xDuoo A35. 13 amplifiers - 12 PP EL34 + Acronos RS-1 Single-Ended KT150 EN
xDuoo A35 and 13 push-pull tube amplifiers, plus Acronos RS-1 KT150 SE
14 amplifiers compared — 13 EL34/KT88 PP + Acronos RS-1 SE KT150
Primary sources: xDuoo official spec sheet and APOS Audio, xDuoo website, Audiophonics, Tharbamar video review, OCR analysis of internal photographs, specialised dealer spec sheets and cross-checked compatibility. Reference system: 3-way LspCAD 6.20 cabinets, ~7 Ω nominal / 5.3 Ω minimum (at 5 kHz, phase 0°) / 89 dB/2.83 V/m (calculation hypothesis, not measured), listening at 3 m.
xDuoo A35 — 45 W from EL34s in ultralinear push-pull (C1: manufacturer-declared power, test conditions missing), ESS ES9018K2M DAC, Bluetooth 5.1 aptX HD, 12AX7 + 2× 12AU7 tubes, custom EI transformers, point-to-point wiring. Speaker taps: 4 Ω and 8 Ω. Attributed Zout: 0.65 Ω / 1.20 Ω pending verifiable source or measurement. Nominal headroom: +7.99 dB. Editorial index: 86. Price: ≈ €1,339.
Context — 14 amplifiers compared (13 EL34/KT88 push-pull + Acronos RS-1 single-ended KT150) against your LspCAD cabinets: 89 dB/2.83 V/m, Zmin 5.3 Ω at 5 kHz with phase 0°, listening at 3 m, 80 m³ room. Required power depends on the model: free-field, one speaker, 95 dB peak = 35.8 W; 100 dB = 113 W. Stereo with coherent summing, 95 dB = 17.9 W; 100 dB = 56.5 W. Room with hypothetical +4 dB, 95 dB ≈ 7.15 W; 100 dB ≈ 22.6 W. With DR>14, microdynamics matter more than raw power. Effective sensitivity (89 dB/W/m vs 89 dB/2.83 V/m) remains a hypothesis, not a measurement.
Provisional conclusion — The three budget push-pulls are adequate for your system. The Jazz 80 is the cheapest and the only one with externally cited measurements (pending documentary verification), but it has the smallest peak margin. The Jazz 90 is the most complete and powerful on paper, but without independent verification. The A35 offers point-to-point wiring and 45 W nominal at a modest price, but its attributed Zout is unverified. None of the three push-pulls in this comparison (A35, Jazz 80, Jazz 90) publishes measured Zout or verifiable P@1%: without that data, the comparison against your 5.3 Ω speakers remains open. The Acronos RS-1 is the single-ended alternative: very clean declared first watt and very low Zout on paper, but with limited peak margin and a minimum recommended sensitivity of 92 dB. NFB confirmed by the manufacturer; depth ND.
GETTING STARTED — NO PRIOR KNOWLEDGE NEEDED (Basic guide G1–G5)
GETTING STARTED — NO PRIOR KNOWLEDGE NEEDED
This guide provides the keys to understanding the technical concepts in the report without being an engineer. The aim is for the reader to be able to interpret the data, tables and analyses in order to make a well-informed decision.
Basic guide: reading this report without being a tube expert
Central idea. This guide provides the keys to understanding the technical concepts in the report. You don't need to be an engineer. The aim is for the reader to be able to interpret the data, tables and analyses in order to make a well-informed decision.
- G1 · how a push-pull amplifier works.
- G2 · which data is reliable and which is not.
- G3 · how many watts you need (decibels and distance).
- G4 · impedance, damping factor and power supply.
- G5 · conclusions and worked exercises.
Minimum glossary
- Class A: the tube conducts current during the full signal cycle. Very linear, but inefficient.
- Class AB: the tube conducts more than half a cycle but less than a full one. A balance between efficiency and linearity.
- AB1 vs AB2: two sub-types. In AB1 the grid never goes positive; in AB2 it does at the peaks. This is an important technical detail for the driver design, but does not change the sound by itself.
- Ultralinear (UL): the output tube's screen is connected to a centre tap on the output transformer (typically 20–50% of the primary). Halfway between triode and pentode.
- Triode: the screen is tied to the anode. Less power, less distortion, lower Zout.
- NFB (negative feedback): part of the output signal is fed back to the inverted input. Reduces distortion and Zout, at the cost of gain.
- THD+N: total harmonic distortion plus noise, expressed as a percentage of the signal.
- Output impedance (Zout): the equivalent internal resistance in series with the speaker. Not to be confused with the speaker taps (4/8 Ω).
- Sensitivity: sound level (SPL) produced by a speaker with 1 W at 1 m (dB/W/m) or with 2.83 V at 1 m (dB/2.83 V/m). Not interchangeable if the nominal impedance is not 8 Ω.
- Crest factor: difference between peak and RMS level of the music. Typically 12–18 dB.
- DF (damping factor): DF = Z_load / Z_out. Reference only: both load and Zout change with frequency.
An amplifier takes the weak signal from the source and turns it into a signal with the voltage and current needed to drive the speaker. In the amplifiers in this comparison, that task is carried out by tubes arranged in three blocks: preamp, driver and power stage. All of it is supplied by a high DC voltage called B+.
Source → Preamp → Driver → Power stage → Output transformer → Speaker
(≈ 1 V) (voltage) (voltage) (current) (impedance matching) (4–8 Ω)
↑
Power supply (B+)Most of the equipment compared here (Cayin, xDuoo, PrimaLuna, etc.) is push-pull. In this configuration the power tubes work in pairs: while one "pushes" (conducts more current), the other "pulls" (conducts less). Their currents merge in the output transformer. This gives two advantages:
- DC current cancels in the transformer core (prevents saturation).
- Many even harmonics cancel, provided the two halves are well balanced.
Cancellation is not perfect: it depends on the degree of symmetry between the halves, tube matching and operating conditions (level, frequency, bias).
The exception: Acronos RS-1. This amplifier is single-ended (SE). Instead of a pair of tubes it uses just one (one KT150 per channel). This simplifies the circuit and eliminates crossover distortion, but is less efficient and usually has a higher output impedance. It is another design philosophy, analysed in detail in §18.
AB1 and AB2. Within class AB (the most common in push-pull) there are two sub-types. In AB1 the tube's grid never goes positive: the driver only has to supply voltage, not current. In AB2 the grid does go positive at the peaks and demands real current, so the driver has to be much more powerful. Public documentation does not allow confirming which exact regime each of the 14 amplifiers operates in. AB1 is assumed because it is usual in domestic equipment, but it is PROBABLE, not CONFIRMED.
This report is not a measurement but a compilation and analysis of information. It combines manufacturer data, third-party data, visual observation and derived calculations. That is why every important claim carries an evidence label. The full table is in §2.
To know whether an amplifier "has enough" power, start from the speaker and the listening distance. The decibel (dB) is a logarithmic scale: +10 dB multiply power by 10 and +3 dB multiply it by 2.
Starting point. The reference speaker is modelled at 89 dB/2.83 V/m. If the reference impedance is 5.3 Ω, that value is equivalent to ≈87.2 dB/W/m, not 89 dB/W/m. This is a HYPOTHESIS for calculation, not a measurement. The calculations below use 89 dB/W/m as the base scenario, and indicate when the conclusion changes with the correction.
At 3 m, the free-field distance loss is 20·log10(3) = 9.54 dB. Therefore: 89 − 9.54 = 79.46 dB with 1 W at 3 m (one speaker, free field).
Three scenarios, not one:
- Free field, one speaker. Theoretical model without a room. P = 10^((L − 79.46)/10).
- Stereo, coherent summing. The two speakers radiate the same signal; total power is split between channels. −3 dB per channel is applied versus the mono case (half power per channel).
- Room estimate. A hypothetical room gain of +4 dB is added for reflections. Not a measurement: a planning scenario.
| Scenario @3 m | Continuous SPL | Peak SPL | W/ch (one speaker, free field) | W/ch (stereo, coherent) | W/ch (room, +4 dB) | Current @5.3 Ω (room model) |
|---|---|---|---|---|---|---|
| Domestic | 65 dB | 80 dB | 1.13 W | 0.57 W | 0.23 W | 0.21 A RMS |
| Movie / active listening | 80 dB | 95 dB | 35.8 W | 17.9 W | 7.15 W | 1.16 A RMS |
| Monitoring | 85 dB | 100 dB | 113 W | 56.5 W | 22.6 W | 2.07 A RMS |
Supply. Tubes need a high DC voltage (B+). In Spain the mains is 50 Hz, so ripple appears at 100 Hz. Capacitors charge on the peaks and discharge between them. Since Q = C·ΔV and the time between peaks is 1/f, we have ΔV ≈ I / (f · C). This calculation underpins the bias adjustment analysed in §6.
Speaker impedance. A "6 Ω" speaker does not always measure 6 Ω: its impedance changes with frequency. The reference speaker drops to a minimum of 5.3 Ω at 5 kHz with phase ≈0° (according to the LspCAD model in §13), and rises to about 11 Ω around 800 Hz. An amplifier that behaves as a voltage source delivers more current when impedance drops.
This is where output impedance (Zout) comes in, acting as a resistor in series with the speaker. The damping factor is DF = Z_load / Z_out. It is a reference: both load and Zout change with frequency. That is why a "catalogue" DF (typically measured at 1 kHz into a pure resistive load) does not describe behaviour at 20 Hz or into a reactive load. A DF of 6 at 1 kHz can be very different at 30 Hz. Which is why measuring real Zout is priority number one.
Thresholds used in this report (at 1 kHz into a resistive load):
- Design target: Zout = 0.5 Ω.
