The following information is a technical guide to the design of the device. Most people will probably want to skip it – our devices are designed to be plug-and-play, and dead simple to use. For those who are interested in what’s happening inside, please read on.
Changes in Version 2.1
- Increased capacitance on power supply lines, with fast tantalum capacitors and “flip-type” ceramic bypass capacitors.
- Revamped power implementation of the USB receiving chip (PCM2706) results in additional jitter reduction in the recovered digital audio signal.
- To accommodate these changes the enclosure is slightly large than the previous version, however, the connecting USB cable is thinner and has strain relief.
- Enclosure < 2 cubic inches
- Cryogenic treatment
About the devilsound DAC
The devilsound DAC is a high quality 16-bit, non-oversampled (NOS), audio digital to analog converter. Audio playback occurs at the recorded frequency (32, 44.1, 48 kHz), with no resampling, upsampling, or oversampling.
In addition, the DAC contains special circuitry to regenerate clean power from the USB port. This combines the convenience of USB power with the performance of an external power supply.
DAC technical design
For those who want to know the technical details about the architecture and chipset used in the devilsound DAC, and also for those interested in digital audio from a more general or DIY perspective, we hope the following information is helpful.
A typical audio DAC can be broken down into four basic ingredients:
- power supply circuitry
- a digital audio receiver
- a digital to analog converter (DAC)
- an output stage
Though there are many factors that go into designing a good DAC, most of it comes down to getting these four ingredients right. Below are more details about each element, and information about how they are implemented in the devilsound DAC.
Power. Clean power is essential for audio, and is especially critical in the mixed signal world of audio digital to analog conversion.
Power from the USB port is extremely convenient, but notoriously dirty – it can fluctuate within a 10% range (4.75 to 5.25 V), and is shared by other peripherals, such as external hard disks, which can inject large amounts of noise into the power line.
In the devilsound DAC, clean power is regenerated from USB power using a small, fast DC-DC converter chip, the MAX8614. The MAX is used to regenerate positive and negative 6 V, which is then filtered (to remove any switching noise) and down-regulated. This provides extremely clean and stable positive and negative 5 V power rails.
The positive 5 V rail is further down-regulated to supply power to other chips on the board, which require lower voltages.
These chips can provide a low-jitter output signal, but often have only limited jitter reduction capabilities. For instance, in the CS1816, jitter up to several kHz is passed straight through to the output (or even amplified slightly), so generating a low-jitter word clock is highly dependent on having a clean input source.
The USB receiver chip world is dominated by SpAct, a technology developed by Burr-Brown (now Texas Instruments) to recover a low-jitter clock signal from USB. There is an interesting article on the web about the development of SpAct, which goes into the difference between USB and traditional digital audio transmission, and how exactly SpAct works.
The devilsound DAC uses the PCM2706, a Burr-Brown chip utilizing SpAct, as the USB receiver. The devilsound DAC bypasses the chip’s built-in digital to analog circuitry, and just uses the PCM2706 to generate a clean, low-jitter digital audio stream.
Stable power is critical for the digital receiver stage. Fluctuations in voltage can easily modulate the clocks and phase-locked loops (PLLs) in this stage, resulting in unacceptable jitter performance. The clean regenerated power in the devilsound DAC helps insure that jitter is kept to a minimum.
Digital to Analog Conversion. The devilsound DAC uses a stereo pair of Analog Devices AD1851s, run in non-oversampled mode.
We use the 1851RZ-J version of the chip. The “R” means surface mount (SOIC), the “Z” means the chip is lead-free / ROHS compliant, and the “-J” specifies the highest quality version of the chip.
The AD1851 is in the same family as the AD1861 and the more popular (especially for NOS) AD1865. (The AD1865 is currently being phased out of production.) These three chips share virtually identical THD specifications, and the AD1851 is a good match for new designs with 16-bit resolution.
Glue Logic. There are actually several different ways of specifying a decoded digital audio signal. The output of the PCM2706 is in I2S, whereas the AD1851 requires a data signal that is right justified.
The solution is to use a shift register, which aligns the data bits, and splits them into left and right channels. The word clock from the PCM2706 is connected directly to the AD1851 DAC chips, to keep the jitter low.
The bit-shifting / channel-splitting circuitry is implemented using a Xilinx CPLD. The devilsound DAC circuit board contains an JTAG header connected to the CPLD, which allows the firmware to be written to the chip after it has been assembled onto the board.
Output Stage. There are many different ways of designing an output stage for an audio DAC. The technique used depends a lot on the digital to analog converter chip, various design trade-offs, and of course personal preference.
The devilsound DAC uses a single-stage op-amp output, which integrates a current to voltage (I/V) converter with a gentle low-pass filter. The circuit is implemented with the theoretical minimum number of components (two resistors and one capacitor per channel).
Integrating the low-pass filter into the I/V stage allows the op-amp to operate far below its bandwidth limit, and eliminates any problems with ringing when driving a capacitive load (i.e., cables and an amplifier). The op-amp is further isolated from the outputs with a snubber resistor.
Also, this circuit allows the output to be DC coupled, eliminating output coupling capacitors. Not only does this save a lot of space (critical for a DAC this size), but it keeps the signal path cleaner, and eliminates any time smearing on the bottom end from the low frequency roll off.