Analog vacuum-tube amplification for a three-way loudspeaker
Analog vacuum-tube amplification for a three-way loudspeaker
Documentary audit of fifteen amplifiers and a characterization program for six more plus one externally measured reference · SPL and output impedance calculators.
Sources: official spec sheets from Fezz Audio, Cayin, Line Magnetic and PrimaLuna, the MR-A-01 crossover report, the xDuoo A35 comparison and Stereophile's measurement of the Silver Luna Prestige. Reference system: 3-way MR-A-01, ≈6 ohms nominal / ≈4.9 ohms minimum (simulated in LspCAD) / 89 dB/W/m (working hypothesis, not measured), listening at 3 m.
Vacuum-tube amplifiers for the three-way loudspeaker with MR-A-01 crossover
Zoom inFigure 1. Three integrated vacuum-tube amplifiers on the test bench: Fezz Titania MK2, Cayin Jazz 90 and PrimaLuna EVO 200. The arrangement allows comparison of dimensions, tube layout and control accessibility before the electrical characterization for the MR-A-01 crossover network.
Summary for deciding — what this report is and is not
What it is: a documentary audit of fifteen vacuum-tube amplifiers to drive the three-way MR-A-01 loudspeaker (6 ohms nominal, 4.9 ohms minimum, 89 dB/W/m as a working hypothesis pending measurement). The MR-A-01 is designed and modeled in LspCAD, with a frozen crossover, but has not yet been built as a measured cabinet. This report publishes no ranking or overall scores because the decisive measurements are still missing.
The evaluation uses five criteria with frozen weights: interaction with the real load (35 %), clean power at a defined distortion (25 %), quality at the real listening level between 7 milliwatts and 1 watt plus residual noise (20 %), supervision, protection and ease of maintenance (10 %), and sustained power continuity (10 %). The tap instruction is an admissibility condition that is not scored, and A-D confidence reports separately on the quality of each model's evidence.
Six models selected to measure first — operational priority, not a ranking
Titania MK2 — manual specifies the 8-ohm tap for 6-ohm loudspeakers, automatic bias and individual protection per tube. Jazz 90 — the same unit in ultralinear 50 W stated and triode 28 W stated, allowing study of two modes and two different output impedances. EVO 200 — Adaptive AutoBias and an architecture contrast, but with ambiguity in the damping factor notation. CS-55A — accepts 4 or 8 ohms for 6 ohms per the manual, a good mid-range candidate with KT88/EL34 variants. LM-216IA — 32 W web vs 38 W European manual discrepancy pending closure by measurement. RS-1 — the program's only single-ended class A, a reference of a different architecture.
What is still to be measured and why it matters: output impedance per tap at 100 Hz, 1 kHz and 10 kHz; clean power at 0.1 %, 1 % and 3 % into 4 and 8 ohms; response into a resistive load and into the MR-A-01 RLC load; distortion and noise at the real listening level; and continuity at 10 W for 60 minutes. Without that, nobody knows how much the response is colored or how much clean power reaches the cabinet.
READING GUIDE — NO PRIOR KNOWLEDGE REQUIRED (G1–G6)
Methodological Manifesto — read once
This report publishes no ranking or overall scores. The matrix is frozen at 35/25/20/10/10 (C7 35% real-load interaction, C2 25% clean power, C4 20% low-level quality, C5 10% reliability, C6 10% continuity) with C1 as a non-scoring admissibility condition and independent A-D confidence.
K-measurement = 0% for the six candidates: measured C7 and C4 on the MR-A-01 are missing. Without §12 there are no definitive scores. The documentary datasheets inform, they do not score electrically. The cover recommendations (Titania MK2, Jazz 90, EVO 200) are indicative, based on documentary clarity and maintenance, not a sonic classification.
In the rest of the report you will find short references to this manifesto: [Manifesto §1] — not redundancy, traceability.
This guide explains, without prior knowledge, how a vacuum-tube amplifier is evaluated for a specific loudspeaker. The central idea: a small, verifiable datum is better than a spectacular figure without conditions. Everything that follows can be re-checked with the report's tables and calculators.
Reading guide
The idea that organizes the whole report. No amplifier has been demonstrated to be the electrically optimal one for the MR-A-01. What can be built is a reproducible method for deciding which ones to try first and what data to demand, based on criteria with explicit weights (35/25/20/10/10) and a separate assessment of evidence quality.
- G1 · How evaluation works: the matrix of §1.
- G2 · What the evidence labels and A-D confidence mean.
- G3 · Why a "not determinable" datum is penalized rather than left neutral.
- G4 · How much power is really needed (and why so little).
- G5 · The first practical decision: 4-ohm or 8-ohm tap?
- G6 · What to do before buying.
G1. How evaluation works: the matrix
The evaluation is not an author's judgment. It is a matrix of five weighted criteria summing 100 %, plus C1 (taps) as a non-scoring admissibility condition, and an A-D confidence that does not add up: it reports separately on the quality of the available evidence. Until C7 and C4 are measured, the matrix produces no definitive scores: first measure, then rank. Confidence does not modify the electrical score.
Frozen matrix — five criteria, no ranking formula C7 Interaction with the real load · largest weight (one third) C2 Clean power at defined distortion · one quarter C4 Quality at the real listening level · one fifth C5 Supervision, protection and maintainability · one tenth C6 Sustained continuity · one tenth C1 Taps · admissibility condition, not scored A-D confidence · evidence quality, independent
G2. Evidence and A-D confidence
Every datum in the report carries an evidence label (CONFIRMED, OBSERVED, CALCULATED, MEASURED (external)…) that says where it comes from. On top of those labels, the A-D confidence is built for each model: whether it is verified or merely stated.
G3. Why an ND datum is penalized (rather than neutral)
When a datum is ND (not determinable), the temptation is to treat it as "neutral". But then the least documented amplifier would benefit merely from not publishing conditions. This report applies the opposite rule: the absence of a datum is a real disadvantage. The documentary penalty rule (what it penalizes and how much) is defined in §8 · C2 Clean power.
G4. How much power is needed (and why so little)
The working scenario is 80 dB SPL continuous with +15 dB crest factor, i.e. 95 dB peaks at 3 meters. With the working sensitivity of the loudspeaker (89 dB/W/m, a calculation hypothesis), the tally is surprisingly small: a 60.9 dB whisper is ≈7 mW, 80 dB continuous is ≈0.57 W, and 95 dB crests are ≈17.9 W per channel (100 dB peak: ≈56.6 W). The exact figures, the full derivation and the interactive calculator are in §2.2.
The didactic conclusion: with this loudspeaker, almost every candidate covers normal listening level with room to spare. That is why nominal power matters less than low-power quality (C4) and interaction with the load (C7). You can recompute these figures with the calculator in §2.2.
G5. The first practical decision: 4-ohm or 8-ohm tap?
Vacuum-tube amplifiers have several speaker taps (4 ohms, 8 ohms…). For a loudspeaker of ≈6 ohms nominal, the choice changes the available power, the effective output impedance and the response variation. There is no universal rule: the manufacturer's manual governs. The complete table of documented instructions per model lives in exactly one place, §2.3 · Which tap do I connect?, and each candidate's taps also appear in the master table of §4.
G6. What to do before buying
- 1. Measure the built MR-A-01: impedance and phase vs frequency, real sensitivity, and distortion or compression at 80, 90 and 95 dB SPL.
- 2. Decide the tap (4 or 8 ohms) following the chosen amplifier's manual.
