Hybrid Class D End system   
Apr 2026

Some PCBs available for free if anyone is interested. 

INTRODUCTION

After the incredible performance of our last Hybrid Class D system, we’ve reached the next level.
We’ve combined a pure, single-ended Class A tube stage with an interstage transformer to deliver a perfectly balanced signal directly to the Class D chip.
It’s the ultimate tube-stage upgrade for any balanced Class D amplifier.

The main features are simplicity, extreme sonic performances and perfect instrumental measurements.

• No opamp

• No global feedback

• No noise

• Only 1 capacitor on the signal path (ELNA Silmic II)

• Very high slew rate 

• Very low distortion until 170W on 4ohm

• Only one tube on the signal path

• Very short internal wiring

• Vaccum tube pre-ampl with very low distortion also at 2 x 6Vrms

• Vaccum tube pre-ampl with a voltage gain from 13 to 21dB

• Frequency band 20Hz - 20Khz at 0dB

• Single chip class D section

SCHEMATIC

Is this a simple project? Technically, yes, it is remarkably straightforward. However, achieving flawless sonic and instrumental performance requires high-end components with no compromises.
As detailed below, I have analyzed the specifications of nearly 50 interstage transformers from various manufacturers. In the end, I selected the one that delivered the most superior results.
I was looking for an inductance value higher than 50H with an air gap capable of supporting 14-20mA with a primary DC resistance not exceeding 1/10 of the internal resistance of the tube used, therefore < 300ohm.

WORKING IN PROGRESS

phase 0 - first listening test to make a direct comparison with the previous differential preamplifier

phase 1 - the pcb are ready so second listening test waiting the power supply transformer

phase 2 - final power supply transformer installed now we can test the power supply capacitors

In this project, I have chosen to use the E180F in a triode-connected configuration.
Instrumental tests and sonic performance proved it to be the best-in-class, despite its excessively high voltage gain.
To manage this, a dedicated attenuator stage is required, similar to the solution implemented in my previous design.
These tubes are readily available as New Old Stock (NOS), with high-quality options from manufacturers like Siemens and Mullard.
The primary alternative is the 5687, which is significantly more expensive and harder to source.
However, the 5687 offers a distinct advantage: a single tube can serve both channels, and its lower gain profile eliminates the need for an attenuator.
I am not a fan of the sound produced by the Russian equivalents of the E180F; although they are quite detailed, they tend to sound harsh.

   E180F - CV3898 old Mullard (single pentode)

  6Z9P - 6ZH9P - E180F (single pentode)

  ECC99 by JJ (dual triode)

5687 Jan Philips (dual triode)

5842 Raytheon (single triode)

INTERSTAGE TRANSFORMERS

In this project, I utilized the Lundahl LL2763/7mA for the first time.
After an extensive market review, this component was chosen to satisfy the following technical requirements:

• High Inductance: To ensure a good low frequency band

• Air gap designed for a minimum of 14mA to use high transconductance tubes

• DC resistance no more than 1/10 of the tube's internal resistance to minimize voltage drop, improve efficiency and obtain a good slew rate

The LL2763/7mA with primaries in serie have a primary inductance of 280H but if we connect the primaries in parallel we have 280 / 4 = 70H, air gap for 14mA and Rdc=270ohm.

This specifications are better than LL1671/20mA with 52.5H  Rdc=312ohm and than LL1689/9mA primaries in parallel with 45H  Rdc=327ohm

Ft(-3dB) = R / (2 * 3.14 * L) = (2400+ 327) / (2 * 3.14 *70) = 6.2Hz

This Lundahl LL2763/7mA is available in 3 versions:

• Silicon-iron (STANDARD)

• Amorphous-iron (AM)

• Nanocrystalline (NC)

Now I'm using the standard version, but I've also tried the AM version in the past, and I think the increased sound performance justifies the price difference.

I've never tried the NC version, so I can't comment on that.

CAPACITORS

Regarding the Ck capacitors, I have chosen the ELNA Silmic II since the Nichicon Muse series has been discontinued.