- Minimum acceptable: Zout = 0.9 Ω (variation ≈ 1.0 dB between 5.3 Ω and 11 Ω).
- 1 dB loss over 5.3 Ω: Zout ≈ 0.65 Ω.
- There are no independent measurements for the A35 or the Jazz 90. The Jazz 80 has powers attributed to Stereo Magazin but pending documentary verification.
- The amplifiers declare different nominal powers (45 W, 50 W UL, 26 W @3%) but none publishes verified Zout or P@1% THD.
- Neither a photo nor a table decides whether an amplifier "is better". Measurements are needed: THD+N vs power, Zout, response into the real load, noise.
Exercises to practise.
- A speaker has 92 dB/W/m. What level does it produce with 1 W at 2 m?
- With the 89 dB/W/m speaker, what power is needed for 100 dB peak at 3 m with one speaker in free field?
- A filter has 1830 µF and the amplifier draws 200 mA. What is the ripple at 100 Hz?
- How many dB apart are 26 W and 45 W?
- If the 89 dB are really dB/2.83 V/m over 5.3 Ω, what is the equivalent sensitivity in dB/W/m?
See solutions
- 92 − 20·log10(2) = 92 − 6.02 ≈ 86 dB.
- 10^((100 − 79.46)/10) ≈ 113 W (free field, one speaker).
- 0.2 / (100 · 0.00183) ≈ 1.09 V.
- 10·log10(45/26) ≈ 2.38 dB.
- 89 − 10·log10(5.3/8) = 89 − 10·(−0.179) ≈ 89 − 1.79 = 87.2 dB/W/m.
PART I — COMPARATIVE REPORT: XDUOO A35 + 13 PUSH-PULLS + ACRONOS RS-1
Comparative technical analysis of 14 tube amplifiers aimed at a specific system: 89 dB/2.83 V/m speakers, Zmin 5.3 Ω at 5 kHz with phase 0°, listening at 3 m, 80 m³ room. All information is classified using the label system in Appendix A.
0. Executive summary
20-second read
A35: 45 W EL34 PP UL (C1), 12AX7 + 2× 12AU7, ES9018K2M DAC, BT 5.1, 4/8 Ω taps, attributed Zout 0.65/1.20 Ω (source pending). ≈ €1,339
Jazz 90: 50 W UL / 28 W TR KT88 (D), 12AX7 + 6SN7, ESS DAC + LDAC + MM phono + triode, 27.5 kg. VIP25: €2,093
Jazz 80: 40 W KT88 UL declared; powers at 3% attributed to Stereo Magazin (pending documentary verification). 1 H/180 mA B+ choke. VIP25: €1,323
Black Ice F22: 50 W EL34 (USA ex-works ≈ €1,605); landed in Tarragona ≈ €4,400 including shipping + customs + 21% VAT on the version actually imported. No longer a bargain.
Conclusion. The three budget push-pulls (Jazz 80, A35, Jazz 90) are ample for 80/95 dB in any room model and tight for 100 dB peak (22.6 W in room model, 113 W free-field). Zout = ND or unverified on all three: this is the critical measurement for this system.
Summary
The xDuoo A35 is an integrated tube amplifier. Its power stage uses 4× EL34 in ultralinear push-pull. The preamp combines 12AX7 + 2× 12AU7. Transformers are custom EI, and the digital block includes ESS ES9018K2M DAC and Bluetooth 5.1. It is compared here against 13 other amplifiers, including the two Cayin Jazz (80 and 90), which share the push-pull tube philosophy.
Provisional conclusion
The A35's architecture is serious and coherent. The point-to-point wiring and custom EI transformers are positive observations, but do not by themselves demonstrate electrical or sonic superiority. Commercial specifications leave unanswered real distortion, effective power, Zout, high-power response, noise and behaviour of the digital block.
See master table and suitability criterion (S12) · See modification budget (S17)
1. Introduction and scope
Reference images — visual basis
The photographs of the Tharbamar unit are used as visual reference. They are the basis of the observed evidence for this analysis. Video: xDuoo A35 EL34 Tube Amp Review & Comparisons
Objective and criterion of this report
This document analyses the xDuoo A35: an integrated tube amplifier with an EL34 push-pull power stage, ultralinear operation and integrated ESS DAC and Bluetooth blocks.
The objective is not to repeat the marketing but to cleanly separate confirmed data from visual observations, measurements, calculations and inferences. A total of 14 amplifiers are compared: 13 EL34/KT88 push-pulls (including the two Cayin Jazz) plus the Acronos RS-1 single-ended KT150.
2. Evidence and specifications
Compact summary of the evidence state
Every row can be traced to its section. This table avoids having to walk the whole report to know the real state of each datum.
| Parameter | Value | Evidence |
|---|---|---|
| Topology | EL34 push-pull ultralinear | CONFIRMED S3 |
| Declared power | 45 W/ch @ 8 Ω (nominal) | CONFIRMED S2 (manufacturer) — test conditions to be specified |
| Power tubes | 4× EL34 | CONFIRMED S2/S3 |
| Driver / preamp | 1× 12AX7 + 2× 12AU7 | CONFIRMED S2/S3 |
| Transformers | Custom EI | CONFIRMED + OBSERVED S3 |
| Construction | Point-to-point in the main analog section | OBSERVED S1/S3 |
| Speaker taps | 4 Ω and 8 Ω (load matching via output transformer) | CONFIRMED Audiophonics / xDuoo S2 |
| A35 Zout | 0.65 Ω / 1.20 Ω attributed to the amplifier (not the speaker taps) | PENDING TRACEABLE SOURCE OR MEASUREMENT S4 |
| Jazz 90 / Jazz 80 Zout | Not published | ND S4/S12 |
| A35 THD | ≤1% declared (does not equal measured 45 W at 1%) | SPEC — no independent verification S4 |
| A35 frequency response | 20 Hz–60 kHz ±3 dB | CONFIRMED — ±1 dB DISCREPANT S2/S4 |
| A35 SNR | ≥93 dB A-weighted | SPEC S4 |
| DAC | ESS ES9018K2M (A35, Jazz 90, Jazz 80) | OBSERVED / CONFIRMED S5 |
| Bluetooth | BT 5.1 (A35: aptX HD; Jazz 90/80: LDAC) | CONFIRMED S5 |
| THD+N vs power (all three) | Not published at 1% | MEASUREMENT PENDING S4/S9 |
| FFT @ 1 W / 7 mW (all three) | Not published | MEASUREMENT PENDING S4/S9 |
| Residual noise (all three) | Not published | MEASUREMENT PENDING S5/S9 |
| DAC/BT noise vs. pure RCA | Not measured | HYPOTHESIS S5/S9 |
| B+ at idle / under load (all three) | Not measured | MEASUREMENT PENDING S6/S9 |
| A35 ultralinear tap % | ≈43% estimated | PROBABLE S9/S11 |
| Jazz 80 B+ choke | 1 H / 180 mA confirmed | MEASURED-EXT / CONFIRMED S4 |
| Jazz 80 powers at 3% | 26 W @8 Ω · 34 W @4 Ω · 46 W @3 Ω (attributed to Stereo Magazin) | PENDING DOCUMENTARY VERIFICATION S4 |
3. Architecture and analog stage
General architecture
EL34 power stage
CONFIRMED 4× EL34 push-pull 45 W UL (nominal). NOT DEMONSTRATED: exact UL tap %, plate/screen voltage, idle current, dissipation, exact class (AB1/AB2), real power at 1% / 0.1% THD.
12AX7 and 12AU7
If visual inspection confirms a long-tail pair on a 12AU7, the rigorous wording would be: "The phase splitter appears to be a long-tail pair based on a 12AU7" OBSERVED. CMRR or symmetry should not be quantified without measurement.
Output transformers
Custom EI CONFIRMED per manufacturer. An EI core does not on its own imply high or low quality. Leakage inductance, primary inductance, parasitic capacitances and response into the real load pending measurement; no figure should be assigned without measurement.
The A35 uses 1× 12AX7 for both channels (shared gain stage). The Jazz 80 and Jazz 90 use 2× 12AX7, one per channel. This architectural difference may be relevant for opera, orchestra or any recording with a critical stereo image.
| Aspect | 1× shared 12AX7 (A35) | 2× separate 12AX7s (Jazz 80/90) |
|---|---|---|
| Channel separation | Limited by capacitive coupling inside the tube (between its two sections) | Greater: no intra-tube channel coupling |
| Crosstalk | Worse at high frequency. Typically 40–55 dB at 1 kHz and degrades above 5–10 kHz | Better. Typically 60–75 dB, or the limit of the circuit itself |
| Stereo image | Compromised: image may narrow or shift | Superior: better instrument localisation and depth |
| Cost | Lower (one fewer tube) | Higher (one more tube and its circuit) |
| Noise | Potentially lower (one fewer source) | Potentially higher (two sources, though separate) |
| Matching | N/A (same tube for both channels) | The two 12AX7s must be matched for identical gain |
3.1 Architecture comparison — A35 vs Jazz 90 vs Jazz 80
| Aspect | xDuoo A35 | Cayin Jazz 90 | Cayin Jazz 80 |
|---|---|---|---|
| Power tubes | 4× EL34 DECLARED | 4× KT88 (EL34 compatible) DECLARED | 4× KT88 (EL34 compatible) DECLARED |
| Input stage | 1× 12AX7 (shared) DECLARED | 2× 12AX7 (one per channel) DECLARED | 2× 12AX7 (one per channel) DECLARED |
| Driver / splitter | 2× 12AU7 (probable LTP) OBSERVED | 2× 6SN7 DECLARED | 2× 12AU7 DECLARED |
| Total tubes | 7 | 8 | 8 |
| Output mode | Fixed UL DECLARED | Switchable UL/triode DECLARED | Fixed UL DECLARED |
| Power transformer | Custom EI DECLARED | Cayin custom toroidal DECLARED | EI OBSERVED |
| Output transformers | Custom EI OBSERVED | New design, larger than nominal DECLARED | Cayin's own EI OBSERVED |
| Construction | Point-to-point (signal) OBSERVED | Integrated on PCB DECLARED | Integrated on PCB OBSERVED |
| Weight | 20 kg DECLARED | 27.5 kg / 25.9 kg DECLARED · MEASURED-EXT | 16.8–17 kg DECLARED |
What CANNOT be concluded from S3
- That the A35 has worse stereo image than the Jazz 80/90 because of using a single 12AX7: it depends on layout and actual crosstalk, which have not been measured.