- 3. Demand from candidates: power at 1 % THD+N, output impedance per tap, and response into a complex load.
- 4. Do not discard a model for low confidence or promote it for high confidence: confidence measures what is known, not how it sounds.
Exercises for practice:
- Using the calculator in §2.2, compute the power needed for a 100 dB peak at 3 m in stereo.
- Compute how many dB SPL 1 W produces at 3 m with a sensitivity of 89 dB/W/m.
- Using the calculator in §12.1, compare the attenuation with 4.9 ohms and 6 ohms for an output impedance of 5.4 ohms, and explain why that number does not predict the loudspeaker's real response.
- For a 6-ohm loudspeaker, which tap does the Titania MK2 manual indicate and which does the CS-55A accept, and why the manual governs.
Show solutions
- ≈56.65 W per channel: SPL of 1 W at 3 m = 79.46 dB; 100 dB needs 20.54 dB more, i.e. 113.3 W per cabinet; with the +3 dB stereo sum, 56.6 W per channel.
- 79.46 dB: 89 − 20 * log10(3) = 89 − 9.54.
- ≈0.88 dB difference between the two point attenuations. The real measurement of the Silver Luna into a full simulated load was ±3.4 dB: the difference between two resistors does not replace the loudspeaker's full impedance curve (magnitude and phase) together with output impedance as a function of frequency.
- The Titania MK2 manual indicates the 8-ohm tap for 6 ohms; the CS-55A's accepts both. The manual governs because the manufacturer designed the transformer for that ratio and knows each tap's thermal and response behavior.
1. The evaluation matrix
Why these weights? Because with an already-designed loudspeaker of medium sensitivity listened to at 3 m, what separates a good amplifier from an unsuitable one is not nominal power: it is how it behaves on the real load (C7, 35 %), how it sounds at the real listening level (C4, 20 %), and how much clean headroom it offers for crests (C2, 25 %). Reliability and continuity, 10 % each, discriminate between comparable models. C1 (taps) is not scored: it is an admissibility and safety condition — a manufacturer's manual instruction does not certify better electrical performance, and turning it into a score would only add confusion to the technical result.
| Criterion | Weight | Subject |
|---|---|---|
| C7 Interaction | 35% | Response on an RLC load representative of the MR-A-01 |
| C2 Clean power | 25% | Power bands at THD+N ≤ 1 %, worst relevant load |
| C4 Low power | 20% | THD+N and noise at listening level |
| C1 Taps | admissibility | Admissibility condition: documented taps and instruction for 6 ohms; not scored |
| C5 Reliability | 10% | Bias, protections, service |
| C6 Continuity | 10% | Sustained continuous power, both channels |
| A-D confidence | independent | Does not add: reports on evidence quality (§11) |
C7 is scored with the maximum absolute response variation between 20 Hz and 20 kHz relative to the 8-ohm reference on the same tap and at the same level. Output impedance is measured at each tap and test level.
H_8Ω(f) = 8 / (8 + Zout(f))
Δ(f) = 20·log10 |H_RLC(f) / H_8Ω(f)|
Δn(f) = Δ(f) - Δ(1kHz)
Δmax = max_{20Hz≤f≤20kHz} |Δn(f)|
C2 uses a band scale (§8). C6 requires a reproducible protocol: 10 W for 60 min on both channels, plus 80 % of P@1 % for 10 min, recording power, THD+N, temperature, bias and protections. The RLC load starts from the MR-A-01 simulation and is replaced by real measurement once it exists, documenting the change. There is no publishable electrical ranking until C7 and C4 are complete.
- K-measurement: sum of criterion weights with a valid, comparable measurement. Currently K-measurement = 0% for the six candidates (C2-M, C4-M, C6, C7 ND).
- Scorable documentary K: weight of criteria with sufficient evidence to apply the documentary scale. Currently C5 contributes at most 10%. No ranking is built from these coverages.
What M and D mean: M identifies a normalized measurement result. D identifies documentary information or estimation. C2-D and C4-D may remain on datasheets to describe what the manufacturer publishes, but they do not replace C2-M or C4-M and do not feed the electrical adequacy score.
2. Reference system and required power
2.1 Electroacoustic load
Before comparing amplifiers, the load they will drive must be known. A loudspeaker is not a fixed resistor: its impedance varies with frequency, and that variation is precisely what tests the amplifier.
| Parameter | Working datum | Status and consequence |
|---|---|---|
| Configuration | 3-way; 27 cm woofer, 7.7 cm midwoofer, 5 cm tweeter | CONFIRMED LspCAD 6.20 pro model on .frd/.zma files. |
| Crossover | 5 components; parallel ways | CONFIRMED Values frozen as MR-A-01 before construction. |
| Nominal impedance | ≈ 6 ohms | CONFIRMED Design reference. |
| Minimum impedance | ≈ 4.9 ohms around 1.4 kHz | CALCULATED LspCAD prediction; pending measurement. |
| Maximum impedance | ≈ 8 ohms | CALCULATED Theoretical maximum of the MR-A-01 model ≈7.8 Ω around 300–400 Hz (LspCAD 6.20 pro screenshot 3-ways-opt-2.7 uF, Fig. 2.1). Together with the 4.9-ohm minimum it defines the 4.9–8 ohm excursion over which C7 is scored. See the MR-A-01 crossover report. |
| EPDR | ≥ 3.5 ohms | HYPOTHESIS Model estimate published in the MR-A-01 crossover report; pending verification with the measured impedance and phase of the built cabinet. |
| Sensitivity | 89 dB/W/m | HYPOTHESIS for calculation, not measured. |
| THD/compression | — | ND The model does not allow concluding at what SPL the cabinet shows significant compression or distortion. |
Physical status of the MR-A-01. SIMULATED The reference system is designed and modeled in LspCAD 6.20 pro, with the crossover frozen before construction: today it does not exist as a measured cabinet. All figures in this report describe the model, not a physical unit.
2.2 Listening scenario and SPL calculator
The working scenario is 80 dB SPL continuous with +15 dB crest factor (95 dB peaks) at 3 meters, free field, with an approximate stereo energy sum of +3 dB. The key tally: with 89 dB/W/m, 1 W at 3 m produces 79.46 dB (89 − 20 * log10(3)). From there, every additional 3 dB doubles the power.
| Target SPL (peak) | Power per cabinet | Power per channel in stereo (+3 dB) |
|---|---|---|
| 95 dB | ≈35.8 W | ≈17.9 W |
| 100 dB | ≈113.3 W | ≈56.65 W |
SPL and required power calculator
Formulas: SPL_1W@d = sens − 20 * log10(d) and P = 10 ^ ((SPL_target − SPL_1W@d − factor)/10). Results per channel. The acoustic correction factor represents an assumed scenario, not a guaranteed room gain.