    

POWER SUPPLY FOR DRIVER STAGE

I have tested these power supply for the vacuum stage driver stage:

1) Virtual battery like the previous Class D project

2) Regulated with feedback near to Maida design

3) Pure CLC like the best Single Ended vacuum tube amplifiers

Only with the third type you get a total naturalness of sound so my choice is this one also because we are creating an extreme project without compromises.

Obviously the quality of all the passive components must be of the highest level without looking at the costs.

The Hammond choke 159M with 15H  Rdc=256ohm and able to support a dc current of 100mA are my first choice.

I have identified these models on the market because the Jensen electrolytics are no longer in production.  

In a single-ended stage, the quality of the filter capacitor following the choke is critical, as it significantly impacts the sonic performance.
I have a pair of vintage Jensen capacitors, and I'll be installing them first as they are my absolute benchmark for audio quality.
I tried looking for some reasonably priced alternatives on the market, such as the Kemet (ex BHC) ALS30, Nichicon KX and large 100uF MKP, but the results were disappointing.  In the past I tried also the Mundorf MILytic HV, these are precise and dynamic, but the sound is very cold and hard.
I recovered some vintage Black Gate WK from my own personal stock at home, enabling a rigorous comparative test with a friend.
With both the Black Gate WK and Jensen caps, the output remains remarkably transparent and organic, even under complex signal loads. The Black Gate WK, however, offers a subtle edge in terms of resolution and detail.
With the other capacitors, there is a substantial alteration of the signal. While this might seem positive on some tracks—making them sound more open and spacious—as soon as we switch to complex signals, the result becomes almost unlistenable.
At this point, the only viable option is to install a pair of 220uF 160V Audio Note Kaisei capacitors, which are claimed to be manufactured with a structure very similar to the vintage Black Gates.
I also tested the frequency response and sonic performance with both 470µF and 220µF, but there was no discernible difference.
I noticed a 180V overshoot at power-on during simulation. To protect the 160V capacitors, I increased R8 to 100 Ohms; this fixed the spike at the cost of a minor reduction in B+ voltage (sim0 and sim1).

The first 2 capacitors are ALC70A271BB250  270uF 250V with snap-in chassie.

Residual noise with output cap. 220uF pure 100Hz sine wave at -83dB (215uVpp) and with 470uF -90dB (115uVpp).

About the transformer to use for the driver section I have used a custom toroidal on my specifications.

The primary and secondaries has been calculated for 260V instead of normal 230V to keep the flux very low.

primary 260V

secondaries 0,2A 130V   26VA  and  1,5A 9V  14VA total power 40W  

the real output will be 115V so if you think to use the Audio Note Kaisei with only 160V reduce secondary of 10%.

OUTPUT STAGE

The TPA3255 is used in many modules with significant price differences starting from $30 but only very few modules can we consider serious products for hi-end systems.
In order to integrate theTPA3255 modules with the tube preamplifier without using an opamp it is necessary that the modules use a dual power supply for the original driver so the decoupling capacitors are oriented in the correct direction (positive connected to TPA input).
      

The best module I found for this project is the D400M P05 produced by Sylphaudio with these main specifications:

• TPA3255 in a mono PBTL with or without PFFB

• input balanced or un-balanced

• opamp with split supply design so dual power supply +12V and -12V

• 2Ω load capability

• Footprint: 130mm x 100mm x 45mm (L * W * H)

• Overcurrent /short protection

• Over voltage protection

• Overheat protection:

• Turn on-off pop suppression  

The main differences from the Fosi Fox v3 are these:

• the use of one chip for each channel in PBTL

• the use of dual power on the driver stage

• Output filter with Coilcraft SER3018H 2.56mohm instead of Sumida DEP1519 16.6mohm

• Output filter with 1uF Panasonic MKP ECWFE  instead of Wima MKS 2

• Interstage capacitor 10uF ELNA Silmic II instead of ELNA RA3

POWER SUPPLY FOR POWER STAGE

Regarding the power supply, there are numerous companies that build switching power supplies but I chose Meanwell products because I needed a reliable product that does not cause problems over time.
The first tests were done with LRS-150-36 power supplies and then I switched to the UHP-200-48 resonant design (LLC) model with also PFC.

There is a huge difference in the sound by switching to the new Mean Well UHP, many harshnesses disappear and you feel a sensation of extreme cleanliness with absolute silence in the background.

