Acronos P52 vs Cayin Jazz 90
Technical comparison between two push-pull valve amplifiers built at opposite ends of the market: European craftsmanship versus Asian industry.
Sources: official Cayin manual, JC Valvular website and video, reviews from fairaudio.de and lite-magazin.de. Reference system: 35-year-old loudspeaker — Crossover A v2.4 MR-A-01, 6 Ω nominal / 4.9 Ω minimum / 89 dB/W/m, listening at 3 m.
What it is about. This report compares two medium-power push-pull valve amplifiers built in opposite worlds: the Acronos P52, handcrafted and European, with no commercial network; and the Cayin Jazz 90, industrial and Asian, with global distribution. The basis of the analysis is public documentation and the study of high-resolution photographs.
What is known. Both carry eight valves and four power valves. The P52 uses 6550s and declares 40 W per channel; the Jazz 90 uses KT88s (or EL34s) with 50 W in ultralinear and 28 W in triode. Both incorporate quality capacitors in the signal path and an active regulation or filtering block on their power supply board. The Jazz 90 is documented down to the last detail by the manufacturer; the P52 is not.
What is not known. Neither publishes output impedance, full-power distortion or global feedback. Nor the exact function of the active power supply blocks, nor the unit value of the P52's three supply capacitors. Its mass circulates as an informal figure of ~40 kg, but a visual estimate gives 19–30 kg. And a key sonic unknown remains: the output transformer parameters of neither of the two.
How it is decided. By two independent axes: signal modernity (how it treats the music) and control modernity (how it manages itself). On signal, a technical tie. On control, the Jazz 90 wins clearly.
Technical glossary · fifteen terms in one line
GETTING STARTED — WITH NO PRIOR KNOWLEDGE (Basic guide G1–G5)
GETTING STARTED — WITH NO PRIOR KNOWLEDGE
This report compares two valve amplifiers. You do not need to be an electronics engineer to follow it, but it helps to have five basic ideas clear. We set them out here with numerical examples, and throughout the report they will be used without further explanation.
Basic guide: reading this report without being an expert
Central idea: a valve amplifier is not a black box. It is a circuit with three clearly differentiated blocks —input, driver and power— and a power supply that feeds them. Understanding that structure is half the work.
Imagine you want to make a speaker sound. The player delivers a very weak electrical signal —on the order of one volt— that by itself moves nothing. The amplifier has to take that signal and turn it into another capable of delivering tens of watts to the speaker. To do so it uses valves.
In a push-pull design, the power valves work in pairs: while one conducts more current, the other conducts less. That "push and pull" has two enormous advantages. The first is that it cancels the DC magnetism in the output transformer, which would otherwise saturate the core. The second is that it cancels the even harmonics of the distortion.
It is worth pausing here because it is a nuance that is often explained backwards. Even-order harmonics (2nd, 4th, 6th) are the most benign to the ear: the second sounds at the perfect octave above the fundamental and is perceived as warmth. Odd-order harmonics (3rd, 5th, 7th) are the ones that give that "metallic" sensation when they appear in quantity. Cancelling the even ones in push-pull is technically desirable, but it has a classic consequence: the residual distortion is dominated by the odd ones, which the ear tolerates worse. That is precisely the traditional criticism of push-pull versus single-ended, and it is worth bearing in mind so as not to confuse what push-pull contributes with what single-ended contributes.
Player → Preamp → Power stage → Output transformer → Speaker (≈ 1 V) (voltage) (current) (impedance matching) (4–8 Ω)
The output transformer is a key component. Valves work with high voltages and small currents; the speaker needs low voltages and large currents. The transformer acts as a translator between both worlds, and its quality influences the sound so much that in handcrafted amplifiers it can cost more than the rest of the circuit combined.
Each statement in the report carries a label indicating its level of evidence, visible in the side legend. The rule is simple: no statement rises above the category of its source evidence. If a data point comes from a photograph, it is "observed" and not "confirmed"; if it comes from a deduction, it is "probable"; if there is no basis to support it, it is "ND" (not determinable).
This is important because one of the two amplifiers —the P52— has very little public documentation. The temptation to fill the gaps with assumptions is great; the report resists it and explicitly records what is not known.
With the reference speaker (89 dB/W/m sensitivity) listening at 3 metres, the distance loss is about 9.5 dB. That means that with 1 W of power the acoustic level at that distance is 79.46 dB, a quiet conversation volume.
To reach a peak of 100 dB —a high but not painful volume— we would need, with a single speaker:
In stereo, the sum of two speakers adds between +3 and +6 dB depending on how the signals correlate, which reduces the required power. The room, if lively, can add another 2-4 dB. In practice, for normal domestic use, 15-20 W per channel well delivered is enough. The P52's 40 W and the Jazz 90's 50 W are ample margin.
Here comes one of the most important issues in the report. A "6 Ω" speaker does not present 6 Ω at all frequencies. Its impedance varies with frequency: it can drop to 4.9 Ω in one zone and rise to 16.5 Ω in another, drawing a wavy curve.
The amplifier and the speaker form a voltage divider. If the amplifier's output impedance (Zout) is very low, the voltage reaching the speaker is almost the same at all frequencies and the response is flat. If Zout is high, the voltage reaching the speaker follows its impedance curve, and that translates into a coloured sound.
The design criterion in solid state is that Zout be equal to or less than 0.5 Ω. In valve amplifiers the practical threshold is somewhat looser: between 0.5 and 1.0 Ω many respectable devices remain linear with reasonable loads. Above 1.5 Ω the alteration starts to be clearly audible. And here is the problem: neither amplifier publishes its Zout. It is the number-one pending measurement.
To fix the ideas, here are four simple calculations with their solutions:
- A 92 dB/W/m speaker: what level does it produce with 1 W at 2 m?
- With 89 dB/W/m, what power is needed for a 100 dB peak at 3 m with a single speaker?
- A 2000 µF filter with 300 mA consumption: what ripple does it present at 100 Hz?
- How many dB are there between 28 W and 50 W?
See solutions
- 92 − 20·log₁₀(2) = 86 dB
- 10^((100 − 79.46)/10) ≈ 113 W
- 0.3 / (100 · 0.002) = 1.5 V
- 10·log₁₀(50/28) ≈ 2.52 dB
0. Executive summary
This report compares two medium-power push-pull valve amplifiers. The first, the Acronos P52, is a Spanish handcrafted product built to order by JC Valvular. The second, the Cayin Jazz 90, is a Chinese industrial product with an international distribution network.
Both belong to the same conceptual family —push-pull with four power valves and an output transformer— but they sit at opposite ends of the market: one relies on tradition, manual component selection and the absence of intermediaries; the other on exhaustive documentation, the integration of digital functions and industrial quality control.
20-second read
What they are. Two medium-power push-pull valve amplifiers. P52: 2×40 W with 6550 valves. Jazz 90: 2×50 W in ultralinear or 2×28 W in triode with KT88 valves. CONFIRMED
What is known. Both use eight valves (input, driver and four output). Both incorporate an active regulation or filtering block on their power supply board. Both probably use an LTP-type phase inverter. PROBABLE
What is not known. The output impedance, the full-power distortion, the global feedback and the real supply voltage of neither of the two. The exact function of the active blocks on the power supply boards. The exact unit value of the P52's supply capacitors. The output transformer parameters of neither of the two. The real mass of the P52: an informal figure of about 40 kg circulates and a visual estimate by eye gives 19–30 kg, both unverified. ND
1. Method, evidence and limits
We start from a specific listening system: the 35-year-old loudspeaker — Crossover A v2.4 MR-A-01, with 6 Ω nominal, 4.9 Ω minimum and 89 dB/W/m sensitivity. The listening distance is 3 metres. The question guiding the report is: which of these two amplifiers fits that system better?
