In a digital era where audio technology continues to evolve, there are still those who value the simplicity and coherence of classic solutions. This project fits precisely into this perspective, adopting a NOS (non oversampling) approach based on the Philips TDA1543 chip, used here with an I2S stream at 44.1 kHz and 16 bit, that is the “Red Book” format typical of most CD productions.

The philosophy behind this project is clear: a large part of the historical digital catalog, and not only material produced before 1995, was created at 44.1 kHz and 16 bit. The idea is not that upsampling can “recreate lost information”, because that is not possible, but that it can simplify some aspects of filtering and the management of spectral images. Here I deliberately chose the opposite path, an essential signal path with direct conversion and sober analog filtering, to obtain a coherent result that is also easy to read from a design point of view.

The choice to incorporate the TDA1543 chip was also guided by the advice of Ivo Calabrese, who appreciates its timbral rendering and naturalness, while remaining a simple and inexpensive component compared to more celebrated solutions such as the TDA1541. The goal is not to “copy” the past, but to use a classic component as a solid base to build a hybrid and experimental project, with targeted and verifiable choices.

This DAC is not limited to the main chip alone. A hybrid approach has been adopted, introducing a phase splitting interstage transformer that creates the interface between the section immediately after the DAC and the tube stage. The phase splitter then drives a push pull stage built with the two triodes of an ECC88 tube. In this configuration the tube section is not an “effect”, but a real amplification and driving stage, with a sonic signature defined by the circuit and the transformers used.

The prototype board presented in this article is equipped with a versatile interface featuring an I2S data input. This allows the DAC to accept signals from various sources, from a PC via a USB interface, to CD players through an SPDIF to I2S module, up to integration with a Raspberry Pi, provided that the source supplies a compatible I2S signal in terms of frequency and format.

In the embryonic phase of the project, here is a snapshot from the laboratory showing the first experiments on the breadboard. In this image you can see the small but crucial phase splitting interstage transformer, developed to create a stable link between the adaptation stage immediately after the DAC and the tube section, while maintaining the balanced configuration required by the following stage.

This transformer is a key component of the hybrid approach, because it performs signal and phase transfer between two very different worlds, the one immediately downstream of the DAC and the tube domain. To explore the characteristics and the fundamental role of this component in the project, I invite you to visit the dedicated page for the phase splitting interstage transformer here.

During the development path I also used the I60KPP88 transformer, a crucial component in the signal chain. In the image below you can see this transformer, designed to be driven by a push pull stage built with the two internal triodes of a single ECC88, in order to obtain a robust output well coupled to real loads.

Its presence ensures accurate signal transfer with low sensitivity to interference, routing it both to headphones and to a possible power amplifier, depending on the chosen configuration. In this way it significantly contributes to defining the character of the sound, not as a generic “color”, but as the direct result of coupling and the impedance seen by the stages involved. To further explore the characteristics and importance of this transformer, I invite you to visit the dedicated page here.

Photos of the Completed Prototype

Here is a series of detailed photos that offer an overview of the completed prototype mounted on a 1000 hole breadboard. This assembly shows the experimental phase of the project and clearly reveals the layout choices, wiring, and separation between the sections.

Note for DIY Builders: A key point of my approach to DIY construction, highlighted in the images of the prototype, concerns the use of tubes on a 1000 hole board. It is important to note that many online builds use similar boards with PCB mount sockets, but in this case the choice was to use standard noval sockets for point to point wiring, the type normally mounted on a panel.

In the images you can see that I drilled a hole in the 1000 hole board and directly screwed two standard panel mount noval sockets. This approach differs from the use of PCB sockets, which are often awkward and mechanically weak on 1000 hole boards.

My recommendation for DIY builders is to consider this solution, because it offers concrete advantages. Mounting noval sockets directly on the board, fixing them mechanically and keeping the connections to the pins as short as possible, ensures a safer and more reliable connection. Compared to solutions with forced leads or pads held by very little solder, this choice reduces the risk of detachment and breakage during tube insertion and replacement. In addition to simplifying assembly, it improves mechanical stability and contact quality, contributing to a more solid and durable build.

For those who share a passion for DIY, I hope this note can be useful as a practical hint and not just as a construction curiosity.

Technical Information

The prototype is characterized by a clear separation between the digital part and the adaptation stage based on an operational amplifier, both powered by stabilized voltages. The op amp receives a dual supply of ±12V, a condition that allows linear operation with good dynamic headroom. The anode section, after rectification, uses a capacitance multiplier to obtain an unstabilized but heavily smoothed voltage. This choice avoids feedback regulators on the HV line, reducing the risk of introducing wideband noise associated with specific regulation topologies.

After the TDA1543 chip there is a passive reconstruction filter, built entirely with discrete components, with the aim of attenuating high frequency components and images typical of NOS conversion, while maintaining a simple and easily verifiable approach.

As for the op amp used in the first instrumental tests, a TL082 was initially employed, around which the interstage transformers were developed. However, in headphone listening tests, preference went to the sound rendered by a NOS MC1458 produced by Motorola, which seemed to me more “full” and less dry compared to the TL082 (non NOS), with the same circuit context.