2.3 Which tap do I connect? — 4 ohms or 8 ohms
This is the first practical decision with a loudspeaker of ≈6 ohms nominal, and it changes the available power, the effective output impedance, the damping factor and the response variation with the load.
| Tap | Transformation ratio | Zout and power | Response | Recommendation |
|---|---|---|---|---|
| 4 ohms | Lower turns ratio; the impedance reflected to the primary drops with the square of the ratio | Reflected Zout ≈ half that of the 8-ohm tap; available power is usually equal or lower, depending on the output stage's current margin | Lower divider attenuation on the speaker's Zmin (4.9 ohms), slightly lower coloration in theory | Consult the manual; measure both taps on the MR-A-01 |
| 8 ohms | Higher turns ratio; the 6-ohm loudspeaker loads the primary above what was intended | Reflected Zout ≈ double; the power delivered to the loudspeaker can be higher if the stage retains margin | Greater attenuation on Zmin; potentially greater coloration with high Zout | Does not presuppose which tap performs better |
3. Characterization program 6+1
Not all amplifiers deserve the same measurement effort. The program selects six candidates covering the architectures and philosophies relevant to the MR-A-01, plus one already measured external reference that serves as a control for the method.
| Order | Amplifier | Role in the program |
|---|---|---|
| 1 | Fezz Titania MK2 | Main candidate: documented instruction 6 ohms → 8 ohms and automatic bias |
| 2 | Cayin Jazz 90 | Power and documentation contrast |
| 3 | PrimaLuna EVO 200 | Architecture and protection contrast (Adaptive Auto Bias) |
| 4 | Cayin CS-55A | Power and tap contrast (4-ohm or 8-ohm accepted) |
| 5 | Line Magnetic LM-216IA | Classic push-pull comparison |
| 6 | Acronos RS-1 | Single-ended class A contrast |
| — | Fezz Silver Luna Prestige | Measured external reference (independent laboratory) |
Each candidate undergoes the same protocol of §12, with the identical MR-A-01 reference RLC load, the same test level and the same measurement conditions. The tap used will be the one expressly recommended by the manufacturer for a nominal 6 Ω load (Titania MK2 → 8 Ω per manual; CS-55A accepts 4 Ω or 8 Ω; EVO 200 pending manual verification). Where the manufacturer accepts both taps, both may be tested, recording power, Zout and response separately. Results from different taps will not be compared as if they were identical electrical conditions. The main comparison is between admissible configurations for the loudspeaker, not between arbitrary selector positions. Only after that trial does the matrix produce real scores and, with them, a defensible evaluation.
What stays out and why. The Jadis Aria —top choice of the xDuoo A35 comparison— and the Cayin Jazz 80 —the only one with attributed external powers— stay out for economy of effort, not as a rejection: the Aria repeats the class AB push-pull architecture already represented among the candidates and publishes no verifiable P@1 %; the Jazz 80 is subsumed by the Jazz 90, which shares the platform and adds the ultralinear/triode switch. No exclusion is a quality judgment: it allocates measurement effort where a measurement can change the picture most.
4. Master table and candidate notes
This table gathers the available documentary information on the candidates analyzed. The 6+1 program members are identified by experimental interest; the 9 documentary alternatives are kept as unscored backup. No datasheet by itself generates an electrical adequacy score. Comparable results will be incorporated into the matrix only after completing the §12 trials (C7 and C4). The nuances that do not fit in a cell (pending verifications, confidence corrections, documentary discrepancies) are in the notes by model; A-D confidence is defined in §11. The measured external reference (Fezz Silver Luna Prestige) has its own contrast case in §6 and does not appear here.
Master table of candidates — a single view
Fifteen candidates in a single table: topology, stated power with its condition, taps, bias, European service, corrected confidence and —where it exists— nauscopio's own report on that model. Price does not score, but the cost of ownership (tube set, service life) informs the purchase decision and is detailed below.
| Model | Topology / tubes | Stated power / condition | Taps / 6-ohm instruction | Bias | EU service | Conf. | nauscopio report |
|---|---|---|---|---|---|---|---|
| Titania MK2 | PP KT88 ultralinear | 60 W nom; 1 % at 1 kHz, no load stated | 8-ohm tap for 6 ohms (manual) | Auto, individual protection | PL/EU yes | C | Titania MK2 |
| Jazz 90 | PP KT88 UL / triode | 50 W UL / 28 W triode; 0.2 % at 1 W/1 kHz, <5 % full power | 4/8 ohms, no 6-ohm instruction | Auto | CN/EU yes | C | Jazz 90 |
| EVO 200 | PP EL34/KT88 switchable | 44 W EL34 / 45 W KT88; <0.1 % at 1 W, <2 % full power | 4/8/16 ohms, no 6-ohm instruction | Adaptive Auto Bias | NL/EU yes | C* | — |
| CS-55A | PP KT88/EL34 | 40 W KT88 / 38 W EL34; 1 % at 1 kHz, no load stated | 4-ohm or 8-ohm accepted | Auto | CN/EU yes | C | — |
| LM-216IA | PP EL34 | 32 W web / 38 W EU manual (discrepancy); 1 % at 1 kHz | 4/8 ohms | Manual | CN/EU yes | C | — |
| RS-1 | SE class A KT150 | 25 W max; 0.5 % at 25 W, 18 W at 0.01 % (PND) | 4/8 ohms | Manual | ES yes | D | RS-1 |
| Galion TS34 V2 | PP EL34, switchable class A/AB (KT88 optional) | 25 W A / 40 W AB; ≤0.32 % A / ≤0.8 % AB at 20 W/1 kHz | 4/8 ohms, no 6-ohm instruction | Manual | US/EU distribution | C | — |
| Jazz 80 | PP EL34 | 2×40 W, no condition stated | 4/8 ohms | Auto | CN/EU yes | C | P52 vs Jazz 80 · P52 vs Jazz 90 |
| Jazz 70 | PP EL34 AB1 UL/triode | 35 W UL / 17 W triode; <5 % full power (loose) | 4/8 ohms | Auto | CN/EU yes | C | — |
| Black Ice F22 V2 | Compact PP | ≈50–60 W (press) PENDING | 4/8 ohms | Exact Bias (panel) | US/EU | C | — |
| Icon Stereo 40 III | PP EL34/KT88 UL/triode | 40 W UL / 20 W triode (distributor) | 4/8 ohms | Manual | UK/EU yes | C | — |
| xDuoo A35 | Compact | ND | ND | ND | CN/EU yes | D | xDuoo and 14 amps |
| Willsenton R8 | PP KT88/EL34/6550 AB1 | 45 W UL / 25 W triode, no THD stated | 4/8 ohms | ND | CN/EU | D | — |
| Jadis Aria | PP EL34, stated class A | 30 W, no THD/load/frequency | ND | Manual | FR/EU yes | D | — |
| Acronos P52 | Handcrafted ultralinear PP | 40 W; P@1 % ND | ND | Manual | ES yes | D | P52 vs Jazz 80 · P52 vs Jazz 90 |
Operating modes — one row per mode (measure power and Zout per mode)
| Model | Mode | Power | Expected Zout | Impact |
|---|---|---|---|---|
| Jazz 90 | Ultralinear | 50 W | Lower | C2 1.00 if 50 W at 1 % THD+N is verified (today unverified) |
| Jazz 90 | Triode | 28 W | Higher (hypothesis, pending measurement) | C2 drops and C7 worsens (hypothesis) |
| Jazz 70 | Ultralinear | 35 W | Lower | C2 0.85 if 35 W at 1 % THD+N is verified (today unverified) |
| Jazz 70 | Triode | 17 W | Higher | Below the 17.9 W threshold |
| Galion TS34 | Class A | 25 W | Higher | — |
| Galion TS34 | Class AB | 40 W | Lower | C2 0.90 if 40 W at 1 % THD+N is verified (today unverified) |
| Willsenton R8 / Icon 40 III | Switchable UL/triode | 45/25 W and 40/20 W | Triode: hypothesis of higher-Zout architecture, pending measurement | Measure per mode |
Cost of ownership — tube set and service life (not scored)
| Model | Power tubes | Indicative service life | Indicative replacement | Note |
|---|---|---|---|---|
| Titania MK2 | 4× KT88 + 4× 12AX7 | ≈3000 h power / ≈10,000 h small-signal | ≈€400–500 (KT88) | Automatic bias, simple replacement |
| EVO 200 | 4× EL34 or KT88 + small-signal | ≈3000 h (Adaptive AutoBias extends life) | ≈€150 (EL34) / ≈€400 (KT88) | AutoBias: better C5 |
| Jazz 90 | 4× KT88 | ≈3000 h | ≈€400 | — |
| CS-55A | 4× KT88 or EL34 | ≈3000 h | €150–400 | Accepts both families |
| LM-216IA | 4× EL34 | ≈3000 h | ≈€150 | — |
| RS-1 SE | 2× KT150 + drivers | ≈3000 h (KT150 expensive) | ≈€300–400 per pair | SE class A: higher consumption |
Galion, Black Ice, Icon, Willsenton, xDuoo, Jadis and P52: pending completion with the distributor (real service life and spare-parts availability in Europe).