The initial limitation is important. Neither of the two amplifiers has independent measurements published in specialist magazines. For the Jazz 90 the manufacturer's specifications and its manual are known; for the P52, only what can be deduced from the photographs. That forces us to compare architecture, documentation and construction, not measured performance. It is an honest but limited comparison.
The following table summarises the most relevant parameters of each device, with an indication of the level of evidence. The valve complement is expressed per channel and total. It is the report's quick-reference table: if you only want to know "who has what", reading it is enough.
| Parameter | Acronos P52 | Cayin Jazz 90 |
|---|---|---|
| Configuration | Push-pull class AB with ultralinear connection: both traits CONFIRMED by the manufacturer. AB1 subtype PROBABLE, inferred from the driver and topology | Push-pull class AB1, switchable ultralinear/triode CONFIRMED |
| Topological lineage | Kinship with the Mullard 5-20, doubled to stereo with 6550s instead of EL34s. Not a replica PROBABLE | Modern Cayin in-house design |
| Power valves | 4× 6550 (2 per channel). The visible Tung-Sol brand corresponds to the Russian New Sensor reissue, not the original American one OBSERVED | 4× KT88 (2 per channel); accepts EL34, 6CA7, KT77 CONFIRMED |
| Input stage | 2× EF86 (low-noise pentode) OBSERVED | 2× 12AX7 CONFIRMED |
| Driver | 2× ECC803S OBSERVED | 2× 6SN7 CONFIRMED |
| Active block on power supply board | MOSFET+TL431 block on the power supply board. The combination is compatible with capacitance multiplier, B+ series regulator or bias reference. Exact function ND OBSERVED | Component with heatsink (TO-220) and auxiliary semiconductors on the power supply board, which is more complex than the P52's. Same type of block, one among several circuits on the board. Exact function ND OBSERVED |
| Phase inverter (LTP) | Probable LTP with ECC803S and tail resistor PROBABLE | Probable LTP with 6SN7 and tail resistor PROBABLE |
| Soft start | Blue SRD-type relay and 10 W ceramic resistor; components OBSERVED, soft-start function PROBABLE | High-voltage delay managed by MCU, about 60 seconds CONFIRMED |
| Rectifier | Solid state (KBU1010) OBSERVED | Solid state OBSERVED |
| Total valves | 8 CALCULATED | 8 CONFIRMED |
| Nominal power | 40 W per channel, without specified THD CONFIRMED | 50 W per channel UL / 28 W per channel TR RMS; THD 0.2 % at 1 W CONFIRMED |
| Full-power distortion | ND | ND |
| Output impedance (Zout) | ND | ND |
| Output transformer parameters | EI type with varnish OBSERVED. Primary inductance, leakage inductance, distributed capacitance and UL tap quality ND | Core type PROBABLE (EI hand-wound in the Jazz series). Sized above nominal power CONFIRMED. Primary inductance, leakage inductance and step response ND |
| Frequency response | ND | 9 Hz – 45 kHz (±3 dB, UL) CONFIRMED |
| Signal-to-noise ratio | ND | 100 dB (line) / 68 dB (phono) CONFIRMED |
| Bias | Internal blue Bourns 3296-type trimmers, visible in two zones. Total number and distribution not determined; bias could be shared per pair or individual OBSERVED / ND | Individual per valve, with VU-meter and BMI CONFIRMED |
| Supply capacitors | 3× TDK/EPCOS B43645 series (105 °C and Y vent visible). Order code not fully legible; unit value not determined OBSERVED / ND | 3× Nichicon 390 µF / 400 V (partial 400 V reading), 105 °C OBSERVED |
| Capacitor service life | > 5,000 h at 105 °C (B43645 series) CONFIRMED | ND |
| Global feedback (NFB) | ND | ND |
| Integrated phono | Not documented | Solid-state MM phono, RIAA ≤ ±0.25 dB, S/N 68 dB CONFIRMED |
| Bluetooth and DAC | Not available | BT 5.1 (QCC5125) with LDAC and aptX HD; ES9018K2M DAC only via BT CONFIRMED |
| Subwoofer output | Not available | Yes, stereo RCA at full frequency CONFIRMED |
| Headphone output | ND | 6.35 mm jack, 85 mW at 32 Ω CONFIRMED |
| Dimensions | ND | 420 × 379 × 215 mm CONFIRMED |
| Mass | Unverified. Informal figure of ~40 kg; visual estimate by eye of 19 to 30 kg. Neither comes from a scale ND | 27.5 kg according to manual; 25.9 kg according to lite-magazin |
| Maximum consumption | ND | 310 W CONFIRMED |
| Price | ND, built to order | €2,990 RRP; ~€2,093 with VIP25 discount CONFIRMED |
It is worth distinguishing four situations that are sometimes confused under the generic label "not known":
- Missing data. The manufacturer does not publish it. This is the case of Zout.
- Deduced data. Inferred from observation, with uncertainty. For example, the function of the P52's MOSFET+reference block.
- Unknown data. There is no basis for inferring it. For example, global NFB or the B+ voltage.
- Conflicting data. Different sources disagree. This is the case of UAT support in the Jazz 90.
| Label | P52 | Jazz 90 | Total |
|---|---|---|---|
| CONFIRMED | 3 | 18 | 21 |
| OBSERVED | 9 | 3 | 12 |
| CALCULATED | 1 | 0 | 1 |
| PROBABLE | 4 | 2 | 6 |
| ND | 14 | 6 | 20 |
| Total | 31 | 29 | 60 |
How the labels are counted. Each cell of the master table containing a label contributes one unit; a cell with two labels contributes two, one for each statement. Purely descriptive rows (without a label) do not count. Under this rule, the distribution shows a clear documentary asymmetry —the Jazz 90 has 18 data points confirmed versus 3 for the P52— but the P52 partly compensates with 9 observed data points versus 3 for the Jazz 90, thanks to the photographic analysis. That said: "observed" describes the presence of a component, not its function. Seeing a resistor is not the same as knowing what it is for. The counter measures quantity of public information, not quality of the equipment.
2. Identity, architecture and valves
Before going into details, it is worth seeing both devices from the inside. The following photographs show the interiors as each source captured them: the P52 through Juan Calatayud's manufacturing video, the Jazz 90 through the fairaudio.de review.
An integrated valve amplifier consists of three blocks: the input stage, which receives the weak signal from the player; the driver stage, which amplifies it and inverts it to feed the two branches of the push-pull; and the power stage, with the output valves and the output transformer.
Before comparing block by block, it is worth looking at the coupling capacitors, which are the ones that let the signal pass from one stage to another while blocking the DC component. Both devices use quality capacitors, although with different distributions and counts that are not directly comparable.
P52. It carries up to eighteen visible films in its signal tray: approximately sixteen yellow polypropylene capacitors (CEC according to the OCR) plus two gold K40Y-9 PIOs. Not all of them necessarily perform a coupling function; some could be snubbers or compensation networks, and separating one from the other would require the schematic.
Jazz 90. It shows four gold capacitors next to the power sockets, two white Realcaps (0.33 µF at 630 V) next to the central pair of small valves and two silver ones on the power supply board. The other pair of small valves shows no visible capacitor. With the current evidence, the two Realcaps are the only confirmable coupling films in the signal path; the gold ones could perform bias, cathode or snubber functions, and the silver ones are in the supply. No WIMA is observed in the photograph: the red stretches that are sometimes confused with capacitors are internal wiring thread.
