Cayin Jazz 90, PP Ultralinear valve with KT88 or EL34 EN

INDEX ESPAÑOL

Acronos P52 · Report V6.8.0 · Oct 2026

by [Meta + ChatGPT + DeepSeek + Claude + Gemini] + maty @nauscopio

Acronos P52, 40-watt ultralinear PP valve amplifier

Reverse-engineering report based on visual inspection of photographs and cross-referencing with manufacturer catalogues. Images from the video: Amplificador ACRONOS P52 Una obra de arte. Cómo se fabricó. and additional interior images provided by the owner of the unit (2026-10).

Traffic light — 15-second read

In short: 40 W nominal valve amplifier (manufacturer data), about 40 kg (personal comment, unconfirmed). Clearly seen in the photos: 3 EPCOS B43645 capacitors of 610 µF / 450 V (they would add up to 1,830 µF only if in parallel: conditional calculation); a vertical board (PCB-3) with 4 blue components, probably relays, of unknown function; a 10W10KJ resistor of 10 kΩ / 10 W (does not seem compatible with a permanent B+ bleeder at 430 V under that hypothesis; its real function is unknown); and two blue trimmers (adjustment potentiometers) on the signal board PCB-S, one at each end. Probable, but unconfirmed: Hongfa HF115F relay, NE5532 chip and TL431 voltage reference; there are two visible 8-pin chips (DIP-8) and the NE555 is not confirmed. Unknown: the function of the auxiliary magnetic element, the valve supply voltage (B+), its quiescent current (bias), the negative feedback (NFB), the ultralinear percentage (%UL), the output impedance (Zout), the damping factor (DF) and the power at 1 % distortion (P@1 %). Absolute priority: measure the P52's Zout and obtain photos of the underside of PCB-S.

Getting started

▸ Basic guide (G1–G6) G1 What is a push-pull amp G2 Which data is reliable G3 How many watts are needed G4 Capacitors and ripple G5 Impedance and current G6 Conclusions and exercises

Part I — Acronos P52 report

0. Executive summary 1. Scope, method and limitations 2. General architecture 3. Small-signal board (PCB-S) 3.1 Observed components 3.2 Probable topology and Mullard 5-20 lineage 3.3 ECC803S driver: current capability to verify 3.4 Probable function of the trimmers 3.5 Driver substitutions 3.6 Cross-reference 4. Power supply board (PCB-P) 4.1 Observed components 4.2 HV bank: B43645 610 µF / 450 V 4.3 Toroids and 10W10KJ 4.4 Auxiliary regulation block 4.5 Blue vs black trimmers 4.6 B+ supply architecture 4.7 PCB-3 and physical partitioning 5. Power stage 5.1 Valves and dissipation 5.2 Class of operation: AB probable 5.3 Bias mode 5.4 6550 substitutions 5.5 Open questions 5.6 Output transformers 6. AC input and RF/EMI 7. Conclusion and traceability

Part II — Special edition

▸ Part II — Special Edition

Appendices

▸ Appendix — Questionnaire for JC Valvular ▸ Unvalidated external proposals ▸ References and credits

Evidence legend

CONFIRMED — data established by external documentation

OBSERVED — directly visible in the photographs

CALCULATED — mathematical result from known data

PROBABLE — reasonable physical inference; pending tracing/measurement

HYPOTHESIS — possibility not sufficiently supported

ND / NOT DEMONSTRATED — not determinable with current evidence

MEASURED — instrumentation on the owner's physical unit

MEASURED-EXT — published external laboratory measurement

REPORTED-EXT — data from a reliable third party without own measurement or official confirmation

GETTING STARTED — WITH NO PRIOR KNOWLEDGE (Basic guide G1–G6)

GETTING STARTED — WITH NO PRIOR KNOWLEDGE

If you have never opened a valve amplifier, start here. Only high-school physics is needed (logarithms, electrical power, capacitors, impedance) to understand why the figures in this report are expressed with such care. Technical details begin in Part I.

Basic guide: reading this report without being a valve expert

Central idea: no one has measured the Acronos P52, so we do not know how it performs. What we do know is what can be seen inside and what a specific speaker in a specific room requires. With those two pieces and a bit of physics you can reason without inventing data.

• G1 · how an ultralinear push-pull amplifier works.
• G2 · which data is reliable and which is not.
• G3 · how many watts are needed (decibels and distance).
• G4 · capacitors and power supply ripple.
• G5 · impedance, current and damping factor.
• G6 · conclusions and exercises with solutions.

G1. What a push-pull valve amplifier is

An amplifier takes the weak signal from the player (on the order of 1 V) and converts it into a signal with the voltage and current needed to drive the speaker. In this device, thermionic valves do the job in two blocks: a preamp, which increases the signal voltage, and a power stage, which supplies the current. Everything is fed by a power supply that produces a high DC voltage called B+ (the "valve supply").

Player → Preamp → Power stage → Output transformer → Speaker
 (≈ 1 V) (voltage) (current)    (impedance matching)  (4–8 Ω)
                          ↑
                 Power supply (B+)

In the power stage the valves work in pairs in push-pull ("push-pull"): while one conducts more current, the other conducts less, and their currents combine in the output transformer. This has two physical advantages: the DC components of the two valves cancel in the core (they do not permanently magnetise it) and many even harmonics are cancelled, which reduces distortion. The actual class of operation (A, AB, B) of the P52's 6550s is classified as AB probable with AB1 probable subtype (§5.2); the exact bias mode (how the quiescent current is set) remains ND.

The output transformer acts as an "adapter": a power valve needs to see a load of thousands of ohms and a speaker has only 4–8 Ω. With a turns ratio n, the impedance is multiplied by n². Illustrative example (not a P52 data point): with n = 25, an 8 Ω speaker is seen from the valves as 25² · 8 = 5000 Ω.

In an ultralinear (UL) arrangement, the screen grids of the valves are connected to intermediate taps of the transformer. It is a compromise: pentode mode gives more power and triode mode gives less distortion. The manufacturer declares PP UL with 6550 (CONFIRMED (manufacturer)); the P52's %UL tap is ND.

G2. Which data is reliable and which is not

This report is not a measurement: it is based on photographs from the manufacturer's video, component catalogues and public documentation. That is why every statement carries an evidence label (the full legend is in the margin, and the method in §1):

Label	What it means	Example in this report
OBSERVED	Directly visible in the photographs.	Silkscreen B43645-S7617-M1 on the HV capacitors.
PROBABLE	Reasonable physical inference, pending verification.	That the three capacitors are in parallel.
CALCULATED	Comes from a formula based on known data.	35.8 W to reach 95 dB at 3 m.
REPORTED-EXT	Data from a reliable third party without own measurement or official confirmation.	Nominal service life of the B43645 series.
MEASURED-EXT	Measured by an external laboratory.	Published figures for other devices (context).
ND	Cannot be stated with current evidence.	P52's Zout and P@1 %.

Why 40 W nominal is not equivalent to a power measured at a specific THD. The maximum power of an amplifier depends on how much distortion is accepted: as clipping approaches, distortion rises very quickly. The P52's 40 W is a commercial figure with no THD, frequency or load condition. Comparing that figure with any published measurement of another device as if they were equivalent would be a methodological error.

G3. How many watts are needed: decibels and distance

To know whether an amplifier "has enough", you must start with the speaker and the listening. The decibel is a logarithmic scale: +10 dB multiply power by 10 and +3 dB multiply it by 2 (10^0.3 ≈ 1.995). It is calculated as follows, with a generic reference system:

1. Sensitivity. The reference speaker produces 89 dB with 1 W at 1 m. HYPOTHESIS of calculation, not measured.
2. Distance. The level falls by 20·log₁₀(d) dB with distance. At 3 m: SPL(1 W, 3 m) = 89 − 20·log₁₀(3) = 89 − 9.54 = 79.46 dB.
3. Power. To reach a level L: P = 10^((L − 79.46)/10). For 95 dB that gives 35.8 W with a single speaker.
4. Stereo. With two speakers and a decorrelated stereo signal (the normal case in recorded music), the acoustic powers add incoherently in a reverberant field: +3 dB. Each channel contributes half: 17.9 W/channel. Declared assumption: decorrelated signal. With a centred mono signal and on-axis listening, the sum would be coherent (+6 dB) and the power per channel would be halved (~9 W), but that case is not the usual one in musical listening.
5. Peaks. Music has peaks above its average level. A crest factor of 15 dB multiplies the power by 10^1.5 ≈ 31.6. That is why you must size for the peak and not for the average: 80 dB average with 95 dB peaks.

Uncertainty weighs. If the real speaker sensitivity were 2 dB lower or higher, the required power would be multiplied by 1.6 or 0.63: an error of the same order as the differences between amplifiers. The interactive calculator in Part I lets you try your own values.

G4. The power supply: capacitors and ripple

Valves need a high DC voltage (B+) obtained by rectifying the mains. In Spain the mains is 50 Hz and, with a full-wave rectifier (a KBU1010 bridge is observed in the P52), the ripple appears at 100 Hz. The filter capacitors charge at the voltage peaks and discharge between them. Since Q = C·ΔV and the time between peaks is 1/f, the voltage drop is approximately:

ΔV ≈ I / (f · C) (approximation: assumes constant current throughout the period and somewhat overestimates the real ripple)

Example with 200 mA and a 610 µF capacitor at 100 Hz: 0.2 / (100 · 610·10⁻⁶) ≈ 3.28 V peak-to-peak ripple. Three equal capacitors in parallel would add up to 1830 µF and give ≈ 1.09 V, but that only happens if they are in parallel, something that has not been verified ND.

Beware of 450 V. 450 V is the capacitor's maximum rated voltage; it does not prove that the P52's B+ is 450 V. A 450 V capacitor can operate at lower voltage, and in real conditions there are margins, transient overvoltages and tolerances. The P52's real B+ is ND.

In addition to the capacitors, the P52 carried two toroids on the Rev.A board (the earlier revision of the board) compatible with an inductive filter (C-L-C type). A coil opposes rapid current variations with a reactance X_L = 2π·f·L. Illustrative example (not a P52 data point): a 1 H coil presents 628 Ω at 100 Hz. Without knowing L, its resistance and the load, it is impossible to quantify how much ripple they actually eliminate. Details in §4.3.

G5. Impedance, current and damping factor

A "6 Ω" speaker does not always measure 6 Ω: its impedance changes with frequency. The reference speaker drops to a minimum of 4.9 Ω around 1.4 kHz. An amplifier that behaves as a voltage source delivers more current when impedance drops: at constant voltage, the current at 4.9 Ω is 6/4.9 ≈ 1.22 times that at 6 Ω.

This is where the amplifier's output impedance (Z_out) comes in, acting as a series resistance with the speaker. The damping factor is DF = Z_load / Z_out. Illustrative example (not a P52 data point): with Z_out = 1.5 Ω, the voltage reaching the speaker is the divider one V_spk = V · Z_load / (Z_load + Z_out). Over 6 Ω it loses 1.94 dB and over 4.9 Ω it loses 2.32 dB: the frequency response partly follows the speaker's impedance curve (0.38 dB difference between the two points). That is why measuring the P52's Z_out is the absolute priority.

Z_out threshold used in the report. The report works with two different criteria:

• Absolute loss over Zmin ≤ 1 dB: requires Z_out ≤ 0.60 Ω (exact calculation over Zmin = 4.9 Ω: with 0.60 Ω the loss is 1.01 dB; with 0.50 Ω it is 0.84 dB). The report adopts 0.5 Ω as a recommended design target, not as an exact equivalence.
• Relative variation between Zmin and Zmax ≤ 1 dB: depends on Zmax (the impedance peak of the cabinet in the bass). Without measuring Zmax, the threshold is not reproducible. Reference table for Zmin = 4.9 Ω:

Zmax (Ω)	Max Zout (Ω) for Δ ≤ 1 dB relative
10	1.33
12	1.10
14	0.98
16.5	0.90
20	0.82
30	0.73

The report adopts Z_out ≤ 0.9 Ω as the minimum working threshold, corresponding to a Zmax ≈ 16.5 Ω. That value is a working reference, not a measurement of the owner's cabinet.

G6. Conclusions and exercises

• There are no published measurements of the P52. It appears to be a careful construction, but that does not equate to good performance: B+, bias, feedback, Z_out and P@1 % remain unproven.
• The weight (≈ 40 kg) suggests larger transformers, but it is not proof of quality or power, and it is also unverified.
• Neither a photograph nor a table decides whether the amplifier "is better": measurements are needed (THD+N vs power, Z_out, response with real load, noise). They are proposed in the Appendix.

Exercises to practise:

1. A speaker has 92 dB/W/m. What level does it produce with 1 W at 2 m?
2. With the 89 dB/W/m speaker from this guide, what power is needed for a 100 dB peak at 3 m with a single speaker?
3. A filter has 2000 µF and the amplifier draws 300 mA. What is the ripple at 100 Hz?

See solutions

1. 92 − 20·log₁₀(2) = 92 − 6.02 ≈ 86 dB.
2. 10^((100 − 79.46)/10) ≈ 113 W.
3. 0.3 / (100 · 0.002) = 1.5 V.

PART I — ACRONOS P52 REPORT

Map of what we know. From the general to the specific: architecture → boards → power stage → AC input → synthesis. The evidence levels are declared in §1 and applied in each section.