Notes by model — nuances the master table does not capture
- Titania MK2. The <0.05 % THD at full power is self-declared and unusually good for a vacuum-tube push-pull: it requires verification. The DF >40 has no MK2 measurement; a DF ≈1 measurement attributed to an earlier generation does not refute the current specification, but justifies measuring its Zout.
- Jazz 90. The most complete datasheet in the program: test conditions stated (1 W, 1 kHz) and no discrepancies between sources. DAC, Bluetooth and phono are out of scope.
- EVO 200. The official sheet's "Damping Coefficient 4 KD (1 kHz)" has ambiguous notation and is not interpretable without measurement: hence its C* confidence, not B.
- CS-55A. The 1 % THD is stated at 1 kHz without identifying the test load: the datum is CONFIRMED without complete conditions.
- LM-216IA. The practical difference between 32 and 38 W is ≈0.75 dB, smaller than the documentary uncertainty. Until the exact product version is identified, C2 is scored with the lower limit: 32 W → 0.79 (if 38 W is confirmed → 0.88 and the score rises pending comparable measurement).
- RS-1. The 18 W at 0.01 % figure verified in nauscopio's own RS-1 report, includes the manufacturer's response citing a damping factor of ≈50. 25 W maximum power. The program's only single-ended unit: pending C7 measurement on an RLC load for interaction with the MR-A-01.
- Galion TS34 V2. The THD bands are stated at 20 W (not full power, not at 1 % THD+N); SNR ≥90 dB and 19 Hz–65 kHz response stated. No tap instruction for 6 ohms and no bias data.
- Jazz 80 / Jazz 70. 40 W with no test condition at all (shop-window ceiling); the Jazz 70's <5 % THD at full power does not allow deriving power at 1 %.
- Black Ice F22 V2. The ≈50–60 W circulate via the press without closed conditions: PENDING VERIFICATION.
- Jadis Aria. Pure class A at 30 W is self-declared and would require thermal and continuity verification (C6): that is why it stays off the list.
- Acronos P52. C1 ND; no electrical parameters published with conditions. A nauscopio P52 vs Jazz 80 report exists.
Pending measurements and their impact on the order
This table summarizes what remains to be measured and how much each measurement can move the list when it arrives. "Impact" indicates the potential for change in the evaluation result, never the amplifier's quality.
| Pending quantity | Status | Potential impact on the order |
|---|---|---|
| C7 · Response on the MR-A-01 RLC load | ND in all 15 candidates | High — weighs 35 % and today does not discriminate; can change the evaluation result once a measured matrix exists |
| C2 · Power at 1 % THD+N, both channels, 4/8 ohms | Partial — nominal power documented, threshold ND | High — distinguishes stated power from clean power; affects the distinction between models with partial documentation |
| Zout per tap (100 Hz, 1 kHz, 10 kHz) | ND in all 15 candidates | High — it is the input to C7 and to the real damping factor |
| C4 · THD+N at 7 mW–1 W and residual noise | ND almost total (only THD at 1 W stated in some) | High — weighs 20 % and today is ND — pending measurement per §12 |
| C6 · Continuity (10 W/60 min; 80 %/10 min) | ND — no trials on any model | Medium — currently outside the calculation (ND); only matters if a model fails its continuity trial |
| Real MR-A-01 sensitivity (at 2.83 V/1 m and converted) | HYPOTHESIS 89 dB/W/m | High — moves the 17.9 W threshold and with it the whole C2 scale |
| LM-216IA 32/38 W discrepancy | PENDING VERIFICATION | Medium — pending comparable measurement |
5. Trial priority and selection criteria
This report publishes no ranking or overall scores. The matrix is frozen at 35/25/20/10/10 and K-measurement = 0 % for the six candidates (C7, C2-M, C4-M, C6 ND; meaning of M and D in the "Coverage" callout of §1). In this documentary phase only C5 contributes at most 10% of informative coverage; it generates no electrical adequacy. It describes maintenance and protection, not electrical adequacy. No performance positions are assigned.
The selection of six amplifiers responds to experimental interest: covering architectures (KT88/EL34 push-pull, KT150 single-ended), bias contrasts and having verifiable tap instructions for 6 Ω. It is an operational priority, not a classification.
Operational priority — where a measurement can change the picture most
If you could only try three: 1) Titania MK2 — clear 6 Ω → 8 Ω instruction per manual (§2.3), automatic bias with individual protection; measure first Zout at 100 Hz/1 kHz/10 kHz per tap and response on the RLC load at 1 W (§12). 2) Jazz 90 — the program's most complete datasheet (0.2 % at 1 W/1 kHz, <5 % at full power, conditions stated), UL (50 W) / triode (28 W) contrast in the same unit; measure power and Zout separated by mode. 3) EVO 200 — Adaptive AutoBias, architecture contrast and ambiguous "4 KD" DF notation; measure effective protection, 6 Ω tap confirmation (§4) and low-level behavior (C4). Then CS-55A (accepts 4 Ω or 8 Ω for 6 Ω per manual, good mid-range candidate with KT88/EL34 variants), LM-216IA to close the 32/38 W discrepancy (notes §4) and RS-1 as the single-ended class A contrast.
5.1 How to minimize the purchase error before the test bench
The report abandons any previous global index and assigns no performance positions or overall scores. The earlier documentary sub-scores were non-reproducible estimates with editorial penalties lacking a formal rule; they are not kept as matrix data. Current result: six candidates selected for experimental interest. Buying before measuring the built MR-A-01 means accepting risk. This section reduces that risk with a verifiable checklist.
| Work order | Model | Reason to measure it | Main uncertainty |
|---|---|---|---|
| 1 | Titania MK2 | Main reference; documented 6Ω→8Ω connection; stated individual protection | Real clean power, Zout and RLC response |
| 2 | Jazz 90 | UL/triode contrast in the same unit | Power and Zout separated by mode |
| 3 | EVO 200 | Adaptive AutoBias; architecture contrast | Effective protection, 6Ω tap and low level |
| 4 | CS-55A | Per manual accepts both taps for 6Ω; KT88/EL34 variants | Clean power and per-tube/per-tap differences |
| 5 | LM-216IA | EL34 push-pull; 32/38 W discrepancy | Exact version, P@1% and Zout |
| 6 | RS-1 | Single-ended class A contrast | Continuous power, Zout and thermal stability |
🧰 Buyer's Toolbox — visible without clicks
5.1.1 Checklist and 5.1.5 email template: the complete versions, without cuts, live in §5.1.1 (documentary checklist: manual, power with closed conditions, Zout/DF with method, bias, EU service, tubes, warranty, right of return) and in §5.1.5 (email template to copy/paste to the distributor). They are not repeated here to keep a single authoritative copy.