- That the A35's point-to-point wiring produces lower distortion or lower noise than the Jazz PCBs.
- That the A35's custom EI transformers are electrically superior to Cayin's toroidals without measuring leakage inductance and load response.
- That the declared 45 W correspond to the exact ultralinear tap % (43% is an assumption).
4. Power: Pnom, P@1% and evidence
Why not all power figures are equivalent
The preferred reference for comparing powers is P@1% THD/THD+N, but only when it is explicitly measured or declared under sufficiently identifiable conditions. A nominal power accompanied by "THD = 1%" without complete conditions does not become a measurement: it stays at C1.
| Evidence | Definition | Example | Use |
|---|---|---|---|
| A — MEASURED | Independent measurement of a specific unit; P@1% determined. | Fezz Silver Luna: ≈26 W, graphical read from Stereophile. | Measured/graphically estimated P@1% |
| A2 — MEASURED-EXT at 3% | Published external lab measurement at 3% THD. | Cayin Jazz 80: 26 W @3%/8 Ω · 34 W @3%/4 Ω (attributed to Stereo Magazin; locatable citation pending). | Measured P@3%; not P@1%. |
| B — DECLARED @1% | Manufacturer specifies power and 1% THD with identifiable conditions. | PrimaLuna 44 W; CS-55A 38/40 W; LM-216IA 38 W; R8 30 W FTC. | Declared P@1% |
| C1 | Power + THD/THD+N = 1%, but relevant conditions missing. | xDuoo A35: 45 W, THD+N = 1%. Jazz 90: 50 W UL, THD 0.2% @1 W. | Declared working Pnom; not measured P@1%. |
| C2 | THD specified at another power point or condition; does not determine P@1%. | Icon Audio: 0.15% at full power; Galion: ≤0.8% at 20 W. | Working Pnom; do not extrapolate. |
| D — NOT DETERMINED | No sufficient P@1% nor independent measurement. | Jadis Aria; Cayin Jazz 90 (for P@1%); Cayin Jazz 70. | Nominal Pnom, with explicit uncertainty. |
| ANALOGY | Inference based on another model; not an evidence category. | Jadis Aria ≈20 W as a conservative scenario by analogy with Orchestra Reference Mk.II. | Separate scenario, never primary P@1%. |
Jadis Orchestra Reference Mk.II: measurement and limits of the inference
The relevant data comes from the Orchestra Reference Mk.II, not the Aria. After the documented bias adjustment, Stereophile places clipping at its usual definition (1% THD) at 26 W/ch, vs 40 W nominal, with an output impedance of about 3 Ω. The 26/40 = 0.65 ratio is correctly calculated.
Jadis Aria: two scenarios, not a false P@1%
Master power table — Pnom / P@1% / P of calculation
| Amplifier | Pnom | P@1% / @3% | Evidence | P of calculation | Nominal headroom |
|---|---|---|---|---|---|
| Black Ice F22 | 50 W EL34 | <1% declared at 20 V | B | 50 W* | 8.45 dB |
| Cayin Jazz 90 | 50 W UL / 28 W TR | Not determined | D | 50 / 28 W* | 8.45 / 5.93 dB* |
| xDuoo A35 | 45 W | THD+N = 1% | C1 | 45 W* | 7.99 dB |
| PrimaLuna EVO 200 | 44 W | 44 W @1% | B | 44 W | 7.89 dB |
| Cayin CS-55A KT88 | 40 W UL | 40 W @1% / 1 kHz | B | 40 W | 7.48 dB |
| Icon Audio Stereo 40 III | 40 W | 0.15% at full power | C2 | 40 W* | 7.48 dB |
| Galion TS34 AB | 40 W AB | ≤0.8% at 20 W | C2 | 40 W* | 7.48 dB |
| Cayin Jazz 80 (KT88 UL) | 40 W KT88 / 36 W EL34 UL | 26 W @3%/8 Ω · 34 W @3%/4 Ω (attributed to Stereo Magazin; locatable citation pending) | PENDING DOC. VERIF. | 26 / 34 W | 5.61 / 6.77 dB |
| Cayin CS-55A EL34 | 38 W UL | 38 W @1% / 1 kHz | B | 38 W | 7.26 dB |
| Line Magnetic LM-216IA | 38 W UL | 38 W @1% / 1 kHz | B | 38 W | 7.26 dB |
| Willsenton R8 UL | 45 W RMS | 30 W @1% FTC | B | 30 W | 6.23 dB |
| Jadis Aria EL34 | 30 W | Not determined | D | 30 W* | 6.23 dB |
| Fezz Silver Luna | 35 W | ≈26 W @1% THD+N | A† | ≈26 W† | 5.61 dB† |
| Acronos RS-1 (SE) | 25 W | 25 W @0.5% | B | 25 W | 5.44 dB |
| Cayin Jazz 70 | 35 W UL | Not determined | D | 35 W* | 6.90 dB* |
What CANNOT be concluded from S4
- That the A35's 45 W are achievable with both channels driven: the figure does not specify it.
- That the 45 W are achievable at 1% THD just before clipping, rather than being a nominal figure with margin.
- That the 45 W are maintained across 20 Hz–20 kHz or only at 1 kHz.
- That the headroom difference between A35 (7.99 dB) and Jazz 80 (5.61 dB) is audible: 2.38 dB is within the model's uncertainty.
5. Digital section and noise
Digital section
ES9018K2M + BT + MCU + selector + remote. It is not a purely analog amplifier. "Pure Analog Mode" should only be used if only RCA inputs are used and the digital block has been shown to be disconnected.
Bluetooth
The A35 includes Bluetooth 5.1 with SBC, AAC, aptX and aptX HD CONFIRMED. The Jazz 90 and Jazz 80 add LDAC (QCC5125) on the same ES9018K2M DAC. Versus amplifiers with UAT, the practical difference depends mostly on the transmitter, the operating system and the negotiated bitrate. It should not be automatically translated into an audible difference.
Can the digital circuit inject noise into the analog section?
Technically possible via shared supply, ground return, capacitive/inductive coupling or electromagnetic radiation. Claiming that the 45.1584 MHz clock is contaminating the signal without measuring it is NOT DEMONSTRATED.
FB1 modification: decoupling analysis
- Identify with a continuity meter which node it supplies.
- Check whether the input relays require digital block power before desoldering FB1.
- Do not make a permanent modification at first. Reversible test with FB1 lifted and jumpered with a clip.
- Measure FFT at 7 mW and 1 W with shorted RCA input (Pure Analog). Save trace.
- Repeat with DAC active, not playing. Save trace.
- Repeat with BT paired but not playing. Save trace.
- Repeat with BT playing digital silence. Save trace.
- Acceptance criterion: if the noise floor difference at 7 mW is < 3 dB across the band, the physical FB1 modification has no justification. If it is > 3 dB, study a reversible modification.
What CANNOT be concluded from S5
- That the A35's digital block introduces audible noise into the analog section: it has not been measured.
- That the ES9018K2M's 45.1584 MHz clock contaminates the audio signal: it is a hypothesis, not a fact.
- That lifting FB1 improves the sound: the "Pure Analog" hypothesis has no measurement backing it yet.
- That the Bluetooth difference (aptX HD vs LDAC) is audible: depends on the transmitter and the negotiated bitrate.
6. Bias and safety
EL34 and KT77 bias adjustment — electrical analysis
The circuit includes 10 Ω shunt resistors on the EL34 cathodes. By Ohm's law: I_K = V_K / 10 Ω CALCULATED.
| Tube | V_K on 10 Ω | I_K | I_P estimate (92-95% I_K) | P_P at B+ 420 V | Note |
|---|---|---|---|---|---|
| Stock EL34 | 350-370 mV | 35-37 mA | 32.2-35.2 mA | 13.5-14.7 W | Conservative PROBABLE |
| NOS Mullard/Siemens EL34 | 340-360 mV | 34-36 mA | 31.3-34.2 mA | 13.1-14.4 W | Cold for NOS PROBABLE |
| Gold Lion KT77 | See B+ calc | See calc | See calc | <17.5 W (70% 25 W) | Measure B+ first CALCULATED |
Bias on Jazz 90 and Jazz 80 — architecture contrast
| Aspect | xDuoo A35 | Cayin Jazz 90 | Cayin Jazz 80 |
|---|---|---|---|
| Adjustment | Trimmers + 10 Ω shunts OBSERVED | 4 potentiometers, one per tube DECLARED | Individual per tube DECLARED |
| Bias VU meters | No | Dual function level + bias OBSERVED | Switchable as bias monitor OBSERVED |
| Active protection | MCU + 5 relays OBSERVED | BMI with MCU + comparator DECLARED | 2× Hongfa HF115F OBSERVED |
Safety: high voltage — important warning
- The A35 (and the Cayin Jazz) contain 400-450 V potentials that can be lethal even when disconnected from the mains. Electrolytic capacitors can remain charged.