The class of operation is confirmed in the case of the Jazz 90 (AB1 according to the official manual) and inferred in the case of the P52 (AB with AB1 subtype probable). In AB1 the valve grid never goes positive with respect to the cathode, which means that the driver only has to deliver voltage, not current. It is the most common mode in domestic amplifiers.
| Device | Class | Status |
|---|---|---|
| Cayin Jazz 90 | AB1 | CONFIRMED by the official manual. The grid never goes positive; the driver only delivers voltage. |
| Acronos P52 | AB, AB1 subtype PROBABLE | Based on the observed driver, the visible topology and the kinship with the Mullard 5-20. AB2 cannot be ruled out without a schematic. |
Why do we believe the P52 is AB1 and not AB2? For three reasons. First, its driver is an ECC803S, a voltage-gain valve but with a maximum cathode current of 8 mA, insufficient to drive a grid in conduction. Second, the resources typical of AB2 are not observed (additional valves, interstage transformer). Third, the declared power of 40 W per channel is perfectly compatible with AB1.
Control architecture: relays and selectors
It is also worth looking at the control architecture of each device: the relays that manage input switching, soft start and protections.
| Aspect | P52 | Jazz 90 |
|---|---|---|
| Input selector | Four relays on a vertical PCB, one per signal cable, in a layout compatible with an input selector. Next to a metal partition that separates it from the rest of the circuit PROBABLE | MCU with 18–28 pin DIP + 2× CD4052; two Hongfa ET1C-A012V relays for switching OBSERVED |
| Relays on rear board | Two relays: one blue SRD-type (probable soft-start) and another transparent one of larger size. Exact function of the second ND OBSERVED | Not observed in that location ND |
The P52 uses four 6550 valves from Tung-Sol, two per channel. The Jazz 90 uses four KT88s and allows switching to EL34 via a rear switch, which greatly expands the substitution options. Here it is worth understanding the differences between these four valve types, because they are not equivalent.
| Feature | 6550 | KT88 | EL34 | KT77 |
|---|---|---|---|---|
| Type | Beam tetrode | Beam tetrode | Pentode | Beam tetrode |
| Max plate dissipation | 35 W | 42 W | 25 W | 32 W |
| Max plate voltage | 600 V | 800 V | 800 V | ≤ 800 V |
| Max screen voltage | 400 V | 600 V | 425 V | ND |
| Heater current | 1.6 A | 1.6 A | 1.5 A | 1.5 A |
The KT88 is the most robust of the four, with the greatest margin of dissipation and screen voltage. The 6550 is its American cousin, with less margin but highly appreciated for its sound. The EL34 is an older pentode, with a classic British sound, with less available power but also less dissipation. The KT77 is a modern version designed as a direct substitute for the EL34.
The input stage is the first to see the weak signal from the player. Its mission is to amplify it without adding noise. Here the two amplifiers choose different paths.
| Aspect | EF86 (P52) | 12AX7 (Jazz 90) |
|---|---|---|
| Type | Low-noise audio-frequency pentode | High-gain double triode |
| Stage gain | Up to ≈ 185 with 100 kΩ load | μ ≈ 100 |
| Transconductance | ≈ 1.85 mA/V | ≈ 1.6 mA/V |
| Natural use | Low-level input, phono preamps | Preamp, voltage gain |
The EF86 was designed expressly for low-level, low-noise input stages, and it is the valve that the Mullard 5-20 used in its first stage. The 12AX7 is the most widely used double triode in history, an excellent all-rounder for many functions. The P52 bets on a valve specific to the task; the Jazz 90 uses the generic but proven solution.
To understand the P52's architecture it is worth knowing the Mullard 5-20, a reference design published by Mullard Ltd in the 1950s in its booklet «Circuits for Audio Amplifiers». The name comes from its five valves (EF86, ECC83, two EL34s and a GZ34 rectifier) and its twenty watts of output.
The P52 is not a copy of the 5-20: it is a modern evolution of its concept, doubled to stereo and with different power valves. The following table shows the correspondences.
| Stage | Mullard 5-20 (mono) | Acronos P52 (stereo) | Correspondence |
|---|---|---|---|
| Input | 1× EF86 | 2× EF86 | Functionally identical OBSERVED |
| Driver | 1× ECC83 | 2× ECC803S | Functional equivalent OBSERVED |
| Power | 2× EL34 push-pull AB1 | 4× 6550 (2 per channel) | Similar topology, different valve |
| Rectification | GZ34 (valve) | KBU1010 (solid state) | Different |
| Feedback | Global | ND | No automatic transfer |
| Nominal power | 20 W over 7.5–15 Ω | 40 W per channel CONFIRMED | Doubling to stereo does not imply 40 W per channel |
The P52 preserves the valve family of the British design but incorporates a modern MOSFET+reference block whose exact function is not known, and adds bias adjustment with trimmers. It is therefore a conceptual descendant, not a historical reproduction.
3. Driver, LTP and power supply block
This section is the most technical in the report, but also the one that clarifies why the comparison is won by nobody in the signal terrain.
Both the P52 and the Jazz 90 incorporate an active block on their power supply board. In the case of the P52 it is formed by a MOSFET and a TL431 chip, the classic combination of a regulation or active filtering circuit. In the case of the Jazz 90 a component with a heatsink (TO-220) is observed with auxiliary semiconductors, within a power supply board considerably more complex than the P52's: it carries more transistors, integrated circuits and small capacitors. It is a block of the same type, but one among several circuits on that board.
It is worth pausing on a question that is sometimes posed badly: could it be a constant current source (CCS) integrated in the signal stage?
What it could be, then
A MOSFET+reference block located on the power supply board responds to one of these three patterns, all of them known in well-designed valve amplifiers:
| Hypothesis | Function | Consequence |
|---|---|---|
| Capacitance multiplier | MOSFET with a capacitor on the gate that multiplies the effective capacitance and filters the B+ ripple. | Real improvement: very low ripple without enormous capacitors. A mark of sophistication. |
| B+ series regulator | The reference fixes the voltage; the MOSFET acts as a pass element and keeps B+ constant against load or mains variations. | Real improvement: stable B+ in all conditions, less dynamic intermodulation. |
| Bias reference | The reference generates a stable voltage for the power valve bias circuit. | Neutral with respect to the signal. Improves bias stability. |
The first two hypotheses would be good news for both devices. A regulated or multiplied B+ is unusual in amplifiers at this price, and in a push-pull the quality of the supply rail reaching the driver and the inverter largely determines the cleanliness of the sound, because those circuits work with very small signals and any ripple leaks amplified into the output.
Consequence for the comparison
Since both amplifiers have an active block on their power supply board, its presence cannot be presented as a differential feature of either. The technical advantage that could be attributed to the P52 on this count disappears, and the balance is levelled.
What is clear is that both manufacturers have decided to invest in power supply quality. In the case of the P52, the block coexists with an EPCOS 105 °C bank and a metal partition between supply and signal. In the case of the Jazz 90, it coexists with three 390 µF Nichicons at 105 °C (1,170 µF in total, less than the P52's bank), a toroidal transformer of its own design, an MCU control system and a board with more transistors, integrated circuits and small capacitors. The fact that the Jazz board is more complex does not imply a better or worse supply. They are two different paths to the same goal.
A technical detail that is sometimes posed badly. In class AB1 the power valve grid never goes positive with respect to the cathode, so there is no DC grid current: the driver does not have to supply DC current. The problem is another one, and it is high-frequency: charging the Miller capacitance of the output valve demands instantaneous alternating current. The formula is I_peak = 2π · f · C · V. With C ≈ 100 pF, f = 20 kHz and V = 20 V peak, that gives approximately 0.25 mA peak.
