gallery/2 ecpsoftware articles diy audio links contact

HPDAC v0.1b2

This is the project page for the HPDAC v0.1b2. (Updated 12/24/2006)


The problem, for me, with the available portable amps is that they aren't really that portable. They all need batteries, and, even worse, they all need sources which also need batteries as well as cables to connect them. I wanted something that I could take to a coffee shop (or the sofa) connect to my computer, and listen without a pile of boxes and wires next to me. That is, I wanted a usb dac with a built in headphone amp that was all powered via USB, and I wanted it to sound good. As there seemed to be nothing of the sort available, I made it myself.

The HPDAC is a non-oversampled (NOS) USB digital to analog converter (DAC) with a built-in headphone amp. The DAC is based on earlier work of mine, though it turns out to be a very similar circuit to several other NOS USB DAC's. The amplifier section is essentially the same circuit as the Mini3'ified version of the now discontinued Pint. The entire system is powered via the USB supply, so there are no batteries or power cords necessary. The project is also very small and will fit comfortably into the smallest Hammond extruded case.

Not only are the amplifier and DAC sections of the HPDAC of very high quality, great care has been taken to ensure that the power derived from the USB source is clean and free of noise. The power supply utilizes a highly efficient DC-DC boost converter to deliver enough voltage to the amplifier and DAC. Further, this boosted power is regulated with high quality regulators and further cleaned with ferrite beads. This arrangement delivers clean strong power which translates into clean strong sound.

Compared to similar projects, the HPDAC provides extraordinarily high quality sound in a small inexpensive package.

Technical Background:

In order to power the entire project via the USB supply, the HPDAC uses a DC-DC boost converter to convert the approvimate 5V input to a higher voltage which is then filtered and reregulated. The exact voltage is determined by the builder. This power supply scheme presents several challenges which must be considered while building. Primarily, since the DC-DC converter is about 70% efficient, and USB can supply about 500mA at 5V, this means that at 10V you have about 170mA at most to work with. Practicality means that 100 to 120mA is probably generous, and since the TDA1543 draws up to 60mA, this is a real limit. Second, the USB specs limit the amount of capacitance on the USB rail unless there is current limiting. The DC-DC converter does have a slow start feature, and the voltage regulators do limit current, but this is still worth being careful with. It is not worth damaging your computer. That said, I have had no problems, but prudence suggests being aware. As a DIY project, you are the responsible party. Finally, it should be mentioned, this is a fairly challenging DIY project. It involves a considerable amount of SMD soldering and is intened for advanced DIY'ers. As such, the only build instructions are to select the parts, and assemble from shortest to tallest. I like to do the tricky stuff first (the PCM2707, the DC-DC converter, and the REG101) but YMMV.

At the request of some car audio guys, I added one last option: the HPDAC can be used as a source/preamp in a speaker amp setup. To do this, you need to remove the option for the buffered ground by not populating L12 (which is on the bottom of the board) and adding a small jumper between J1 and J2. To reduce noise a bit, you might try using a ferrite instead of a wire jumper, though I have not tried this myself to know if it is a good option -- try it and see. I should warn you that there are no output coupling caps from the amplifier, so obtaining a vanishingly low DC offset or using a power amp with input coupling caps is essential. Using OPA2132 opamps mine had an offset of 0.4mV on each channel. Note that you should still be able to drive headphones with this setup, but perhaps not quite as well, though the difference is minimal.

View the schematic
View the top of the board
View the bottom of the board

Late Breaking News

There appears to be a small bug in the schematic and hence in the board. To date, there have been no reports of problems being caused by it, but if you are concerned, it is an easy fix. Pin 11 (SSPND) on the PCM2707 is powered while it should not be (it actually provides voltage to turn on some DAC chips.) To disable this, you can either cut the small trace that connects pin 10 to pin 11 (it is right between the two pads -- this is most easily done before the board is built) or, once assembled, use an X-acto knife to cut pin 11 off of the chip.

Other late breaking news is that if a short occurs, the DC-DC converter is susceptible to damage. The most likely cause of this is a short at the headphone jack during insertion or removal of the headphones. As such, it is recommended that you only plug or unplug the headphones while the unit is powered off, or at least the volume is turned down. It might be that this only occurs if the computer does not have a built-in current limiter to the USB port, but I am not sure about that.