0. Executive summary

Acronos P52 — general view. Image: JC Valvular. Click to enlarge.

20-second read

What it is: an ultralinear push-pull valve amplifier with two 6550 valves per channel and 40 W nominal. At the input it has a 6Ж32П (EF86 equivalent) and an ECC803S (double triode). They most likely amplify the signal and invert its phase to drive the two 6550s PROBABLE; the exact schematic of that inverter is ND until the board is traced. The transformers are of varnished EI laminations and sit in a chassis with a metal partition OBSERVED. New terms? Start with the basic guide.

What has been identified in the photos: three EPCOS/TDK B43645 capacitors of 610 µF / 450 V OBS (1,830 µF in total CONDITIONAL CALCULATION only if in parallel); a vertical board (PCB-3) with 4 blue components OBS, probably relays PROB, of function ND; a 10W10KJ resistor of 10 kΩ / 10 W OBS; two blue trimmers (adjustment potentiometers) on the signal board PCB-S, one at each end; and 14 MKP CEC C2N capacitors, 2 PIO K40Y-9 and 5 snubber-type MKP capacitors (they damp voltage spikes) OBS.

Identified with less certainty: the Hongfa HF115F relay PROB, the NE5532 operational amplifier PROB and the TL431 voltage reference PROB. There is no confirmation of the NE555 timer ND. There are two visible 8-pin chips (DIP-8) on the power supply board PCB-P. Two 0.2–0.6 mH toroids (Rev.A board, visual estimate) do not appear in the photographed current unit.

What is NOT demonstrated: the real B+ voltage (the valve supply), the effective capacity of the capacitor bank, the bias mode of the 6550s (how their quiescent current is set), the global negative feedback (NFB), the %UL, the power at 1 % distortion (P@1 %), the output impedance (Zout) or the damping factor (DF). Nor is it known what the blue trimmers, the 10W10KJ resistor or PCB-3 are for, where the low voltage for the NE5532 (if confirmed) would come from, or how the start-up timing circuit works.

Probable topological kinship: the P52 follows the same valve family as the Mullard 5-20 (design of c. 1956, published in 1959), doubled to stereo and with solid-state rectification. Confirmation requires the manufacturer's schematic or measurements (§3.2). "Probable kinship" is not "replica".

General description

The Acronos P52 is an ultralinear push-pull amplifier of 40 W nominal, with 6550 valves in the power stage. In the interior photos at least three boards can be distinguished: the main signal board (PCB-S), very dense with components; the power supply and control board (PCB-P), located behind; and an independent vertical board (PCB-3) with four blue components, probably relays. In the rear right corner there is also an auxiliary magnetic element of unknown function and a longitudinal metal partition separating the power and control area from the signal area.

Where the "40 W nominal" figure comes from: it is a commercial figure from the manufacturer JC Valvular (P52 spec sheet: 40 W PP UL with 6550). It does not indicate at what distortion, frequency or load it is measured, so it is classified as C1/D — nominal power without a distortion criterion, not equivalent to P@1 %. In other words: it is a catalogue figure, not a comparable measurement.

Provisional conclusion

The design appears more elaborate than a minimum valve schematic, especially in the control and auxiliary supply section, which is more complex than it appears at first glance. But none of the data that most influence real behaviour (B+, 6550 bias, NFB, %UL, Zout, P@1 %) is demonstrated. That is why the report remains provisional reverse engineering: it describes what is seen and clearly marks what remains to be measured.

1. Scope, method and limitations

1.1 Source of the evidence

• Visual inspection of photographs from the JC Valvular manufacturer's video: Amplificador ACRONOS P52 Una obra de arte. Cómo se fabricó.
• Additional interior photos (2026-10) provided by the owner of the unit. They allow reading the silkscreen of the three EPCOS B43645 capacitors, identifying the Hongfa HF115F relay and confirming the blue Bourns trimmers on PCB-S. They also show the vertical PCB-3, the auxiliary magnetic element and the longitudinal metal partition.
• Comparison with manufacturer catalogues: TDK/EPCOS, Bourns, TI (TL431, NE5532), JJ Electronic (valves), Hongfa, Schaffner, Würth, TAD (ECC803S), Mullard (5-20, EF86, EL34, ECC83), Genalex/New Sensor (KT77, KT88), RCA/GE (6SN7, 6CG7, 12BH7, 5687).
• Owner's reference equipment (the speaker with which it is listened to): Crossover A v2.4 MR-A-01 (6 Ω nom / Zmin 4.9 Ω / 89 dB/W/m), listening at 3 m.

1.2 Catalogue context — JC Valvular range

To see where the P52 fits, this table orders the manufacturer's models and their topology (how the circuit is assembled):

Model	Topology	Declared power	Category
Acronos M1 / RS-1	Pure Single-Ended Class A (KT120/KT150)	18–20 W	Pure valve, low power
Acronos P52	Pure valve Push-Pull Ultralinear (4× 6550)	40 W nominal	Pure valve, medium power. Manufacturer's ceiling in pure valve topology
Acronos M2	Hybrid (valve preamp + transistor output stage)	60 W (field data, 2026-10)	Hybrid, high power, high native DF without dependence on the output transformer

Note on the M2's power (PROBABLE / field data): distributor spec sheets give 50 W (High End Korner, 2024) or 40 W (Amigoshifi, 2021, possibly a prototype), but the 2026-10 field data places the current unit at 60 W per channel. The M2 is not comparable to the P52 due to its topology. Among the manufacturer's pure valve topology models, the P52 is the highest power.

1.3 Roles glossary

Role	Who	What they contribute
Reviewer	Report author (maty @nauscopio)	Reference system, calculations, report structure
Reference system owner	Owner of the Crossover A v2.4 MR-A-01	Sensitivity, Zmin, listening conditions (89 dB/W/m, 3 m)
P52 unit owner	Person with physical access to an Acronos P52 unit	Additional interior images (2026-10) revealing PCB-3, auxiliary magnetic element and metal partitioning
Manufacturer	JC Valvular (Juan Calatayud)	Manufacturing video, 40 W commercial spec sheet, design and components
Weight informant	Blog reader	Personal communication: P52 ≈ 40 kg, unverified

No measurements on the equipment. It has not been confirmed that there is access to a real P52 to measure it with instruments. That is why the report relies only on photographs, catalogues and public documentation.

2. General architecture — overview

2.1 Provisional physical architecture

The interior photos allow drawing how the P52 is organised internally: at least three boards and a metal partition separating two zones. The following diagram is a location map, not an electrical circuit.

                       P52 (provisional physical architecture)
                                  │
          ┌───────────────────────┴───────────────────────┐
          │                                               │
    POWER / HV ZONE                              CONTROL / AUXILIARY ZONE
          │                                               │
       PCB-P                                          PCB-3 (vertical)
          │                                               │
   KBU1010 · C-HV EPCOS ×3                        4 blue components
   relays · R power 10W10KJ                       probably relays
   TO-220 semiconductors                          function ND
   MOSFET + TL431 block (PROB)
   two DIP-8 (function ND)
          │                                               │
          └───────────────────────┬───────────────────────┘
                                  │
                          PCB-S (front)
                    high density of passive components
                    and semiconductors · function by block ND
                                  │
                          valve stages
                                  │
                          OPT · speaker outputs

   Longitudinal metal partition physically separates both zones.
   Auxiliary magnetic element in rear right corner (function ND).
   Twisted wiring bundled in several runs.

   PROVISIONAL PHYSICAL ARCHITECTURE DIAGRAM.
   Does not represent a traced electrical schematic.

2.2 Functional block diagram with evidence levels

230 V MAINS
 │
 +-- [EMI cond. obs.], NTC, RF-EMI filter (no L/N choke identified)
 │
 +-- Main power transformer
       │
       ↓
  KBU1010 RECT
       │
       ↓
 C HV EPCOS B43645 610/450 ×3  --- OBSERVED (verified silkscreen)
                                   1,830 µF CONDITIONAL CALCULATION if in parallel
                                   Service life >5,000 h @ 105 °C (TDK, REPORTED-EXT)
       │
       ↓
    L1 toroidal Rev.A   --- FUNCTION ND (absent in the current photographed unit)
       │
       ↓
 C / HV branch           --- PROBABLE
    +-------------------+
    │                   │
 power                 auxiliary
 6550 / OPT            MOSFET + TL431 (PROB)
                        blue Bourns
                        FUNCTION ND
                        (application not traced)

 10W10KJ (10 kΩ / 10 W) --- OBSERVED
   Does not seem compatible with a permanent B+ bleeder at 430 V under that hypothesis.
   Function ND: bypassed inrush limiter, auxiliary bleeder, divider or load.

 Hongfa HF115F relay   --- PROBABLE (probable soft-start/inrush)
 Songle/Finder auxiliary relays --- PROBABLE

 PCB-3 vertical with 4 blue components --- BOARD OBSERVED
   4 components --- PROBABLE relays · FUNCTION ND

 NE5532 --- PROBABLE (function ND). Two DIP-8 visible.
 NE555  --- ND / pending visual confirmation

 Auxiliary magnetic element (rear right corner) --- ND
   Open hypotheses: auxiliary transformer, mains module, other.

 Longitudinal metal partition --- OBSERVED
   EMI/isolation intent --- PROBABLE
   Real effectiveness --- ND

 Blue Bourns trimmers on PCB-S --- TWO OBSERVED, one at each end
   Total number and function ND. Favours one per channel over four bias.

 6550 with no clearly associated trimmer on PCB-S --- OBSERVED
   Exact bias mode, quartet matching --- ND

 CLASS OF OPERATION: AB probable (subtypes A, AB1, AB2 not observed)
   AB1 is the most coherent reading with the observed topology (§5.2).
   Indirect argumentation; without schematic the subtype cannot be closed.

2.3 Global topology

The Acronos P52 is a stereo ultralinear push-pull amplifier: two 6550 valves per channel and 40 W nominal per channel. The input stage (small signal) uses a 6Ж32П (EF86), a pentode, and an ECC803S (double triode from JJ Electronic) to provide gain and invert the signal phase. The chassis houses three varnished EI lamination transformers, with a metal partition between the power supply and PCB-S OBSERVED. The valve combination functionally recalls that of the Mullard 5-20 (1959, design c. 1956), doubled to stereo and with the substitution of EL34s by 6550s (§3.2).

2.4 Functional blocks

Block	Contents	Section	Global status
AC input and RF/EMI	Capacitors after IEC, NTC, PE, no L/N choke identified	§6	PROBABLE general structure; X/Y class ND
PCB-P — Power supply and control	KBU1010, EPCOS B43645 HV, 10W10KJ, MOSFET + TL431 block (PROB), Hongfa relay (PROB), auxiliary relays, blue and black Bourns, two DIP-8 (NE5532 PROB, second ND)	§4	PROBABLE by components; exact functions ND
PCB-S — Small signal	EF86, ECC803S, 14 MKP CEC C2N, 2 PIO, two visible blue Bourns 3296Y trimmers, precision resistors, signal diodes	§3	PROBABLE layout; function by block ND
PCB-3 — Independent vertical	4 blue components probably relays + transistors/diodes + electrolytics + resistors + wiring	§4.7	OBSERVED the board · PROBABLE relays · ND function
Power stage	4× 6550, EI OPT, bias	§5	Class AB probable · AB1 subtype probable (§5.2). Bias mode, NFB, %UL ND

2.5 What determines the real behaviour and why it remains unknown (ND)

These are the data that would decide how the P52 sounds and behaves. None can be deduced from photos alone:

• Real B+ voltage and its behaviour under load.
• Bias mode of the 6550s: how their quiescent current is set.
• Operating subclass within AB (AB1 or AB2). It is an indirect inference, not an observed data point.
• Global feedback (NFB): whether it exists and how much there is (it returns part of the output to the input to reduce distortion).
• Ultralinear percentage (%UL): at what height of the output transformer (OPT) the taps are connected.
• P@1 % (power at 1 % distortion) and distortion-versus-power curve (THD+N).
• Zout (output impedance) and damping factor (DF): how much control the amp exercises over the speaker.
• Effective capacity of the HV bank.
• Matching of the 6550 quartet (that the four valves are similar).
• B+ supply architecture: number and separation of branches (§4.6).
• Exact function of the blue trimmers on PCB-S: total number, destination and adjustment method.
• Exact function of the 10W10KJ: does not seem compatible with a permanent B+ discharge at 430 V under that hypothesis; other alternatives remain open (§4.3).
• Function of PCB-3 and of the four blue components (§4.7).
• Origin of the low-voltage rail for the control electronics (if NE5532 is confirmed).
• Topology of the start-up timing circuit (specific component ND).
• Driver dynamic reserve: whether the ECC803S delivers sufficient voltage and current margin to the 6550 grid under worst-case conditions (high band, maximum excursion). It is a question to verify, not an established defect (§3.3).

In summary: this report describes what is seen and proposes reasonable hypotheses, but does not close any of the above data points. To do so, measurement will be needed.

3. Small-signal board (PCB-S)

3.1 Observed components

PCB-S is a JC Valvular FR4 board. It has four noval holes (9-pin valve sockets) for the EF86 and the two ECC803S triodes, and a fifth circular hole at the right end, with no apparent socket (function ND). It carries fourteen MKP CEC C2N capacitors (630-1000 V), two paper-in-oil (PIO) capacitors and five MKP snubber capacitors. The resistors are green MOX 10K28 of 2 W and white ones of 1 W. There are two blue Bourns 3296Y trimmers (100 ppm/°C, 25 turns), one at each end; their total number and electrical function are ND.