- Disconnect the power cord, wait the time indicated by the service procedure, measure voltage before touching, use instruments and probes rated for high voltage.
- Never discharge a high-voltage capacitor by shorting it with a screwdriver.
- Correct method: discharge resistor rated for maximum voltage, stored energy, instantaneous power and isolation voltage.
7. Modifications and tubes
7.1 Methodological principle
It is not "the more components, the better". Phase 1: documentation. Phase 2: measurement (noise, THD+N, response, power, bias, temperature, Zout). Phase 3: change one variable at a time. Phase 4: measure again.
7.2 Preamp tubes (12AX7 + 12AU7)
| Position | Options | Evidence | Note |
|---|---|---|---|
| Central 12AX7 | Stock / Tung-Sol / NOS Mullard Blackburn / Telefunken Ulm / PSvane 12AX7-S Art TII | Stock CONFIRMED; rest PROBABLE | Prioritise low noise / low microphony. |
| 2× 12AU7 driver | Stock / JJ ECC82 / Gold Lion / NOS Amperex / NOS Mullard CV4003 / NOS Brimar CV4003 APOS | PROBABLE; Brimar CONFIRMED APOS spec | Matched pair within 2-3% mu/gm improves long-tail symmetry. |
7.3 Power tubes: EL34 / KT77 / 6CA7 / KT88
| Tube | Compatibility | Anode dissipation | Note |
|---|---|---|---|
| EL34 (stock) | CONFIRMED | 25 W | Base reference of the A35 design. |
| 6CA7 | PROBABLE | 25 W | Functional equivalent to the EL34. |
| KT77 | PROBABLE | 32 W | Drop-in documented by Genalex. Requires bias re-check after measuring real B+. |
| PSvane EL34PH | PROBABLE | 25 W | Mullard replica. |
| PSvane Summit EL34 | PROBABLE | 25 W | PSvane flagship line. |
| KT88 | NOT DEMONSTRATED (in A35) | 35 W | Compatible octal socket, but the A35 design is not documented for KT88. In Jazz 90 and Jazz 80 it is the stock tube. |
7.4 Coupling capacitors
- Stock white (xDuoo) — MKP 630 V OBSERVED. Medium ESR. Correct for signal, not electrolytic.
- Mundorf EVO Oil — MKP oil / 450 V. Low ESR, low DF. Measurable improvement in DF, audible NOT DEMONSTRATED without measurement.
- Jantzen Superior Z-Cap — MKP / 800 V. Very low ESR.
- ClarityCap CSA / MR — MKP / 630 V. Low dielectric absorption.
7.5 Supply and bias
Extreme caution. Check Vmax, capacitance, ripple current, ESR, inrush current and transformer capacity.
7.6 Digital modification / FB1
The Pure Analog hypothesis is interesting as a controlled experiment. Procedure: compare FFT with digital inactive, DAC active, BT active, BT connected, digital playback. Only if the difference is significant, study a reversible physical modification.
7.7 What not to modify without schematic or measurement
8. Recommended tubes
Comparison: Genalex KT77 vs PSvane EL34PH vs PSvane Summit EL34
| Characteristic | Genalex KT77 | PSvane EL34PH | PSvane Summit EL34 |
|---|---|---|---|
| Matched quartet price | ~€200–250 | €399 | ~€403 |
| Type | Beam tetrode | Mullard-replica pentode | Flagship pentode |
| Max anode dissipation | 32 W | 25 W | 25 W |
| Warranty | Standard | Standard | 24 months |
Final tube recommendation for the A35
9. Measurement protocol
The 8 priority measurements
The first five constitute the base protocol. The last three are what distinguishes this report from a conventional review.
| # | Measurement | What it resolves | Minimum instrument | Priority |
|---|---|---|---|---|
| 1 | Zout on 4 Ω and 8 Ω taps | What is the real Zout of the A35, Jazz 90 and Jazz 80? | True-RMS multimeter + 4 Ω and 8 Ω resistive load ≥100 W | CRITICAL |
| 2 | THD+N vs power @4/8 Ω | At what THD is the declared power delivered? | Sound card + calibrated attenuator + REW/ARTA | CRITICAL |
| 3 | FFT at ~7 mW and ~1 W | Harmonics at the real night-listening level. | Same as (2) | CRITICAL |
| 4 | Response 20 Hz–20 kHz @1 W and near nominal power | ±1 dB or ±3 dB? Maintained at high power? | Same as (2) | CRITICAL |
| 5 | Residual noise: pure RCA vs active DAC/BT | Does the digital block add noise at 7 mW? | Same as (2), shorted input, output loaded | CRITICAL |
| 6 | Noise at 7.1 mW (65 dB night) | This is the level at which 80% of listening happens. It shows whether the DAC/BT contaminates the first watt. | Same as (2), into real load | HIGH |
| 7 | Zout into the real LspCAD load (not pure resistive) | The catalogue DF (1 kHz, resistive) does not describe behaviour at 20 Hz or into a reactive load. With Zmin 5.3 Ω @5 kHz and peak 11 Ω @800 Hz, this measurement characterises real interaction. | Simulated reactive load (RLC network) + oscilloscope | HIGH |
| 8 | B+ at idle and under load | Determines whether the bias calculation is speculative. Without real B+, the S14 table is hypothetical. | HV multimeter + 1000:1 attenuator probe | HIGH |
What information is missing
Complete schematic, high-resolution photos of both PCB sides, transformer identification, bias measurement on the 4× EL34/KT88, B+, ripple, output impedance, THD+N vs power, output FFT and unambiguous identification of FB1 and rails.
10. Evaluation and traceability
Provisional technical evaluation
Very interesting architecture: 4× EL34 push-pull ultralinear, 12AX7 + 2× 12AU7, custom EI transformers, point-to-point. NOT DEMONSTRATED yet: exceptionally low noise, high DF, especially low distortion, exceptionally linear transformers, flat response at full power, absence of digital interference, optimal bias.
Conclusion
The xDuoo A35 presents a serious and coherent architecture. The point-to-point wiring and custom EI transformers are interesting elements, but do not by themselves prove sonic superiority. The Jazz 80 has externally cited measurements (pending documentary verification). The Jazz 90 has no published independent measurements.
What CANNOT be concluded from this report
- That the A35 sounds better than the other 13 amplifiers compared.
- That the custom EI transformers are electrically superior to those of another manufacturer.
- That point-to-point construction produces lower distortion, lower noise or better sound than a well-designed PCB.
- That replacing capacitors with Mundorf, Jantzen or ClarityCap produces an audible improvement.
- That a specific tube (PSvane, Brimar, Mullard NOS, KT77) is objectively better than another.
- That the A35's DAC/Bluetooth block (or the Jazz 90/80's) introduces noise into the analog section.
- That the current factory bias is optimal without characterising the specific unit.
- That the declared 45 W are achievable at a defined THD with both channels driven.
- That the Zout attributed to the A35 (0.65 Ω on 4 Ω) is necessarily better than that of another amplifier without measurement.
- That the ultralinear tap is exactly 43%: it is a comparison with typical EL34 topologies.
- That the A35 is suitable for any speaker: the evaluation depends on the LspCAD model.
- That the index scores reflect objective sound quality: they are an editorial index.
- That the Jazz 80's 3% THD powers are verified measurements until the original citation is located.
- That the Acronos RS-1 has a global feedback loop or a specific topology: NFB confirmed as existence, depth ND.
Traceability of claims
| Area | Status |
|---|---|
| 45W/4xEL34/push-pull/UL/12AX7+12AU7/DAC/BT/100k/300mV/63dB/232W/60s/380x340x200/20kg | CONFIRMED Apos Audio — primary source |
| Point-to-point main board / custom EI | CONFIRMED per doc + OBSERVED |
| MKP 630 V / blue trimmers / green 10 Ω shunts / ceramic sockets | OBSERVED |
| 20 Hz-60 kHz / THD ≤1% / SNR ≥93 dB | CONFIRMED ±3 dB — ±1 dB DISCREPANT / NOT DEMONSTRATED |
| Speaker taps 4 Ω / 8 Ω | CONFIRMED Audiophonics / xDuoo |
| A35 Zout 0.65/1.20 Ω | PENDING TRACEABLE SOURCE OR MEASUREMENT |
| THD@1W/@45W/PowerBW/43% UL/Lleakage/DAC-BT noise | NOT DEMONSTRATED / pending measurement — UL 43% PROBABLE |
| Cayin Jazz 80 (26/34/46 W @3%; 1 H/180 mA choke; consumption 148/231 W; weight 16.8-17 kg) | Choke, consumption and weight: CONFIRMED manufacturer. Powers at 3%: PENDING DOCUMENTARY VERIFICATION |
| Cayin Jazz 90 (50 W UL / 28 W TR; SS MM phono; BT 5.1 LDAC; weight 27.5 kg) | CONFIRMED manufacturer manual + MEASURED-EXT lite-magazin (25.9 kg) |
| Mundorf/Jantzen/ClarityCap electrical improvement | PROBABLE / sonic NOT DEMONSTRATED |
11. Power and listening levels
Acoustic model: three explicit scenarios
The figures below are calculations from the modelled sensitivity of the cabinet (89 dB/2.83 V/m per the design history; ≈87.2 dB/W/m if Z = 5.3 Ω; 89 dB/W/m is used as the base hypothesis), at a 3 m listening position. Effective sensitivity remains a hypothesis, not a measurement.
| Scenario | Peak SPL | Free field, 1 speaker | Stereo, coherent summing | Room, +4 dB hypothetical | Current @5.3 Ω (room model) |
|---|---|---|---|---|---|
| Night critical listening, continuous | 80 dB peak | 1.13 W/ch | 0.57 W/ch | 0.23 W/ch | 0.21 A RMS |
| Movie critical peak 80 + 15 | 95 dB peak | 35.8 W/ch | 17.9 W/ch | 7.15 W/ch | 1.16 A RMS / 1.64 A peak |
| Movie peak 85 + 15 | 100 dB peak | 113 W/ch | 56.5 W/ch | 22.6 W/ch | 2.07 A RMS / 2.92 A peak |
SPL and required power calculator — 89 dB / 5.3 Ω
Formulas: SPL_1W@d = sens − 20·log10(d) and P = 10^((SPL_target − SPL_1W@d − factor)/10). Results per channel.