In absolute terms it seems little, but the question is not the magnitude: it is the linearity with which the driver delivers that current. A driver working with 0.5–1 mA quiescent (ECC803S) can enter the curved zone when delivering 0.25 mA peak, while one with 5–8 mA quiescent (6SN7) does so in the linear zone. This is the real difference, and it is a problem of slew rate and high-frequency linearity, not of direct current.
It is worth not declaring a winner on that single data point, however. The ECC803S has a higher open-loop gain (μ ≈ 100 versus μ ≈ 20 for the 6SN7) and a much higher internal resistance (62.5 kΩ versus 7.7 kΩ). If the designer takes advantage of the higher gain to apply more global NFB, he can partially compensate for his current disadvantage and reduce the resulting Zout. That is: the current data point favours the Jazz 90 in the driver, but the available gain data point favours the P52 in the NFB. Neither of the two assets is decisive without measuring the final result.
| Aspect | P52 | Jazz 90 |
|---|---|---|
| Driver valve | ECC803S (μ=100, Ri=62.5 kΩ, maximum cathode current 8 mA) | 6SN7 (μ=20, Ri=7.7 kΩ, maximum cathode current 20 mA) |
| Probable inverter type | LTP PROBABLE | LTP PROBABLE |
| Quiescent current advantage | Low (0.5–1 mA). Narrow linear zone for Miller loading. | High (5–8 mA). Ample margin for Miller loading. |
| Gain advantage | High (μ = 100). More NFB available if the designer uses it. | Low (μ = 20). Less margin for global NFB. |
The following diagram summarises the signal and power chains of each device in a single stroke. The ND labels indicate unavailable parameters. The annotations in parentheses indicate where the photographic evidence stops: the underside of the printed circuit boards is not documented in either device.
╔══ ACRONOS P52 · kinship with Mullard 5-20, modern evolution ══╗
SIGNAL
[RCA in] → [EF86 ×2] → [ECC803S ×2] → [4× 6550 PP UL] → [EI output trafo] → [Speaker]
(pentode) (driver / LTP) (2 per channel) (impedances ND)
· possible CCS in LTP tail (underside of
· possible CCS in EF86 plate load PCB not
photographed)
POWER SUPPLY
[AC] → [EI trafo] → [SS rect. KBU1010] → [3× EPCOS 105 °C · 520–760 µF/u] → [B+ ND]
(not valve) ripple ≈ 0.88–1.28 Vpp (I = 200 mA)
├→ [SRD-type relay + 10 W ceramic resistor · soft-start PROBABLE]
├→ [Relays: 4 on vertical PCB (function ND) + 2 on rear board]
└→ [MOSFET + TL431 block · function ND: multiplier, regulator or reference]
[Global NFB: ND] · [Class AB1: PROBABLE] · [B+ delay: PROBABLE]
[Estimated total bank: 1,560–2,280 µF] · [Capacitor service life: > 5,000 h at 105 °C]
╔══ CAYIN JAZZ 90 ══╗
SIGNAL
[RCA / BT / Phono] → [12AX7 ×2] → [6SN7 ×2] → [4× KT88 PP UL/TR] → [Output trafo] → [Speaker 4/8 Ω]
(preamp) (driver / LTP) (2 per channel) (core type EI PROBABLE)
probable LTP with tail resistor
POWER SUPPLY
[AC] → [Toroidal trafo] → [3× Nichicon 390 µF / 400 V · 105 °C] → [B+ ND]
ripple ≈ 1.71 Vpp (I = 200 mA)
├→ [MCU + BMI · bias monitored per valve]
├→ [Speaker protection relays · B+ delay ~60 s]
├→ [MOSFET + reference block · function ND]
└→ [ES9018K2M DAC · BT 5.1 · Solid-state MM phono]
[Global NFB: ND] · [Class AB1: CONFIRMED] · [6SN7 driver: 20 mA quiescent]
[50 W UL / 28 W TR per channel] · [THD 0.2 % at 1 W] · [Total bank: 1,170 µF]
Three observations that the diagram makes visible at a glance:
- Both devices have a MOSFET+reference block on their power supply board, not in the signal path. The consequence is discussed in §3.1: by location, neither can be a signal CCS, although the underside of the boards leaves that possibility open.
- The P52's capacitor bank is larger than the Jazz 90's at any value in the 520–760 µF per unit range (1,560–2,280 µF total versus 1,170 µF). The calculated ripple is consequently lower in the P52 under the same consumption hypothesis, although the presence of active blocks in both devices can reduce the real ripple far below these figures.
- The control architecture diverges completely. The P52 has six relays distributed across two boards with partially unknown functions; the Jazz 90 concentrates control in an MCU with BMI, confirmed speaker protection and documented B+ delay.
4. Power supply, bias and protection
The power supply converts the alternating current from the mains into the direct voltages the valves need. The quality of that conversion directly influences the sound: a poorly filtered ripple leaks into the signal and is heard as background hum.
The critical element is the high-voltage electrolytic capacitors. The total capacity of the bank determines the ripple level: the more capacity, the less ripple. The approximate formula is this:
| Aspect | P52 | Jazz 90 |
|---|---|---|
| Code and unit value | 3× TDK/EPCOS. Only a repeated line of the type «LL … 7?17-A1» and the category «…/105/56» can be read from the silkscreen; the full order code is not visible. The unit value cannot be deduced; the range of 520 to 760 µF is a working hypothesis OBSERVED / ND | 3× Nichicon, 390 µF / 400 V (partial 400 V reading), 105 °C, read on the edge of the capacitors OBSERVED |
| Total bank capacity | Estimated, unconfirmed: between 1,560 and 2,280 µF depending on unit value ND | 1,170 µF (3 × 390) CALCULATED |
| Estimated ripple (200 mA, 100 Hz) | Between 0.88 and 1.28 Vpp depending on unit value CALCULATED | ≈ 1.71 Vpp for 1,170 µF CALCULATED |
| Nominal temperature | 105 °C OBSERVED | 105 °C OBSERVED |
| Service life at maximum temperature | > 5,000 h at 105 °C (B43645 series, TDK) CONFIRMED | ND |
Applying high voltage to a valve whose cathode has not yet warmed accelerates its ageing. That is why well-designed amplifiers wait between 30 and 60 seconds before applying the working voltage. This function is called soft start or B+ delay.