Note on the count. The figure "14 MKP + 5 snubber" comes from earlier photos. It does not contradict the current photo, but a single oblique view is not enough to confirm it. For a definitive count a photo from above, perpendicular to the board, is needed.

Note on the CEC C2N. There are many film capacitors distributed symmetrically. They may serve several purposes: coupling stages, bypassing signal, local decoupling, compensation, damping spikes (snubber) or forming frequency networks. Assigning each C2N to a specific function without tracing the board would be a methodological error.

The large number of components and the symmetrical arrangement of the two channels are OBSERVED and are consistent with careful construction. In the photos no trimmer is clearly associated with each 6550, so bias remains unresolved (§3.4).

3.2 Probable topology and Mullard 5-20 lineage

The ECC803S is a double triode (A.F. twin triode from JJ Electronic): two independent triodes within the same noval envelope. With an EF86 and an ECC803S per channel, there are three common ways to arrange them:

1. EF86 (input) → ECC803S as LTP (long-tailed pair) with both triodes. It is the classic combination: the EF86 provides voltage gain and the ECC803S pair inverts the phase with high gain. MAIN HYPOTHESIS, not observed in the photos.
2. EF86 (input) → one triode of the ECC803S as a gain stage → the other as cathodyne or driver (simple phase inverter). ALTERNATIVE HYPOTHESIS.
3. EF86 (input) → one triode of the ECC803S as a gain stage → the other as an asymmetric driver. It is unusual. MINOR HYPOTHESIS.

Probable reading (common to the three hypotheses):

RCA → EF86 (pentode, Ra 100 kΩ)
     → C → ECC803S (double triode)
            [hypothesis 1: LTP with both triodes]
            [hypothesis 2: gain + cathodyne/driver]
            [hypothesis 3: gain + asymmetric driver]
     → 2× C → 6550 grids

Gain budget (estimate, not measurement). Suppose that with 1 V_rms at the input maximum power is reached and that the 6550 grids need about 30–50 V_peak (≈21–35 V_rms). The total gain needed would be 21–35×, that is, 26.5–31.0 dB. This fits with an EF86 (typical stage gain ≈100–150×, see note) followed by an ECC803S LTP (≈10–20×), with a global feedback reducing the final gain. The calculation is compatible with hypothesis 1 (EF86 → LTP), but does not prove it. PROBABLE.

Note on the EF86 gain. A pentode does not publish a μ comparable to that of a triode: its real μ (gm × ra) would be on the order of thousands. The "185" value that circulates in the literature is the maximum stage gain — gm (1.85 mA/V) times the plate load (≈100 kΩ) —, not a datasheet parameter. The real stage gain in a specific design is lower (typically 100–150), because the effective plate load falls due to the internal resistance and the load of the next stage. When this report cites "gain ≈100–200×" for the EF86, it refers to that order of magnitude of the stage, not to a catalogue μ.

What is the Mullard 5-20?

The Mullard 5-20 is a reference push-pull amplifier design published by Mullard Ltd in its booklet «Circuits for Audio Amplifiers» (the design has been circulating since c. 1956; revised edition in 1959, reprinted in 1962). The name is descriptive: 5 valves (EF86, ECC83, 2× EL34, GZ34) and 20 watts of output. The 5-20 documentation went through several revisions; it is advisable not to treat "Mullard 5-20" as a single, invariable schematic.

It was the reference amplifier of the British high-fidelity school in the 1950s. It served as a template for numerous later commercial amplifiers and influenced the British school of which Leak, Rogers and Radford were part. It became the canonical standard of British push-pull.

Historical context. In the late 1950s domestic high fidelity was in full expansion. Mullard published the 5-20 as a didactic example so that amateur builders and small manufacturers could assemble an amplifier of contrasted quality with components and valves of the brand. The success of the design turned it into a de facto standard.

Analogy with the Mullard 5-20 (PROBABLE / ANALOGY). The combination of EF86 (low-microphony input pentode) + ECC803S (high-gain double triode, μ=100 and pins equivalent to the ECC83) as LTP phase inverter coincides with the architecture of the classic Mullard 5-20, published by Mullard Ltd in Circuits for Audio Amplifiers (1959, design circulating since c. 1956; reprinted 1962). Its advantage is the high open-loop gain, which allows global NFB to be applied without losing sensitivity. In the P52, the original EL34s are replaced by a 6550 quartet. Status: PROBABLE / analogy. It is the main indirect argument available to formulate the small-signal stage hypothesis, but it does not constitute direct evidence that the P52 reproduces the Mullard circuit. It remains to trace the tracks and confirm the %UL and the NFB depth in dB.

Mullard 5-20 → Acronos P52 correspondence

Stage	Mullard 5-20 (original, mono)	Acronos P52 (observed, stereo)	Correspondence
Input	1× EF86 (low-noise pentode)	2× EF86 / 6Ж32П (one per channel)	Functionally identical
Driver / Inverter	1× ECC83 (double triode, μ ≈ 100)	2× ECC803S (one per channel)	Functional equivalent (ECC803S is a long-plate variant of the ECC83 family)
Power	2× EL34 in push-pull class AB1	4× 6550 (2 per channel)	Similar topology, different valve
Rectification	1× GZ34 (valve)	KBU1010 (solid state) OBSERVED	Different (solid-state rectification)
Feedback	Global NFB — see note	ND (not measured)	No automatic transfer to the P52
Configuration	Mono	Stereo	Doubled
Nominal power	20 W over 7.5–15 Ω	40 W per channel DECLARED by the manufacturer	Doubling to stereo does not imply 40 W per channel: that depends on the valves, the supply, the bias and the transformers
THD at 20 W	≈ 0.1 % at 1 kHz (Mullard, Circuits for Audio Amplifiers, 1959)	ND	—
B+	≈ 400 V for EL34	ND	Probably higher for 6550

On the Mullard 5-20's feedback. Different figures circulate in the literature for the 5-20's global NFB: some sources cite 20–26 dB, others approximately 30 dB. It has not been confirmed that they correspond to the same revision of the circuit, so this report does not fix a single figure. More importantly: the Mullard 5-20's NFB does not automatically transfer to the P52. The P52's global feedback remains ND until traced or measured.

What this correspondence implies (all consequences are PROBABLE, not DECLARED):

1. It reinforces the AB1 inference. The Mullard 5-20 operated in AB1. If the P52 follows a close schematic, it is coherent that it also does, although it cannot be stated as certain.
2. It explains why there is no cathode follower or interstage transformer. The 5-20 does not use them; the P52 neither. They are consistent with the British school of the time.
3. It explains the choice of EF86. It is not arbitrary: it is the input valve of the reference design.
4. It explains the choice of ECC803S. It is the premium variant of the 5-20's ECC83, available today as a high-end option.

Limitations of the analogy. The correspondence with the Mullard 5-20 is useful but does not replace the schematic. The following remain ND: the exact topology of the phase inverter (LTP, cathodyne, Schmitt or other), the exact component values, the real B+, the NFB depth and the specific modifications with respect to the original (modern capacitors, SS rectification, bias adjustment with trimmers, etc.). Nor can it be overlooked that doubling the 5-20 to stereo would give two 20 W channels, not one 40 W channel: the P52's 40 W per channel is DECLARED by the manufacturer and may be consistent with the change from EL34 to 6550, but that relationship is not automatic.

How to detect global NFB visually (practical rule for future macro photos). Look for a medium-high value resistor (typically between 4.7 kΩ and 100 kΩ) connected from the output transformer secondary winding (8 Ω tap) to the EF86 cathode or to a point on the driver. If it exists, there is global NFB. If there is also a capacitor in series with that resistor, the NFB has a DC cutoff and avoids shifts in the operating point. The absence of that resistor does not rule out local NFB (for example, in the driver cathode).

Indicative model — not an extrapolation of the P52. The following table relates global NFB depth with typical orders of magnitude of Zout and DF, as a pedagogical reference. The real values depend on open-loop Zout, frequency, transformer, load, stability and phase. It does not allow the P52's Zout to be estimated from a supposed NFB.

Configuration (indicative)	Typical Zout (order of magnitude)	Typical DF over 6 Ω
No global NFB	2 to 4 Ω	≈ 1.5 – 3
6 dB global NFB	1 to 2 Ω (÷2)	≈ 3 – 6
12 dB global NFB	0.5 to 1 Ω (÷4)	≈ 6 – 12
20 dB global NFB	0.2 to 0.4 Ω (÷10)	≈ 15 – 30

The values are indicative and are not an extrapolation of the P52. The P52's effective NFB is ND.

6550 bias — open hypotheses. The 2026-10 photo shows two blue trimmers on PCB-S, one at each end. That points more to "one per channel" (signal stage adjustment or inverter DC balance) than to "four trimmers, one for each 6550". But the underside of PCB-S is not photographed, so it is not ruled out that there are two other bias trimmers. The correct status is:

Hypothesis	How it works	Evidence
Two trimmers, one per channel (signal stage adjustment or inverter DC balance)	Two blue trimmers, one at each end of PCB-S, adjust the symmetry of each channel.	FAVOURED BY CURRENT OBSERVATION — two visible trimmers, one at each end
Individual bias with 4 trimmers	Four trimmers individually adjust the bias of each 6550. Two could be on the underside.	HYPOTHESIS — not ruled out; requires underside inspection
Fixed bias by resistive divider	Fixed negative voltage by resistive divider (the blue trimmers would have another function).	PROBABLE; mode ND.
Cathode bias	Power cathode resistors with decoupling.	ND — not identified.
Bias from auxiliary block	Adjustment from PCB-P (MOSFET + TL431 + blue Bourns).	HYPOTHESIS — requires tracing.

Exact bias mode: NOT DEMONSTRATED. It is necessary to see the underside of the board or ask the manufacturer. Question P14 of the Appendix.

3.3 ECC803S driver: current capability to verify

The ECC803S is a voltage-gain valve (μ = 100, Ri = 62.5 kΩ). That nature is a datasheet data point, not a limitation. What is worth making clear is that a voltage valve is not the same as a current valve, and that the question "does the ECC803S deliver enough current to the 6550 grid?" cannot be answered from the photos. This section orders why that question matters, what can be estimated and what cannot be concluded.

Why the driver current can matter

Charging the input capacitance of a power valve requires current. The peak current needed to charge a capacitance C at a frequency f with a peak voltage V is:

I_peak = 2π · f · C · V

With C ≈ 100 pF (conservative hypothesis: includes Miller effect, grid capacitance and parasitics), f = 20 kHz and V = 20 V peak: I_peak ≈ 0.25 mA. It is a peak, instantaneous figure, not a continuous current.

The JJ 6550 datasheets circulating among distributors give considerably lower capacitances: on the order of 14 pF grid and 0.85 pF grid-plate. The Miller effect multiplies the latter by the gain of the power stage and the resulting value depends on the operating point. The 100 pF figure used here is an order-of-magnitude ceiling, not a measurement of the P52.

If the P52 operates in AB1 —the most coherent with the observed topology (§5.2)—, the control grid does not go into positive conduction, and there is no DC grid current. The driver only has to deliver voltage and a small peak current to charge parasitic capacitances. In that scenario, the question of whether the ECC803S is "enough" is a matter of margin, not of insufficiency.

Current capability comparison

Valve	μ	Ri	Icat, max (datasheet)	Socket	Natural role
ECC803S (current)	100	62.5 kΩ	8 mA (TAD)	Noval (9 pins)	Voltage gain
12AX7 / ECC83	100	62.5 kΩ	8–9 mA	Noval	Voltage gain
12AT7 / ECC81	60	~15 kΩ	~20 mA	Noval	Moderate-gain driver
12AU7 / ECC82	20	~7.7 kΩ	~20 mA	Noval	Current driver
6SN7	20	~7.7 kΩ	20 mA	Octal (8 pins)	Robust driver, excellent linearity
6CG7	20	~7.7 kΩ	20 mA	Noval (9 pins)	Identical to 6SN7 in noval format
12BH7	17	~7 kΩ	~30 mA	Noval	Maximum-current driver in noval
5687	17	~1 kΩ	~35 mA	Noval (different pinout)	Minimum impedance

Note on the ECC803S data point. The TAD datasheet specifies a maximum cathode current of 8 mA, not a "peak current available for AB2". The two are not equivalent. The typical operating current of the stage (quiescent current) is much lower, on the order of 0.5–1 mA in a voltage design. This data point does not serve to demonstrate that a stage cannot work correctly; it documents the valve's profile.

V6.6 reformulation — no demonstrated limitation. Earlier versions of this report described the ECC803S as "the weakest link in the small-signal chain". That statement is not supported by the current evidence: the calculation of I_peak = 0.25 mA is well done, but it does not demonstrate insufficiency. In probable AB1 there is no DC grid current, and the real input capacitance of modern 6550s is lower than the 100 pF used as a ceiling. The correct conclusion is more cautious:

• The ECC803S is the element whose current delivery capability deserves verification if one wants to evaluate the driver's dynamic reserve.
• It cannot be identified as a performance limitation without measuring signal excursion, high-frequency distortion, response over complex load and grid current.
• If the characterization showed an anomalous increase in THD with frequency or an inability to maintain the required excursion, then it would make sense to study lower-impedance drivers (§3.5).

