<< back to page 1 page 2
In this application, a stable masterclock signal is extremely important. It's the 'heartbeat' of the circuit: the conversion of the digital signal to the analog output signal is controlled by it. Any small deviation in time (this is in the range of tens of pico seconds) could cause a sample to be converted a bit too early or a bit too late, which results in small errors in the analog output signal. This will then deviate from the value that was originally intended. The result of this is a decrease of the detail and space in the soundstage.
For a stable clock signal, a good oscillator is required, which is powered bij a low-noise power supply. The Flea is such a supply. On the Flea's PCB, the 24.576MHz oscillator module is mounted and the Flea provides it with clean power. The originally specified oscillator module can be used, but it is recommended to select a type that has good jitter specifications, for example a Tentlabs XO.
The ASRC of the ezDAC operates on a 3.3V supply voltage. To prevent overdriving of the clock-input, the voltage of the Flea has to be set to 3.3V also. This makes sure the amplitude matches the input range of the ASRC. How this can be done can be found on the page of The Flea. With thin coax the Flea is connected to the ezDAC PCB, on the empty pads of the oscillator module.
The Flea is fed from the mains power by a dedicated power supply module. This consists of a modified ready-made module from Conrad (190-835) and this supplies a regulated 18VDC. By using a separate power supply, the coupling of noise through the power lines is prevented as much as possible.
Big thanks to 'Mags' from the diyAudio forum for supplying this Flea PCB!
...and other tweaks
There are quite some deviations from the original partslist of the ezDAC. Wherever possible, the decoupling capacitors around the IC's and voltage regulators have been enlarged, and better types are used. The tantalum electrolytics have been replaced by Rubycon ZA low-ESR types. In the +/-12V section for the opamps, the capacitors after the voltage regulators have been replaced by Black Gate N-types of 33uF/16V. Near the D/A-converter (PCM1794A), the two 47uF electrolytics that decouple the internal bias voltage (Vcom), have been replaced by these 33uF/16V Black Gates also.
In the analog output circuit, for R25 to R28 and R29 to R32, higher value resistors are used around the opamps. These resistors set the amplification-factor, but were chosen too low with respect to the 25Ω current resistors in the original design. This way, the voltage present on the inverting input is influenced too much by the output voltage swing of the opamp. Through the low impedance output, the 1k5 and 100Ω resistors are actually in parallel with the 25Ω resistor. When the output voltage changes, with this the resistance that the output of the DAC 'sees' also changes. The symmetry around the opamp is disturbed and the calculated output voltage of 2Vrms is not reached. That's why the 100Ω resistors are enlarged to 1k5, and the 1k5 resistors are changed to 22k. The original amplification factor stays the same this way.
A list with the used components and the ordernumbers can be found here:
The ezDAC v1.5
By the great succes of the ezDAC and the request for more PCBs after the first series, the creator of this project (Evan) decided to initiate a further development of this DAC. Together with some members of the Shiny Metal forum, which includes the writer, a number of improvements has been implemented in the circuit, and the PCB lay-out has been optimized. Also, a soldermask has been added, which significantly simplifies the mounting of the SMD components.
Besides the standard 'half-brick' oscillator module, a 5x7mm SMD oscillator can also be fitted now. These are more widely available and often have better jitter specifications. A good example is the CWX813 from Connor-Winfield. This is a 25ppm oscillator with a jitter specifiaction of 5ps (rms). Measurements will have to prove if such an oscillator can compete with a Tentlabs XO or The Flea.
Between the outputs of the opamps and the negative supply voltage, a resistor or CRD (Current Regulator Diode) can be mounted to bias the opamps into class-A. Applied with some opamps this has a positive effect on the sound. For extensive information regarding this subject: see Tangentsoft Audiologica. Furthermore, a number of capacitors has been added to improve the decoupling of the IC's.
The partslist of the version 1.5 PCB is available here:
Pictures of The ezDAC v1.5
Screenshots of the analog output without signal on an Agilent E4440 spectrum analyzer. It is clearly visible that the audio-band is very 'quiet'. Most of the noise is way below -100dBm. Because this analyzer has a 50 ohm input, the measurements are only a representation of the real specifications, but it gives a nice indication. With some math, this comes down to 2,24uVrms. In the 150kHz picture, the reference-level is lowered to -50dBm. At 700kHz, the rising noise-floor is clearly visible. This spectrum is typical for this kind of 'delta-sigma' or 'noise-shaper' converters.
<< back to page 1