🔀 Decision flowchart — what do I do?
├─ YES → Is it 8Ω (Titania) or does it accept both (CS-55A)? → Note the recommended tap = valid C1
└─ NO → Ask the manufacturer for written confirmation → If no answer = ND (do not buy without measuring)
EU technical service + tube stock?
├─ YES → Trial priority +1
└─ NO → High maintenance risk (C5 penalizes)
P@1% with closed conditions published?
├─ YES → Compute headroom with the §2.2 calculator → ≥7 dB headroom over the power needed for a 95 dB peak at 3 m (editorial threshold of Fig. 5.1)?
└─ NO → ND → Demand the datum or measure yourself (§12)
Result: Titania MK2 (clear docs) / Jazz 90 (power) / EVO 200 (auto bias) = indicative 1-2-3 priority
Price versus dynamic headroom: a complementary perspective
The following chart compares estimated purchase price and the available power margin for reaching 95 dB SPL peak. It is not a sound-quality ranking and does not replace the MR-A-01's electrical adequacy matrix. It adds an economic reading the trial-priority table does not offer: visualizing the trade-off between investment and stated headroom.
- It is not a quality ranking. The technical priority of §5 answers "what is worth measuring first"; this chart answers "what stated margin do I get per euro invested".
- Two acoustic hypotheses: 7.149 W/channel (chart, with +4 dB room) vs 17.9 W/channel (§2.2, no room gain). Headroom here cannot be transferred directly to the main report's SPL calculator.
- Black Ice F22 V2: ≈€4,400 delivered in Tarragona, including estimated import, customs and shipping costs. Do not use the ex-works USA figure for an EU comparison.
- 7 dB threshold: an editorial visual guide to group budget / intermediate / high candidates. It does not imply that below it sounds worse, or that above it there is real measured margin.
- Validity: the power figures are stated or estimated with C/D confidence per §11. The final decision still depends on the pending measurements of C2-M, Zout per tap and C7 on the MR-A-01 RLC load (§12).
5.1.1 What to demand from the seller before paying — documentary checklist
No datum not in the manual or an official sheet with conditions is comparable. If the seller does not provide it in writing, it is ND.
- Complete PDF manual for the exact model and revision (not a brochure). Verify: available taps and the explicit instruction for 6 Ω — the Titania MK2 says 6Ω→8Ω, the CS-55A accepts 4 or 8Ω. If absent, demand written confirmation from the manufacturer/distributor. This is C1.
- Power with closed conditions: P@1% THD+N, at 1 kHz and in a 20 Hz–20 kHz sweep, with load (4Ω/8Ω), both channels, and what "nominal" means. If there is only "60 W" without THD/load/frequency, it is a shop-window ceiling.
- Zout or DF with method: How was DF measured? Zout at 100 Hz / 1 kHz / 10 kHz per tap? A DF>40 without an MK2 measurement refutes nothing, but certifies nothing either. Ask for the two-load method if it exists.
- Bias and protection: Auto, supervised auto, manual with test points? Individual protection per tube with indication? What happens when a tube fails? The Titania declares individual protection, the EVO 200 Adaptive AutoBias. Ask for the replacement procedure.
- European service: distributor with stock of transformers, boards and matched tubes, repair lead time, shipping cost. Without documented EU service, the model moves to the waiting list per §13. Note country and contact in the "EU service" column of §4.
- Stock tubes: brand and matching grade of the included KT88/EL34/KT150, whether new or factory burned-in. Ask for a matched spare quartet/pair with Ip measurement.
- Real warranty: years on electronics, months on tubes, whether it covers bias and transformers, and whether the warranty requires servicing at an official center.
- Right of return: days, who pays shipping, whether they accept returns with hours of use (<30 h), and the state of the original packaging. In writing.
5.1.2 What to check at home in the first 14 days — without laboratory instruments
You do not need an analyzer. You need method and a notebook. Home inspection is limited to what you can check without opening the unit or touching the high-voltage circuit.
- Initial inspection (15 min). Unbox, photograph the serial number, tubes and factory bias. Check there is no shipping damage, no crackling or burning smell after switch-on. Let it warm up 30 min with no signal.
- Noise and hum — C4. Inputs shorted or with no source, volume at minimum and then at 9 o'clock/12 o'clock. Get within 1 m of the loudspeaker: 50 Hz hum? Note per channel.
- Balance and gain. Low-level mono signal (60–70 dB). Centered image? If the imbalance is >1 dB, ask for a replacement.
- 4Ω vs 8Ω tap — subjective observation, not an electrical decision. Only if the manual accepts both for 6 Ω (e.g. CS-55A) and following instructions, listen to both. Note impressions, but the correct tap is the one documented by the manufacturer.
- UL/triode modes. Verify the mode change per the manual (off/standby). An SPL app is only an indicative reading of acoustic level, not an electrical test.
- Thermal stability — C6. After 60 min at conversation level (80 dB ≈0.57 W), observe without touching: any smell? Does the bias drift on the LEDs? Do not touch transformers, tubes or chassis: they can reach high temperatures and pose a burn hazard.
- Document everything. Date, hours of use, tap, tubes, incidents.
5.1.3 Common mistakes that make the purchase more expensive
- Buying by nominal watts. With a hypothetical 89 dB/W/m and listening at 3 m, 17.9 W/channel gives 95 dB stereo peaks. Beyond 35–50 W there is no audible gain in this scenario. C7 (35%) and C4 (20%) weigh more.
- Confusing the 4Ω tap with "more control". Zout depends on the transformer and the feedback, not on the printed figure. Only the manual governs, and then the measurement on an RLC load (§12).
- Assuming EL34 and KT88 sound the same. Bias, power and Zout change. In the EVO 200 and CS-55A, each family needs its own adjustment and replacement (€150 vs €400). Do not change family without re-biasing if it is manual.
- Immediate tube rolling. Do not change tubes in the first week. First establish the noise and balance baseline with the stock tubes.
- Ignoring the cost of ownership. A KT88 quartet costs €400–500 every ≈3000 h. A single-ended class A like the RS-1 consumes more and needs more expensive tubes. See the table in §4.
- Buying without confirming EU service. An amplifier without spare parts in Europe is a waiting-list item, not a candidate, per §13.
5.1.4 Total cost of ownership — quick template
Price does not score in the matrix, but it decides whether you can keep the amplifier for 5 years.
- Initial price — does it include VAT, shipping, matched tubes?
- Power set — 4× matched KT88? 2× KT150? (≈€150 EL34 · €400–500 KT88 · €300–400 KT150 pair)
- Service life — expected hours with correct bias? (≈3000 h power · ≈10,000 h small-signal)
- Bias — auto with individual protection? Manual with test points? (Titania 1.00 · EVO 200 0.60 · manual points 0.40)
- Repair — stock of transformers and boards in the EU? Lead time? No datum = ND = risk.
- Resale — does it hold value with stock tubes? Models with EU service and a complete manual resell better.
5.1.5 Email template to the distributor — copy/paste
Subject: Technical enquiry [exact model + revision] for a 6Ω nominal / 4.9Ω minimum loudspeaker
If the answer does not include conditions (load, THD, frequency, tap), it feeds neither C2 nor C7 and remains ND in the matrix. It is still useful information: it tells you that you will have to measure that datum yourself per §12.