Boards are sold out. As of now, there are no plans to order another run. It is possible that another revision may appear, but not any time soon. Thanks to everyone who ordered one and participated in this project.

Samgotit's comments at Head-Fi (regarding the v0.1a board with NJR4556 opamps)...
The first revision was a huge surprise considering its size and power supply. I found Bass and "Weight" I have never heard before out of the K701. This includes through all of the non-portable gear I own! There were moments when the bass was... "concentric rings in the water cup"... spectacular. Understand, I'm not a huge bass head, but the new sound was very addictive.

Parts List
Part Value/Notes Size/#
C1 This is the main power decoupling cap for the DAC chip. 47uF is probably a decent value to use though some people like to use larger. Quality counts in this spot, so an Os-Con, Cerafine, or Blackgate is a good idea. Digikey has the Silmic II or the Panasonic FM series which are also good choices. 3.5mm pin, 8mm width
C2 (A, B, C, D) These are the main power supply caps. They provide some filtering, but mainly seem to provide instantaneous power for the amplifier section. There is a dilemma here where higher capacitance will offer better sound, more bass, etc., but the USB standard limits the total amount of capacitance that can be used safely The HPDAC does limit the inrush current both with a slow starting DC-DC converter, and a voltage regulator that limits the current. Exactly what value capacitance to use is the responsibility of the builder to determine. 470uF is probably on the high end, 4x47uF high quality caps is perhaps on the low end? I would say that 4x 47uF Silmic II caps from Digikey (part #604-1054-ND) is probably a decent place to start, but this might also be a good place to experiment. You only need to populate one position but can do all 4 if you want. up to 4 caps with 3.5mm pin, 8mm width
C3, C4 C3 is the input cap for the DAC's regulator. C4 is the regulator's output cap as well as the decoupling cap for the DAC. Wima MKP2 0.1uF/63V caps (or their equivalent from other manufacturers) are recommended, but polyesters will do in a pinch. The Mouser part number for these caps is 505-MKP20.1/100/5 5mm pin, 5.5mm x 7.5mm
C5, C6 33pF ceramic 1206
C7-C19, C26 0.1uF (100nF) ceramic (but see the note about C8 below in the VR3 section as in some cases it needs to be significantly higher in value.) 1206
C20 4.7uF Tantalum
Assembly Note: Tantalum caps are polarized, and unlike electrolytics have their positive side marked. In this case, the positive side is the one furthest from the USB jack side of the board.
C21 .01uF (10nF) ceramic, but see the note about this cap below in the VR3 section. 1206
C25 4.7uF Tantalum
Assembly Note: The positive side is the one nearer the board's edge (furthest from the label.)
C27, C28 Optional, 10pF Silver Mica or Film. In general, these should only be used if there are oscillation problems, and then only as a last resort. The amp sounds better without them. 1210
CR, CL, CG These are the AC coupling caps that connect the DAC and the amplifier sections. High quality is essential as the entire signal passes through them. The cheapest I would use would be 3 x 4.7uF Nichicon ES bipolar caps which you can get from me. You could also use 4.7uF Blackgate N series caps here. The Blackgates cost significantly more. They do sound better (more detailed, less grain, maybe a little less bass but more balance) but not necessarilly a ton better. (You can see my comments about these caps at My sense is if you have $200 + headphones, go with the blackgates, otherwise don't worry about it. If you do use the Blackgate caps here, give them minimum 100 hours to break in. I know this sounds silly, but it took mine in this design at least that to develop any bass at all. Both Panasonic and Nichicon make cheaper bi-polar electrolytics that are available from Digikey/Mouser, but they are not recommended as they sound awful. If you choose to use polar caps, the "+" on the board is meaningless, so be sure to get the direction right -- either by measuring or by following the schematic. 2,5mm pin 6mm case
L1 There is no L1. .
L3-L12 These are ferrite beads used to help clean up the power supply. Specific parts are not essential. MI1206K260R-10 works fine. 1206
The following parts (DC1, C22, C23, C24, R5, R6, L2, D1) make up the boost switching power supply. I have included a suggestion for parts which provides about 9.5V and which has proved to be stable. However, there is no substitute for reading the DC-DC converter's datasheet.
DC1 This is the DC-DC converter. Use a TPS61040. As this is an important element of the project, it is a very good idea to read over the datasheet. Also note that Digikey shows this as out of stock, but if you add it to your cart you will see that they have several thousand, so no worries. TPS61040
C22 (Cff on the datasheet) 100pF ceramic 0805
C23 (Co on the datasheet) 1uF tantalum or ceramic.
Assembly Note: If you use a polarized tant here, the positive side is the one furthest from the board's edge. I have only used ceramic caps, so I can't comment on the difference.
C24 (Cin on the datasheet) 4.7uF Tantalum
Assembly Note: This is a very small part. The positive side is the one furthest from the "C24" label.