Bias is the adjustment of the quiescent current flowing through the power valves. A correct setting is decisive for their service life and for the amplifier's behaviour. Here the difference between the two devices is notable.
| Aspect | P52 | Jazz 90 |
|---|---|---|
| Bias adjustment | Internal blue Bourns 3296-type trimmers, visible in two zones. Their total number and distribution are not confirmed: if there were only two for four valves, the bias would be shared per pair and would require matched quartets; with more trimmers, it could be individual or per channel. Exact function ND | Individual per valve, with four potentiometers accessible from above and a VU-meter switchable as bias monitor CONFIRMED |
| Procedure documentation | No public procedure exists | Official manual with detailed procedure CONFIRMED |
| Valve protection | No valve protection circuit identified. The four relays on the vertical board are probably for input selection, not protection ND | BMI system with MCU, distributed hardware, overcurrent comparator and relay actuation CONFIRMED |
| Speaker protection (DC) | A single candidate relay on the rear board. No visible DC detection circuit. It is ND, not confirmed absence: the underside of the PCB is not photographed | BMI system with MCU; protection relays in series with the speaker output CONFIRMED |
| Recommended burn-in | ND | 50 h minimum, 100 h recommended CONFIRMED |
| Minimum off/on interval | ND | 2 minutes CONFIRMED |
| Matched valves | ND | Yes CONFIRMED |
The physical construction of a valve amplifier greatly influences its behaviour. The separation between the power supply and the signal section, the type of transformer, the wiring routing: everything counts.
| Aspect | P52 | Jazz 90 |
|---|---|---|
| Power transformer | EI type OBSERVED. An unidentified auxiliary transformer appears; function ND | Toroidal, of its own design and in-house manufacture CONFIRMED |
| Output transformers | EI type with varnish OBSERVED | New development, sized above nominal power and capable of working with loads below 4 Ω CONFIRMED. Core type PROBABLE: EI for the Jazz series and documented construction, not toroidal (the toroidal data point refers to the mains transformer) |
| Internal separation | Metal partitions in several zones: longitudinal partition between supply and signal, and partition between the vertical selector PCB and the rest OBSERVED. Real effectiveness not measured. | Compact layout with transformers in a row; shielded metal box for BT and DAC OBSERVED |
| Heater wiring | Twisted blue/red/green bundles OBSERVED | Brown/orange twisted pair to cancel hum OBSERVED |
| Maximum consumption | ND | 310 W CONFIRMED |
The mass of a valve amplifier is a data point that is sometimes handled imprecisely, but it matters: it gives an idea of the volume of iron and copper (transformers, chassis, heatsinks), and also of the ease of placing it in a cabinet. In the case of the P52, the figure that circulated in forums and comments was approximately 40 kg, but it did not come from a scale but from informal communication. The analysis of the photographs of the assembly process allows that data point to be bounded with more criteria.
From the image, three considerations allow the weight to be bounded:
- The three EI transformers. In view of their relative size with respect to the operator's hand and the chassis, each one probably weighs between 4 and 6 kg, without reaching the values of a 100 W monoblock. The three together would add up to between 12 and 18 kg.
- The chassis. Folded sheet metal with a glossy finish, sides not visible, top cover perforated for the sockets. A chassis of this size in 1.5–2 mm sheet metal weighs between 5 and 8 kg.
- The electronics. PCBs, capacitors, resistors, sockets, wiring, potentiometers and connectors: between 2 and 4 kg.
5. Functionality and connectivity
This section is the most practical in the report for the buyer. It does not talk about sound but about what each amplifier allows you to do without buying anything else.
| Function | P52 | Jazz 90 |
|---|---|---|
| Line inputs | Probably four relay-switched inputs, managed by the vertical selector PCB PROBABLE | 2× RCA line, 100 kΩ / 300 mV CONFIRMED |
| Phono input | Not documented | Solid-state MM phono, 47 kΩ / 3 mV / S/N 68 dB / RIAA ≤ ±0.25 dB CONFIRMED |
| Bluetooth | Not available | BT 5.1 (QCC5125) with LDAC, aptX HD, aptX LL, aptX, AAC, MP3, SBC CONFIRMED |
| Integrated DAC | Not available | ESS ES9018K2M (Sabre32, 32-bit) only via Bluetooth CONFIRMED |
| Subwoofer output | Not available | Yes, stereo RCA at full frequency CONFIRMED |
| Headphone output | ND | 6.35 mm jack, 85 mW at 32 Ω, 16–300 Ω CONFIRMED |
| Remote control | Not documented | Full infrared CONFIRMED |
| VU meters | Not equipped | Dual function: level and bias OBSERVED |
| Speaker outputs | Gold terminals visible; impedances ND | Screw terminals: 4 Ω and 8 Ω CONFIRMED |
6. Power, SPL and load matching
For the reference system —89 dB/W/m sensitivity, listening at 3 m— the distance loss is about 9.5 dB. With 1 W the resulting level is 79.46 dB from one speaker. The sum of two speakers in stereo adds between +3 dB (energetic sum) and +6 dB (coherent sum); the room can add between 2 and 4 dB more depending on its reverberation.
| Scenario | Target peak SPL @3 m | W/ch with +3 dB | W/ch with +6 dB | W/ch with +8 dB |
|---|---|---|---|---|
| Domestic | 95 dB | 18.0 W | 9.0 W | 5.7 W |
| Elevated | 100 dB | 56.8 W | 28.5 W | 18.0 W |
| Monitoring | 105 dB | 179.6 W | 90.0 W | 56.8 W |
SPL and required power calculator
Formulas: SPL_1W@d = sens − 20·log10(d) and P = 10^((target SPL − SPL_1W@d − factor)/10). Results per channel.
This is the most important technical section of the report. It addresses the only parameter that can audibly alter the response of the complete system: the amplifier output impedance in relation to the varying impedance of the speaker.
How amplifier and speaker interact
The voltage divider formula is this:
The smaller Zout is, the less the frequency response varies when traversing the speaker's impedance curve. The reference thresholds depend on the type of amplifier:
| Criterion | Threshold | Meaning |
|---|---|---|
| Ideal design Zout | ≤ 0.5 Ω | Spectral variation ≈ 0.58 dB between 4.9 and 16.5 Ω. It is the target handled in solid state and in the best valve amplifiers with high NFB. |
| Respectable Zout in valve amplifier | 0.5 – 1.0 Ω | Many well-regarded valve amplifiers work in this range. DF between 6 and 12. The spectral variation is around 1 dB, audible in direct comparison but not "coloured". |
| High Zout | > 1.5 Ω | Variation greater than 1.5 dB over the real curve. Here one can speak of audible coloration. |
Attenuation table by Zout
| Hypothetical Zout | DF @6 Ω | Attenuation @6 Ω | Attenuation @4.9 Ω | Alteration |
|---|---|---|---|---|
| 1 Ω | 6.0 | −1.34 dB | −1.61 dB | 0.27 dB — negligible |
| 2 Ω | 3.0 | −2.50 dB | −2.97 dB | 0.47 dB — marginal |
| 4 Ω | 1.5 | −4.44 dB | −5.18 dB | 0.75 dB — perceptible under critical conditions |
Spectral alteration chart
Negative feedback: the link with Zout
Global negative feedback (NFB) is the main tool designers have to lower Zout. The more NFB, the lower the Zout, the wider the bandwidth and the less distortion —at the cost of losing some naturalness and gain.
| Configuration | Typical Zout | DF over 6 Ω |
|---|---|---|
| No global NFB | 2 to 4 Ω | ≈ 1.5 – 3 |
| 6 dB global NFB | 1 to 2 Ω | ≈ 3 – 6 |
| 12 dB global NFB | 0.5 to 1 Ω | ≈ 6 – 12 |
| 20 dB global NFB | 0.2 to 0.4 Ω | ≈ 15 – 30 |
Response variation calculator by Zout
7. Axis A — Signal modernity
We call signal modernity the degree to which each amplifier incorporates contemporary technical solutions for treating the musical signal: active loading of the input stage, the phase inverter tail, the quality of the coupling capacitors, the type of driver and the bias management.
It is a different axis from the one seen in the following section. Here it is not about "more or fewer functions" but about how the path through which the music passes is built.