3.4 Probable function of the blue Bourns on PCB-S

Two blue Bourns 3296Y trimmers are seen (100 ppm/°C, 25 turns, 0.5 W), one at each end of the board, which fits with one per channel. What exactly they adjust (the signal stage, the inverter DC balance or something else) remains ND without tracing the board. If there were two others on the underside, individual bias per valve would be possible again.

3.5 Driver substitutions — proposal pending measurement

Current status: do not modify yet. There is no measurement demonstrating that the P52's ECC803S falls short. Before considering any substitution, Phase 0 of characterization must provide data (THD+N vs frequency and power, response over complex load, stability with square wave). What follows is a conditional technical proposal, not a justified improvement.

If the characterization showed a real driver limitation —for example, anomalous increase in THD with frequency, inability to maintain the required excursion over complex load or unstable behaviour under NFB— then it would make sense to study lower-impedance valves with higher current capability. This section catalogues the candidates and the technical conditions so that evaluation, if it comes, is orderly.

Technical conditions to verify before any substitution

• Socket and pinout compatibility. The ECC803S is a variant of the 12AX7: noval socket with dual filament (pins 4/5/9). The 12AT7, 12AU7 and 12BH7 share that pinout and filament configuration: they fit physically without adaptation. The real risk is not in the pins, but in the operating point (bias, cathode resistor and plate load will be displaced). The 6CG7 matches the signal pins but has a single 6.3 V filament between pins 4 and 5 with screen on 9: in a socket wired for 12AX7 it receives no filament and requires rewiring. The 6SN7 (octal) and the 5687 (noval with a different pinout) require socket adaptation.
• Total loop gain. Going from an ECC803S (μ = 100) to a 6SN7 (μ = 20) or a 12BH7 (μ = 17) reduces the stage gain by some 14.0 to 15.4 dB. The EF86 at the input (typical stage gain 100–150×) would largely compensate for that loss, but if the P52 uses global NFB, that NFB might need readjustment.
• Operating point. Each valve has a different transconductance curve. The change alters the quiescent current, plate dissipation and the symmetry of the push-pull pair. In the P52, with two visible trimmers, it is probably possible to readjust the bias, but the adjustment range is ND.
• Subsequent verification. Square wave at 1 kHz and 10 kHz, response at 20 kHz over complex load, THD+N vs power, stability with NFB.

Technical warning. None of the proposed substitutions is plug-and-play. All require verifying pinout, plate voltage, bias range and heater consumption. Any intervention must be done by a technician with access to the P52 schematic (which is ND).

Candidates ordered by practical suitability

Candidate	Type	Notes
6CG7	The most balanced	Electrically identical to 6SN7 but in noval format. Requires verifying the filament wiring (single 6.3 V vs dual 12AX7). Used by Conrad-Johnson Premier Eleven.
6SN7	Technical reference	Excellent linearity and mechanical robustness. Requires socket adapter (noval → octal).
12BH7	Maximum performance in noval	~30 mA available. 12AX7-compatible pinout. Used by VTL S-200 (with 6550). Requires readjusting bias and verifying filament consumption.
5687	Minimum impedance	~1 kΩ. The best for handling parasitic capacitances. Noval pinout different from 12AX7. Used by Sonic Frontiers Power 2 (8× 6550).
12AU7 / ECC82	Intermediate	~20 mA, 12AX7-compatible pinout. μ = 20. Requires verifying remaining gain with the EF86.

Conservative alternatives (same socket, no topology change)

If one decided to improve the driver quality without changing family, there are alternatives within the long-plate 12AX7:

• TAD ECC803S Premium Selected — long plate, premium selection with noise, microphony and shock tests.
• Mullard Reissue 12AX7 — long plate, warm and three-dimensional profile.
• Sovtek 12AX7LPS — long plate, spiral heater to reduce hum.
• Genalex Gold Lion 12AX7 — gold-plated pins, very low microphony.

These valves do not increase the driver's current capability (they are still 12AX7s). They may bring improvements in noise, microphony and consistency without altering the topology or socket, but the character descriptions ("warm", "crystalline") come from audiophile folklore, not from measurements; they are cited only as commercial orientation.

Classification. The whole of this section is a TECHNICAL PROPOSAL — PENDING MEASUREMENT. It is not a justified improvement. It would only be activated if the electrical characterization demonstrated a real driver limitation, with a measurable delta.

3.6 Cross-reference

The bias mode is discussed in §3.2 and §5.3. The trimmers are catalogued in §4.5 and in Part II §7. The B+ supply is in §4.6 and PCB-3 in §4.7. The class of operation is developed in §5.2. The driver substitutions, in §3.5 (proposal pending measurement).

PCB-S images

PCB-S — MKP CEC C2N and PIO capacitors, two visible blue Bourns trimmers (one at each end), precision resistors, signal diodes. Four noval holes with wiring and a fifth hole without an apparent socket (function ND).

Coupling crop Bourns between sockets, MKP and 2 PIO

Front JJ ECC803S + 6Ж32П (EF86) + Tung-Sol 6550

4. Power supply and control board (PCB-P)

4.1 Observed components

PCB-P macro — KBU1010, input MKP, EPCOS B43645 610/450 (verified silkscreen ×3), 10W10KJ, TO-220 heatsinks, MOSFET+TL431 block (PROB), Hongfa HF115F relay (PROB), auxiliary relays, blue and black Bourns, two DIP-8 (NE5532 PROB, second ND).

On PCB-P the following are clearly visible: the KBU1010 rectifier (10 A), input MKP capacitors, the 10W10KJ resistor, three large EPCOS/TDK B43645 (silkscreen B43645-S7617-M1 read on all three), a 100 µF / 450 V capacitor and TO-220 heatsinks. With less certainty, a regulation or control block with TL431 PROB and its trimmer, the Hongfa HF115F relay PROB and Songle and Finder auxiliary relays (function ND) are identified. There are also two DIP-8 chips whose marking is not legible in the photos: one could be a NE5532 PROB and the other is ND. The NE555 is not confirmed.

The photos also show diodes and transistors in the right zone, an encapsulated magnetic element in the rear right corner (function ND, see §4.7), several auxiliary electrolytics blue and black distributed on the board and a fuse holder next to the mains input. On the rear wall gold terminals with nuts and blue, red and green twisted wiring are visible; they could be speaker outputs or inputs, but their function is ND.

Evidence rule. If a component is identified only by its shape or a partial reading of the marking, it is classified as PROBABLE. That is why: NE5532 → PROBABLE; NE555 → ND; TL431 → PROBABLE, function ND. Thus a doubtful reading does not become the basis for deducing a topology.

Interior — metal partition separating supply / PCB-S, three varnished EI transformers, twisted bundles. Vertical PCB-3 can be seen in the rear left corner.

4.2 HV bank: EPCOS B43645 610 µF / 450 V — verified silkscreen

The silkscreen is clearly read on all three units of the high-voltage (HV) bank:

Observed marking (×3): B43645-S7617-M1 · 450V · 610µF

The three electrolytics are together, forming a single HV bank on PCB-P. That supports them being connected in parallel, but does not prove it: the tracks on the underside must be traced to confirm.

Epistemological chain on the HV bank — separation of levels

Statement	Status
Three HV electrolytics physically grouped forming a bank	OBSERVED
Printed code B43645-S7617-M1 visible on all three units	OBSERVED on the silkscreen
610 µF / 450 V per unit, as printed	OBSERVED on the silkscreen
Exact correspondence of the code with the TDK datasheet (ordering codes table)	PENDING VERIFICATION — not yet cross-checked against the full ordering code
Series identified as B43645 (not B43655, a closely related family)	PROBABLE from the visible code · requires confirmation
Parallel connection of the three	PROBABLE — simplest reading, coherent with grouped bank · ND without tracing
1,830 µF effective	CONDITIONAL CALCULATION under the parallel connection hypothesis
Real equivalent capacity of the bank	ND until the solder side is inspected
Nominal service life >5,000 h at 105 °C	Data for the B43645 series according to TDK documentation REPORTED-EXT — not measured on this unit
Arrhenius extrapolation to 65 °C (hypothesis)	≈ 80,000 h CALCULATED under the hypothesis of operation within the ripple rating
Arrhenius extrapolation to 55 °C (hypothesis)	≈ 160,000 h CALCULATED under the hypothesis of operation within the ripple rating

Ripple of a filter capacitor: Vpp ≈ I / (f · C), with f = 100 Hz (the ripple frequency after rectifying the 50 Hz mains). Example: 610 µF with 200 mA → Vpp ≈ 3.28 Vpp; with 1,830 µF and 200 mA → ≈ 1.09 Vpp. Note: the formula exaggerates the real ripple, because it assumes constant current throughout the period.

450 V is the capacitor's maximum rated voltage; it does not prove that the P52's B+ is 450 V. A 450 V capacitor can operate at lower voltage, and in real conditions there are margins, transient overvoltages and tolerances. The P52's real B+ is ND. As its own guideline, this report assumes that the B+ should not exceed 85–90 % of that figure (≤ 382–405 V); it is a longevity criterion of the report, not a manufacturer's standard.

Methodological note on the B43645 / B43655 family.

• The nominal service life of the B43645 series according to current TDK documentation is >5,000 h at 105 °C. TDK identifies the 8,000 h series as other families.
• TDK has closely related families B43645 and B43655, with non-identical specifications. Until the full inscription is confirmed against the ordering codes table, the exact identification of the P52 component must be classified as PROBABLE, not as CONFIRMED.
• The Arrhenius extrapolations (≈80,000 h at 65 °C, ≈160,000 h at 55 °C) are order-of-magnitude hypotheses, not guarantees, and presuppose operation within the ripple current rating.
• Factors preventing a closed quantitative comparison with other capacitors: real internal chassis temperature, effective ripple current and quality of the exact series. None are measurable with current evidence.

4.3 Toroids and 10W10KJ resistor

Rev.A (earlier board revision): two brown copper toroids of 15 mm, with fine wire of 0.3–0.5 mm (2 wires each), in series between the KBU1010 and the EPCOS bank. Rev.B (the board revision in the currently photographed unit): those toroids are no longer in their historical position. They are not filament transformers nor a 1 H choke.

On the "0.2–0.6 mH" value. It is a VISUAL ESTIMATE; it has never been measured. The correct statement is: Rev.A: two toroidal elements observed in the past. Their inductance, resistance, saturation current and function are ND. In the currently photographed unit they are not seen in their historical position.

Revision / Element	Observed	Value / Function	Status
Rev.A toroids	2× copper 15 mm series KBU1010→EPCOS (in historical images)	Toroidal magnetic element. L, DCR, I_sat and function ND.	OBSERVED in Rev.A · value ND
Rev.B (current unit)	The toroids are not observed in their historical position	Confirmed absence in the photographed unit	OBSERVED
10W10KJ	Silkscreen 10W10KJ	10 kΩ / 10 W. Does not seem compatible with a permanent B+ bleeder at 430 V. Exact function ND.	OBSERVED the value · ND the function
Blue component PCB-P	Blue Bourns + blue electrolytics + SONGLE SRD	Negative bias relay (hypothesis)	OBSERVED the components · ND the exact function

Thermal note on the 10W10KJ — conditional closure.

• Direct calculation: with B+ = 430 V, the permanent dissipation between B+ and ground would be P = V² / R = 430² / 10,000 ≈ 18.5 W, above its 10 W rating. Maximum continuous voltage admissible for 10 W over 10 kΩ: V_max = √(10·10,000) ≈ 316 V.
• Logical closure: if it were directly subjected permanently to 430 V, it would dissipate ≈18.5 W against a 10 W rating. That configuration does not seem compatible with permanent operation at 430 V under that hypothesis.
• Quiescent current if it were in B+: I = 430 / 10,000 = 43 mA.

What CANNOT be deduced:

• That the 10W10KJ is a pre-charge resistor. That is PROBABLE, not demonstrated.
• That the bank it acts on is 1,830 µF.
• That the relay switches at the end of the RC transient of the complete bank.
• That there is a dominant time constant τ = R·C. If the above hypotheses held, τ = 10,000 · 1.83·10⁻³ ≈ 18.3 s, but that calculation is conditional.
• That the resistor really sees 430 V. If it operates on an auxiliary branch at 200–300 V, it would dissipate 4–9 W and be within its rating.

Open alternatives for the function of the 10W10KJ (all ND):

1. Inrush limiter bypassed by the Hongfa relay.
2. Bleeder of the regulated auxiliary branch (if that branch operates at 200–300 V, it would dissipate 4–9 W, within rating).
3. Voltage divider or regulator load.
4. Other function not considered.

Status: ND until traced. Question P15 of the Appendix.

4.4 Auxiliary regulation/control block: MOSFET + TL431 + Bourns — PROBABLE

On PCB-P there are TO-220 heatsinks, a TL431 PROB, associated blue Bourns trimmers and auxiliary electrolytics. That set fits several possible functions:

• Regulate the negative bias voltage of the 6550s.
• Provide an auxiliary low-voltage supply (Vaux) for the control electronics.
• Monitor and protect (compare thresholds, detect overcurrent or DC at the output).
• Provide a reference voltage to another circuit.