L2 (L1 on the datasheet) ELJ-FA100KF 1206/1210
R5 (R2 on the datasheet) R5 and R6 set the ouput voltage if the DC-DC converter. The chosen values will give you about 9.5V. If you want more or less, read the datasheet for how to determine these resistors' values.
R6 (R1 on the datasheet) 634K (see note above) R1206
D1 (D1 on the datasheet) MBR0530T1G, or perhaps MBR0530TPMSCT
Assembly Note: D1 is polarized. The cathode (the marked end) is the one furthest from the USB jack side of the board.
The following parts (R1L and R to R4L and R and Pot) are the resistors in the amplifier section. If you are using an unstable "cranky" opamp, several things need to be balanced:
  • The gain of the amp section is determined by R3* and R4* (R4/(R3 + 1))
  • R1* in parallel with R2* should equal R3* in parallel with R4*
  • R2* should be 10X the value if the pot
The suggested values below are simply one such configuration. If you are using a stable opamp (usually a FET input part) then think of the amp section as a basic CMoy to determine part values.
Pot This is the potentiometer. It sets the input impedence of the amplifier section as well as adjusting volume. 10K is the default value here. Increasing the size of the pot means also increasing R2L and R2R which could increase the noise level a little. Probably not enough to worry about, so 20K or 50K is fine but 10K is preferred. EVJ-C20F02D14 is the default recommendation.
Assembly Note: The pot has very small spacers built in to it. Using a wire cutter, you should trim these off as it will fit the case a little better.
Panasonic EVJ
R1L, R1R 1K This may be jumpered in some configurations -- if in doubt, leave it in. 1/8W Axial
R2L, R2R 100K for a 10K pot, 200K for a 20K pot, etc. (this may be optional in some configurations, but if used it should be 10X the value of the volume pot.) 1/8W Axial
R3L, R3R 1K2 1206
R4L, R4R 6K2 1206
R1, R2 1M 1206
R3, R4 These form the voltage divider for the amp's virtual ground. 47K is appropriate. A little higher or lower should be fine. 1206
R7, R8 22R 1206
R9, R10 1K5 1206
RBCK, RD, RWS These connect the USB receiver chip to the DAC chip and are used to help clean the digital signal. Anything from 22R to 100R should be fine with 22R being the default recommended value. 1206
RIVL, RIVR, REF These are the so-called I/V resistors that convert the current output from the DAC to voltage. Quality counts here more than usual with different resistors offering very different sound. If you use metal film, use high quality resistors like PRP or older Holco. Dale's are fine though they are not my favorite as are Draloric. You can also use Allen Bradley or Stackpole for a fuller sound. Rikens are a great choice, though they will be very tight so must be flush with the board -- Kiwame's won't fit. There is also space on the bottom of the board for 1206 sized resistors which might be a decent option.
For what it's worth, my impression of various resistors follows: I find the PRPs to be very clean and sweet sounding, but to lack a little bass and be a little cold. They make the HPDAC sound a little more solid-statey with a lot of crisp detail. The PRPs (and really any metal film resistors) will sound a litle bit sheeny in the way that a lot of high end solid state gear does. If you like that sort of sound, then this is what you want here. Steel string acoustic guitars sound very real to me with these. Other metal films like Dales or Panasonics, or other generic resistors are similar but lack the sweetness.
Allen Bradleys (and other vintage carbon composition resistors) are the polar opposite. They are warm and pillowey and very euphonic. They offer a thick warm sound. They are a little noisier and you might hear it as background hiss at high volumes, though there is some hiss anyway due to the PS, so the addition is probably not noticable.. You can get these at various places, including Handmade electronics, Nebraska Surplus, and a few other places.
Riken Ohms are in the middle. While perhaps closer to the ABs in tonal balance, they have all the detail of the PRPs and also do not have the noise problems of the vintage resistors. They are, however, expensive.
One last resistor to think about are Tantalum resistors. These are my favorite I/V resistor offering a very natural organic sound. I use them in my personal HPDAC (though I use the ABs in my secondary one.) They are really expensive, though and are available from Parts Connexion and Angela.
Also, the REF resistor is considerably less important to the overall sound, so a generic metal film is okay there.
Optimal value is tricky here. This is in part because it depends upon what voltage you give the DAC chip (I'll assume 5V here) but also because they set the output impedance of the DAC section and influence the amplifier section. If you are using a cranky high current opamp, then you will probably be using a 10K pot which means that you want the output impedence of the DAC to be 1K or lower. Thus, in this case, 1K is recommended. If you are using a less cranky opamp, then you can use a higher value pot and thus a higher value I/V resistor. If this is the case, then you will need to do some reading and or experimenting to find the optimal value. 2.7K for the I/V's and 1.4K for the Ref is probably a decent place to start, though you may want 8V for this option. Anyhow, as noted, 3 x 1K work well with a 5V supply.