6. Contrast case — Fezz Silver Luna Prestige
The Silver Luna Prestige is not a candidate: it stays outside the program. Even so, it is the report's most instructive case, because an extensive independent laboratory measurement exists, performed by John Atkinson at Stereophile. That said, the tested unit exhibited a significant anomaly in the right channel (about 3 dB less gain and more distortion), so the later power measurements were taken on the left channel only. It is the report's only A confidence.
- Measured Zout: ≈5.4 ohms at 1 kHz from the 8-ohm tap; ≈2.8 ohms from the 4-ohm tap. MEASURED (external)
- Response variation on a simulated load: ±3.4 dB (8-ohm tap); ±2.1 dB (4-ohm tap). MEASURED (external)
- Power: the laboratory did not reach 35 W with its usual definition of clipping at 1 % THD+N; that power arrived at 10 % THD+N in pentode mode. MEASURED (external)
- Anomaly of the measured unit: gain imbalance of about 3 dB between channels and more distortion in the right one, probably a fault of that unit.
| Tap | 20 Hz | 1 kHz | 20 kHz |
|---|---|---|---|
| 8 ohms | 6.5 ohms | 5.4 ohms | 6.25 ohms |
| 4 ohms | 3.2 ohms | 2.8 ohms | 3.0 ohms |
Response variation on Stereophile's standard simulated load: ±3.4 dB 8-ohm tap, ±2.1 dB 4-ohm tap. It is not an MR-A-01 prediction. Source: Stereophile measurements.
Didactic reading. On the scale of §7, a ±3.4 dB variation would correspond to C7 = 0.20 as an illustration of the scale on another simulated load — exactly what is to be expected from a Zout of 5.4 ohms on a loudspeaker whose impedance swings between 4.9 and 8 ohms. The case demonstrates two things. First: output impedance and nominal power are distinct variables, and both must be verified — a Zout of 5.4 ohms can modify the response on the load even if the power is sufficient. Second: the real variation on a complex load cannot be deduced from the attenuation between two point resistances; for that, the full impedance curve (magnitude and phase) and Zout versus frequency are needed.
Note: The Silver Luna's measured behavior on Stereophile's simulated load confirms the importance of measuring a complex load; it does not validate a direct prediction for the MR-A-01, which does not yet physically exist.
CRITERIA AND A-D CONFIDENCE — §7 C7 · §8 C2 · §9 C4 · §10 C1/C5/C6 · §11 A-D CONFIDENCE
7. C7 · Interaction with the load (35%) — complex formula
C7 evaluates the amplifier's response variation on an RLC load representing the MR-A-01. The quantity that scores is the measured response, not an estimate obtained solely from a damping factor figure or Zout at 1 kHz.
| Maximum absolute variation Δmax 20 Hz–20 kHz | C7 |
|---|---|
| ≤1 dB | 1.00 |
| >1 to ≤2 dB | 0.70 |
| >2 to ≤3 dB | 0.40 |
| >3 dB | 0.20 |
Closed mathematical definition of C7 — normalized RLC vs 8 Ω comparison
For complex output impedance Zout(f) and complex load Z_speaker(f):
A(f) = 20·log10 |H(f)|
Δn(f) = Δ(f) - Δ(1kHz)
Δmax = max_{20Hz≤f≤20kHz} |Δn(f)|
- H_RLC(f): on the RLC load representing the built MR-A-01. H_8Ω(f): the same expression on an 8 Ω reference resistor.
- 1 kHz normalization: Anorm(f)=A(f)-A(1kHz) and Δn(f)=Δ(f)-Δ(1kHz).
- Δmax: maximum absolute deviation of the normalized curve.
- Silver Luna: ±3.4 dB 8 Ω tap Stereophile — do not automatically convert to 6.8 dB peak-to-peak for the new definition.
- Levels and phase: measure at 1 W and high level per §12. The worst relevant case scores. Magnitude and phase must be retained.
Status: C7 = ND for the six candidates. Measurement pending per §12.
8. C2 · Clean power (25 %)
C2 measures usable power, not catalogue power: per channel, with both channels active, at THD+N ≤ 1 %, into 4-ohm and 8-ohm loads, with frequency and duration stated. Both are published and the worst relevant load for the MR-A-01 scores (normally 4 ohms).
| Power at ≤1 % THD+N (P) | C2 score (linear interpolation per band) |
|---|---|
| <10 W | 0.00–0.20 (0.02·P) |
| 10–17.9 W | 0.20–0.50 |
| 17.9–35 W | 0.50–0.85 |
| 35–50 W | 0.85–1.00 |
| >50 W | 1.00 (capped) |
0.02·P if 0 ≤ P < 10
0.20 + 0.30·(P-10)/7.9 if 10 ≤ P < 17.9
0.50 + 0.35·(P-17.9)/17.1 if 17.9 ≤ P < 35
0.85 + 0.15·(P-35)/15 if 35 ≤ P < 50
1.00 if P ≥ 50
}
P in watts per channel, worst relevant load (normally 4 Ω), both channels active, THD+N ≤1% at 1 kHz and in a 20 Hz–20 kHz sweep.
The 17.9 W threshold is the power the scenario of §2.2 requires per channel for 95 dB crests at 3 m in stereo (hypothetical 89 dB/W/m). The scale saturates at 50 W because, with this loudspeaker and scenario, more nominal power adds no audible gain. If there are no closed test conditions, C2 is ND and produces no score.
Note on the test load: resistive 4 Ω is a conservative laboratory load, not an exact representation of the MR-A-01's 4.9 Ω minimum. The real interaction is evaluated with the RLC load in C7.
9. C4 · Low power (20 %)
This is where the real listening level lives. The scenario of §2.2 places average listening at fractions of a watt: 80 dB continuous is ≈0.57 W per channel and a 60.9 dB whisper is ≈7 mW. C4 requires THD+N at 7 mW, 100 mW, 570 mW and 1 W, together with residual noise and background hum.
10. C1, C5 and C6 · Taps, reliability and continuity
C1 · Taps — admissibility condition (not scored)
C1 does not enter the score: the manual's tap instruction is compatibility and safety information, not an electrical merit. The scale is kept only as an admissibility criterion to verify that the amplifier can be connected without risk to the ≈6-ohm MR-A-01. A model without verifiable taps or instruction is not thereby excluded, but requires confirmation with the manufacturer before testing.
| Situation | C1 score |
|---|---|
| Manual with a specific instruction for 6 ohms (e.g. 6 ohms → 8-ohm tap) | 1.00 |
| Manual accepting 4 or 8 ohms for 6 ohms | 0.70 |
| Taps available but no instruction | 0.40 |
| No verifiable information | ND |
| Physical absence of taps confirmed | 0.10 |
C5 · Supervision, protection and maintainability (10 %)
C5 does not measure statistical reliability or failure rates: it assesses, on paper, the ease and safety of supervising, protecting and maintaining the amplifier. It is useful because a tube amp is calibrated and ages with use, but it is not an MTBF.
| C5 score | Documentary condition | Example |
|---|---|---|
| 1.00 | Automatic bias with individual protection per tube, status indication | Titania MK2 (conditional) |
| 0.70 | Automatic bias without individual protection | Cayin Jazz 90 |
| 0.60 | Supervised automatic bias (Adaptive AutoBias) | PrimaLuna EVO 200 |
| 0.40 | Manual bias with accessible test points | — |
| 0.30 | Manual bias without documented facilities | Acronos RS-1, P52, Jadis Aria |
| 0.20 | Manual bias without any documented facility | — |
Correction: European service does not score in C5. It is an admissibility condition per §13 (distributor with stock of transformers, boards and matched tubes, verifiable repair lead time — not merely commercial distribution). This avoids double-counting with §13. Pure scale without EU service.