Status: the components (MOSFET/TO-220, TL431 if confirmed, trimmers, electrolytics) are OBSERVED. That they form an auxiliary regulation or control block is PROBABLE. Their exact application and the voltage they regulate are ND. Without tracing the board it cannot be stated that it is a "bias regulator".

On the DIP-8 and the NE5532. If the chip identified as NE5532 is confirmed, it cannot be powered from the 400 V of B+, because TI specifies a total supply of 10–30 V. Therefore a low-voltage rail must exist. It could come from an auxiliary winding of the main transformer, from a small separate transformer, from another winding or from a regulator fed from an intermediate rail. The existence of the rail is PROBABLE and its origin is ND. The second DIP-8, of function ND, is not enough to deduce a specific timer (NE555 or other): it only documents that there is one more chip whose function cannot be read.

4.5 Trimmers: blue vs black

Colour / position	Temp. coefficient	Probable function	Evidence
Blue (PCB-S)	100 ppm/°C	One per channel: small-signal stage adjustment or inverter DC balance. If there were two others on the underside, they could be individual 6550 bias.	OBSERVED two, one at each end · total number and function ND
Blue (PCB-P)	100 ppm/°C	Auxiliary regulation/control block (with MOSFET + TL431); exact application ND	PROBABLE
Black (PCB-P)	ND	Possible threshold/timing or other control function	HYPOTHESIS

Blue 100 ppm: the Bourns 3296 series is documented as 25 turns, 0.5 W, ±100 ppm/°C. Black: series and temperature coefficient ND.

4.6 B+ supply architecture — HYPOTHESES

4.6.1 The question

In a 40 W push-pull, the four 6550s draw a current that changes with the music, and those changes make the B+ rail voltage oscillate. If that oscillation reaches the small-signal stages, it causes intermodulation and unwanted coupling. A classic technique to avoid it is to split the B+ into branches.

This question (how the B+ is distributed) is independent from that of the control and auxiliary circuits (§4.7). The existence of a relay board or an auxiliary magnetic element does not tell whether the B+ goes on one branch or several, nor vice versa.

4.6.2 The three possible architectures

(a) BRANCHES WITH SERIES R OR L

Main B+
   ├── POWER B+ ──► 6550
   ├── R1 + C3 ──► DRIVER B+ (ECC803S)
   └── R2 + C4 ──► INPUT B+ (EF86)

(b) SINGLE RAIL WITH LOCAL BYPASS

Main B+ ──► all valves, without R or L in series
              with C_local next to each socket

(c) SINGLE RAIL WITHOUT LOCAL BYPASS

Main B+ ──► all valves, without C_local

Important physical note. Architecture (b) is not a "partial separation", but another category. A capacitor in parallel with the rail, without resistance or inductance in series, does not isolate at audio frequencies.

4.6.3 What is observed in the P52

Evidence	What it suggests	Status
Common filter bank on PCB-P (three EPCOS B43645 with the same silkscreen, grouped)	Single main B+ rail or parallel banks per channel	OBSERVED
In the available photographs no power resistors in series with B+ derivations are observed on PCB-P	Probable absence of architecture (a)	OBSERVED limited · inference PROBABLE
Two small electrolytic capacitors near the noval sockets of PCB-S	Most probable reading: cathode bypass, not rail decoupling	HYPOTHESIS — neutral evidence regarding branch separation
Presence of 10W10KJ (10 kΩ / 10 W) incompatible with a permanent bleeder of the main B+ at 430 V	Suggests auxiliary branch or start-up circuit, not main B+ rail	OBSERVED the value · ND the function
Cable bundle between PCB-P and PCB-S	Cannot count wires or trace route	ND

4.6.4 The three architecture hypotheses

Architecture	Description	Status
(a) Branches with series R or L	Power B+ → 6550; driver B+ → ECC803S with RC network; input B+ → EF86 with RC network.	HYPOTHESIS — not elevated by numerical coincidence
(b) Single rail with local bypass	Single B+, with local capacitors next to the sockets (without R or L in series).	PROBABLE — most conservative reading with current evidence
(c) Single rail without local bypass	Single B+, without local capacitors.	MINOR HYPOTHESIS

4.6.5 B+ ripple in push-pull: cancellation in the OPT

In a push-pull stage, the B+ ripple reaches the plates of the two output valves in phase, that is, to both ends of the output transformer (OPT) primary. The OPT largely rejects that common component. What matters is the ripple reaching the driver and phase-inversion stages, because there it does become a spurious signal that is amplified.

4.6.6 How the question would be resolved

1. Photo of the underside of PCB-S (solder side).
2. Photo of the connector/B+ entry point to PCB-S.
3. Count the Bourns trimmers on PCB-S (P14).
4. Trace the connection of the 10W10KJ (P15).
5. Voltage measurement at B+ input to PCB-S and at the EF86 and ECC803S plate pins.
6. Ripple measurement at each point with oscilloscope.

Conclusion of §4.6: with current evidence, the most prudent thing is to assume a PROBABLE single rail with local bypass (architecture b). To resolve it, the underside of the PCB must be traced and voltage and ripple measured at the key points.

4.7 PCB-3 and physical partitioning

4.7.1 The evidence

The general interior view shows at least three boards and a longitudinal metal partition between zones. The new elements are:

• PCB-3: independent vertical board in the rear left corner, with four blue rectangular components in a row and associated electronics (transistors, diodes, electrolytics, resistors and connectors).
• Auxiliary magnetic element: encapsulated component in the rear right corner, separate from the three main transformers and near the mains input and a fuse holder.
• Longitudinal metal partition between the power supply and control zone and the signal and power zone.

4.7.2 Epistemological chain of PCB-3

Statement	Status
There is an independent vertical board in the rear left corner	OBSERVED
The board contains four blue rectangular components aligned	OBSERVED
The four components are electromechanical relays	PROBABLE — geometry and layout compatible, without marking reading
The board performs signal/input switching	HYPOTHESIS — compatible with location next to the rear panel
The board performs mute or output protection	HYPOTHESIS alternative
The board performs power switching	HYPOTHESIS less probable
Specific function of each component	ND

How reasoning works here. It does not jump from "probable relays" to "signal relays". Even if they are, they could switch signal, mute, power, protection or several things at once. Being near a zone changes the probabilities, but does not replace track tracing.

4.7.3 On the metal partitioning

Statement	Status
A longitudinal metal partition exists between zones	OBSERVED
The arrangement aims to reduce electromagnetic coupling and/or separate zones of different voltage/power	PROBABLE — deliberate constructive decision
The partition is effective for EMI, ground return, residual magnetic field or crosstalk	NOT MEASURED

4.7.4 On the auxiliary magnetic element

OBSERVED: an encapsulated element in the rear right corner, separate from the three main transformers and near the mains input and a fuse holder.

ND: its function. Open hypotheses, none confirmed:

• Auxiliary control transformer (would fit with the low-voltage rail that the NE5532 would need, if confirmed).
• Mains input module or fuse holders (would fit with its location next to the mains input).
• Another magnetic or filtering component.

Specifically, it cannot be stated that this component is the source of the low-voltage rail: it is a hypothesis, not an observation.

4.7.5 What it implies and what it does not imply

It implies:

• Describing the P52 inside as at least three boards, a metal partition and an auxiliary magnetic element of ND function.
• Assuming that control is more complex than a quick view suggested.
• Noting as a structural finding the deliberate separation between the power zone and the control and signal zone.

It does not imply:

• That the B+ has more branches (it is another question, §4.6).
• That there is a relay-based input selector (HYPOTHESIS).
• That there is a relay-based protection system (HYPOTHESIS).
• That there is valve warm-up timing with a specific component (HYPOTHESIS).
• That the NE5532 detects DC at the output (HYPOTHESIS).
• That the auxiliary magnetic element is a transformer (ND).

Conclusion of §4.7: the interior photos have expanded the physical map of the P52, but have not resolved the electrical circuit. What PCB-3, the auxiliary element and the metal partition do remains ND until there is a traced schematic and targeted macro photos.

5. Power stage

5.1 Valves and dissipation — 6550 vs KT88 vs EL34 vs KT77

Four 6550s are visible (two per channel), which in the manufacturer's video appear to be Tung-Sol. How much heat each plate dissipates in operation is ND: V_plate-cathode · I_cathode (voltage times current) would need to be measured.

To place the P52's valve in its family, this table compares the fundamental electrical characteristics of the four most relevant octal-socket valves. The limits that condition ultralinear (UL) operation are shown in bold.

Characteristic	6550 (P52)	KT88	EL34	KT77
Type	Beam tetrode	Beam tetrode	Pentode	Beam tetrode
Max plate dissipation	35 W	42 W	25 W	32 W (source-dependent)
Max plate voltage	600 V	800 V	800 V	≤800 V (Genalex) / 600 V (alternative sources)
Max screen voltage (g2)	400 V	600 V	425 V	ND
Filament current	1.6 A	1.6 A	1.5 A	1.5 A (approx.)

On the maximum screen voltage. In ultralinear configuration, the screen (g2) is connected to a tap of the output transformer and is subjected to a fraction of the plate voltage. This limit is a real constraint on the designer and determines what percentage of ultralinear is safe. The three types have very different limits: 6550 = 400 V, EL34 = 425 V, KT88 = 600 V. The KT88 is the one that offers the most margin in UL. The 6550, with 400 V screen, imposes a tighter design constraint on the P52 than a KT88 would have on the same transformer.

Note on KT77 and evidence classification. Sources disagree on the KT77's maximum plate voltage (800 V per Genalex, 600 V per other sources). That discrepancy requires treating any KT77 substitution with prior verification of the amplifier's real B+. Evidence classification for this table: 6550, KT88 and EL34 CONFIRMED by JJ Electronic and Mullard datasheets; KT77 REPORTED-EXT with conflicting sources.

5.2 Class of operation: AB probable; AB1 subtype probable

The class of operation (A, AB, B) describes how the work is shared between the valves of a push-pull pair when there is no signal. It determines the quiescent current, the thermal dissipation and the behaviour near clipping. Within class AB there are two subtypes: AB1 (the control grid never goes positive with respect to the cathode; the driver only delivers voltage) and AB2 (the grid goes positive at peaks and demands real current; it requires a driver capable of delivering it).

P52 classification

For the P52, both the class and the subtype are indirect inferences. The most honest thing is to leave both labels at PROBABLE.

Aspect	Classification	Basis
Class AB	PROBABLE	The visible topology (push-pull 6550 pair, valve driver in class A, typical dissipation of 2 valves per channel) is consistent with AB. There is no explicit manufacturer data point verified against the P52's official spec sheet.
AB1 subtype	PROBABLE	Four convergent observations (see below). Consistent with the Mullard 5-20 lineage and with the absence of resources typical of AB2.
AB2 subtype	NOT RULABLE OUT	Requires schematic, voltages, bias, load impedances and interstage transformer data. Without them it cannot be ruled out by calculation.

AB1 subtype argumentation (four convergent observations)

1. The driver is an ECC803S, a voltage-gain valve (μ = 100, Ri = 62.5 kΩ). Its specified maximum cathode current is 8 mA (TAD datasheet), but that data point must not be interpreted as grid current available for AB2. It is a voltage valve, not a power one.
2. The resources typical of AB2 are not observed. A conventional AB2 implementation requires a considerably higher grid current delivery capability than is usual in an ECC83/ECC803S voltage stage. The photographs do not show an evident power driver stage: there are no additional valves outside the 4× 6550s and the 4 preamp/driver valves; there is no interstage transformer (only the three large EI ones: power and two outputs); the visible MOSFET/TL431 block is low-power and probably manages bias, not signal excursion.
3. The declared power (40 W per channel) fits AB1, not AB2. A push-pull with 2× 6550 per channel in AB1 delivers 40–60 W per channel depending on B+ and transformer. In AB2 with the same B+ one could reach 50–75 W. If the design were AB2, the cost of implementing it would be justified by an extra power that does not appear.
4. Similarity to the Mullard 5-20. The historic reference design (1959, design c. 1956) operated in AB1 with a voltage driver (ECC83). If the P52 follows that line, it is coherent that it also operates in AB1 (see §3.2). This observation is analogy, not direct evidence.

What would change the conclusion. For the P52 to be AB2, one of these three things would be needed, which are not observed but cannot be ruled out without a schematic:

1. That the ECC803S does not drive the 6550 directly but an intermediate stage not visible.
2. That the MOSFET/TL431 block acts as a current buffer towards the grid.
3. That there is a small interstage transformer camouflaged among the components.

Without schematic, voltages, bias, load impedances and interstage transformer data, AB2 cannot be ruled out by calculation.

Operational conclusion. The most honest classification of the P52 is: class AB probable, AB1 subtype probable with technical foundation. Neither of the two labels rises to CONFIRMED without a schematic or measurement, because the manufacturer has not published a complete technical spec sheet for the P52 with these details. Nor does it fall to "indeterminate", because the four convergent observations clearly point to AB1.

5.3 Bias mode

Open hypotheses: two visible trimmers on PCB-S (one per channel), which could adjust the signal or the inverter DC balance; individual bias with four trimmers, if there are two more on the underside; fixed bias by voltage divider; cathode bias; or bias from the auxiliary block. Full development in §3.2. The exact bias mode is ND. The class of operation is classified as AB probable, AB1 subtype probable (§5.2).