It should be noted that, although an MC1458 is present, it is possible to replace it with other 8 pin op amps provided they are compatible with a total supply of 24 volts (±12 volts) and capable of delivering sufficient current to drive the small transformers, while maintaining an effective drive impedance not exceeding 600 ohm.

The presence of an op amp might raise some doubts, but it should be considered that the TDA1543 requires a proper I to V conversion and adaptation stage. In addition, in practice, the difficulty of obtaining a transformer phase splitter that worked well under all test conditions led to the choice of a more energetic and repeatable drive. This setup also allows controlled experimentation with different op amps, including special or high quality components, while remaining within sensible electrical constraints.

The circuit is equipped with a gain control placed between the phase splitting transformers and the ECC88 tubes, acting on the balanced signal and serving as a level control. An important point should also be clarified: when I speak of the absence of feedback, I am referring to the tube section and the analog chain downstream of the transformers, understood as the absence of global feedback on that stage. Where the op amp is present, the local feedback necessary for its operation obviously remains an integral part of the circuit.

The next images offer a close up look at the digital heart of the prototype, the TDA1543. In particular you will see a close up of the chip and the 4 wire I2S bus connector, which exits and connects to a small USB to I2S decoder. This decoder, based on the PCM2706 chip, was suggested by Ivo Calabrese.

The connector features a particular implementation: the wires are not soldered, but plugged in. This choice makes it possible to quickly replace the USB decoder with another compatible I2S source, allowing different configurations to be tested without having to redo the wiring each time.

This modularity of the design makes the project very flexible during the experimental phase and allows rapid comparison of behavior with different sources, while keeping the analog chain unchanged and reducing uncontrolled variables.

Power Transformer

In the next image I show you the heart of the project’s power supply, the custom developed power transformer. This component is responsible for providing energy in a stable way with low sensitivity to mains disturbances, a particularly important condition when audio transformers are nearby and a tube stage is used.

To obtain a clean and repeatable result, several solutions were implemented in the design of this power transformer:

1. Strategic Double Insulation: The transformer is configured with double insulation, where primary and secondary windings occupy separate cavities in the bobbin. This choice is not only related to the electrical safety typical of double insulation bobbins, but was also adopted with the aim of reducing capacitive transmission of disturbances from the mains, limiting the transfer of noise to the secondaries.
2. Oversized Core: The core is oversized compared to the power actually required (just over 10 watts on a core that could handle about 32). This helps reduce stray fields, decreasing the likelihood of interference with nearby audio transformers.
3. Oversized Wire Sections: The wire sections are also oversized, so the transformer operates with lower current density compared to conventional solutions, with lower losses and reduced heating.
4. Different Orientation: The power transformer is oriented differently from the audio transformers on the board, in order to minimize magnetic coupling. In practical tests, no induced hum is observed in the audio circuit already at a distance of about 5 cm.
5. ECC88 Heater Supply in AC: The ECC88 heaters are powered directly with alternating current. With correct layout and tidy wiring, this choice proved to be quiet and free of induced noise, avoiding unnecessary complexity.
6. Different Current Density: The secondary that feeds the heaters is designed with higher current density compared to the other secondaries. This results in a greater voltage drop under load, helping to contain inrush current peaks at power on and reduce overall stress on components over time.

These technical choices reflect a very concrete goal: to reduce the variables that generate noise and interference, while maintaining a simple, robust, and easily serviceable construction.

Update 30/12/2023

While the prototype of the PulsarWave DAC with Philips TDA1543 chip and ECC88/6922 tubes was preparing for its debut, I spent time integrating the electronics on a solid base with improved aesthetics. The project therefore moved from the breadboard to a proper chassis, while maintaining the same circuit logic and improving overall ergonomics.

The chassis is a HiFi 2000 recovered from an old unpresentable unit that I dismantled years ago and brought back to life. After thorough cleaning, I drilled holes in the aluminum panels to house the power switch and position the small PCM2706 module with USB input. Thanks to the 3D printer, I made resin adapters to securely mount the components. The output RCA connectors and the level control were placed on the rear panel to ensure easy access and convenient connection.

The bottom plate was drilled to house the board and the 23S75 power transformer, with a dedicated node for chassis grounding. The VDE inlet with fuse, already present, was properly reused. Inside, a pair of NOS ECC88 tubes found their home, while custom support feet and a three dimensional logo were added for a final touch of aesthetics and functionality.

The crucial moment was the first listening test. Initially the Motorola operational amplifier gave the sound a denser and slightly darker character, prompting me to return to the TL082 for some tests. In headphone listening, however, preference returned to the Motorola, confirming that the choice of op amp must be evaluated in the real context of use and not only with abstract criteria.

Further refinements were made to the circuit to adapt it to the desired nuances, and the overall result was evaluated positively. The various assembly stages are documented in the attached photos, offering a detailed look at the transformation path of the project from prototype to complete build. Here are a couple of measurements:

Harmonic Distortion Spectrum

Square Wave at 1 kHz