C6 · Continuity (10 %)
C6 · Continuity (10 %). No model has yet had a continuity trial per the protocol (10 W for 60 minutes on both channels, plus 80 % of the 1 % power for 10 minutes). That is why C6 stays outside the calculation for everyone, like C7, until the trial exists; no filler value is assigned.
11. A-D confidence of the evidence
A-D confidence answers a different question from the matrix: not "how does the amplifier perform?" but "how much can we trust what we know about it?". It is independent of the score: a model can be impeccably documented (confidence B) and turn out mediocre when measured, or poorly documented (confidence D) and sound very good. Confidence neither rewards nor punishes performance: it organizes uncertainty.
| Letter | Assignment criterion |
|---|---|
| A | Independent measurement of the model itself (MEASURED (external) or MEASURED) covering the decisive parameters: Zout, power at defined THD and response into a load. |
| B | Locatable official documentation with complete test conditions (load, THD, frequency) and no discrepancies between sources. |
| C | Official documentation with incomplete conditions, open discrepancies between official sources, or key self-declared data without external verification. |
| D | Majority ND in the essential parameters, or key data of undetermined provenance. |
12. MEASUREMENT PROTOCOL AND ATTENUATION CALCULATOR
12. Measurement protocol and attenuation calculator
The protocol applies to the six candidates of §3 with identical reference RLC load and level conditions. The tap is the one the manufacturer recommends for 6 Ω — see the correction in §3: first measure on the documented admissible tap and, only if the manufacturer allows it, on the alternative, with results separated by tap. It consists of six amplification tests plus the prior characterization of the loudspeaker, which is an independent phase.
| Test | Minimum condition | What it resolves |
|---|---|---|
| Characterization of the built MR-A-01 | Measure |Z(f)| and phase on the finished cabinet; sensitivity at 2.83 V/1 m with axis, window and environment declared; publish dB/2.83 V/1 m as the primary datum and the conversion to dB/W/m only with an explicit reference impedance. | Replaces the hypothetical 89 dB/W/m sensitivity and the simulated impedance; fixes the real load and calculation SPL. |
| Power vs THD+N | 4 ohms and 8 ohms; 1 kHz and 20 Hz–20 kHz sweep; both channels driven; report at 0.1 %, 1 % and 3 %. | Comparable clean power (C2) and real clipping. |
| Zout per tap | 4/8-ohm taps; at least 100 Hz, 1 kHz and 10 kHz; two known loads method. | Real damping factor under traceable conditions (C7). |
| Response into a load | 1 W and high level; first 8 ohms resistive and then an RLC load based on the MR-A-01. | The variation scored by C7 and stability into a reactive load. |
| Low-level THD/FFT | ≈7 mW, 100 mW, 0.57 W and 1 W; 1 kHz and sweep; harmonic spectrum. | Behavior at the real listening level (C4). |
| Residual noise | Terminated inputs, controls in documented positions, 20 Hz–20 kHz bandwidth and A-weighting. | Audible noise at the listening position and channel difference (C4). |
| Continuity and stability | 10 W for 60 min and 80 % of P@1 % for 10 min; clipping, oscillation, DC at the output, temperature and start-up under a safe protocol. | C6, and detection of faults that do not show up with a simple load. |
12.1 Attenuation calculator for two resistive loads
This calculator does not predict the loudspeaker's real response. It applies the resistive voltage-divider formula Attenuation (dB) = 20 · log₁₀ (Z_speaker / (Z_speaker + Z_out)) for two point resistive loads. The difference between the two attenuations is not equivalent to the frequency-response variation on a complex load: for that, the full |Z(f)| and phase curve of the loudspeaker together with the amplifier's Zout as a function of frequency are required. The Silver Luna example illustrates it: with Zout = 5.4 ohms, the difference between 4.9 ohms and 6 ohms is only ≈0.88 dB, while the laboratory measured ±3.4 dB on a full simulated load.
Resistive divider: attenuation for two loads
12.2 Minimum viable measurement and what to ask the manufacturer
If the full protocol is out of reach, there is an intermediate step between "everything ND" and the test bench. With a signal generator, a pair of power resistors and a decent-quality audio interface —indicative budget of €300–400, similar to the kit in the xDuoo report— one can measure: Zout per tap with the known-load-resistor method at 100 Hz, 1 kHz and 10 kHz; relative power at 1 % THD+N into a resistive load; and relative response into the RLC load against the resistive reference. These are home measurements: they are labeled as such and do not replace the protocol above.
If measuring is not possible, ask the manufacturer or distributor in writing for three things: continuous power at THD+N ≤ 1 % with load, frequency and both channels active stated; output impedance per tap at 1 kHz (and at 10 kHz if published); and stated behavior into a reactive load. The answer —or its absence— feeds the model's A-D confidence.
13. VALIDITY CONDITIONS AND VERDICT
13. Validity conditions and verdict
- Do not modify the weights after seeing the results.
- Do not replace missing measurements with estimates.
- Do not assign scores to ND criteria.
- Do not confuse unknown compatibility with incompatibility.
- Do not present documentary confidence as performance.
- Do not present the trial priority of §5 as a performance ranking.
Methodological verdict. The matrix is frozen at 35/25/20/10/10. C1 is an admissibility condition and A-D confidence reports separately on evidence quality. In the current phase no global score or adequacy ranking is published: C7, C2-M, C4-M and C6 require comparable measurements. Manufacturer specifications are kept as documentary information, but do not replace experimental results. The trial priority of §5 is an operational selection, not a performance classification. Once the protocol is complete, the matrix will be computed with the measured results, without modifying the weights or replacing missing data with estimates.
Admissibility gates and listening protocol. Two conditions are filters, not scores: the tap instruction (C1) and the availability of service or distribution in Europe — without it, the model moves to the waiting list. This report's evaluation is exclusively electrical: the listening protocol (phase 2) will be defined separately, with its own pre-registration, levels and signal set, and will never mix its results with the adequacy matrix.
ANNEX A — HOW TO READ THE EVIDENCE LABELS
Annex A — How to read the evidence labels
Every claim in the report carries an evidence label indicating its level of support. On these labels the A-D confidence of §11 is built: the letter summarizes, for each amplifier, how much the whole of its data can be trusted.
| Label | What it means | What it contributes to confidence |
|---|---|---|
| CONFIRMED | Locatable official documentation. Ideally with identifiable test conditions (load, THD, frequency). | Adds toward B; without complete conditions, stays at C. |
| OBSERVED | Visible in photographs, silkscreen or rating plate. | Not scored: it describes, it does not certify electrical behavior. |
| CALCULATED | Result of applying a formula. | Not scored: declared with its formula and hypotheses. |
| PROBABLE | Reasonable physical inference, unchecked. | At most C; never supports a B. |
| HYPOTHESIS | Possibility without sufficient support (e.g. the MR-A-01's 89 dB/W/m). | Not scored: declared as such. |
| MEASURED (external) | Measurement published by an identifiable independent laboratory, with instrumentation. It may be of another product of the same family; it is explicitly declared if not the exact model. | Adds toward A if it covers decisive parameters of the model itself; if it is of another product, it does not directly certify and is documented as contextual reference. |
| ND | Not determinable with current evidence. | Cumulatively, it leads to D. |
| REPORTED (external) | Datum from a reliable third party, without a closed citation. | Treated as pending until the citation is closed. |
| PENDING VERIFICATION | Datum attributed to a source that could not be closed documentary. | Supports no more than C. |
| MEASURED | Own measurement on the physical unit. | Supports an A if it covers the evaluated parameter. |
ANNEX B — TECHNICAL GLOSSARY AND TUBE TABLE
Annex B — Technical glossary and tubes
Here are, in plain language, the terms the matrix uses. A warning about tubes: their catalogue powers are typical dissipation powers (the heat they can withstand), not limits of the amplifier that uses them.