5.4 6550 substitutions

The P52's octal socket accepts, in principle, other valves from the same family: KT88, KT90 or KT120. The substitution is not automatic; two conditions must be verified before changing the valves.

• Current reserve of the 6.3 V heater winding. The 6550 and KT88 draw 1.6 A per valve; the KT120 draws 1.7 to 1.95 A depending on manufacturer. Four KT120s add up to 7.8 A on power alone, plus the heaters of the four preamp/driver valves. If the EI transformer has no margin, the heater voltage will fall below 6.0 V and shorten the life of all valves (not just the substituted ones).
• Margin of the negative bias circuit. A correctly biased KT120 requires more negative bias voltages than a 6550. If the bias circuit does not reach that value, the valve will operate outside its optimum point and may dissipate beyond its limits.

Without transformer data. The nominal current of the P52's 6.3 V winding and the voltage range of the bias circuit are not available. Any substitution by KT90 or KT120 must be considered conditional on prior verification by a technician. KT88s have the same heater consumption as 6550s (1.6 A), which makes them priority candidates if one wants to increase plate dissipation reserve without touching the transformer; their maximum Vg2 (600 V) is higher than the 6550's (400 V), which widens the margin in UL. The KT88 substitution does not require a heater change, but does require bias readjustment.

5.5 Open questions

Parameter	Status	Section where it is treated
Class of operation / subtype	AB PROBABLE; AB1 subtype PROBABLE; AB2 NOT RULABLE OUT without schematic.	§5.2
Global NFB	ND — the Mullard 5-20 figure is in conflict in the literature (20–26 dB vs ~30 dB) and does not transfer to the P52. Visual detection described in §3.2	Appendix P6
%UL	ND	Appendix P7
Zout and DF	ND — NFB effect tabulated as an indicative model in §3.2	Appendix P2
P@1 % and THD+N vs power curve	ND	Appendix P1
6550 quartet matching	ND	Part II §8
B+ architecture	HYPOTHESIS (a) · PROBABLE (b) single rail with local bypass	§4.6, Appendix P10
Phase inverter topology	HYPOTHESIS (EF86 → LTP main); Mullard 5-20 analogy as indirect argument; ND confirmation	§3.2
ECC803S driver dynamic reserve	Element to verify. I_cat,max = 8 mA (TAD). Without measurement no limitation can be stated. Alternatives catalogued as proposal pending measurement: 6CG7, 6SN7, 12BH7, 5687.	§3.3, §3.5
Function of the blue trimmers on PCB-S	TWO OBSERVED · total number and function ND	§3.4, §4.5, Appendix P14
Function of the 10W10KJ	OBSERVED the value · ND the function (does not seem to be a permanent B+ bleeder at 430 V)	§4.3, Appendix P15
Function of PCB-3	OBSERVED the board · ND the function	§4.7, Appendix P16
Origin of the low-voltage rail	PROBABLE if NE5532 confirmed · ND origin	§4.4, Appendix P16

Methodological note: the UL taps are transformer wires, not board tracks. The %UL is found by measuring the AC signal ratio between the plate (pin 3) and g2 (pin 4) of the OPT.

5.6 Output transformers

Three varnished EI transformers are visible (one power and two output), with a metal partition between the supply and PCB-S OBSERVED. The primary impedance, transformation ratio, leakage inductance and frequency response are ND.

5.7 Calculators — SPL and Zout variation

SPL and required power calculator — P52 40 W

Formulas: SPL_1W@d = sens − 20·log10(d) and P = 10^((target SPL − SPL_1W@d − factor)/10). Results per channel.

Sensitivity (dB/W/m)89.0 dB/W/m

Listening distance3.0 m

Target SPL (peak)95 dB

Acoustic correction factor0 dB — one channel / free field+3 dB — energetic stereo+6 dB — coherent stereo / dry room+8 dB — coherent stereo + live room

Enable JavaScript to see the calculator result.

Methodological note. P52 power: 40 W nominal declared without a THD condition. Hypothetical sensitivity.

Response variation calculator by Zout

Amplifier Zout (Ω)2.0 Ω

Speaker Zmin (Ω)4.9 Ω

Speaker Znom (Ω)6.0 Ω

Enable JavaScript to see the calculator result.

What this calculator measures. It returns the difference between the attenuation at Zmin and at Znom with the load modelled as resistive: it is a pedagogical approximation, not a simulation of a real speaker. P52 Zout: ND, pending measurement.

6. AC input and RF/EMI — observed status

6.1 Observed status of the input filter

In the photos capacitors of about 0.1 µF / 1000 V are visible just after the IEC connector (the mains input), which fit with an EMI (interference) filter. Their X/Y safety class is not established, and no common-mode L/N choke is visible.

6.2 PE and chassis

The chassis appears directly connected to protective earth (PE) PROBABLE.

6.3 Y capacitors

They have not been identified with certainty. Their presence is PROBABLE by position, ND by tracing.

6.4 Soft-start and start-up sequence — open problem

Open problem. The P52 has a Hongfa HF115F relay (PROB) and a 10W10KJ resistor (10 kΩ / 10 W) that would fit with a soft start, that is, with limiting the inrush current. From the photos alone it cannot be known:

• Whether the 10W10KJ is part of an HV bank pre-charge circuit.
• Which bank capacitors are in series with it.
• What circuit controls the relay switching.
• How much time passes between connecting to the mains and the relay switching.
• What component counts that time (NE555 ND, another IC ND, or an RC with comparator ND).

The only thing that could be calculated is this: if the 10W10KJ were connected permanently between B+ (430 V) and ground, it would dissipate ≈18.5 W against 10 W nominal, something incompatible with continuous use. That makes it unlikely to be a permanent bleeder of the main B+, but does not say what its real function is or how the start-up is timed.

The τ = R·C = 18.3 s calculation is conditional: it assumes that the 10W10KJ is in series with the complete 1.83 mF bank and that the relay switches at the end of the RC transient. Neither of the two things is demonstrated. It is an open problem, not a resolved contradiction.

6.5 What this report does NOT propose to modify here

Reading: the input filter appears limited to film capacitors next to the IEC connector PROBABLE. Adding a choke is discussed in Part II §6, and would only make sense if a measurement of conducted EMI or background noise demonstrated a problem caused by the mains.

7. Conclusion and traceability

7.1 Acoustic model — conditional scenarios

The power needed to reach 95 dB depends on how the cabinet sensitivity has been defined. Each row of the table starts from a different reference, and each result is coherent with its own. All values are conditional estimates, not P52 measurements.

Reading / condition	Level at 3 m	Power for 95 dB	Margin over 40 W
89 dB @ 1 W / 1 m; listening at 3 m, free field	≈79.5 dB	≈35.8 W	+0.5 dB
89 dB @ 2.83 V / 1 m over 5.3 Ω (1.51 W); listening at 3 m, free field	≈77.7 dB at 1 W	≈54.1 W	−1.3 dB
89 dB SPL with 1 W measured directly at 3 m	89 dB	≈4.0 W	+10.0 dB

Calculation: power = 10^((95 − level per 1 W)/10) and margin = 10·log10(40/power). Row 2 uses the sensitivity referred to 1 W (77.7 dB at 3 m), which is the magnitude coherent with the rest of the calculation. Note: the table is mono and free field. The three cases use the same reference (level at 3 m with 1 W applied), without mixing 1 W with 2.83 V. None of these scenarios demonstrates the P52's real performance.

7.2 Synthesis of the report

PCB-S shows careful construction, with many coupling and filtering capacitors, two blue Bourns trimmers (one at each end) and blue precision resistors. In the supply there are three EPCOS/TDK B43645 with silkscreen B43645-S7617-M1, 450 V and 610 µF, OBSERVED on all three units. The effective capacity of the set remains ND: the connection admits at least four readings, and only three in parallel would give 1,830 µF. The nominal service life of the B43645 series is >5,000 h at 105 °C according to TDK, and the exact correspondence of the code with the datasheet remains PENDING VERIFICATION against the ordering codes table, because the B43645 and B43655 families are not identical.

The vertical PCB-3 with four blue components is OBSERVED; that they are relays is PROBABLE; their function is ND. The auxiliary magnetic element in the rear right corner is OBSERVED as a physical component, but its function is ND: auxiliary transformer, mains input module/fuse holders and other uses are open hypotheses.

The longitudinal metal partition is OBSERVED; the EMI/isolation separation intent is PROBABLE; its real effectiveness is ND.

The Hongfa HF115F relay is PROBABLE by position and geometry, compatible with soft-start/inrush. The two DIP-8 observed on PCB-P reinforce the presence of control electronics, but do not identify a specific component: NE5532 PROBABLE, NE555 ND.

The 10W10KJ resistor (10 kΩ / 10 W) is OBSERVED. The thermal calculation makes it unlikely to be a permanent bleeder of the main B+ (18.5 W at 430 V > 10 W nominal), but its exact function remains ND: bypassed inrush limiter, auxiliary branch bleeder, divider or regulator load are the open alternatives.

The MOSFET + TL431 + Bourns combination makes an auxiliary regulated branch plausible, but its voltage and destination remain ND; a relationship with the bias of the 6550s is not ruled out.

The B+ architecture admits three physically distinct readings: (a) separate branches with series R or L, (b) single rail with local bypass, (c) single rail without local bypass. The most conservative reading is PROBABLE single rail with local bypass (b). The numerical coincidence "three capacitors / three branches" does not elevate the weight of (a).

The ECC803S is a double triode. With one ECC803S and one EF86 per channel, the classic combination is EF86 (input) → ECC803S as LTP with both triodes MAIN HYPOTHESIS, supported by the analogy with the Mullard 5-20 (1959, design c. 1956). The required gain is 21–35× (26.5–31.0 dB), compatible with EF86 + LTP + NFB, but that does not prove it. On bias, the two visible trimmers favour "one per channel" over "four individual", but the range remains open. On the class of operation: AB probable; AB1 subtype probable with foundation (four convergent observations; see §5.2). The global feedback, ultralinear percentage, P@1 %, Zout and DF remain unknown (ND).

On the kinship with the Mullard 5-20: the P52 follows the same valve family as the British reference design (c. 1956, revision 1959), doubled to stereo and with solid-state rectification. This correspondence is a useful argument for inferring the P52's architecture, but does not demonstrate that it reproduces the internal circuit: voltages, bias, component values and NFB may differ. The specific NFB figure of the 5-20 does not transfer to the P52, and in the literature it is in conflict (20–26 dB vs ~30 dB depending on source and revision).

V6.6 reformulation on the driver. Versions before V6.5 described the ECC803S as "the weakest link in the small-signal chain". That statement is not supported: the I_peak ≈ 0.25 mA calculation is well done, but does not demonstrate insufficiency, and in probable AB1 there is no DC grid current. The correct conclusion is that the driver current delivery capability is an element to verify if one wants to evaluate the dynamic reserve, not an established limitation (§3.3). The driver alternatives (6CG7, 6SN7, 12BH7, 5687) are catalogued as TECHNICAL PROPOSAL — PENDING MEASUREMENT (§3.5), not as a justified improvement.

7.3 What CANNOT be concluded from this report

This list serves to prevent reading more than what is there. With only photos and catalogues, none of the following can be stated:

• That the P52 really meets the 40 W at a given THD.
• That the effective filtering capacity is 1,830 µF — it depends on the connection.
• That the B43645-S7617-M1 code corresponds exactly to the TDK datasheet (ordering code not cross-checked) or that the family is B43645 and not B43655.
• That the B+ voltage is 450 V or any other.
• That the bias mode of the 6550s is fixed by divider, by cathode, from auxiliary branch, or individual with four trimmers. Probable class is not the same as exact bias mode.
• That the AB1 subtype is confirmed: it is probable with foundation, but not demonstrated. AB2 cannot be ruled out by calculation without a schematic.
• That the AB class is confirmed by the manufacturer in verifiable official documentation.
• That the blue trimmers on PCB-S have a bias function, inverter DC balance function or signal stage adjustment function, nor exactly how many there are.
• That the 6550s are matched.
• That global NFB exists or its depth. The original Mullard 5-20 figure is in conflict in the literature and, in any case, is not extrapolable to the P52.
• That the ultralinear percentage is the usual 43 % reference.
• That the exact phase inverter topology is LTP, cathodyne or another. The Mullard 5-20 analogy is PROBABLE / ANALOGY.
• That the Rev.A board toroids had a specific inductance of 0.2–0.6 mH: the value is VISUAL ESTIMATE, not measured. In the currently photographed unit they are not present.
• That there are separate B+ supply branches with series R or L.
• That the 10W10KJ is the main inrush limiter, an auxiliary branch bleeder, a divider or a regulator load. The only thing closed is that it does not seem to be a permanent bleeder of the main B+ under the 430 V hypothesis.
• That PCB-3 performs input selection, mute, protection or any specific function. Only the board and the presence of four probably-relay components are observed.
• That the auxiliary magnetic element is an auxiliary transformer, a mains module or fuse holders, or any other specific component. Only the existence of a physical component in that position is observed.
• That there is a relay-based input selector, a relay-based protection system, or warm-up timing implemented with a specific component.
• That the NE5532 performs DC detection, mute, servo or any specific function.
• That the metal partition is effective for EMI, ground or crosstalk — not measured.
• That the fifth PCB-S hole has a specific function — ND.
• That the gold terminals on the rear wall are speaker outputs, inputs or something else — function ND.
• That the P52's ECC803S is insufficient to drive the 6550s. It cannot be stated without measuring excursion, high-frequency distortion and stability over complex load.
• That the proposed driver alternatives (6SN7, 6CG7, 12BH7, 5687) work better in the P52. They are reasoned technical proposals, not validated. None is plug-and-play.
• That the P52 sounds better or worse than other amplifiers.

