A look at the D2A
SPIDF to analogue converter

4 January 2015

These boxes and similar are ubiquitous on Ebay.

D2A pic I bought one recently from TradeMe to use in a project. It cost only NZ$29. The box takes digital audio coded as stereo SPIDF in either optical or coaxial forms and produces stereo audio. It cannot decode multi-channel coded audio. The following is a brief description of the performance and what was done to make it work properly.

Initial performance

With a full scale sine-wave of 400Hz generated from 'Cool Edit Pro' fed in from the pc using either the optical or coaxial inputs (one at a time), the sine wave produced was distorted. There was clipping on the negative half-cycle. Seen on an oscilloscope as below:

D2A distorted waveThe scope is set to 0.5V/division. Measured on my Praxis system, second harmonic was -28dB, 3rd harmonic was -30dB, 4th was -34dB, 5th was -39dB and many other harmonics were produced to above 10kHz. Audio harmonic distortion was over 5.5%. Reducing the coded audio level down to below -8dBFS was necessary for the output to become undistorted. Given the price and effort involved in doing a return of the item, I removed the cover to see what could be found.

Inside the box -what was found

The box actually contained a 8416CZZ Cirrus logic 192kHz digital interface receiver chip, a Cirrus Logic 344C1315 24bit DAC, a low drop-out regulator and a number of peripheral components. Given the high pedigree of the components, why was the output distorted? I traced the analogue output with the scope and discovered that the output of the D-A converter was a perfect sine-wave. The clipping was produced between the D-A and the output connectors. dc rail was OK at 3.4V.

I partly traced the circuit to find there was a pair of transistors after the D-A outputs, which are intended to mute the analogue output when there is no digital input. This muting arrangement is set to 'short' the audio outputs but was poorly implemented, so that even when the transistors were supposed to be off, they were not fully so. I removed the two transistors entirely and this made the box produce pure sine-waves even with full amplitude digital audio levels.

D2A clean waveClean output with 0dBFS digital input.

Final results

Now, with 0dBFS tones applied from the pc, the output level is 2.4Vp-p and is clean. All harmonics and other products are more than 90dB down; most below -100dB. I did a quick check of frequency response and it is flat down to 40Hz and also to at least 20kHz. The pic below shows the board with Q1 and Q2 (on the left) removed. D2A pcb

The consequences of the board lacking it's mute feature are minor. When there is no digital input present, there is a rising noise floor commencing at a frequency of 15kHz and rising to above 48kHz (limit of my measurements). The noise floor is at least 50dB below full output level, so is barely noticeable in any case. In addition, one does get a fairly loud click when the digital input is first applied, nonetheless a small price for the now good performance.

Axino-Tech January 2015

To make a comment on this, go to my contact page.