Side-by-side comparison
| Element | Acronos P52 | Cayin Jazz 90 |
|---|---|---|
| Input stage | 2× EF86, pentode designed for low noise OBSERVED | 2× 12AX7, generic high-gain double triode CONFIRMED |
| Driver | 2× ECC803S, low quiescent current, high gain OBSERVED | 2× 6SN7, high quiescent current, low gain CONFIRMED |
| Phase inverter | Probable LTP with tail resistor PROBABLE | Probable LTP with tail resistor PROBABLE |
| Coupling capacitors | Up to 18 visible films in the signal tray (16 yellow CEC per OCR + 2 gold K40Y-9 PIO); not all necessarily coupling OBSERVED | 2 confirmable coupling films in the signal path (white Realcaps). 4 gold ones next to the power sockets and 2 silver ones in the supply, function to be determined OBSERVED |
| Power valves | 6550 Tung-Sol (Russian New Sensor reissue), beam tetrode OBSERVED | KT88, British beam tetrode; accepts EL34, 6CA7, KT77 CONFIRMED |
| Rectification | Solid state (KBU1010) OBSERVED | Solid state OBSERVED |
Interpretation
In quantity of visible films, the P52 has up to 18 in its signal tray versus 2 confirmable coupling films in the Jazz 90's signal path. The exact number of coupling films in each device is not confirmed: in the P52 some could be snubbers or compensation networks, and in the Jazz 90 the four gold ones could perform other functions. In any case, the P52 has a clear material advantage in quantity of films, without that equating to better sound.
By contrast, the Jazz 90 has an advantage in the driver current reserve. Its 6SN7 admits 20 mA per cathode versus the 8 mA of the P52's ECC803S, and it works with much higher quiescent currents that allow it to deliver the Miller capacitance charging current with more linearity at high frequency. It is a slew rate problem, not a direct current one, but the effect on the high band is real.
The P52 partly compensates because its ECC803S has much more available gain (μ ≈ 100 versus μ ≈ 20). If the designer takes advantage of it to apply more global NFB, he can reduce the resulting Zout and linearise the response. The net result cannot be declared without measuring.
In phase inverter topology, both probably use an LTP with tail resistor. The MOSFET+reference blocks of both devices are on the power supply board, so they are not part of this axis: any improvement they contribute is a supply one, not a signal one.
A constructive detail worth mentioning: the P52 uses metal partitions in several zones of the chassis to separate functional blocks (supply and signal, and input selector from the rest). It is a detail engineering decision, not a technological advantage, and it does not change the balance of the axis, but it is part of the general care that the manufacturer puts into the mechanical design of the device.
Axis A verdict
Real tie, without a clear advantage for either of the two.
In quantity of visible films the P52 has a clear material advantage. In driver current reserve the Jazz 90 has a clear advantage. In gain available for NFB the P52 has the advantage. In signal topology both are equal. In input stage architecture both have valid solutions by different paths.
The net result is a tie. The P52's quantity of films, the Jazz 90's driver current and the P52's available gain compensate each other approximately, without any of the three assets being decisive on its own.
8. Axis B — Control modernity
The control axis measures how each amplifier manages its own operation: bias monitoring, automatic protections, soft start, maintenance accessibility and integrated connectivity.
Here we are not talking about musical signal but about everything that happens around it. On this axis, the Jazz 90 plays in another league.
Side-by-side comparison
| Element | Acronos P52 | Cayin Jazz 90 |
|---|---|---|
| Bias management | Internal trimmers, not accessible without opening the chassis. Number and distribution unconfirmed: bias could be shared per pair or individual. No public procedure. | Four individual potentiometers accessible from above, with VU-meter switchable as monitor. Detailed procedure in the manual. |
| Monitoring | No visible instrumentation. | BMI system with MCU; vertical boards per valve; overcurrent comparator. |
| Speaker protection | A single candidate relay on the rear board; exact function ND. | Protection relays in series with the output, managed by MCU. |
| Soft start | SRD-type relay and ceramic resistor observed; function probable but not confirmed. | Documented high-voltage delay, about 60 seconds. |
| Active supply | MOSFET+reference block on the power supply board; function ND. | MOSFET+reference block on the power supply board; function ND. |
| Connectivity | Analogue inputs only. No Bluetooth, no DAC, no phono. | Solid-state MM phono, Bluetooth 5.1 with LDAC, ES9018K2M DAC, subwoofer output, headphone output, remote control. |
| Distribution network | Handcrafted manufacturer, built to order. | Established distributors, including Lyric Audio in Spain. |
| Technical documentation | Practically non-existent. | Complete official manual with specifications, procedures and warnings. |
Interpretation
The Jazz 90 does not just do more things: it does them more accessibly to the user. Its BMI system monitors the current of each power valve and acts if it detects deviation, its bias can be adjusted without opening the chassis and with individual control per valve, and its manual explains the procedure step by step. In the event of a valve failure, the protection acts without the user having to do anything.
The P52, by contrast, does not publish its bias procedure, and the layout of its trimmers is not confirmed: if the bias were shared per pair, any valve replacement would require a matched quartet and any drift diagnosis would be harder to isolate. For any operation related to the valves a technician is needed who will probably have to consult the manufacturer. It is the price of handcrafted manufacture: without a support network, every operation becomes a project.
In the active supply section, the two devices are tied: both have a MOSFET+reference block on their power supply board, with function still to be determined. It is not a differential asset of either.
Axis B verdict
The Jazz 90 wins clearly. It is not a question of nuances: it is a structural difference. The Jazz 90 is designed so that the user can maintain it; the P52 is designed so that the manufacturer maintains it. And since the manufacturer has no support network, in practice that means that the P52 requires a direct relationship with the person who made it, while the Jazz 90 can be maintained with the documentation in hand.
This axis weighs heavily in daily use. A valve amplifier requires valve changes every few thousand hours and periodic bias adjustments. In the Jazz 90 the adjustment is individual and free; in the P52 the adjustment is internal and probably goes through a technician. Over five years, the difference is not only monetary: it is one of operational viability.
9. What neither of the two incorporates
So that the reader does not form a mistaken idea, it is worth listing the modern valve amplification techniques that do not appear in either device. Both are evolutions of the classic school, not avant-garde designs.
| Technique | Description | P52 | Jazz 90 |
|---|---|---|---|
| Gyrator (active load) | Simulated inductance with active components. Maximises gain and linearity of a stage. | Not observed ND | Not observed ND |
| MOSFET follower (PowerDrive) | MOSFET source follower between the driver and the output valve grid. Eliminates the coupling capacitor. | Not observed ND | Not observed ND |
| Cathode feedback | Applies a fraction of the output signal to the cathode of the power valves. | Not observed ND | Not observed ND |
| Heater elevation | Connects a positive DC voltage to the heater winding to reduce hum. | Not observed ND | Not observed ND |
| DC heaters | Heater supply with rectified direct current. Eliminates 50 Hz hum. | Not observed; twisted AC bundles OBSERVED | Not observed; twisted AC pair OBSERVED |
| Auto-bias / bias servo | Circuit that automatically adjusts the bias. | No; manual bias with trimmers OBSERVED | No; manual bias assisted by VU-meter and BMI CONFIRMED |
| SMPS (switched-mode power supply) | Switched-mode power supply instead of a linear transformer. | No; linear EI transformer OBSERVED | No; linear toroidal transformer in the supply OBSERVED |
| Maida regulator | High-voltage regulator based on an LM317 floating on a high-voltage MOSFET. | Not observed ND | Not observed ND |
| Direct coupling (DC) | Elimination of the coupling capacitor between stages. | No; coupling capacitors observed OBSERVED | No; coupling capacitors observed OBSERVED |
10. Total cost of ownership
The purchase price is only part of the real cost of a valve amplifier. Over its useful life one must add replacement valves, maintenance labour and electricity consumption. Let us estimate it over five years.