General rule: the entire report is based on photos and catalogues. No statement should be read as a verified measurement, unless it carries the label CONFIRMED, OBSERVED, MEASURED or MEASURED-EXT.

PART II — SPECIAL EDITION: WHICH COMPONENTS TO CHANGE OR ADD?

PART II — SPECIAL EDITION: WHICH COMPONENTS TO CHANGE OR ADD?

This part proposes an order of work. Phase 0 — Characterization (essential before touching anything): measure what the P52 is really like. Then four blocks: A. Safety (high voltage and differential probe). B. Supply and start-up (B+ at rest and under load, ripple, ESR, toroids, auxiliary block, start-up sequence, inrush, relays and cathode current of each 6550). C. Performance (NFB, stability, distortion, P@1 %, Zout and DF, frequency response, noise, THD+N at 4 and 8 Ω). D. Components (C2N, resistors, Bourns, wiring, EMI, brand change). Finally, S4.

20-second read

What it is: a conditional protocol for a possible special edition of the Acronos P52. It is not a list of upgrades nor a catalogue of high-end components.

What it contains: in each area, first how the P52 currently is and what remains to be known. If specific components were proposed, they appear separately as AUDIOPHILE PROPOSAL — S3 (listening hypothesis, not demonstrated).

What it is NOT: it is not recommended to change anything in the P52 without measuring first.

Central idea: the special edition is not defined by a list of expensive components, but by what characterization demonstrates is worth changing and what can then be verified.

Safety warning before any measurement

Stored energy. If the three 610 µF capacitors were in parallel (connection ND), the stored energy would be:

• At 400 V: E = ½ · 1830·10⁻⁶ · 400² ≈ 146.4 J.
• At 430 V: E = ½ · 1830·10⁻⁶ · 430² ≈ 169.2 J.
• Maximum possible if B+ = 450 V: E ≈ 185.3 J.

With B+ between 400 and 430 V, the working range would be 146 to 169 J. It is not known through what path the bank discharges (ND).

Procedure

1. Disconnect from the mains and ensure it cannot be re-energised.
2. Wait, without assuming it has already discharged.
3. Measure the remaining voltage with an instrument and probes rated for the task.
4. If voltage remains, discharge with a suitable resistor, never with a screwdriver.
5. Measure again after a few minutes.
6. Only then, with the device unplugged, connect the measurement clips.
7. To measure how the voltage falls: switch on, switch off and read the instrument without touching anything.
8. Always use a high-voltage differential probe and rated probes.

S0–S4 scale — Definition

Label	Meaning	Purpose
S0 — Reference	No intervention	A valid outcome of the process
S1 — Reliability	Changes justified by reliability or safety	Verifiable problem
S2 — Performance	Changes with demonstrable electrical improvement	Measurable Δ
S3 — Listening	Substitutions evaluated by controlled listening	Reproducible perceptual difference
S4 — M&M	Characterization, matching and documentation	Traceability

0.1 — Minimum instrumentation

• High-voltage differential probe with its specifications in view.
• Rated probes for the voltage and measurement category.
• Oscilloscope of at least 20 MHz (100 MHz or more for RF/EMI).
• Well-dimensioned discharge resistor.
• IR or contact thermometer.
• LCR/ESR meter (capacitance and series resistance), always with the bank discharged.

1. Coupling capacitors

There are 14 × MKP CEC C2N and 2 PIO — OBSERVED. What each C2N does is ND: so many capacitors suggest coupling, bypass, decoupling, compensation, snubber or a mixture.

When to stop: if a C2N has a cutoff frequency (fc) of 2 Hz or less and shows no leakage or anomalous ESR, it is kept.

Audiophile proposal — S3 · LISTENING HYPOTHESIS

• Mundorf MCap EVO Oil / Supreme, ClarityCap CSA / MR, Jupiter HT, V-Cap OIMP / CuTF, Solen Fast Cap, Auricap.
• Note: they are S3 and require a controlled test. They are not presented as a demonstrated improvement.

2. Power supply capacitors

There are 3 × EPCOS/TDK B43645 610 µF / 450 V — OBSERVED. Their connection is ND. The 1,830 µF are CONDITIONAL CALCULATION (they are only valid if the three are in parallel). Series nominal service life: >5,000 h at 105 °C (TDK, REPORTED-EXT).

When to stop: if the EPCOS work within their specifications, they are kept. Own guideline of the report: B+ ≤ 382–405 V (85–90 % of 450 V). It is not a manufacturer's standard.

Audiophile proposal — S3 · LISTENING HYPOTHESIS

• Mundorf MLytic AG / HC — check that they exist in 610 µF, ≥450 V and with the right form factor.
• Nichicon KG / Fine Gold / KW — DISCARD DUE TO VOLTAGE (≤100 V).
• Rubycon MXG / MXH — S3; MXH not verified.
• Jensen / ASC — check voltage, capacitance and form factor.

3. Resistors

There are green 2 W MOX resistors (10K28) and white 1 W ones — OBSERVED. Their function and the power they actually dissipate are ND.

It only makes sense to replace them if it is demonstrated that they are marginal in power (little thermal margin), that their tolerance is insufficient or that they have degraded.

Audiophile proposal — S3

• Plate/cathode signal: Takman Rey, Vishay Dale CMF / RN.
• Feedback and grid stoppers: PRP, Shinkoh, Amtrans AMRG.
• Power: Mills, Ohmite non-inductive.

4. Decoupling and bypass

Their function and values are ND. What can be seen: a 100 µF / 450 V capacitor OBSERVED, small electrolytics near the sockets and several auxiliary ones distributed on PCB-P.

Decoupling is not added if it is not first demonstrated to be needed. The reference is §4.6: the probable B+ architecture (b) already includes local bypass. An additional bypass is only justified if the ripple measurement at the EF86 and ECC803S pins demonstrates it is insufficient.

Audiophile proposal — S3

• Add or improve local decoupling of the signal stage with MKP or low-inductance polypropylene capacitors.
• If the B+ is a single rail with local bypass (b) or without bypass (c), study series RC networks for the driver and input (§4.6). It is classified as S2 if a measurable difference (Δ) is demonstrated.

5. Internal wiring

Twisted wire bundles are visible — OBSERVED. Their gauge, material and routing are ND.

The wiring geometry matters more than the conductor metal. What matters is the gauge, length, ground returns, separation between power and signal, loop areas, voltage drop, temperature and insulation.

Audiophile proposal — S3

• Change the signal wiring to OCC copper or silver.
• Use adequate-gauge cable with teflon or silicone insulation for high voltage and filaments.
• Keep the current twisting and routing.

6. Supply and protections — PSU V2 and DC Blocker

Definition of the PSU V2

The PSU V2 is the power supply proposed for a possible P52 revision, provided the equipment has been characterized first.

ORIGINAL EI TRANSFORMER → SOLID-STATE RECTIFICATION → MODERATE C1
   → HT CHOKE → MAIN C2
       ├── POWER B+ ──► 6550
       ├── R1 + C3 ──► DRIVER B+
       └── R2 + C4 ──► INPUT B+

Status: CONDITIONAL PROPOSAL. It is not implemented until Phase 0 clarifies three things: whether the toroids are or are not audio-frequency chokes, what the current B+ architecture is (a, b or c) and how much ripple remains at the EF86 and ECC803S pins.

HT filtering choke

If the measurements demonstrate that the toroids are not audio-frequency chokes, adding a 5–10 H choke between C1 and C2 (the first two filter capacitors) would be studied.

Common-mode choke

It is only studied if a conducted EMI measurement demonstrates a problem caused by the mains.

X and Y capacitors

No X/Y class could be assigned to the capacitors seen after the IEC connector. If Y capacitors were added, the leakage current to earth (touch current) must be taken into account.

Relays

The Hongfa HF115F is probably the soft-start relay, and the Songle and Finder ones are auxiliary. Relays with gold contacts are for low-level signals; they are not suitable for switching power.

Soft-start and start-up sequence — open problem

Before modifying anything, the following must be measured: the inrush current, the time it takes to charge the HV bank, the delay between connecting to the mains and the relay switching, and how the Hongfa HF115F and the auxiliary relays behave.

Open problem (see §6.4): from the photos alone it cannot be known whether the 10W10KJ limits the inrush current (with the relay bypassing it), discharges an auxiliary branch, forms a divider or is the load of a regulator. The τ = R·C = 18.3 s calculation is conditional. Before adjusting any delay, the circuit must be traced and the real sequence measured.

10W10KJ resistor

Marking 10W10KJ; most probably it is 10 kΩ / 10 W / ±5 %. Its function is ND. The thermal calculation makes it unlikely to be a permanent discharge of the main B+ (18.5 W at 430 V vs 10 W nominal). The following alternatives remain open: bypassed inrush limiter, auxiliary branch discharge, voltage divider or regulator load.

DC Blocker at the mains input — S1 proposal

A DC Blocker at the IEC input is a circuit of two diodes and two electrolytics in antiparallel that prevents the DC component of the mains from reaching the transformer. EI transformers are sensitive to that DC current: a few hundred mV can shift the core hysteresis cycle, and that produces mechanical hum, more quiescent current and heating.

Classification: S1 — Reliability. It is not an audiophile proposal (S3), but a reliability measure that extends the transformer's life.

Cost: two silicon rectifier diodes (1N5408) and two 10,000–22,000 µF / 10 V electrolytics in antiparallel.

When to install it: the DC current actually present in the owner's mains must first be measured. If it is negligible (<50 mV), the DC Blocker adds nothing. If it is significant (>200 mV), or if the transformer hums mechanically for a cause that is not the supply ripple, installation is justified.

Note: installing the DC Blocker does not require touching the transformer or the amplifier circuit: it is interposed between the IEC connector and the primary. It is an external, reversible and inexpensive intervention.

ECC803S driver substitution — proposal pending measurement

Current status: do not modify yet. There is no measurement demonstrating that the P52's ECC803S falls short. What follows is a conditional technical proposal, not a justified improvement. See §3.3 and §3.5 for the development.

If the characterization (Phase 0) showed a real driver limitation —for example, anomalous increase in THD with frequency, inability to maintain the required excursion over complex load or unstable behaviour under NFB— then higher-current-capability valves would be studied. Among the candidates catalogued in §3.5: 6CG7, 6SN7, 12BH7, 5687.

Provisional classification: S2 — Performance if a measurable improvement is demonstrated (Δ in high-frequency distortion, stability under NFB or response at 10 kHz over complex load). Requires:

• Verification of the socket pinout and heater consumption.
• Readjustment of the bias and driver operating point.
• Verification of total loop gain if NFB exists.
• Stability test with square wave at 1 kHz and 10 kHz.

It is not plug-and-play. None of the proposed substitutions fits directly without verification. The intervention must be done by a technician with access to the P52 schematic (which is ND).

Bias servo — out of scope

A bias servo would require changing the bias topology. It does not consist in replacing components.

NFB switch — discarded due to redesign

Changing the NFB depth simultaneously modifies Zout, DF, bandwidth, distortion, stability and gain. It is not a component upgrade, but a redesign of the amplifier, and would require the complete P52 electrical schematic (which is ND). It is out of scope.

Proposals classified by purpose

Proposal	Classification
DC Blocker at IEC input	S1 — conditional on measuring DC offset
Adjust the start-up circuit delay (if it is confirmed to exist and its topology)	S1 — conditional on resolving the open problem of §6.4
Replace the ECC803S with 6CG7 / 6SN7 / 12BH7 / 5687 as driver	S2 — only if measurement demonstrates limitation · not plug-and-play · see §3.5
PSU V2: HT filtering choke (5–10 H) + branch separation (R1/C3, R2/C4) if the toroids are not audio-frequency chokes and driver/input modulation is high	S2 if a measurable Δ is demonstrated
Common-mode choke at mains input (Würth WE-CMB or equivalent)	S2 if a measurable Δ is demonstrated
Improve or add Y safety capacitors	S1 (safety) — verify touch current
Replace relays with Panasonic, Omron or gold-contact ones	Low-level signal: S3 · Power/protection: S1. Gold not suitable for power.
NFB switch (0/6/12 dB)	DISCARDED — topology redesign

7. Trimmers (Bourns)

The blue ones (3296 series) are documented with a 100 ppm/°C temperature coefficient; the black one's is ND. There are two blue trimmers visible on PCB-S, one at each end. How many there are in total and what they do is ND (see P14).

Colour / position	Coef.	Probable function	Evidence
Blue (PCB-S)	100 ppm/°C	One per channel: small-signal stage adjustment or inverter DC balance. If there were two others on the underside, they could be individual bias.	TWO OBSERVED one at each end · total number and function ND
Blue (PCB-P)	100 ppm/°C	Auxiliary regulation/control block — exact application ND	PROBABLE
Black (PCB-P)	ND	Possible threshold/timing or other — function ND	HYPOTHESIS

Audiophile proposal — S3

• Blue on PCB-S: Do not fix while its function is unknown.
• Blue PCB-P: S1 or S2 depending on function.
• Black: S1 or S2 depending on function.

