A Portable DAC with Headphone Amplifier.

by Andrija Ifkovic


The idea behind the portable DAC is to get better sound than from the line out or headphone out of portable CD player, a pc sound card or other digital audio source. Many of us want to listen to music on the go or on PCs, and would like to get better quality than most of audio rigs today can provide. The quality of the headphones is the main contributor to the overall sound quality from a rig. Quite a few higher-end headphones require more powerful amplifiers than what’s supplied with the average portable device. When it comes to audio sources (i.e., PCDP or other), these devices are mass-market, optimized for a set of parameters that doesn’t include audio quality.

I set my goal high and wanted to get superior sound quality from both portable and non-portable equipment. This PDAC is my fourth design. It has fewer parts (and fewer SMD parts), is smaller, and is easier to build than the other three prototypes. For a change, this one is both a DAC and a headphone amplifier. It has a 3.5mm input jack with breakout, so that the PDAC can be used as a regular headphone amplifier. The portable DAC is battery-powered, but can run off a good quality wall-wart or better, a separate regulated power supply. The PDAC’s compatibility with digital sources is excellent. It works with any optical out I’ve tried: DVD player (at 24-bit/96kHz, although internally the PDAC is 16-bit/96kHz), portable CD player, Sony CDP-CX300 CD MegaChanger, and the SoundBlaster Audigy Platinum sound card.

The Circuit

There are obviously compromises to be made in order to make a normally largish device fit a portable case. What does the “portability” imply? What features can be sacrificed and to what level? What is the target performance level?

  • Size – device must be of “portable” size – something that can fit into a pocket or at least into one of those bags meant for carrying around PCDPs and the associated CDs, amps, headphones etc.
  • Power source – device must be operating on batteries.
  • Audio quality target – Obviously, the device has to sound significantly better than most PCDPs (when used in line-out mode) in order for this project to make any sense.
  • Other features – a built-in headphone amplifier, optical/coax digital input.
  • Cost

Figure 1

Figure 1 is block diagram of the final PDAC. The most critical parameter in designing a portable DAC is power consumption. The digital receiver and DAC chips draw the most power. The research on these chips was hard work that lasted for weeks and then months. Every semiconductor manufacturer has readily available, cheap, simple to use, low to ultralow power consumption chips, but most of these won’t work for this project because the sound quality would be the same as that from PCDPs. The first prototype used the Crystal CS8412 receiver and CS4397 DAC chips, but was not successful because it ate batteries way too fast.

Figure 2

Beginning with the second prototype, I used the Burr-Brown DIR1703 receiver and Analog Devices AD1866 16-bit R2R DAC. The DIR1703 has very low jitter (75ps) and will handle up to 24 bit and 96kHz but output 16/44 to the AD1866. It needs only 30mA at 3.3V. Setting up the chip is somewhat cumbersome and involved. The biggest issue is its inability to handle S/PDIF directly from the line without any conditioning, and worse yet, an inability to handle both coax and optical in without a switch. Crystal chips allow that, although of course optical and coax cannot be active at the same time.

The DIR1703 requires the external crystal or oscillator, but this shouldn’t be counted against it as that’s probably why it can claim much lower jitter since it doesn’t extract master clock and recycle it but uses two separate PLLs. It also does sample rate conversion, which is quite neat, even though not necessary. The receiver is reset at system turn-on by the Analog Devices ADM707 signal generator. The DIR1703 is available from Digikey, and it’s cheap. The biggest negative has got to be its size – SSOP, potentially tough to solder. If soldering is successful, however, the gain is small footprint of the chip, and the space is at premium in this project.

Figure 3

As to the new choice for DAC, I don’t know how I was persuaded to try AD1866. I think it must have been an email from Rickcr42, and at any rate he is certainly the one responsible for making me try that chip. I considered its numbers not very appealing at first, but when I became desperate, it started looking good. The AD1866 is an R2R DAC, which means in principle it is a network of resistors using values R and 2R in a ladder-like formation, and each of 16 bits controls one of the nodes, contributing 2 times less voltage than the previous more significant bit at the output of the network. It accepts a digital stream with 2x to 16x oversampling at 44.1 kHz. The oversampling can be as high as 192 kHz, but is not supported by the DIR1703. And the power consumption of this R2R chip is by far the best considering the sound quality it claims.

The AD1866 DAC called for different analog stage as it has single-ended output with 2.5V as common voltage. I used for two AD823 opamps and one CLC5602 buffer to make a combined fourth-order Bessel filter and buffered amplifier for the DAC’s output. The buffered amplifier was based on the multi-loop circuit of the Toni Kemhagen’s Lindesberg headphone amplifier.