Purchase price
| Item | P52 | Jazz 90 |
|---|---|---|
| Purchase price | ND, built to order. The manufacturer does not publish a price list. | Official RRP €2,990; with the 30 % VIP25 discount, approximately €2,093 CONFIRMED |
Replacement valves
| Component | P52 | Jazz 90 |
|---|---|---|
| Power quartet | 4× 6550 (New Sensor or other current manufacturers) ≈ €160–220 | 4× KT88 ≈ €240; EL34/6CA7/KT77 ≈ €120–250 |
| Input/driver valves | 2× EF86 + 2× ECC803S ≈ €80–120 | 2× 12AX7 + 2× 6SN7 ≈ €60–100 |
| Total replacement in KT88 | ≈ €240–340 | ≈ €300–340 |
| Total replacement in EL34/6CA7/KT77 | Not applicable | ≈ €180–350 |
Labour and maintenance
| Aspect | P52 | Jazz 90 |
|---|---|---|
| Bias procedure | Requires technician with access to the manufacturer. Estimated cost €50–80 per visit. | The user can execute it with the manual and the VU-meter. Cost €0. |
| Overcurrent protection | A single candidate relay on the rear board; function partially ND. | Active BMI system. Protection against valve failure at no additional cost. |
| Spare parts availability | Handcrafted manufacturer, built to order. Lead times ND. | Established distributor network. |
Five-year estimate
Acronos P52: ≈ €240–340 (valves) + €100–160 (two technician visits for bias) ≈ €340–500, plus the purchase price (ND). Energy consumption is not included because the manufacturer does not publish it.
Cayin Jazz 90 in KT88: ≈ €300–340 (valves) + €0 (autonomous bias with VU-meter) + ≈ €170 (energy: 310 W × 2 h/day × 365 × 5 = 1,131.5 kWh × €0.15/kWh; 310 W is the maximum consumption, so it is an upper bound) ≈ €470–510, plus the purchase price (€2,093–2,990).
Cayin Jazz 90 in EL34/6CA7/KT77: the valve cost drops to €180–350, so the total over five years would be €350–520 before the purchase price. The P52 does not accept this switching: its power valve is the 6550 and there is no documented switch to change family.
11. Conclusions and decision table
What we know
- The Jazz 90 is far better documented. Topology, class of operation, bias procedure, protections and electrical specifications are published by the manufacturer.
- The P52 shows signs of good construction and careful topology. Metal partitions in several zones of the chassis, low-noise EF86 at the input, 105 °C EPCOS capacitors in the supply, MOSFET+reference block on the power supply board.
- Both devices have an active block on their power supply board. It is not a differential feature of either. The exact function is to be determined in both.
- Both use quality capacitors in the signal path. The "P52 boutique, Jazz 90 basic" narrative is incorrect: the Jazz 90 uses four gold capacitors, two white Realcaps and two silver ones in the supply.
- The difference in quantity of films is real but not decisive. The P52 has up to 18 visible films in its signal tray versus 2 confirmable coupling films in the Jazz 90's path, but the exact number of coupling films in each device is not confirmed and quantity does not measure quality.
- The P52's mass is not verified. The visual estimate (19–30 kg, order of magnitude) does not support the ~40 kg, but does not rule it out either.
- The P52's bias architecture is not confirmed. Internal trimmers are seen in two zones; if there were two for four valves (bias per pair), the maintenance difference with the Jazz 90's four individual controls would be real; with more trimmers, less so.
The two axes, summarised
| Axis | P52 | Jazz 90 | Winner |
|---|---|---|---|
| Axis A · Signal | Probable LTP with tail resistor; low-noise EF86; high-gain ECC803S; up to 18 visible films on board (not all necessarily coupling); shared MOSFET+reference block | Probable LTP with tail resistor; high-current 6SN7; 2 confirmed coupling films in the signal path; shared MOSFET+reference block | Tie |
| Axis B · Control | Internal bias with trimmers (number and distribution ND); probable soft start; no MCU; no documented protection; no connectivity | BMI with MCU; individual bias per valve; documented soft start; speaker protection; DAC/BT/phono; distributor network | Jazz 90 |
What we do not know
- Zout of neither of the two. Priority number one.
- Output transformer parameters of neither of the two. The component that most determines the sound.
- Full-power distortion of neither.
- Exact function of each device's MOSFET+reference block. Requires comparative tracing of the two power supply boards.
- Unit value of the P52's EPCOS bank. The code does not allow resolving the exact value.
- Exact nominal voltage of the Jazz 90's Nichicons (400 V, partial reading).
- Real mass of the P52 without packaging. The visual estimate of 19–30 kg is only indicative.
- Number and distribution of the P52's bias trimmers.
- Global feedback, B+ voltage and supply ripple of both.
What would change the conclusion
- P52 Zout ≤ 0.9 Ω: it would become a clear alternative.
- Substantial differences in the two devices' supply blocks: they could break the Axis A tie, if one of the two blocks turns out to be a complete regulator and the other only a bias reference.
- Square-wave step response of the OPTs: it would reveal the relative quality of the two output transformers, which is where a good part of the sound is decided.
- Confirmation of the P52's bias architecture: if it were individual with four trimmers, it would shorten the Axis B difference; if it were per pair with two, it would confirm it.
- Re-reading the TDK code at macro resolution: it would close the P52's bank value and ripple.
Quick decision table
| If your priority is… | Guidance | Reason |
|---|---|---|
| Lower documentary uncertainty | Jazz 90 | Extensive official spec sheet, documented bias procedure, declared protections. |
| All-in-one integration | Jazz 90 | DAC, Bluetooth, MM phono and subwoofer output, without external equipment. |
| Greater quantity of quality capacitors | P52 | Up to 18 visible films in the signal tray versus 2 confirmable coupling films in the Jazz 90's path. Quantity does not measure quality. |
| Purist philosophy · handcrafted construction | P52 | Handcrafted construction, high-voltage bank with three identified TDK/EPCOS capacitors, metal partitions in several zones of the chassis. |
| Kinship with the classic British school | P52 | Shares the Mullard 5-20 valve family, doubled to stereo with 6550s. |
| Driver with higher quiescent current | Jazz 90 | The 6SN7 admits 20 mA versus the 8 mA of the P52's ECC803S, and delivers the Miller capacitance charging current with more linearity at high frequency. |
| Autonomous maintenance and individual bias | Jazz 90 | Official manual, VU-meter, BMI and individual bias per valve. The P52 has internal trimmers of unconfirmed number and distribution. |
| Active speaker protection | Jazz 90 | The Jazz 90 has confirmed speaker protection with relays in series with the output. In the P52 there is only a candidate relay on the rear board, probably for soft start; no confirmed speaker protection. |
| Supply capacitor longevity | No clear difference | Both banks are 105 °C (TDK/EPCOS and Nichicon). Only the TDK series service life is documented (over 5,000 h at 105 °C); the Nichicon's is ND. |
| Power valve flexibility (tube rolling) | Jazz 90 | Declared KT88/EL34 switching with individual bias. In the P52, if the bias were per pair (unconfirmed), tube rolling would be limited to factory-matched quartets. |
| Integrated MM phono | Jazz 90 | Solid-state stage with RIAA error ≤ ±0.25 dB and S/N 68 dB. |
| Greater declared gross mass | Neither (tie) | The Jazz 90 declares 27.5 kg. The P52 has no verified weight, but a visual estimate by eye gives 19–30 kg, without a scale. There is no firm basis either to say the P52 is heavier or to say it is not. |
| Long-term support continuity | Jazz 90 | Established distributor network. The P52 depends on a single person and has no public bias procedure: if the manufacturer ceased its activity, the device would be left without documented support. |
| Optimise matching without measuring Zout | Jazz 90 with the 4 Ω tap | With a 6 Ω nominal speaker, the Jazz 90's 4 Ω tap gives a higher effective damping factor (better speaker control and less interaction) at the cost of losing approximately 1 dB of maximum power. The 8 Ω tap does the opposite: more available power but more interaction with the cabinet's impedance curve. Since the reference system has 89 dB/W/m and listening is at 3 m, the 1 dB of power is not critical, and the 4 Ω tap is the conservative option. It is a practical recommendation that can be applied tomorrow without an oscilloscope. |
| Use with speakers of complex impedance | Pending | Requires measuring Zout in both and obtaining the cabinet's impedance curve. |
| Monitoring-level listening (105 dB at 3 m) | Neither | With 89 dB/W/m and 3 m, a 105 dB peak demands about 180 W per channel. Neither reaches it. |
12. Pending measurements
To close the unknowns of this report, a series of measurements on physical units would be needed. Here they are ordered by priority and grouped when the operation is the same on both devices.