8. S4 — Matched & Measured

S4 does not change components. It is a measurement and documentation layer that can be added to any edition, including S0.

What it includes

• Select and measure the four 6550s one by one.
• Check the quiescent current with the valves already installed.
• Estimate the plate dissipation margin (how much heat each valve tolerates).
• In ultralinear, the screen current cannot be measured without intervening in the circuit, so the total cathode current is documented.
• Compare the two halves (branches) of the push-pull.
• Compare the regulated auxiliary voltages of the left and right channels.
• Document the relevant electrical measurements.

What "matched" means

• Valve matching: select the valves beforehand, according to measured parameters.
• Installed current matching: check the real current with the valves already installed.
• Branch balance: compare the two halves of the push-pull.
• Individual adjustment: only if the circuit allows it (depends on what the blue trimmers on PCB-S actually do).

"Matched" does not mean changing valves. If the installed quartet is already sufficiently balanced, S4 documents it without replacing anything.

APPENDIX — TECHNICAL QUESTIONNAIRE FOR JUAN CALATAYUD (JC VALVULAR)

Appendix — Technical questionnaire for Juan Calatayud (JC Valvular)

Purpose. These sixteen questions are intended so that the designer of the Acronos P52 can answer with his own design and measurement data. With his answers almost all the ND parameters of the report would be closed.

By the type of answer they request:

• Design decisions (no measurement): P3, P4, P5, P6, P7, P10, P11, P14, P15, P16.
• Bench measurements (need a production unit): P1, P2, P8, P9, P12.
• Operational and commercial data: P13.

Block 1 — Power and dynamic behaviour

1. Real power at low distortion.

   The published 40 W nominal, at what harmonic distortion (THD) percentage do they correspond? And what is the maximum clean continuous power at 1 % THD over 4 Ω and over 8 Ω?

2. Output impedance and damping.

   What is the output impedance (Zout) measured in the audible band and the damping factor (DF) with 4 Ω and 8 Ω loads? How does the frequency response vary over the real impedance curve of a speaker with a 4.9 Ω minimum?

3. Output transformer taps.

   Does the secondary have dedicated selectable 4 Ω and 8 Ω taps, or a single optimised output? What is the optimum load impedance for each tap?

Block 2 — Valves, bias and small-signal architecture

1. Bias system and class of operation.

   How is the bias of the four 6550s adjusted? Are there accessible trimmers, individual or per-pair adjustment, or fixed bias by divider or cathode? What is the target quiescent current value per valve? Are the two blue trimmers visible at the ends of PCB-S associated with the signal stage adjustment or with the 6550 bias? Are there other trimmers on the underside? Additional confirmation: is the class of operation AB? And the subtype: AB1 (grid never positive with respect to the cathode) or AB2 (positive grid at peaks with grid current)?

2. Phase inverter topology.

   What topology does the ECC803S use to invert the phase: long-tailed pair (LTP, or cathode-coupled pair), cathodyne/concertina or another? If LTP, does it use both triodes of the ECC803S?

3. Global feedback (NFB).

   Does the P52 have global feedback (NFB)? How much, in dB? How does it contribute to the Zout value requested in question 2?

4. Ultralinear percentage (%UL).

   At what percentage of the output transformer winding are the screen grids connected (ultralinear)? Is it the usual 43 % reference or different?

Block 3 — Power supply, rails and safety

1. Real B+ voltage and dielectric margin.

   What is the real B+ DC voltage at the 6550 plates and what margin does it keep with respect to the 450 V rating of the EPCOS B43645 capacitors? Does the B+ vary between channels?

2. HV capacitor bank discharge.

   Does the supply have an automatic discharge circuit for the high-voltage bank when switching off? How long does it take for the residual voltage to drop below 50 V?

3. B+ supply architecture and branch separation.

   Is the supply of the 6550s, the ECC803S and the EF86 separated into independent branches with series resistors or chokes and their own RC networks, or do they share a common B+ rail with local bypass capacitors? If separated, how many branches are there?

4. Soft-start and start-up.

   How does the circuit manage the high-voltage delay to protect the cold cathodes and avoid the inrush current at switch-on? What function does the Hongfa HF115F relay perform and what specific component performs the timing?

Block 4 — Compatibility, noise and operational data

1. Residual noise and speaker compatibility.

   What is the level of audible residual noise (hum at 50/100 Hz and hiss) with the input shorted? What sensitivity and minimum speaker impedance do you recommend for obtaining maximum dynamics without clipping?

2. Operational data and support.

   What is the real net weight, the consumption at rest and at full power, the warranty and technical service procedure, and is the electrical schematic with test points delivered to the buyer?

3. PCB-S trimmers.

   How many Bourns trimmers are there on PCB-S and what function does each have? If there are two, are they intended for the signal stage or for the phase inverter? If there are two others on the underside, are they the individual bias for the four 6550s?

4. Function of the 10W10KJ.

   Is the 10W10KJ resistor (10 kΩ / 10 W) connected in series with the main B+ (bypassed by the Hongfa relay after start-up), or is it on the regulated auxiliary branch? What function does it exactly perform —inrush limiter, auxiliary branch bleeder, divider or regulator load—? What is the programmed delay and what component performs the timing?

5. Control architecture and auxiliary elements.

   How many functional boards does the P52 have? What function does the vertical PCB-3 with four blue components (probably relays) perform? What is the encapsulated element in the rear right corner: auxiliary transformer, mains input module/fuse holders, other? If there is an NE5532 IC (or similar), where does it get its low-voltage supply from? What are the two DIP-8 on PCB-P? What is the function of the gold terminals on the rear wall?

The nine critical purchase points

Point	Associated question	Why it is decisive
1	P1 — Power at 1 % THD	Determines whether the 40 W are usable with clean dynamics
2	P2 — Zout / DF	Defines the stability of the response over the real impedance
3	P4 — Bias adjustment and class of operation	Discriminates between bias topologies and confirms the class
4	P10 — B+ architecture	Confirms whether the pulsating power current modulates the sensitive stages
5	P9 — HV bank discharge	Determines physical safety when handling the interior
6	P13 — Warranty and schematic	Ensures operational continuity and repairability
7	P14 — PCB-S trimmers	Discriminates between alternative bias topologies
8	P15 — 10W10KJ and timing	Closes the resistor function and the start-up topology
9	P16 — Control architecture and PCB-3	Closes the function of PCB-3, the auxiliary element and the origin of Vaux

APPENDIX — UNVALIDATED EXTERNAL PROPOSALS

Appendix — Unvalidated external proposals

Mundorf, ClarityCap, Jupiter, V-Cap, Solen, Auricap, MLytic, Jensen, ASC, Nichicon KW (discarded due to voltage for the HV bank), Rubycon MXG/MXH (MXH not verified), Fine Gold, etc. They are kept for traceability of the research process, but they do not constitute a Nauscopio recommendation. Their eventual evaluation is fully subject to the Phase 0 → A → B → C → D protocol and to a level-matched, reproducible S3 test.

REFERENCES AND CREDITS

References and credits

Technical sources

• TDK/EPCOS — Aluminum electrolytic capacitors, snap-in, B43645 / B43655, 610 µF / 450 V. Nominal service life >5,000 h at 105 °C (B43645 series according to current documentation). Verify the full code B43645-S7617-M1 against the ordering codes table. (accessed 2026-10)
• Bourns — 3296 series, 3/8" square trimpot, 25-turn, ±100 ppm/°C. (accessed 2026-09)
• Texas Instruments — TL431 adjustable precision shunt regulator. (accessed 2026-09)
• Texas Instruments — NE5532 dual low-noise operational amplifier. Total supply specified 10–30 V. (accessed 2026-10)
• Hongfa — HF115F miniature high power relay. Family with capacity up to 16 A and 5 kV coil/contact isolation; there is an HF115F-I variant for high inrush loads. (accessed 2026-10)
• JJ Electronic — ECC803S, A.F. twin triode. μ=100, gm=1.6 mA/V, Ri=62.5 kΩ. (accessed 2026-09)
• JJ Electronic — 6550 beam pentode. Wa=35 W, Ua=600 V, Ug2=400 V, If=1.6 A. (accessed 2026-09)
• JJ Electronic — KT88 beam pentode. Wa=42 W, Ua=800 V, Ug2=600 V, If=1.6 A. (accessed 2026-09)
• Mullard Ltd. — Circuits for Audio Amplifiers, 1959 (reprinted 1962; the 5-20 design has been circulating since c. 1956). Contains the reference design of the 5-20 amplifier: EF86 + ECC83 + 2× EL34 push-pull class AB1 + GZ34. THD ≈ 0.1 % at 20 W / 1 kHz. Cited as topological analogy in §3.2. (historical document)
• Mullard — EF86, Z729, CV2901, CV4085, small signal pentode. Historical documentation from the mid-1950s. Source of the "low hum, low-microphony pentode" and of gm = 1.85 mA/V. (historical document 1953)
• Mullard — EL34 Data Sheet. Maximum screen voltage Vg2 ≈ 425 V. (historical document)
• Mullard — ECC83 Data Sheet. Ri = 62.5 kΩ, μ = 100, maximum cathode current ≈ 8–9 mA. (historical document)
• Tube Amp Doctor (TAD) — Technical Data ECC803S (PDF document). Maximum cathode current 8 mA; maximum DC plate voltage 330 V; maximum plate dissipation 1.2 W; maximum positive grid 0 V. Source of the data point in §3.3. amptek.es · TAD ECC803S (accessed 2026-10)
• TAD (Tube Amp Doctor) — EF86 / 6267 Premium Selected. (accessed 2026-09)
• Genalex / New Sensor — KT77 Gold Lion and KT88 Gold Lion datasheets. KT77 plate voltage: 800 V according to Genalex, 600 V according to alternative sources. (accessed 2026-10)
• RCA Receiving Tube Manual RC-30 (1975) and GE Essential Characteristics (1954) — Data for the 12AU7, 6SN7, 6CG7, 12BH7 and 5687 valves. Maximum 6SN7 cathode current ≈ 20 mA. Source of the table in §3.5.
• VTL S-200 — Commercial reference for the use of 12BH7 with 6550.
• Sonic Frontiers Power 2 — Commercial reference for the use of 5687 with 8× 6550.
• Conrad-Johnson Premier Eleven — Commercial reference for the use of 6CG7 as driver.
• Schaffner — FN9244B / FN9280 series IEC inlet filters. (accessed 2026-09)
• Würth Elektronik — WE-CMB common mode chokes. (accessed 2026-09)

Sources on the Acronos P52

• JC Valvular — manufacturing process video: Amplificador ACRONOS P52 Una obra de arte. Cómo se fabricó.
• Manufacturer's commercial spec sheet — 40 W PP UL with 6550 (without THD condition).
• Blog reader — personal communication: approximate P52 weight ~40 kg. Unverified.
• Interior images provided by the owner of the unit (2026-10) — they verify the silkscreen B43645-S7617-M1 on the three EPCOS units, identify the Hongfa HF115F relay, confirm the presence of blue Bourns trimmers on PCB-S and document the existence of vertical PCB-3, auxiliary magnetic element and longitudinal metal partition. They also provide the partial readings that support the PROBABLE identifications of the NE5532 and TL431, and the visual non-confirmation of the NE555.
• Field data (2026-10) — current power of the Acronos M2: 60 W per channel (hybrid). Public distributor spec sheets (High End Korner 2024: 50 W; Amigoshifi 2021: 40 W, possibly prototype) outdated with respect to the field unit.
• Acronos P52 vs Cayin Jazz 90 technical comparison (nauscopio, V3.5.0, October 2026) — source of the expanded development on the Mullard 5-20 (§3.2), the conflict of NFB figures for the 5-20 in the literature and the reinforcement of the four-convergent-observation AB1 argumentation (§5.2).

Image credits

• Acronos P52 images — JC Valvular manufacturer video, reproduced for technical analysis purposes.
• Additional images (2026-10) — provided by the owner of the unit, reproduced for technical analysis purposes.

Note on this V6.8.0. Layout revision with nauscopio V5.10 FINAL GSC + SOCIAL template, applied to V6.7 without technical content changes:

• JSON-LD updated: author.split with name/alternateName, author.url → https://nauscopio.wordpress.com/about/, about → Thing with @id (avoids offers/review/aggregateRating), isPartOf, isAccessibleForFree, ISO 8601 dates with +02:00 zone, top-level image with caption, citation array with 5 sources, mainEntityOfPage with @id, version 6.8.0.
• First image (§0): migrated to eager + fetchpriority=high, S1600, 1600×924, alt with og:image note, <a> wrapper for lightbox.
• Wrapper: <main> → <div class="report-main"> (compatible with theme V8.5+).
• aria-pressed of the theme toggle to false (the JS corrects it on render).
• Badge and footer updated to V6.8.0.
• GSC+SOCIAL checklist at the start (commented).
• No changes: basic guide G1–G6, §0–§7, complete Part II, questionnaire appendix, proposals appendix, references. All technical content, calculations, tables and evidence classifications are preserved in full with respect to V6.7.

Acronos P52 · Report V6.8.0 · Oct 2026 · nauscopio