The second PDAC prototype had deep, deep bass and shrill-free highs, natural-sounding strings and wide, distortion-free dynamic range. In listening sessions with friends we eventually found out some weaknesses. The sound was not as transparent as it is on my bigger DACs or as it is on friend’s high end Marantz SR-19EX receiver. Cymbal clashes didn’t reverberate as long, ambience was not quite there and a friend commented that it was harder to distinguish voices of two singers on one of my test tracks (from Verdi’s Don Carlo). Also, this prototype did not have volume control, so it cannot be used for driving headphones directly. This limits the usability as a portable device by a large factor.

The third prototype made the analog stage DC coupled by adding virtual ground driver, courtesy of a BUF634 buffer. It added a volume control, although this was not the ideal solution and was somewhat experimental, and a fully functional coaxial input. It could power headphones although the volume wasn’t close to deafening due to the limited voltage swing.

For the fourth (and final prototype), I wanted sound that would equal or beat all audio sources I have at home, such as my upgraded CD player, which presented a challenge that I got close to in the third prototype. The sound of my friend’s high-end receiver (Marantz SR-19EX) was better, even to untrained ears, although it took some listening time to determine that. The other thing prompting this revision was Chu Moy, the owner and moderator of HeadWize, who asked me to simplify the DAC so that it was easier to build.

Figure 4

The first thing I did was to tie a separate headphone amp using the Elantec EL2001 buffer and two 9V batteries directly to the output of DAC chip. It did sound better and quite different. To increase the voltage available to analog stage, I went full out and implemented the ultimate multiloop circuit with the Elantec EL2001 and new Analog Devices AD8610 opamps. I threw out the Bessel filter mostly because I wanted to simplify build, and I didn’t want to waste expensive and hard-to-find AD8610 opamps on filtering (it was replaced with a first-order RC filter). I did however manage to implement fully-functional volume control (a shunted design for highest fidelity).

At the same time, I removed some parts from the digital section, mostly 1K resistors, and replaced many SMD parts with through-hole equivalents. I also utilized the underside of the board and moved some voltage regulators there, and also got rid of the regulator for the analog section that was now not usable. These changes have reduced both the complexity and size of the board, and allowed two extra batteries to be placed in the box, after I also got rid of never-used battery charger.

Figure 5

Prototype 4 was first built with separate battery supplies for the digital (VCC) and analog sections (VA): 6 AAs in series (VCC) and two 9V batteries or two 7.2V Li-ion batteries in series (VA). They all still fit in the box! However, the power supply and headphone amplifier schematics – figures 4 and 5 – are wired for a convenient single battery supply (eight AA batteries). Rx is the current limiting resistor for the zener diode. The main advantage of having two separate battery supplies is that the input of the headphone amplifier can be direct-coupled (the capacitor at R16 becomes a 1K resistor and R16a is removed). The full-size schematic of prototype 4 shows the wiring for the two battery supply (see below).

Note: Because the analog section uses a virtual ground, the grounds for the digital and analog sections are separate. They are ac-coupled through two capacitors shown in figure 4.

I got overly obsessed with the issue of voltage regulation for the digital section. I went through every possible scenario, including ultra-precision references powering transistors or opamps, all the way up to two-terminal simple shunt regulators. In order to minimize jitter, I wanted ultra low noise regulators and good load regulation for very stable voltages. Line regulation was somewhat less critical, since battery voltage is probably fairly stable in short term. Good behaviour into MHz region was desired as well for feeding fast digital chips.

I settled on the Analog Devices ADP3303. Actually, I wanted ADP3303A, which is adjustable and can get to 5.5V if operated on the very edge of its specifications, but they are not available in small quantities. Luckily, the 3.3V and 5V versions are easily available. These regulators can supply 200mA, have excellent load and line regulation (0.8% at 25°C) and very low noise (30uV RMS). They also have a low dropout voltage of 180mV at 200mA (important for battery operation), an indication of going out of regulation when voltage is too low, current and thermal limiting, and can tolerate just about any capacitor and stay stable. The ADP3303 can hold up well into the MHz region and consume very little power (only 1.5 to 4mA at full 200mA load). Their transient response and PSRR are very good (the PSRR doesn’t look so good in the 10kHz range but it is supposed to be much better when you use decent capacitors). There exists a lower power version of the of the ADP3303, which is pin compatible, called the ADP3103.