The report assumes the 4 Ω tap is the correct one. But with these speakers the analysis has nuances. The impedance data from S13:
| Frequency | Impedance | Phase | Note |
|---|---|---|---|
| 20 Hz | ~6.5 Ω | — | Bass region |
| 200 Hz | 6.5 Ω | — | Low-mid region |
| 800 Hz | 11 Ω (maximum) | — | Midwoofer crossover |
| 1 kHz | ~10 Ω | — | Mid region |
| 5 kHz | 5.3 Ω (minimum) | 0° | Critical region |
| 20 kHz | 7.2 Ω | — | Treble |
| Average impedance | ~7 Ω | — | Closer to 8 than to 4 |
The two trade-offs
| Aspect | 4 Ω tap | 8 Ω tap |
|---|---|---|
| Optimised for | 4 Ω load | 8 Ω load |
| Matching with real load (~7 Ω) | Moderate mismatch: real load is higher than the tap nominal | Smaller mismatch: real load is closer to the tap nominal |
| Typical Zout | Lower (more feedback, lower turns ratio) | Higher (less feedback, higher turns ratio) |
| Damping factor | Better over 5.3 Ω | Worse over 5.3 Ω, better over 11 Ω |
| Current demand | Higher (load demands more current) | Lower |
| Behaviour at Zmax (11 Ω @800 Hz) | Amplifier sees a load larger than expected | Better matched |
| Behaviour at Zmin (5.3 Ω @5 kHz) | Amplifier sees a load larger than expected, but with lower Zout | Higher Zout worsens the variation |
Which one to choose?
- Zmin 5.3 Ω is closer to 4 Ω than to 8 Ω. The 4 Ω tap is the one that deviates least from minimum impedance, which is where the amplifier suffers most.
- Zout is usually lower on the 4 Ω tap. All else being equal, lower Zout = less interaction with the speaker's variable impedance = flatter response.
Response variation due to Zout — calculator
12. Master table and suitability criterion
Master table — power, evidence and suitability
| Model | Price (Oct 2026) | Pnom | P@1% / @3% | Evidence | P of calc | Headroom | Index* |
|---|---|---|---|---|---|---|---|
| Jadis Aria | €4,900 MSRP · €3,430 VIP25 | 30 W EL34 / 35 W KT88 | Not determined | D | 30 W* / 35 W* | 6.23 / 6.90 dB* | 94 |
| Acronos RS-1 (SE) | €2,400 | 25 W | 25 W @0.5% | B | 25 W | 5.44 dB | 93 |
| Cayin Jazz 80 | €1,890 MSRP · €1,323 VIP25 | 40 W KT88 / 36 W EL34 UL | 26 W @3%/8 Ω · 34 W @3%/4 Ω pend. doc. verif. | PEND. DOC. VERIF. | 26 / 34 W | 5.61 / 6.77 dB | 93 |
| Cayin Jazz 90 | €2,990 MSRP · €2,093 VIP25 | 50 W UL / 28 W TR | Not determined | D | 50 / 28 W* | 8.45 / 5.93 dB* | 92 |
| PrimaLuna EVO 200 | ≈ €3,990 | 44 W EL34 | 44 W @1% | B | 44 W | 7.89 dB | 91 |
| Cayin Jazz 70 | €1,598 MSRP · €1,119 VIP25 | 35 W UL / 17 W triode | Not determined | D | 35 / 17 W* | 6.90 / 3.76 dB* | 89 |
| Fezz Silver Luna Prestige | ≈ €2,299 | 35 W | ≈26 W @1%† | A† | ≈26 W† | 5.61 dB† | 89 |
| Icon Audio Stereo 40 III | ≈ €1,630 | 40 W | Not determined | C2 | 40 W* | 7.48 dB* | 88 |
| Cayin CS-55A | €1,690 EL34 · €1,760 KT88 · VIP25: €1,183 / €1,232 | 40 W KT88 / 38 W EL34 UL | 40/38 W @1% / 1 kHz | B | 40 / 38 W | 7.48 / 7.26 dB | 87 |
| xDuoo A35 | ≈ €1,339 | 45 W | THD+N = 1% | C1 | 45 W* | 7.99 dB* | 86 |
| Galion TS34 V2 | ≈ €3,300 | 25 W A / 40 W AB | ≤0.8% at 20 W | C2 | 25 / 40 W* | 5.44 / 7.48 dB* | 86 |
| Willsenton R8 UL | ≈ €1,450 | 45 W RMS | 30 W @1% FTC | B | 30 W | 6.23 dB | 85 |
| Line Magnetic LM-216IA | ≈ €1,599 | 38 W UL | 38 W @1% / 1 kHz | B | 38 W | 7.26 dB | 84 |
| Black Ice F22 V2 | ≈ €1,605 ex-works USA · ≈ €4,400 landed Tarragona | 50 W EL34 | <1% at 20 V | B | 50 W* | 8.45 dB* | 83 |
Breakdown of the real landed cost in Tarragona:
| Item | Estimated amount |
|---|---|
| EL34 base ex-works USA | ≈ €1,605 |
| 25 kg volumetric shipping + insurance USA→Spain | + €300–600 |
| TARIC duty 8518.40 (0–4.9% on CIF) | + €0–180 |
| VAT 21% on CIF + duty | + €400–500 |
| DUA / FedEx / UPS handling | + €60–90 |
| Realistic total landed Tarragona | ≈ €2,400–2,900 for the EL34 version |
| KT88 65 W version (ex-works $2,501–5,000) | ≈ €4,300–4,500 landed Tarragona |
Three separate evaluations (complementing the editorial index)
The numerical index is an editorial tool. It must not be read as an objective measure of sound quality. To decide, three independent evaluations are offered:
| Evaluation | What it assesses | Current status |
|---|---|---|
| A. Demonstrated electrical performance | Power at defined THD, measured Zout, response into real load, noise, stability. | None of the 14 has the complete set. Jazz 80: partial (attributed). Fezz: partial (Stereophile). |
| B. Suitability for the system | SPL required at 3 m, dynamic margin, speaker Zmin, electrical phase, sensitivity. | A35 and Jazz 90: pending measured Zout. Jazz 80: pending documentary verification. RS-1: NFB confirmed, depth and measured Zout ND. |
| C. Purchase value | Verifiable current price, equipment, connectivity, maintenance, warranty, accumulated cost of justified modifications. | Tied at the top: Jazz 80 (€1,323 VIP25) and A35 (€1,339). Next tier: Jazz 90 (€2,093) and RS-1 (€2,400). Outside the "bargain": imported F22 (€4,400) and PrimaLuna (€3,990). |
Editorial index formula
13. Cabinet model and LspCAD
Latest crossover iteration — dedicated page
The crossover design has evolved since the version summarised in this report. The latest iteration is documented on its own page:
3-way LspCAD 6.20 cabinets and behaviour with EL34 at 3 m
Cabinet: 3-way LspCAD 6.20, ≈89 dB modelled (reference to be confirmed: dB/W/m or dB/2.83 V/m?), Zmin 5.3 Ω @5 kHz phase 0°, Zmax 11 Ω @800 Hz, 6.5 Ω @200 Hz, 7.2 Ω @20 kHz.
Crossover: Woofer L3 1.5 mH, Mid C1 15 µF + R2 3.3 Ω + L4 0.62 mH, Tweeter C2 2.2 µF + R3 1.5 Ω.
13.1 Why these cabinets are an ideal test bed for 30–50 W tube amps
1. Stable driver: paper with mica, no foam, Fs stable for 35 years.
2. Reference impedance: Zmin 5.3 Ω phase 0° @5 kHz favourable for EL34 PP.
3. First watt: 65 dB night = 7.1 mW (room model).
4. Peak: 95 dB peak = 7.15 W RMS (room model) or 35.8 W (free field, 1 speaker).
5. Cost: replace dried electrolytics with MKP + Zobel <€120.
LspCAD design (full-size captures)
Horizontal 0-60°: 88.6 @200, 88.0 @800, 87.2 @1k dip, 89.8 @5k, 90.0 @9k. Attenuation: 10° −0.4 dB @10k, 30° −2.8 dB @5k, 60° −10 dB @5k.
Vertical: 100 cm ear = 100 cm driver = 0° @2.7 m. Head movement ±10 cm = ±2.1° → <0.3 dB in the flat zone.
Optional RLC correction
RLC notch 10 Ω + 4.7 mH + 8.2 µF. Resonant frequency: F = 1/(2π√LC) ≈ 811 Hz nominal.