up to 1/2W Axial (lower W rating is fine) or 1206 SMD
RLED Sets the LED current. Anything from 470R to 1K is fine, lower will lead to a brighter LED and will drain more power. 1206 SMD
PCM This is the chip that converts from USB to something the DAC chip can use (I2S). The board was designed with the PCM2707 in mind, but the PCM2706 also works just fine. PCM2707
TDA1543 This is the DAC chip and is supplied with the board. In theory, you could stack these*, but if you decide to do this be very cautious about current draw. Personally, I wouldn't do it. Also, don't put the chip in a socket. Considerable attention has been paid to layout such that the digital signal to the chip is as clean as possible. The socket is counterproductive and will audibly degrate the sound quality.
* Some people think this sounds better, though I don't. It seems to trade some midrange magic for greater extension at the frequency extremes. The idea is that on any given bit a DAC can make an error. By stacking chips you increase the chance that there is some error on any given bit, but decrease the amount of that error.
8 pin DIP
OP1, OP2 These are the opamps for the amplifier section. OP2 provides voltage gain while one half of OP1 creates a virtual ground while the other half buffers the ground. In general, since there are no buffers, at least for low impedence headphones, you want high current opamps. This typically means something that has some amount of instability.* The amp section is designed to handle this instability in all but the worst cases. Thus, opamps like the NJR4556 of the LM6172 are appropriate. (For what it is worth, I have found the NJR4556 to have very good bass and dynamics but to be a bit harsh on the top end. Conversely, the LM6172 is very smooth and open on the top end but does not have quite as much bass slam, though there is plenty for most purposes.) The default recomendation is thus the LM6172. It is cheap and sounds very very good. My sense is you should have a good reason for using something else, but then, it is your amp so you maybe shouldn't listen to me. For headphones that do not need as much current, less cranky opamps (OPA2132, AD8620) may work well. Also, if you plan to set this up as a preamp (this is a note to the car audio guys!), using a less cranky fet input opamp like the OPA2132 or AD8620 might be a good idea as the offset should be slightly lower.
* A full vetting of this topic is well beyond the scope of this parts list. The amp section is based upon the Mini3'ified version of the now discontinued Pint amp. However, while those amps concentrate in the AD8397, that particular part draws enough current that its use here might be problematic. If you want to use it, be sure you know what you are doing. I have not tried it, so I can't really offer any advice. I have, however, tried the AD8397 in other circuits and to my ear the LM6172 is the better sounding opamp anyway. The high current opamps (the ones mentioned above) that are perhaps more appropriate here are a bit more stable than the 8397 making them a little easier to work with.
2x Dual 8 pin SMD
VR1 This is the voltage regulator for the amplifier section. My suggestion is that you want something that limits current to avoid damaging the USB bus in the computer either due to the inrush of C2 or due to a short somewhere. You also want a LDO regulator so that it can be as close to the output of the DC-DC converter as possible to avoid wasting voltage. So, if you used the suggested values for R5 and R6, then an 8V LDO regulator with 150mA current capability is perfect here. TL750L08CLP works quite nicely. TL750L08CLP
VR2 This is the voltage regulator for the TDA1543. The part you use will depend on a number of factors. First, you must give it a high enough voltage on the input to work and it's output voltage must fall within the range specified by the TDA1543. This means that the output must be between 5 and 8 volts. It also means that, depending upon where you set the DC-DC converter's output you might need a LDO regulator here. While high voltage does change the sound of the TDA1543, it is not necessarilly for the better, though it is not necessarilly for the worse either. The default choice would be the LM78L05 as it is stable and easy to use and cheap. However, some people swear by the AN8008 and AN8005 Vregs for the position next to the 1543. My own experience is that these do indeed sound pretty good. However, they are only rated for 50mA and the 1543 datasheet says that it can draw up to 60mA, so they might not be the best choice. On the other hand, they might be worth considering. LM78L05ACZFS-ND
VR3 This is the voltage regulator for the PCM2707. There are a couple of ways to set this up. The PCM2707/2706 has built in internal voltage regulators. They are enabled by supplying the Vbus pin with 5V. When this is done, then the other power supply pins need to be decoupled externally, and have a measured 3.3V on them whic his internally supplied. Alternatively, you can diasable the the internal regulators by supply 3.3V to the Vbus pin and then supply 3.3V to all of the power pins youself.

