This project was conceived as a thoughtful evolution of a series of experiences developed during the second lockdown period, when the available time made it possible to return to experimentation without compromises, setting aside “clever” shortcuts and focusing exclusively on real sonic results. The goal was to build a tube DAC preamplifier with transformer-coupled outputs, capable of combining transparency, three-dimensionality, and soundstage control, while avoiding both the artificial sonic aesthetic of certain zero feedback circuits and the spatial compression typical of designs with excessive feedback. The result is a device designed to interface directly with high-quality power amplifiers, without masking anything and without imposing a caricatured sonic signature.
For this project I drew inspiration from several sources and previous experiences. In particular, I deliberately blended what I had learned with the granny27 and with the Nibiru. The idea was to obtain a transformer-coupled DAC preamplifier, similar in concept to the old Euridice project, but with an extra edge.
This “extra edge” comes from the experience gained precisely with the Nibiru, where I had successfully experimented with a controlled three-dimensional spatial effect, obtained by deliberately combining the behavior of an inductive element with a very mild but extremely precise amount of negative feedback. In the Nibiru, the key element was an inductor, whereas here that role is played by the transformer. The feedback was carefully calibrated to avoid two equally wrong extremes: on one side, the loose and unfocused sound of true zero feedback circuits, and on the other, the flattening and compression of the soundstage typical of designs with high feedback.
The result is a balance that is difficult to achieve but extremely rewarding, a virtuous middle ground between three-dimensionality, focus, and naturalness, making listening engaging without introducing coloration or artifices. It is exactly that kind of moderate, thoughtful, and beneficial feedback that those who dogmatically wave the banner of zero feedback will likely never understand. This project is published as premium.
To avoid the issues encountered with the granny27, which eventually had been loaded with a transistor CCS, sounding very good but not fully meeting my original objective, I decided to start from a tube with significantly lower internal resistance. The choice fell on the 6BX7, a tube often underestimated but extremely interesting for this type of application.
The 6BX7, as well as its equivalent 6BL7, is a power double triode with the same socket as the 6SN7, but with decidedly different electrical characteristics. It can dissipate 6 watts per plate, has an internal resistance of about 1300ohm, and a Mu of 10. Functionally, it closely resembles the power section of a 6EM7. It is particularly well suited for headphone amplifiers, line preamplifiers, very low power amplifiers, or as a driver for large tubes such as 300B or 845.
For this project I designed a new interstage transformer, the i12K600, specifically intended to work correctly with tubes having low internal resistance.
This transformer has a 12kohm primary and a 600ohm secondary with center tap, and it can handle a DC current of 10mA. It can be used with tubes having an internal resistance from about 1700ohm downward, such as 6BX7, 6BL7, but also 5842 or 417A. In my case, already having a robust signal coming from the DAC, I preferred the 6BX7. Thanks to its low Mu and the transformation ratio, I managed to achieve nearly unity gain, while still working with a genuinely amplifying tube, capable of imparting an authentic and natural character to the sound.
This approach is the exact opposite of many inexpensive buffers found on the market, where a follower tube is inserted merely for show, while the real work is done by op amps hidden under the hood.
To further experiment, I also included in the preamp, in addition to the DAC, a JFET preamplifier based on BF256 and 2SK170. A selector switch is therefore present, allowing switching between the DAC, the phono preamp, or an AUX input for external sources.
As a first step, I built the power supply in a separate chassis, in order to keep the power transformer away from the interstage transformers and avoid the introduction of hum, a problem I had already encountered with the granny27.
The anode supply was implemented using a 6X5GT. The 6BX7 stage draws about 20mA constantly while operating in class A, so the 6X5GT is perfectly sized, being capable of delivering up to 70mA. It is not stressed in the slightest, will last a long time, and is easily available in NOS versions at reasonable cost. This is the exact opposite of what certain “gurus” claim, insisting that to supply a few milliamps one necessarily needs oversized rectifiers and huge banks of capacitors, without even distinguishing between current, voltage, and output resistance of a stage.
After the 6X5GT, I used a CLC cell, as in the granny, with an old NOS Philips screw-terminal capacitor whose electrical characteristics put many modern capacitors, even so-called audio grade types, to shame. Purely for curiosity, and to achieve absolute instrumental silence, I then suppressed the residual ripple with a capacitance multiplier, visible on the left after the green rectifier bridge that instead powers the RIAA preamp at 24 volts.
A small personal satisfaction in the use of vintage components: in the capacitance multiplier, next to the BUX85G, one can notice a NOS BC207. In the video below it is possible to observe the operation of the multiplier, with the oscilloscope scale set to 20mV per division. The lower trace shows the signal before the multiplier, the upper one, almost flat, the signal after. The slight low frequency undulations are simple variations in the mains voltage and should not be confused with residual ripple. The multiplier is not a regulator.
Moving on to the chassis of the preamp itself, the anode supply is further decoupled by an RC cell with a 22ohm resistor and a 2.5uF NOS paper-in-oil Sprague capacitor, of the type intended for aviation use, equipped with mounting screws to withstand mechanical acceleration.
Paper-in-oil capacitors in audio applications give their best in power supplies. When inserted into the signal path, or worse still in crossovers, they do tend to smooth roughness, but at the cost of a loss of detail, especially in the high frequencies.
The trimmer visible in these images is used to perfectly balance the two channels. With a very low NFB rate, about 2dB, even tubes with well matched sections will never be perfectly identical. Balancing is carried out under oscilloscope by acting exclusively on the feedback network, without touching the input signal, thus avoiding the introduction of noise.
The 6X5GT
Measurements: total harmonic distortion is 0.78% with an output of 8Vpp, a value higher than the maximum normally used, which is around 6Vpp in operation with the DAC. Bandwidth is 30Hz to 35kHz at -1dB, measured at the same amplitude.
How it sounds
I experimented with various bypass capacitors on the 6BX7 cathode electrolytics, trying polypropylene, cellulose acetate, and other solutions. In the end I found my balance with polypropylene, which favors detail without impoverishing body. The sound is rich and pleasing, the high frequencies are lively but never aggressive, instrument separation is excellent, voices are well focused at the center of the stage, and the soundstage is wide and credible. The final result is fully satisfying.