B1. Power tubes
| Tube | Type | Typical dissipation | Where it appears in this report |
|---|---|---|---|
| EL34 | Pentode | 25 W | PrimaLuna EVO 200 (factory configuration), CS-55A (EL34 mode) |
| KT88 / 6550 | Beam tetrode | 35–42 W | Titania MK2, CS-55A, EVO 200 (option) |
| KT150 | Beam tetrode | 70 W | Acronos RS-1 (single-ended class A) |
B2. Matrix terms
| Term | Operational definition | Criterion |
|---|---|---|
| Zout | Output impedance per tap, measured at least at 100 Hz, 1 kHz and 10 kHz. Not the same as the stated damping factor. | C7 (35 %) |
| DF = Znom / Zout | Damping factor. A stated DF without measured Zout certifies nothing: a Zout of 5.4 ohms means DF = 1.48 against 8 ohms (1.11 against the MR-A-01's nominal 6 ohms), which on the C7 scale is worth 0.20. Formula: DF = Znom / Zout. | C7 |
| P@1 % THD+N | Output power at 1 % distortion, both channels driven, at 1 kHz into a specified load. It is the only power datum comparable between models. | C2 (25 %) |
| THD+N at low power | Distortion plus noise at 7 mW, 100 mW, 570 mW and 1 W: the range where real listening happens with this loudspeaker. | C4 (20 %) |
| Bias | Biasing of the power tubes. Automatic with individual protection per tube is the top level of C5. | C5 (10 %) |
| RLC load | Simulated load with resistor, coil and capacitor reproducing the loudspeaker's impedance; used to measure response variation without risking the real loudspeaker. | C7 |
| A-D confidence | Letter summarizing each model's evidence quality, independent of the score. | §11 |
ANNEX C — DETAILED MEASUREMENT PROTOCOL
Annex C — Detailed measurement protocol
This annex develops the §12 protocol step by step. The rule common to all measurements: any Zout or THD datum must always state frequency, load, tap, power and thermal state; without those conditions the figure is not comparable.
1. Characterization of the built MR-A-01
- Measure |Z(f)| and phase on the finished cabinet.
- Measure sensitivity at 2.83 V/1 m with axis, window and environment declared.
- Publish dB/2.83 V/1 m as the primary datum; the conversion to dB/W/m only with an explicit reference impedance.
- Measure distortion and compression at 80, 90 and 95 dB SPL.
2. Power vs THD+N
- 8-ohm and 4-ohm loads.
- 1 kHz and 20 Hz–20 kHz sweep.
- Both channels driven.
- Report at 0.1 %, 1 % and 3 %.
3. Zout per tap
- 4-ohm and 8-ohm taps.
- 100 Hz, 1 kHz and 10 kHz.
- Two known loads method (Annex B and §12.1).
- Record the variation with frequency.
4. Response into a load
- 1 W and high level, first with 8 ohms resistive.
- Then with the RLC load based on the MR-A-01.
- Record the variation caused by Zout and the output transformer.
5. Low-level THD/FFT
- ≈7 mW, 100 mW, 570 mW and 1 W.
- 1 kHz, 100 Hz and 10 kHz.
- Record the harmonic spectrum.
6. Residual noise
- Terminated inputs and controls in the documented position.
- 20 Hz–20 kHz bandwidth, A-weighting.
7. Continuity and stability
- 10 W for 60 min on both channels; 80 % of P@1 % for 10 min.
- Clipping, oscillation, DC at the output, temperature and bias.
- Start-up under a safe protocol.
Safety notice
ANNEX D — SAFETY, LIMITS AND ELECTRICAL PROTOCOL
Annex D — Safety and limits
This annex collects what the matrix does not score but conditions any real test: safety and the physical limits of working with high voltage.
- The B+ holds charge for minutes or hours after power-off. Verify 0 V with a CAT III instrument.
- Incorrect bias causes overcurrent and can damage the output transformer.
- Never operate without a load or into a short circuit: the transformer generates overvoltages.
- Tubes exceed 150 °C and the chassis 60 °C: leave 20 cm of ventilation.
Limits of validity of the report
| Limit | Status | Consequence |
|---|---|---|
| MR-A-01 sensitivity | 89 dB/W/m: calculation hypothesis, not measured | All SPLs in §2.2 are hypotheses. Measure at 2.83 V/1 m. |
| MR-A-01 impedance | ≈6 ohms nominal, ≈4.9 ohms at 1.4 kHz (LspCAD simulation) | The conversion between voltage and power references is not traceable until measured. |
| Zout of the fifteen candidates | ND for all; only the Silver Luna has 5.4/2.8 ohms measured | C7 cannot be scored: that is why there is no ranking. |
| P@1 % THD+N | With defined load, documented only for the LM-216IA (32/38 W with discrepancy) | C2 is not comparable between models until measured. |
ANNEX E — SOURCES AND TRACEABILITY
Annex E — Sources and traceability
Primary sources
- Nauscopio · 35-year-old three-way loudspeaker: crossover A MR-A-01. Open source
- Nauscopio · xDuoo A35 and 14 amplifiers comparison report. Open source
- Fezz Audio · Titania Mk 2. Open official page
- Stereophile · Fezz Audio Silver Luna Prestige, Measurements. Open measurement
- Cayin · Jazz 90, Jazz 80, Jazz 70 and CS-55A. Open official site · Jazz 80 page
- PrimaLuna · EVO 200. Open official page
- Line Magnetic · LM-216IA (32 W, official website). Open official page
- Line Magnetic · LM-216IA (38 W, European manual). Open PDF manual
- Nauscopio · Acronos RS-1 single-ended KT150. Open report
- Galion Audio · TS34 V2, official specifications. Open specifications · Product page
- Black Ice Audio · Fusion F22 V.2, official page. Open official page · PDF manual
- Icon Audio · Stereo 40 MKIII, instruction manual. Open manual
- Willsenton · R8 Series, official website. Open official site
- Jadis · Aria, official page. Open official page
Complementary and contrast sources
- Tube datasheets. EL34, KT88, 6550, KT150, 12AX7, 12AU7, 6SL7, 6SN7, ECC83, ECC82.
- Nauscopio · index of technical reports (English). Open index
- Nauscopio · index of audio technical reports (English). Open index
- Darko.Audio · Cayin Jazz 70, presentation. Open article — REPORTED (external), launch context only; the technical data are taken from the official website.
- PMA Magazine · Galion TS34, review. Open review — PENDING VERIFICATION for data not published by the manufacturer.
- HomeTheaterHifi · Black Ice Fusion F-22, review. Open review — PENDING VERIFICATION: attributes 50–60 W without test conditions; not reclassified as a specification.
- Speaker Decision · Cayin Jazz 70, specifications sheet. Open sheet — secondary contrast of the 35 W UL / 17 W triode figure.
About the making of this report
This report was produced with cross-auditing by several AI tools (Meta, ChatGPT, DeepSeek, Claude, Mistral) and manual verification of primary sources. The final statements are the author's responsibility.