So, there are 3 ways to set this up:

  • The simplest (and cheapest) is to use the PCM's internal voltage regulator. To do this, do not populate C7, C8, C21, L6, and L10. Instead of using a regulator for VR3, simply jumper from the input to the output pins.
  • To supply 3.3V externally, use a 3.3V REG101 for VR3 and populate C7, C8, C21, L6, and L10 as per the normal instructions.
  • If you like to live on the edge and want to use dual regulation of the PCM, you can use a low dropout voltage regulator for VR3 to supply something on the order of 4.5 to 4.75V (the choice may depend on what voltage your USB supplies). In this case you should populate C7, C8, and C21 according to the voltage regulator's datasheet, but you should not populate L6 and L10 as you will be using the PCM's internal regulators.
The $20,000 question here is, does this make any difference, and the answer is, I have no idea. I have been using the 3.3V REG101 option as my default, but the simple cheap option seems fine even though it is theoretically noisier. I have not tried the 3rd option, so I have no opinion there.
SOT23-5 foot print
LED1 Any 3mm LED should be fine. This part can be left out if you don't like LED's. It only indicates that usb power is connected, not that anything is working.
Assembly Note: The anode (the longer lead) goes closer to the "RLED" label.
Q1 The 12MHz crystal oscillator for the PCM2707. Part number MP120 (CTX058-ND) from Digikey is fine here. HC49
USB Part #571-7877801 from Mouser is suggested. I have tried a lot of USB jacks and these have become my favorite as the USB plug clicks in and stays stable. 571-7877801
OUT CP1-3513-ND (SJ1-3513) from Digikey is the best option. However, 161-3507 from Mouser also fits, though it it a little crooked. I was also at Radio Shack the other day and noticed that they have a jack that should fit -- it looked the same as the Digikey jack to me. CP1-3513-ND
Case The board was designed to fit in a Hammond 1455C802. Digikey seems to have some of these in stock, but if they run out Mouser stocks them as well. 1455C802
Knob You will probably want a knob of some sort. Smaller is better here as otherwise it will extend beyond the edge of the case. Digikey has a bunch to choose from. (226-1025-ND, 226-1029-ND, 226-1033-ND, 226-1041-ND) .

Some build notes and options:

  • In general, this project like most PCB based projects should be built from shortest to tallest parts. In addition, I have found that a good order is to first add the most difficult parts (the PCM2707, the REG101, the DC-DC converter, and the opamps), then add the smd parts on the top of the board, and then add the rest of the surface mount parts to the bottom, then do the tall stuff.
  • If desired, one can build this project as only a NOS USB DAC leaving off the headphone section. In addition, CG should be left out and the output ground should be from true ground. There will be no cap bleeding resistors on the output, so the next stage will need resistors (10K to 100K) from input to ground (usually a volume pot.)

Special thanks go to Tangent without whose Pint this project would not have happened, at least not in the form it happened in. Similarly, Amb and Morsel's modification of the Pint and work on the Mini3 proved very helpful. It is probably worth looking at the Pint project page for understanding the issues with the amplifier section though since I suggest you stay away from the AD8397, some of those issues are not that important.