CABLES

The internal cables used are the Canare L-2T2S High performance Microphone cable shielded 0.26mm² and 4S11G STAR QUAD Speaker cable OFC Copper 4x2.08mm², both with with insulation PE Polyethylene.

PCB

You can use this little module to connect the tube preampl. to the opamp output of D400 (use only with TPA3255 modules with dual power supply, ex. +12V / -12V or invert the interstage capacitors in the correct direction).
The first version include an attenuator about -6dB to reduce the voltage gain of the final system, to verify if it is necessary.

R1,R2,R3,R4     2700ohm 1/4W  SMD 1206
IC              2 x 4 Single row connectors 2,54mm

• Here the files to produce this PCB for the op-amp connection

• Here the files to produce this PCB for the op-amp connection 

Follow the relay module to switch-on the amplifier with a low current button.
I have used a 22mm Waterproof Ring Illumination Black Anti Vandal Push Button Switch, Blue, 1NO1NC Latching, LED voltage 12V.

• Here the to produce this PCB  

R1 82ohm  1/4W

RLED 680ohm 1/4W in series to one pin to reduce light.

The PCB layout (16x17 cm) incorporates strategic cutouts directly beneath the transformers.
This design choice is implemented for 2 technical reasons:

• Vibration & Microphonics Isolation: The cutout creates a structural break that stops transformer mechanical hum from reaching the vacuum tubes, preventing microphonic noise interference.

• Thermal Management: The openings facilitate an "entry-effect" for air convection.

R42,R43,R18,R19       220K    1/4W   Vishay Dale RN55
R20,R21,R44,R49         2K    1/4W   Vishay Dale RN55
R2,R38                100ohm  1/4W
R1,R27                 47ohm    1W

C1,C3                 330uF  6.3V  ELNA Silmic II
C5,C6                  47uF  6.3V  ELNA Silmic II
C2,C4                 470uF 250V  Kemet ALS30 or                     
                      220uF 500V Jensen elect.
                      220uF 160V Audio Note Kaisei (to use this reduce the transf. secondary)

TR1,TR2               LL2763/7mA
U$2,U$5               E180F +  noval socket

The connections are 63862-1 (CUT STRIP) by TE Connectivity / AMP (cod. Mouser  571-63862-1-CT, cod. RS 718-7987)

• Here the files to produce this PCB 

R45              68K    1/4W
R46             220ohm  1/4W
R48               1K    1/4W
R83              12ohm    3W   PR03
TR1             100ohm  multi turns trimmer

CY5,6,7,8      0.01uF   300VAC  WIMA MP3-X1 10N 
C7             4700uF    16V
C10,C11,C8      220uF    16V
C5,6            270uF   250V or any good cap. with min 220uF

D1,D2,D3,D4     UF5407  
U$19,U$16       1N4007
B1              diode bridge 100V 2A  2KBP01
T3              2N2904, 2N2905, 2N2907

U$15            LM350

The connections are 63862-1 (CUT STRIP) by TE Connectivity / AMP (cod. Mouser  571-63862-1-CT, cod. RS 718-7987)

• Here the files to produce this PCB

KK2 Heatsink type SK104 or EA-T220-38E compatible L=38mm  Thermal resistance = 7.5K/W

CABLES 

The internal cables used are the Canare L-2T2S High performance Microphone cable shielded 0.26mm² and 4S11G STAR QUAD Speaker cable OFC Copper 4x2.08mm², both with with insulation PE Polyethylene.

CHASSIE

I will use this Hi-Fi2000 chassie.

• Slim Line 02/350 10mm BLACK front panel - 3mm aluminium covers

• Supplement for substitution of fully vented aluminium cover 3mm Slim Line 350mm

TUBE DRIVER MEASUREMENTS

The E180F in triode connection and the 5687 give a good low distortion < 0.3% and a perfect 20Hz - 20KHz at 0dB.
Using the E180F is necessary an attenuator after the interstage transformer to reduce the voltage gain.

In this test I added a temporary capacitor to the output just to observe the high-frequency roll-off.
This component is not used in the final version.

DETAILS

Ensure the interstage transformer chassis are connected to the ground point.