Construction of my sound reproduction system

My interest in Hi-Fi dates back to the 1980s. Until 2010 I used several commercial stereo systems, the last one comprising a Sony XA7ES cd player, a Mark Levinson No:331 amplifier, and a pair of Klipsch LaScala speakers. Afterward, I designed and built my audio equipment myself – first a pair of speakers, then several amplifiers and a digital crossover. I use a NAS Server based on Openmediavault to store ripped music files in a lossless format and a computer as the player.

Speakers

The speakers I built consist of bifurcated, folded bass horns, and 250Hz tractrix horns.

The folding method of the bass horns are identical to the Klipschorn. However, the horn profile is different. The back air chamber is smaller, the profile is exponential (no “rubber throat”) and the front of the speaker is narrower. According to simulations made with Hornresp this horn’s frequency response is flatter and extends higher than Klipschorn. It is also designed to be placed against a wall, as well as in the corner. It is usually difficult to find two appropriate, symmetrical corners in a living room. Indeed, this is why I needed to design these horns. Although the low frequencies are not as well loaded against a wall as in a corner, these speakers still reach down to 29Hz at -3dB in my living room with Eminence Delta 15A drivers.

The tractrix horns are paired to concentric, two-way BMS 4592ND drivers. The crossover frequency between the low and mid drivers is 380Hz, while the signal to the tweeter is filtered by the stock BMS crossover at 6,500Hz.

Below are some photos showing the construction of the speakers:

Amplifier

I initially built a 4-channel amplifier based on Hiraga Le Monstre with passive line-level crossover (PLLXO).

This amplifier’s design is quite simple but requires huge amounts of power supply capacitance. I first built a CRC filter with a capacitance multiplier after the rectifiers and two 100,000 uF capacitors after the resistors for each channel. Then, I increased the number of capacitors to four per channel making the total capacitance 1.6 Farads.

A PLLXO is simpler and much cheaper to produce than an active crossover, but for the crossover frequency to be stable, the combined input impedance of the attenuator and the amplifier must be constant regardless of the volume setting. To this end, I designed a custom attenuator requiring at least an 8 deck switch with adequate number of positions. I found a suitable 10 deck switch with silver contacts originally made for CNC machines and built the attenuator.

Below is the first iteration of the amplifier with the PLLXOs, the capacitance multipliers, and the custom attenuator:

This amplifier worked as I expected and I was quite happy with the sound. But after a couple of years, the silver contacts started to make noise. It wouldn’t be difficult to restore them, but I was already considering to build a digital crossover at the time. So, I built one and also made several improvements to the amplifier. The PLLXO and the attenuator were discarded in favor of the digital crossover. Most parts of the power supply were moved to another chassis both to make room for additional capacitors and to separate noisy rectifier diodes from the audio circuitry. The amount of capacitance for each low-frequency channel was increased to 600,000uF and the total power supply capacitance of the amplifier reached 2 Farads. Cabling was also improved.

Below is the power supply of the amplifier:

And the amplifier’s final configuration:

The red switches next to the RCA jacks are two-level attenuators. With the DACs’ maximum output and the high settings of the attenuators, the amplifier delivers 2 x 5 Watts and 2 x 0.5 Watts RMS at 8ohms per low and high frequencies, respectively. This amplifier is capable of delivering more than 10 Watts RMS per channel, but I always include attenuators that prevent any possibility of clipping in my designs. At the low settings of the attenuators the signal level is -10dB with respect to the high settings. I use low settings when making measurements or experiments to prevent extreme sound levels in case of a mishap.

Before upgrading the amplifier, I could hear a buzzing noise at 50cm from the mid horns which was already a very respectable noise level given the drivers’ sensitivity of 118dB/m. After the upgrades, the system became dead silent even with the ear inside the mid horns.

Digital Crossover

When it comes to making adjustments to audio signals there are important limitations in analog circuitry. Some adjustments are either very difficult or simply impossible to accomplish. For example, it is not possible to alter frequency response without affecting phase, and vice versa. But, with digital signal processing, especially with finite impulse response (FIR) filters, the possibilities are endless.

The digital crossover I built is made up of a miniSHARK DSP board, a Digi-FP SPDIF connectivity board, a VOL-FP board for remote volume control, an XMOS U8 based USB streamer, two stereo ES9018 DAC boards, and power supplies.

Measurements

For acoustical measurements, I use a calibrated UMIK-1 USB microphone and HOLMImpulse freeware software. In the 1990’s I was using a Behringer measurement microphone with a mic amplifier I designed and built, and Cooledit software with Aurora Plug-ins. HOLMImpulse is much more practical than these old software.

For the design of the filters, I use RePhase, again a freeware software. I tried several configurations and ended up with 96dB/octave Linkwitz-Riley linear filters. I prefer 48kHz plug-ins of miniShark to be able to use more taps compared to 96kHz. Since my living room is not symmetrical, I do slightly different equalizations for left and right channels but exactly the same phase equalization for both, in order not to alter the timbre of the sound.

The frequency response at the listening position, phase and impulse responses at 1m as of this writing are as follows:

Of course, what matters most is the sound quality and I am quite happy with it. Yet, I think that slight improvements should be possible with bigger mid horns, 6 channel amplification in lieu of 4, and further fine-tuning of the FIR filters.