Appendix — Horizontal and vertical spinorama
14. Bias: two static operating points
Two static configurations, not "bias 65 dB" and "bias 80 dB"
Listening level does not by itself determine bias. You can listen at 65 dB or 80 dB with exactly the same operating point.
| Tube | Point A (low dissipation) | Point B (greater margin) | Note |
|---|---|---|---|
| KT77 Gold Lion Genalex | 33 mA / 330 mV / 50.3% / 12.57 W / ~8000 h | 38 mA / 380 mV / 57.9% / 14.48 W / ~5000 h | Recommended Point A |
| 6CA7 EH/JJ | 35 mA / 350 mV / 57% / 14.3 W / ~7000 h | 40 mA / 400 mV / 66% / 16.4 W / ~4500 h | — |
| EL34 Mullard NOS xf2 / JJ E34L | 35 mA / 350 mV / 57% / 14.3 W / ~6000 h | 40 mA / 400 mV / 66% / 16.4 W / ~4000 h | — |
| KT88 Genalex Gold Lion | 40 mA / 400 mV / 47% / 16.4 W / ~8000 h | 48 mA / 480 mV / 56% / 19.7 W / ~6000 h | Recommended Point B (Jazz 90/80) |
| PSvane Art Series TII / S | 33-35 mA / 54-57% | 38-40 mA / 62-66% | — |
15. Personalised Top 3 — second, complementary amplifier
Chart — Price vs headroom at 95 dB peak
The direct answer
Which three amplifiers are the best as a second system — complementary to a good solid-state chain — for these 89 dB / 5.3 Ω / 3 m 3-way speakers, in an enclosed 80 m³ room, at 80 dB continuous with 95 dB peaks, with classical, baroque, early music, opera, jazz and folk, based in Tarragona?
1. Jadis Aria — 30 W EL34 / 35 W KT88 / KT120
Brings: the only one with an explicit sound profile aimed at opera and baroque. Point-to-point with premium components, 20 kg of iron, 5 Hz–40 kHz.
Dealer: Lyric Audio Élite (Valladolid). Service: Hifi Center Madrid + Lyric Audio's own service centre.
2. Cayin Jazz 80 EL34 — 36 W UL / 18 W Triode
Brings: the most complete at a rock-bottom price. With VIP25 discount at €1,323: preamp 2×12AX7 + 2×12AU7, supply with documented 1 H / 180 mA LC choke, KT88-ready toroidal, triode mode, 16-300 Ω headphones, bias VU meters without a multimeter, Bluetooth QCC5125 with LDAC/aptX HD.
Dealer: Lyric Audio Élite (Valladolid).
2.b Cayin Jazz 90 — 50 W UL / 28 W TR, DAC + MM phono + triode
Brings: the most complete of the current Cayin range. With VIP25 discount at €2,093: 4× KT88 with EL34 compatibility, 2× 12AX7 + 2× 6SN7, own toroidal, ESS ES9018K2M DAC, BT 5.1 with LDAC/aptX HD, solid-state MM phono (RIAA = ±0.25 dB, S/N 68 dB), subwoofer output, 85 mW @32 Ω headphones, switchable triode mode, BMI with MCU, 27.5 kg.
Dealer: Lyric Audio Élite (Valladolid).
3. Acronos RS-1 — 25 W / 18 W @ 0.01% THD
Brings: the best in the zone where 80% of listening happens. SE class A without crossover distortion, 18 W @0.01% THD (published), DF≈50 declared, Zout ≈0.16 Ω calculated under the 8 Ω convention. NFB confirmed by the manufacturer: all Acronos use feedback. UL confirmed: all their tube amps are ultralinear. RF/EMI IEC C14 filter confirmed: all include it.
The honest bit
All four are at evidence C-D. The Jazz 80 has cited external measurements (pending documentary verification); the Jazz 90 has the most complete official spec sheet; the A35 brings point-to-point wiring and price; the RS-1 is SE class A with NFB confirmed (depth ND). The editorial recommendation is based on declared architecture, manufacturer sound profile, Zout against 5.3 Ω with phase 0°, serviceability and service availability in Spain.
16. FAQ
Eight concrete questions with short answers
| Question | Short answer | Detail in |
|---|---|---|
| Are 45 W enough for these speakers? | Depends on the model. In the room model (22.6 W required for 100 dB): yes, with margin. Free field (113 W required for 100 dB): no. Uncertainty over effective sensitivity (±2 dB) multiplies required power by 1.6. | S11, S4 |
| Which output tap? | 4 Ω tap. Halves interaction with the speaker's variable impedance. | S11.1 |
| Does KT77 make sense? | Yes, if circuit compatibility is verified (supply, transformer) and bias is re-adjusted. Dissipates 32 W vs 25 W for the EL34. | S6, S8 |
| Does it make sense to change preamp tubes? | It is the step with the greatest expected impact at 65 dB (7 mW). But without measuring before/after, the real effect is NOT DEMONSTRATED. | S7, S8 |
| Is it worth modifying capacitors? | Electrical improvement probable. Audible improvement NOT DEMONSTRATED without before/after measurement. Identify the component and its function first. | S7, S9 |
| Is it worth disconnecting/modifying FB1? | Not without measuring first. The "Pure Analog" hypothesis is interesting as a controlled experiment. A/B test in S5. | S5, S7 |
| A35 or Jazz 80 or Jazz 90? | Depends on priority: A35 = price + point-to-point; Jazz 80 = LC choke + price (with documentary reservation); Jazz 90 = complete functional equipment. | S12, S15 |
| Can I use KT88 in the A35? | Octal-socket compatible but NOT DEMONSTRATED for the A35. In the Jazz 80 and 90 it is the stock tube. | S7, S3 |
17. Modification budget
Cumulative cost per step, not just part price
Modification steps
| Step | What you buy | Quoted cost | Changes (electrical) |
|---|---|---|---|
| 0. Stock | Nothing — S9 baseline | €0 | — |
| 1a. Budget tubes | EL34 JJ or Tung-Sol matched quad | ~€120–180 | Factory bias and matching |
| 1b. Mid-range tubes | Genalex Gold Lion KT77 matched quad | ~€250 | Documented drop-in. +7 W dissipation margin — verify circuit limits |
| 1c. PSvane classic | PSvane EL34PH/4 Xtreme Classic matched quartet | €399 | Mullard-replica pentode, metal base, 25 W |
| 1d. PSvane flagship | PSvane Summit EL34 matched quartet | ~€403 | PSvane flagship pentode, 25 W, 24-month warranty |
| 2. Modern preamp | PSvane 12AX7-S Art TII + 12AU7-S Art TII pair | ~$140 | Gain/µ spec CONFIRMED |
| 2b. Premium NOS preamp | Brimar CV4003 NOS APOS Archive Vault, matched 5% pair | ~€430 | Provenance and Amplitrex testing |
| 3. Coupling capacitors | Mundorf EVO Oil / Jantzen Superior Z-Cap | NOT VERIFIED | Lower ESR/DF per datasheet |
| 4. Supply / bias / FB1 | Multimeter and resistive load (see S9) | €0 instruments | — |
PSvane vs KT77 — cost, risk and bias margin
| Option | Quartet cost | Max dissipation | Bias margin vs 70% | Warranty |
|---|---|---|---|---|
| Genalex KT77 | ~€250 | 32 W | +7 W vs EL34 | Standard |
| PSvane EL34PH Xtreme Classic | €399 | 25 W | 0 W (same as stock EL34) | Standard |
| PSvane Summit EL34 | ~€403 | 25 W | 0 W (same as stock EL34) | 24 months |
Cumulative cost scenarios
| Scenario | Composition | Parts cost |
|---|---|---|
| A — Reliability and best price | KT77 + PSvane 12AX7-S Art + JJ ECC82 Gold Pin | ~€390 |
| B — Warm classic | PSvane EL34PH + PSvane 12AX7-S Art + Brimar CV4003 NOS Vault pair | ~€829 |
| C — Greater commercial positioning | PSvane Summit EL34 + PSvane 12AX7-S Art + Brimar CV4003 NOS Vault pair | ~€833 |
| D — Full budget PSvane | PSvane EL34PH + PSvane 12AX7-S Art + 12AU7-S Art pair | ~€539 |
| E — Burn-in and backup | JJ ECC83S Gold Pin + JJ ECC82 Gold Pin pair + EL34 JJ matched quad | ~€180 |
How to use this table
It is not a spending recommendation. It is price information that was already scattered across S7 and S8.
18. Acronos RS-1 — single-ended KT150 — NFB confirmed, depth ND
Documentary notice for this section
Methodological note for this section
This chapter presents a structural hypothesis, not a closed conclusion. The order is: first Hypothesis 0 (the figures may not be reliable or comparable), then hypotheses 1–4 of architecture, and finally the verification plan that discriminates between them. The final label is NFB confirmed as existence, depth ND (Not Determined).
18.1 The central idea is correct
An SE pentode without significant NFB behaves more as a current source than as an ideal voltage source. This explains why an SE with little NFB can interact strongly with the speaker impedance:
If Zspk changes with frequency, so does the voltage applied to the speaker.
The manufacturer confirms that the RS-1 uses NFB. The question is no longer whether there is feedback, but how much and of what type.
18.2 Starting point: the M1 as a confirmed genealogical reference
The Acronos RS-1 review published in Et Sona establishes an explicit relationship with the previous model in the range:
"On the website you'll find the Acronos M1 with KT120 valves, but on this occasion we have the RS-1 which mounts the KT150s; please consider this when comparing technical data." CONFIRMED secondary source
The M1 declares 20 W pure class A ultralinear with KT120. The RS-1 declares 25 W with KT150, keeping the single-ended pure class A philosophy. The genealogical relationship between the two models is useful for orienting investigation, but does not replace a schematic of the RS-1 nor a measurement of a specific unit.