Download PC Board Layouts for Portable DAC (320K)
Download Datasheets for Portable DAC (1.5M)
Download Full-Size Schematics/Price Lists for Portable DAC vers. 3 and 4 (205K)

All of the prototypes except one were assembled on a pc board. I built one on protoboard. However, since several of the digital chips and the AD8610 opamps are SOIC, building this version of the portable DAC with point-to-point wiring or a protoboard is not a very good idea. The pc board for the final version of the portable DAC measures approximately 8cm x 8cm (little over 3″ x 3″) and is double-sided. The downloadable pc board patterns are in postscript formatted files. I’ve kept signal traces as short as possible and tried to have the ground plane as continuous as possible. All pads and traces are on the top (ground) side, surrounded by ground plane whenever possible and applicable.

The number of vias are the absolute minimum that don’t compromise the ground plane continuity too much. Vias (holes connecting lines in different layers) are kept away from parts and traces and are not present on critical signals, and especially not where the traces are thin and dense. Pad and via sizes, as well as trace sizes, are enlarged to allow for imprecision of the manual manufacture of PCB. Wherever possible, layout was done to minimize need for soldering through-hole components on the top side. Through-hole resistors are to be placed horizontally even though the space requirements are stringent, because it’s been shown (in an Analog Devices application note, I think) that vertical mounting will produce temperature differentials, which in turn create loss of precision. Due to my increasing expertise in making PCBs and building this thing over and over again, this prototype was done in two evenings and worked almost right away.


I wanted to use good yet easy-to-get components. Very limited size favors SMD, and I used that a lot in both the digital and analog sections. In the digital section, SMD parts are used, of 1206 size preferably, or less if required by the space constraints. They are superior in digital circuits, as their parasitics are smaller than through-hole components. They are also smaller, cheaper, have very low profile and can be simultaneously present one above another on both sides of the board. As downside, they are somewhat harder to solder, harder to debug, harder to desolder without causing destruction, can tolerate less current and voltage and can have inferior characteristics, while better specified parts may well be considerably more expensive than the through-hole counterparts.

I had tons of problems with my previous builds because the SMD contacts were not really made even though it looked soldered. I recommend having some solder on the pads beforehand, not just puting the chip on the copper (or worse, unremoved photoresist). Solder-coated pads are easy if the board is professionally made and not so hard for the DIYer. Just make them with lots of solder on the iron – the flux in the solder is enough. Then simply use desoldering braid to remove bridges between several pins at the time. At the end, use ohmmeter to check if there are no short circuits and if connections are actually made. A good magnifying glass is almost a must. No special tools are needed whatsoever (other than loupe).

Close to half of the portable DAC price is in semiconductors and capacitors, especially the polymer ones. Total parts price is about $230 Canadian I think ($150 US or so). There was no compromising in the price though, and since all the expensive semiconductors are critical to design it can’t be lowered much either.

Resistors for digital section are usually not critical. I used whatever Digikey had in the SMD form factor. Capacitors, in digital section are SMD ceramics, other than big decoupling ones. NPO or C0G should be used here, or at least X7R. I used X7R as it’s cheap yet relatively stable, but considering there are not many capacitors, one should really get the most stable ones.

Digital decoupling capacitors should behave nice well into the MHz region, so that they can smooth out the noise on the digital lines. The standard choice is Sanyo OS-CON. Unfortunately, greedy vendors ask very high prices for these, maybe even more than for Black Gate capacitors, and those are mostly older types that are now obsolete. The SMD Panasonic polymer capacitors seem to have very nice properties. They are easily available from the most convenient vendor – Digikey. And SMD in their case doesn’t really mean small. No one should have trouble soldering these. The one drawback is that they are even more expensive than OS-CON. The polymer caps are all 33uF, 6V because there is a discount when buying 10 or more from Digikey. The full-size schematic shows some polymers to be 4.7uF or 100uF, but those capacitor values are almost arbitrary.


In the analog stage, I wanted standard size 1/8W resistors. It is hard to impossible, and also expensive to get good resistors in SMD size. Holco and Vishay bulk foil come to mind but Holco is overly expensive and Vishay is not only expensive but also completely unavailable to even a non-US company, let alone a private person. I don’t know how good or bad are higher-grade SMD resistors like those 0.1% 10ppm. I wanted to leave room for experimentation. There are more choices in standard size, plus I’m sure most people will find them easier to solder. So I settled on standard size. Vishay-Dale are good and cheap choice here, but one can use Draloric, Roderstein, Holco or any standard metal film resistor, if one so wishes, even Tantalum resistors.

Non-polar capacitors in analog section are polypropylene – pure, unaltered, simple polypropylene. Since they’re all small capacity, it’s size-wise feasible and cheap to use them. Boutique capacitors would probably be too large. However you might be able to fit Silver Mica capacitors. These are the most stable you’ll find but they’re expensive. If I wanted to go SMD, I would choose Panasonic 2% PPS film SMD capacitors. They have nice tolerances and filters are their intended usage.