| Priority | Measurement | Purpose |
|---|---|---|
| 1 | Comparative tracing of the power supply boards of both devices | Determines whether the MOSFET+reference blocks perform the same function (capacitance multiplier, B+ series regulator or bias reference). |
| 2 | Zout of both by the resistive divider method | Determines the interaction with the real impedance of the cabinet. |
| 3 | Square-wave step response of the OPTs of both devices over an 8 Ω resistive load | Reveals primary inductance, leakage inductance and high-band behaviour. It is the measurement that provides the most sonic information of all the pending ones. |
| 4 | Complete impedance curve of the Crossover A v2.4 | Quantifies the real spectral alteration. |
| 5 | Real weighing of the P52 without packaging | Closes the mass data point. The informal figure of ~40 kg and the visual estimate of 19–30 kg are contradictory and neither is verified. |
| 6 | Photo of the P52's three EPCOS capacitors with diameter and length measured against a visible scale reference (valve socket, KBU1010 bridge), and macro of the printed band on the sleeve if lighting allows | The pair (diameter × length) narrows the catalogue candidate list to one or two specific values, even before reading the band. In parallel, macro of the complete Nichicon marking on the Jazz 90 to confirm the nominal voltage (400 V, partial reading). |
| 7 | 45° side photo of the Jazz 90's gold and white capacitors | Closes brand and value of the coupling capacitors (gold and white). |
| 8 | Distortion curve versus power | Establishes the P@1 % THD. |
| 9 | Frequency response with real load | Quantifies the variation over the impedance curve. |
| 10 | Tracing or inspection of the P52's rear board relay | Determines whether it is for speaker protection, soft start or another function. If there is no speaker protection, the P52 is left without that safety layer against faults. |
| 11 | Supply ripple of both with oscilloscope | Replaces the conditional calculations. |
| 12 | Quiescent and peak grid current of the P52 | Confirms whether it is AB1 or AB2. |
| 13 | Tracing of the Jazz 90 PCB (underside) | Rules out or confirms a circuit not visible in the signal. |
| 14 | Global negative feedback depth | Explains the measured Zout. |
| 15 | Manufacturer consultation | UAT, P@1 % THD, Zout, function of the P52's relays, class AB1, bias procedure, number and distribution of trimmers. |
REFERENCES AND CREDITS
References and credits
Primary sources
- Acronos P52 · manufacturer. JC Valvular. Official spec sheet and website. 40 W per channel in push-pull class AB. Manufacturing process video: «Amplificador ACRONOS P52 Una obra de arte. Cómo se fabricó». Source of the interior and assembly process images, and basis of the photographic analysis in section 4.5.
- Cayin Jazz 90 · manufacturer. Cayin. Official documentation and user manual. Source of THD 0.2 % at 1 W / 1 kHz; UL power 50 W + 50 W RMS and TR 28 W + 28 W RMS; class AB1 push-pull; 4× KT88; substitution groups; Bluetooth codecs LDAC, aptX HD, aptX LL, aptX, AAC, MP3, SBC (no UAT); phono gain 40 dB; SUB OUT full frequency; fuses T1.6AL250V / T3.15AL250V; burn-in 50/100 h; soft-start ~60 s; minimum interval 2 min; matched valves. Note on the OPT: Cayin's documentation describes the power transformer as a toroidal of its own design; the core type of the output transformer is not specified in the official spec sheet.
- Cayin Jazz 90 · Spain distribution. Lyric Audio. RRP €2,990 and VIP25 discount of 30 %. tienda.lyricaudio.com/producto/cayin-jazz-90 (accessed 2026-10)
- Cayin Jazz 90 · photographic analysis. Analysis of the interior photograph (fairaudio.de review). Identification of the distributed BMI system, the shielded Bluetooth+DAC module, the solid-state MM phono and the voltage regulators. Detailed analysis of the power supply board, where a block with a heatsink of the same type as the P52's is identified, within a more complex board.
- Cayin Jazz series · output transformers. Reviews of the Jazz series published in 7Review and other specialist media describe the output transformers of the range as EI hand-wound. Cited here to justify the PROBABLE label for the core type of the Jazz 90 in §1.2 and §4.4, in the absence of direct confirmation by the manufacturer on the specific Jazz 90 model.
- TDK/EPCOS B43645 capacitor. Official datasheet: TDK B43645 (PDF). The family covers values from 160 to 1,470 µF in 450 and 475 V versions, ±20 % tolerance (M) and service life over 5,000 h at 105 °C. The value list does not strictly follow the E24 series: it includes 490, 530, 540, 580, 610, 670, 760, 840 and 890 µF among others. The 760 µF value is available at 450 V with order code B43645A5767M05 and 30 × 55 mm case.
- Reference system. LspCAD simulation of the impedance curve of the 35-year-old loudspeaker — Crossover A v2.4 MR-A-01, provided by the owner (2026-10). Not a measurement.
Complementary sources
- Jazz 90 reviews. lite-magazin.de, «Cayin Jazz 90 – Alleskönner mit Röhren-Power». fairaudio.de and hifi.de, reviews without instrumental measurements.
- Historic reference design. Mullard Ltd. Circuits for Audio Amplifiers (Booklet TP413). First edition April 1959; revised second edition August 1960.
- Valve datasheets. Mullard (EL34, ECC83, EF86), Genalex/New Sensor (KT77 Gold Lion, KT88 Gold Lion), JJ Electronic (6550, KT88, KT90, KT120, ECC803S, EL34), TAD (ECC803S).
- Modern techniques and push-pull physics. Tubelab documentation on PowerDrive. diyaudioprojects documentation on discrete CCS with MOSFET+TL431. Texas Instruments, TL431. Morgan Jones, Valve Amplifiers (chapters on push-pull ripple cancellation and output transformers).
about → array of Thing with @id (avoids Rich Results errors with two compared products), isPartOf, isAccessibleForFree, ISO 8601 with +02:00 zone, top-level ImageObject with width 1600 / height 1067 / caption, citation array added (5 sources), mainEntityOfPage with @id. First image migrated to eager + fetchpriority=high with <a> wrapper for lightbox and 1600×1067 dimensions, alt with og:image note. Second image (Jazz 90 inside) also with <a> wrapper and 1600×900 dimensions (estimated). <main> → <div class="report-main">. aria-pressed theme toggle to false. Badge and footer updated to V3.5.0. GSC+SOCIAL checklist block at the start (commented). The technical contents, calculations, tables, glossary, appendices and references of V3.4 are preserved in full.