18.3 Tube chain: EC86 NOS → ECC803S NOS → KT150
| Stage | Tube | Type | Function |
|---|---|---|---|
| Preamp | EC86 NOS | UHF triode | Initial amplification. µ = 68, gm = 14 mA/V, ra ≈ 4.86 kΩ. |
| Driver | ECC803S NOS | Dual triode | One triode drives; the other is a cathode follower (low Zout toward the KT150). |
| Power | KT150 | Beam tetrode | Pure class A ultralinear. |
Total: six tubes (two KT150, two ECC803S, two EC86). The cathode follower improves high frequency but is not responsible for DF ~50: that is decided in the output stage (KT150 + OPT + NFB).
18.4 Class A ceiling — B+ lower constraint
In a class A SE, the maximum theoretical power is limited by B+ and idle current:
With bias 130 mA, 25 W requires B+ = 384.6 V theoretical. With practical losses (Vknee ≈ 40 V, ohmic losses ~10%): B+practical min ≈ 478 V.
| B+ | Pmax theoretical | Interpretation |
|---|---|---|
| 384.6 V | 25.00 W | Exact theoretical limit, no margin |
| 430 V | 27.95 W | With Vknee, no ohmic losses |
| 478 V | 31.07 W | Practical minimum, no margin |
| 500 V | 32.50 W | Comfortable zone |
Per-tube dissipation: at 450 V, 58.5 W (83.6% of the 70 W max). At 500 V, 65 W (92.9%). High but within specification. Requires mandatory ventilation and 15–20 min warm-up.
18.5 Empirical precedents: Icon ST30SE and Dared Saturn
| Parameter | Icon ST30SE | Dared Saturn | Acronos RS-1 |
|---|---|---|---|
| KT150/channel | 1 | 1 | 1 |
| Topology | SE, switchable UL/triode | SE triode | SE class A UL |
| Weight | 26 kg net / 28 kg boxed | 18 kg | 14 kg |
| NFB | Switchable H/L/0 | −39 dB (declared, inconsistent) | Confirmed, depth ND |
| DF | UL: 7–9 · TR: 15–17 | 6 | ~50 (Et Sona, 1 kHz average) |
| THD near max | 1% @ 17 W · 2% @ 30 W | ND | 0.01% @ 18 W · 0.5% @ 25 W |
18.6 Hypothesis 0: the specifications may not be reliable or comparable
Before building any architecture tree to explain DF≈50 and THD 0.01%, the chapter must consider the most uncomfortable possibility: that the declared figures do not represent the real behaviour of the amplifier under conditions comparable to the precedents'.
Scenarios within Hypothesis 0
- DF measured at 1 kHz into a pure resistive load. The real DF at 20 Hz or into a reactive load could be substantially lower. MANUFACTURER confirms that DF ~50 is an average at 1 kHz into resistive loads.
- THD measured at 1 kHz without sweep. The 0.01% at 18 W may not hold at 20 Hz or 20 kHz.
- THD measured with only one tube conducting or only one channel driven.
- THD measured before full warm-up.
- THD measured at a comfortable operating point. The 0.5% at 25 W is the realistic "at the limit" figure.
- Measurement conditions not equivalent to those of Hi-Fi News.
- The 18 W @ 0.01% datum could be inherited from the M1, not an RS-1 measurement.
The way to discriminate it is the most obvious and the most expensive: own measurement. With a measurement bench replicating the Hi-Fi News methodology, the uncertainty would drop drastically.
18.7 Manufacturer response on DF: the 12 factors
Juan Calatayud (JC Valvular) has responded by email to the report's questions. The material is significant and deserves its own section.
Key points of the response
- DF is complex in tubes ("infinitely easier" in solid state).
- It is calculated during OPT design, not measured at the end.
- The DF ~50 figure is an "average DF", measured at 1 kHz into fixed resistive loads.
- It is not constant: it varies with frequency and with the reflected speaker impedance.
- It does not mention NFB in dB, nor UL %, nor cathode feedback.
The 12 DF factors according to the manufacturer
| # | Factor | Note |
|---|---|---|
| 1 | Internal impedance of the output tubes | Pentode/beam tetrode > triode in Zout |
| 2 | Stage configuration | Pentode, beam tetrode (6550, KT88), triode, UL |
| 3 | OPT turns ratio | n² fixes the primary reflected impedance |
| 4 | Primary winding resistance | Direct ohmic losses |
| 5 | Secondary winding resistance | Lower Rcc → higher DF |
| 6 | Global negative feedback | Reduces Zout by (1 + ß·A). Central factor of §18.8 |
| 7 | Local feedback | Cathode degeneration and other forms |
| 8 | OPT leakage inductance | Limits coupling. Important at HF |
| 9 | Primary magnetising inductance | Relevant at LF. Insufficient → bass loss |
| 10 | OPT parasitic capacitances | Influence HF. Can generate resonances |
| 11 | Wiring and connections | Series resistances between amplifier and speaker |
| 12 | Operating frequency and load | DF varies with frequency and complex impedance |
18.8 The central question: how much NFB is needed? — loop gain calculation
The two extraordinary figures
| Source | Power | THD | Configuration |
|---|---|---|---|
| Tung-Sol KT150 | 20 W | ≈ 14% | Pure pentode, class A1, no NFB |
| Acronos RS-1 (declared) | 18 W | 0.01% | Ultralinear + NFB |
Three orders of magnitude. Not a contradiction: it is what UL + NFB provides.
How much NFB the THD demands:
49–54 dB of loop gain for THD alone. The declared DF of ~50 corresponds to 31.7 dB as an impedance ratio. The two figures are obtained under different hypotheses and are not directly comparable.
Is 31.7 dB achievable?
- OPT attenuation: 20·log10(25) ≈ 27.96 dB.
- Closed-loop gain: Vout = √(25·8) = 14.14 Vrms. With 100 mV: Acl = 141.4 = 43.01 dB.
| Scenario | Raw | After OPT | Closed | Loop | Final DF |
|---|---|---|---|---|---|
| Optimistic | 100 dB | 72.04 dB | 43.01 dB | 29.03 dB | 35.9 |
| Typical | 95 dB | 67.04 dB | 43.01 dB | 24.03 dB | 20.3 |
| Realistic | 92 dB | 64.04 dB | 43.01 dB | 21.03 dB | 14.3 |
At best, 29 dB, below the 31.7 dB needed. A standard global loop is not enough for DF 50.
For THD: with 21–29 dB of loop and 3–5% open-loop in UL, the final THD would be 0.1–0.45%. 20–33 dB short.
- ra in UL. The datasheet gives ra ≈ 3 kΩ in pentode. UL connection reduces ra versus pentode. If raUL ≈ 1.5–2 kΩ instead of 3.5 kΩ, the loop needed for DF 50 drops to ~26–28 dB, within the 21–29 dB available range. The verdict "a standard global loop is not enough" is sensitive to this parameter.
- OPT Raa. §18.8 uses n = 25 (~5 kΩ reflected over 8 Ω); §18.4 uses Raa = 3 kΩ (n ≈ 19.4). With 3 kΩ consistently, the loop needed for DF 50 rises to ~36 dB. These are two different hypotheses about the same transformer.
18.9 If sensitivity were 1000 mV instead of 100 mV
The declared sensitivity could be a typo (1000 mV), an integrated-amp figure at full volume, or the real value. This changes the whole hypothesis tree.
| Sensitivity | Acl | Loop | THD without CF | THD with CF ×5 |
|---|---|---|---|---|
| 100 mV | 43.01 dB | 21–29 dB | 0.1–0.45% | 0.02–0.09% |
| 1000 mV | 23.01 dB | 41–49 dB | 0.011–0.045% | 0.002–0.009% |
The 20 dB difference between scenarios goes entirely into loop gain. With 41–49 dB of loop:
- Reduction factor: 112.2 (41 dB) to 281.8 (49 dB).
- THD without CF: 0.011–0.045%.
- THD with CF ×5: 0.002–0.009%.
- Without CF, final THD lands at 0.011–0.045%. The declared 0.01% enters the best case.
- With CF ×5, final THD is 0.002–0.009%, below 0.01% in all cases.
- The only scenario where H1 revives without cathode feedback.
18.10 Architecture hypotheses (conditioned on H0 = false): four compatible routes
With all available data, and assuming Hypothesis 0 has been ruled out, the tree reduces to four structurally compatible routes:
Hypothesis 1 — Deep-loop global NFB
A feedback loop from the OPT secondary to the driver input, with high loop gain. Compatible with SE architecture, with the M1 as precedent, and with the Dared Saturn as a KT150 SE global-NFB precedent (−39 dB). With 100 mV sensitivity, the model does not reach 31.7 dB of loop. With 1000 mV, H1 revives.
Hypothesis 2 — Cathode feedback + possible light global loop
A dedicated winding in the OPT feeding the KT150 cathode. Lowers Zout 3–5× without deep global loop. Does not contradict the §18.8 calculation. Best fit: with CF ×5 + UL + 29 dB, THD at 0.025–0.089%. Close, not exact.
Hypothesis 3 — Hybrid output with low-impedance buffer/follower
An additional active element in the output stage that lowers Zout without depending on classic loop gain. Ruled out as sole explanation: a buffer does not lower THD.
Hypothesis 4 — Combination of the above
UL + local NFB around the output stage + output buffer. Consistent with both figures, requires complexity not observed.
18.11 What the Icon ST30SE and Dared Saturn teach us about the RS-1
Both amplifiers already use feedback: the Icon switchably up to 25 dB, the Dared fixed at −39 dB. And both stay at a DF of 6–9 (Icon UL) or 6 (Dared). The RS-1 goes much further in declared DF and THD.
