As mentioned earlier, the DIYaudio thread contains many posts regarding my DAC and how it was tweaked and polished with the help of their members. If you would like to go through my design process or want to ask questions, that's the place to go. This article is mainly a summary of what I did (and to save you from reading nine pages of text).
What's a DAC, and do I even need one?
A DAC, or digital-to-analog converter, is a device that converts a digital audio signal, such as that from a MP3 file or CD, into the analog domain. When you connect your headphones to the jack on your computer and listen to a song, the digital bits from the computer enter the DAC in your motherboard and are released as an analog signal, which is amplified and sent out as a sound wave into your ears. Naturally, not all DACs are created equal, and onboard audio chips generally aren't designed for Hi-Fi sound. The switching power supply as well as the computer circuitry generates lots of noise, which ends up in your headphones. Also, motherboards are generally designed without audio performance in mind, and their onboard DACs often have higher distortion and less detail than a standalone unit.
If you can't hear the difference, then it doesn't matter. However, I did - and I didn't want to just get one from a store. Of course, what's the fun in buying it when you can make it yourself... and designing was a better challenge than, say, soldering parts on a pre-made board.
I'll only talk about the chips here. For passives, I just chose general-purpose parts (there's lots of debate about whether or not audiophile parts sound better, and just don't have the money to try them out).
- Power regulation: TI LP5907 for 3.3V output, and a standard 7805 as a 5V pre-reg. The LP5907 has low noise and a high PSRR, crucial characteristics when working with analog signals. The 7805, well, doesn't really matter too much in this context, and has reasonable specs.
- SPDIF receiver: Wolfson WM8804, which basically turns the SPDIF signal into an I2S signal (the format required by my DAC chip). The only reason I chose this over other products is that it uses a reference crystal to remove (possibly audible) jitter in the signal.
- DAC chip: TI PCM5102A. I didn't feel comfortable working with a differential (current output) DAC and opamps for my first design, so I had to choose a chip with a line driver (voltage output) that can be directly connected to other equipment. Objectively, this DAC has pretty transparent specs and isn't too expensive.
I chose to follow the manufacturer's datasheet schematics in most cases, and later designed the power supply with the help of simulation software. By using a linear power supply instead of a USB connection (my original intention was for it to be USB-powered), I do not have to rely on the noisy computer PSU and the USB inrush current specifications. A ground plane is used not only to return ground signals, but also as a EM shield for the circuitry.
Small modifications to the default circuits included the addition of extra capacitors, damping resistors for the I2S lines, and an isolation transformer for the SPDIF input. The latter is designed to prevent common-ground noise.
It might not have the frills present in boutique DACs, but simple is good. And cheap.
Schematic is below:
This is a Creative Commons design after all, and is intended for DIYers to build. The Gerbers needed to make the boards are available at the end of this post. However, I do rely on fairly fine-pitch chips, 0805 SMT passives, and SOT-23-5 regulators. Mistakes are painful, and I broke one of my boards and lost a PCM5102A while attempting to realign the chip.
Personally, I feel that SMT work can be done without specialized equipment, like a reflow oven or magnifier. I built this board with only a Hakko 926 iron and a bottle of flux, and believe that it's entirely doable if you're reasonably competent in soldering.
There are still some problems that could be ironed out, but the current design does work and can be built. However, I would suggest that you read the forum thread first and understand all the issues before attempting to order boards. On first powerup, be sure to check all regulators with a voltmeter to ensure that power is being delivered to the components.
Subjectively, the DAC sounds warmer that I expected. Detail shows up nicely and the sound is smooth, but I wouldn't completely call it transparent or neutral. Whether you like it or not - that's up to you to decide. I won't go into the argument between solid state and tube amp enthusiasts that has gone on for decades.
Due to the warmer sound, genres like classical, jazz, country, etc. sound better than synthesized tunes like hip-hop or electronic. Then again, the differences are very small and are only visible through A/B testing with a Toshiba SD-3950 as a reference. The Toshiba has a lower THD, meaning that it should be a more neutral device than my DAC. However, subjectivity is subjectivity, and all my interpretations should be treated with a grain of salt.
My initial RMAA tests were limited by the Realtek ALC892 CODEC I used, which reported clipping and high distortion. However, the output levels of the DAC exceed the Realtek's specifications, and I do believe that this is what's causing the issues. For more details, as well as the actual RMAA results, start here on this post and work your way down.
I do plan to get a recording interface to make proper measurements, and after that can confirm whether or not my board has problems. I already have a DAC1.1 in mind, which would have minor tweaks (3.6V regulator for the charge pump, support for film caps) and a better layout.
Once I get more information, I'll make a new article. Stay tuned.
I'm sharing this design under the CC-BY-NC 4.0 license (same as my blog).