When bypass capacitors of low electrolytic value range are needed, ELNA Cerafine or Black Gate series N (top grade) can be used, as they are cheap enough in the needed capacity – or Panasonic FC or Nichicon Gold if the price or availability are issue. Due to size limitations, I am using only a small (cheap and easy-to-get Panasonic FC) electrolytic capacitor before every voltage regulator.


The above front panel layout (actually for prototype 3) has both optical and coax inputs. See the full-size schematic for prototype 3 for more information about adding a coax input. The specified 3.5mm input jack has an extra set of “breakout” contacts for switching between two stereo sources. So basically when there’s nothing plugged in, the signal from the DAC passes on to the amplifier. When there IS something plugged into the jack, that input gets fed to the amplifier and the digital section is disconnected (though it is still drawing power, so maybe add an extra switch to disengage the battery supply).


The LEDs are a red/green low-current types, using less than 1mA at resistor values I’ve set. Since IC sockets are not only impractical when there is need to solder from both sides, but also problematic for hosting high bandwidth buffers and opamps, I’ve chosen ultralow profile component carriers for DIP chips. LEDs should be very efficient to facilitate low power consumption. They should also be compact, and PCB mountable. For the rest of the parts such as jacks, diodes and so on, my tendency for this project was to use ones that are in standard sizes and easy to get from the major suppliers, namely Digikey.


The Hammond enclosure (model 1598ASGY) measures 6.1″ x 3.6″ x 1.4″. The schematics show the single 8-AA battery pack version of the portable DAC, but the pictures above show the PDAC with the original power supply: two separate battery packs (six AA batteries and two 9V batteries). If AC power is desired and the supplies for the digital and analog sections are separate, one will have to use something like a 5-pin DIN jack and build a separate dedicated supply/charger. If a single supply powers the digital and analog sections, one can obviously use much simpler jack, and only one, preferably regulated power supply is needed (although I think a wall-wart would work too). I am sure the sound quality would suffer significantly with a wallwart, since there is no voltage regulation on the analog section. One thing I’d like to see is a bigger capacitor bank for the analog stage. That would help the sound clarity in complex passages and improve bass, as well as reduce any degradation from wall-warts.

The Results


I use this portable DAC as my home DAC as well. It beats all the previous DACs I’ve built, probably because the DAC chip AD1866 is superior to CS4390, or maybe because the new receiver chip, DIR1703, has 3 times less jitter than CS8414. The PDAC is a stereo DAC. To use the PDAC with multi-channel sources like DVD players, set the DVDP’s digital output to PCM. If the PDAC is fed a Dolby Digital or DTS signal, it will output a TERRIBLE and LOUD noise.

The PDAC’s sound is incredible. There is plenty of gain now, and the volume control works wonderfully. Great imaging (channel separation), great transparency, great neutrality. For the first time, I could listen to this PDAC and hear POSITIVE differences compared to my (modified) CD player or anything else. Music has more details. The timbre of many instruments has changed – they sound more lifelike than ever before.

The PDAC drives headphones works very well this time around. There is plenty of gain, and it can drive any load to extremely loud level. The sound is transparent and natural. The Beyerdynamic DT931 headphones are a clear winner when used here, and Etymotics sound fantastic as well. On top of this, the PDAC can operate as a standalone headphone amplifier through the 3.5mm input jack. The jack is a line-in that automatically disconnects the digital section from the headphone amplifier, whenever a plug is inserted into the jack.

The SoundBlaster Audigy sound card internally resamples 44kHz to 48kHz and therefore sound quality is not the same on the digital output. It’s not easy for me to hear differences when comparing good DACs that are above certain quality threshold, and the Audigy has a superb internal DAC. However, the Audigy does add noise to analog and digital outputs (muting the line-in does reduce the noise, but does not eliminate it). The PDAC is much quieter than the Audigy DAC, and the PDAC’s headphone amplifier is quieter and more powerful than the Audigy’s amplifier for headphones.

I use PowerDVD (a software DVD player) with the incredible Dolby Headphone. The Dolby Headphone effect gets encoded in the Audigy’s digital stream too, so plugging the PDAC into optical out works the same as plugging the front-out analog into the PDAC. In any case, the effect is fantastic. I’ve also listened with some friends a few times in different apartments on speakers and the conclusion is that this PDAC does beat or equal everything we have (which is mostly high quality, consumer gear like Marantz, not audiophile labels).

c. 2002 Andrija Ifkovic.
From Designing a Portable Digital to Analog Converter. Republished with permission.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.