18.12 Philosophical counterpoint: Terry Gesualdo and La Dolce Audio
"I never do feedback between stages, only local feedback within a stage. […] I design by sound, so I have tried a million different combinations and found what works or is able to produce the sound I like." OBSERVED public statement by the designer
His philosophy can be broken down into four principles:
- Local feedback only.
- "Imperfections" as a virtue.
- Design by sound.
- Paradoxical transparency.
18.13 The methodological correction: NFB confirmed, depth ND
- Confirmed: the RS-1 uses NFB. MANUFACTURER
- Confirmed: the RS-1 is ultralinear. MANUFACTURER
- Confirmed: the RS-1 includes an RF/EMI filter on the IEC C14 inlet. MANUFACTURER
- Observed: two threaded fuse holders on the rear panel. OBSERVED
- It cannot be stated that the declared figures are real or comparable. ND
- It cannot be stated what the NFB depth is in dB. ND
- It cannot be stated that the RS-1 has global NFB. ND
- It cannot be stated that the 8 trimmers are loop compensation. ND
18.14 Verification plan: which checks would discriminate between hypotheses
| Check | Source | What it discriminates | Cost |
|---|---|---|---|
| Own DF and THD measurement replicating the Hi-Fi News methodology | Own measurement bench | Hypothesis 0 vs hypotheses 1–4 | Medium |
| Published Zout vs frequency curve | Spec sheet, review, third-party measurement | Hypotheses 1/2/4 vs 3 | Low (search) |
| Presence of power-semiconductor heat sinks in the output signal path | Interior photos | Hypothesis 3 (hybrid) vs 1/2/4 (loop) | Low (photos) |
| Identification of the 8 trimmers by function | Schematic or wiring inspection | Hypotheses 1/2 vs 3/4 | Medium |
| Feedback loop schematic | Manufacturer or leaks | All hypotheses | High |
| Ask manufacturer: "how much NFB, in dB, and of what type?" | H1 vs H2 vs H4 | Low | |
| Real B+, UL %, OPT Raa, ra in UL | Measurement or manufacturer | Class A ceiling, §18.8 model | Low-Medium |
18.15 RS-1 conclusion
The manufacturer confirms three house traits: all Acronos use feedback, the tube amps are ultralinear, and all include an RF/EMI filter on the IEC C14 inlet. It also confirms that DF is calculated during OPT design from 12 factors and that DF ~50 is an average measured at 1 kHz into fixed resistive loads.
The KT150 in single-ended pentode configuration, without significant feedback and without an active Zout-reduction mechanism, presents a high output impedance. This can interact strongly with the frequency-dependent speaker impedance. But the RS-1 does have NFB, is ultralinear and has an RF/EMI filter.
The RS-1 is specified at DF≈50 and THD 0.01% at 18 W, with a single KT150 per channel. Before building any architecture tree, the chapter acknowledges a Hypothesis 0: the figures may not be reliable or comparable to the precedents' measurements.
Under the model in §18.8, 0.01% THD at 18 W demands 49–54 dB of reduction; the declared DF of ~50 corresponds to 31.7 dB as an impedance ratio. Available loop gain is 21–29 dB at 100 mV sensitivity. Below the threshold.
If Hypothesis 0 is ruled out, then the RS-1 cannot be a KT150 SE in ultralinear with the same feedback level as the Icon or Dared. The compatible hypotheses are four: (1) deep-loop global NFB, (2) high-gain local feedback (cathode feedback), (3) hybrid output with low-impedance buffer/follower —ruled out as sole explanation—, and (4) combination.
Neither H1 nor H2 reaches exactly 0.01% at 100 mV. H1: 0.1–0.45%. H2: 0.025–0.089%. If sensitivity were 1000 mV, the loop would rise to 41–49 dB: THD 0.011–0.045% without CF, 0.002–0.009% with CF. The only scenario where H1 revives without cathode feedback.
The exact nature of the Zout-reduction mechanism remains ND until a schematic, primary documentation or measurement appears. The Zout ≈0.16 Ω in the "Traffic light" is a derivation (DF=50, 8 Ω), not a direct measurement.
Class A ceiling: B+ = 384.6 V theoretical and ~478 V practical. The 450–500 V zone is not optimistic, it is what is needed. KT150 at 79–93% of maximum depending on B+.
Construction indicator: 14 kg vs 26 kg net for the Icon and 18 kg for the Dared. Thermal density 13.9 W/kg, ~2× that of the Icon (6.8 W/kg).
What CANNOT be concluded from S18
- That the Acronos RS-1 has no negative feedback: it does, confirmed by the manufacturer.
- That the topology is pure triode or pentode: it is UL, confirmed by the manufacturer.
- That it does not carry an RF/EMI filter: it does, confirmed by the manufacturer.
- That the NFB depth in dB is determined: NFB = confirmed, depth ND.
- That the NFB type (global vs local vs cathode feedback) is determined: ND.
- That the 8 trimmers are loop compensation: one of several plausible hypotheses.
- That the declared DF≈50 is maintained at 20 Hz and into a reactive load: the catalogue datum is typically measured at 1 kHz into a pure resistive load.
- That the 0.01% THD at 18 W is an exceptional first-watt datum: it is a single point on a curve.
- That the 18 W @ 0.01% is an RS-1 measurement and not an M1 inheritance: ND.
- That the UL % is determined: ND.
- That the 100 mV sensitivity is a power-stage figure: ND. It could be an integrated amp at full volume, or a typo (1000 mV?).
- That the filter is Schaffner: probable, not confirmed.
- That the RS-1 is suitable for 89 dB speakers without checking the manufacturer's recommended minimum sensitivity of 92 dB.
- That the declared figures are false or exaggerated: the model does not prove that the RS-1 cannot meet them; it indicates what would need to be achieved for them to be consistent.
- That the class A ceiling implies the 25 W figure is false.
- That the manufacturer's response closes the NFB question: it adds DF context, not loop context.
APPENDICES AND SOURCES
Support material. Appendix A explains how to interpret the evidence labels. Appendix B gathers the primary sources and traceability documentation.
APPENDIX A — HOW TO READ THE EVIDENCE LABELS
Appendix A — How to read the evidence labels
Every claim in the report carries an evidence label indicating its level of support. This appendix expands the summary legend in the side panel (visible on desktop) for the mobile version and printout, where the aside is hidden.
The eight labels, one by one
| Label | What it means | What would be needed to raise the level |
|---|---|---|
| CONFIRMED | Datum established by official manufacturer documentation (manual, schematic, technical correspondence) or verified silkscreen. | Already at the highest level. Can only be downgraded if a documentary contradiction appears. |
| OBSERVED | Directly visible in photographs or in board silkscreen. | Confirm the interpretation with the manufacturer, or measure the component directly. |
| CALCULATED | Mathematical result obtained by applying a formula to known data. | Experimentally verify the result with an instrument. |
| PROBABLE | Reasonable physical inference, consistent with what is observed, but unverified. | Direct measurement or documentary confirmation. |
| HYPOTHESIS | Possibility consistent with available data but not sufficiently supported. | Additional evidence that turns it into PROBABLE or rules it out. |
| REPORTED-EXT | Datum from a reliable third party (dealer, specialist publication) without own measurement or manufacturer confirmation. | Official manufacturer confirmation or independent measurement. |
| MEASURED-EXT | Measurement published by an independent lab, with identifiable conditions. May refer to another product. | Measurement of the product in question under the same conditions. |
| ND | Not determinable with current evidence. | Direct measurement, manufacturer schematic or additional documentation. |
How several labels combine in the same row
The same claim can carry several labels at once, because different parts of the sentence have different evidence levels.
Each part of the sentence comes from a different source. A reader who reads all three parts with the same label will not understand the real state of the matter.
APPENDIX B — PRIMARY SOURCES AND DOCUMENTATION
Appendix B — Primary sources and documentation
Direct links to the primary sources of the report's critical data. The consultation date is the time of writing (October 2026).
| Source | Type | Link / status |
|---|---|---|
| xDuoo (manufacturer) | Official specifications | xduoo.com |
| APOS Audio | US distributor — primary source of specifications | apos.audio |
| Audiophonics | EU distributor — price and availability; 4/8 Ω tap spec sheet | audiophonics.fr |
| PSvane | Tube manufacturer | psvane.co |
| Genalex / Gold Lion | Tube manufacturer (Reflektor) | genalexaudio.com |
| Tube Amp Doctor (TAD) | European tube distributor | tubeampdoctor.com |
| Lyric Audio Élite | Spain distributor (PSvane, Cayin, Jadis) — VIP25 with 30% discount | lyricaudio.es |
| JC Valvular (Acronos) | Spanish manufacturer — Málaga | Tel. 659 15 37 68 |
| Tharbamar (YouTube) | A35 video review and interior photos | youtube.com/watch?v=o3jCYLKleiY |
| Cayin / Lyric Audio | Manufacturer and distributor documentation for Jazz 80 and Jazz 90 | cayin.com · lyricaudio.es |
| Black Ice Audio | US manufacturer of the F22 — ex-works price vs real landed price Tarragona | blackiceaudio.com |
| Stereo Magazin / fairaudio / 7Review | Jazz 80 powers at 3% THD | LOCATABLE CITATION PENDING (issue, page, conditions) |
| lite-magazin.de | Jazz 90 review (25.9 kg weight, B+ delay) | — |
| Stereophile | Jadis Orchestra Reference Mk.II and Fezz Silver Luna measurements | — |
| Hi-Fi News | Icon Audio ST30SE measurements (January 2020) | — |
| Et Sona | Acronos RS-1 review (M1/RS-1 genealogical reference; DF ~50) | Secondary source |
| LspCAD 6.20 | Acoustic simulation software | — |