distortion falls with level can't follow) and clipping held hard near 76 W ("no dynamic headroom"): full-output reference 73.64 W, knee 0.93; 0.0047% at 1 W, 0.011% at 5 W, 0.017% at 10 W, 0.027% at 20 W, 0.045% at 40 W, 0.063% at 60 W, 0.18% at 70 W, 1% at 76 W. Load exponent 0.6, the least-squares fit to the 8, 4 and 2 Ω curves at equal voltage. Output inductance 2.5 µH from 0.205 Ω and 1.57 Ω at 100 kHz (0.38 Ω at 20 kHz). Noise: the A-weighted 73.5 dB as hiss, no hum. Clipping into lower loads from the dynamic figures (`clipLoads`).

## Not published for any of these amps
Power supply sag and how lopsided the clipping is: the model has no sag and clips both halves of the wave alike (decided 2026-10-25; earlier presets used unsourced values). Nor SMPTE intermodulation (60 Hz and 7 kHz), which the simulator's distortion table reports, so that figure can't be checked against a measurement.

## Model check: high-frequency intermodulation and THD against frequency (2026-10-29)
The amp model (`js/amp-processor.js`, each preset in `js/presets.js`) was run offline at 96 kHz with hum and hiss off, under each test's conditions into 8 Ω. THD sums harmonics 2–9. Twin-tone products are in dB relative to one of the two tones, which is how we read Stereophile's figures; the SIT-3's agreement supports that reading. The two tones together peak as high as a sine at the stated peak power. The Cayin was run at 4 W peak into 8 Ω, as its caption says.

Twin tone, equal 19 and 20 kHz:

| Preset | Level | 1 kHz product, measured | Model | 18 and 21 kHz, measured | Model |
| --- | --- | --- | --- | --- | --- |
| set300b (Cayin) | 4 W peak | −56 dB | −41 dB | not given | −59 dB |
| pushpull (PrimaLuna) | 13 W peak | −56 dB | −49 dB | −60 dB | −59 dB |
| classab (Parasound) | 50 W peak | −86 dB | −77 dB | below −86 dB | −107 dB |
| measurement (Purifi, corrected) | 100 W peak | about −120 dB | −127 dB | about −120 dB | −135 dB |
| sit3 | 1 W peak | −54 dB | −52 dB | "relatively low" | −75 dB |
| f2 | 1 W peak | not published | −42 dB | not published | far below −100 dB |

THD against frequency, model at 20 Hz / 1 kHz / 10 kHz / 20 kHz:

| Preset | Level | Measured | Model |
| --- | --- | --- | --- |
| set300b | 1 V | rises at the top of the audio band | 0.19 / 0.155 / 0.155 / 0.15% |
| pushpull | 4.8 V | about 0.9–1.3 / 0.35–0.4 / 0.5–0.6 / 0.9–1% | 0.92 / 0.18 / 0.18 / 0.17% |
| classab | 20 V | about 0.014% at 1 kHz, 0.05–0.07 / 0.10–0.13% | 0.014% throughout |
| measurement | 20 V | extremely low, small rise above 5 kHz | 0.00005 / 0.00003 / 0.00003 / 0.00003% |
| sit3 | 2.83 V | relatively constant with frequency | 0.25 / 0.245 / 0.245 / 0.24% |
| f2 | 2.83 V | not published | 0.82 / 0.80 / 0.80 / 0.77% |

Where the model is off:
1. **It has no treble rise in THD.** Its distortion is the same at every frequency apart from the output transformer's bass term. Measured THD at 20 kHz is about 8× the model's for the Parasound and about 5× for the PrimaLuna; the Cayin rises too, by an amount not on the page. The SIT-3 and the corrected Purifi barely rise, as modeled. For tones above 10 kHz the missing harmonics fall above 20 kHz; what reaches the audible band is the intermodulation between high tones, which the model already makes at or above the measured level.
2. **Its 1 kHz difference product is at or above the measured one,** by 15 dB for the Cayin, 9 dB for the Parasound, 7 dB for the PrimaLuna and 2 dB for the SIT-3. The Purifi's is 7 dB below, but both are near −120 dB. The PrimaLuna's 18 and 21 kHz products match (−59 against −60 dB).
3. **The PrimaLuna's midband THD is about half the 8 Ω tap's** (0.18% against 0.35–0.4%, which includes hum). The preset takes its clipping point from the 8 Ω tap and its 1 W distortion (0.104%) from the 4 Ω tap.
4. **The bass rise from the output transformer matches** the PrimaLuna (0.92% at 20 Hz against about 0.9–1.3%).

The model doesn't understate what the amps add to the audible band on these tests; the presets were left unchanged.