|
After about a year with the 8-chip version of Doede Douma's DDDAC1543 (a project still available on this site although the DAC itself has been gutted to donate parts for a third generation of this design), I felt like trying something slightly more radical. The new DAC was supposed to a) have 16 TDA1543 chips in parallel, be powered by an even larger 12V SLA battery, feature better components throughout, accept a proper BNC coax SPDIF cable, and drive a pair of headphones via a stepped attenuator that should serve to attenuate the line output as well if so desired.
The project began in November 2004, following several requests to build such a unit for a few interested people who saw my original DDDAC design and heard I was looking into building a new version. Theoretically, the new DAC sounded like a project that should have taken 8-12 weeks at the most, but a number of "issues" slowed down the project quite significantly. It was finally completed in late April 2004. I had totally underestimated the time it would take to prep the chassis, the battery boxes, and then there was the whole issue with that headphone attenuator...
The efforts were well worth it though, since this DAC very clearly outperforms the old version. Much smoother, significantly better bass, dead quiet, longer playing time, and then there's the added benefit of the passive attenuator that lets me completely bypass my preamp, even though there's still some fine tuning to do to make that work well enough for daily use.
It all started with locating the proper enclosure for a DAC that would generate a lot of heat, but should work well without any fan noise. My old 8-way DDDAC worked quite well without a cooling fan, but that DAC had a large hole in the enclosure right above the 8 chip DAC stack. With 16 DACs I had to either build a really tall enclosure or come up with something different.
|
This is where this all started - cutting the machined raw DAC chip heat sink into individual parts. The idea was to create a custom heat sink that allows to transfer the heat from the DAC chips to the enclosure heat sink. This meant having aluminum parts that would go between each DAC chip and then screw together into a 16-chip tower with smooth sides that can be bonded to the chassis walls using thermal epoxy as often used to attach heat sinks to computer video card processors. These strips of aluminum were the most expensive single component of the set of DACs I built, coming in at over $60 each. |
|
The cutting was quite simple - grab the old saw and start hacking away. Rough edges get filed later. |
|
Time to solder the DAC "towers" together. watch for chip orientation, because once they are in there you can't see which way they point, and just one placed in the wrong way will cause some serious problems... I used "arctic silver" to enhance the thermal transfer from chips to the machined aluminum, but again caution is in order, because arctic silver is a conductor. I think the next time around I'll be using a ceramic-based heat sink compound (you'll get to that realization when you desolder the entire tower to find out what is causing a voltage drop in one channel...) |
|
A finished 16-chip TDA1543 tower. All pins must be soldered perfectly (and you have to avoid overheating chips. Disassembly in order to spot a bad chip takes even more time than doing it right the first time around. There are 128 solder points on a tower like this. Not a project for beginners. |
|
And this is the "heat sink enclosure" I chose for this project. Originally designed as an external CD-ROM enclosure, this unit offers 40mm internal height and about 5 1/4" width. Ample room for the PCBs and other components that will need to go inside.The DAC tower will have to be mounted along one of the large heat sinks on the sides of the box. The voltage regulator feeding the DAC tower also gets mighty hot with 0.5 amps being drawn so that I had to find the best layout to allow heat transfer from the regulator to the enclosure as well. |
|
Because it looks so good - 64 DACs. Putting these all into a single DAC would be great, but you'd need a rather beefy power supply (2 amp constant draw) and even better cooling. These towers went into 4 separate DACs I was building at the same time. Right now, while typing this report, I am already building the next evolution of this design. This time around it took me exactly 1 hour to solder one of these towers. |
|
I wanted to use some computer PCB standoffs to space the board from the enclosure bottom. Had to drill holes and tap them to prep the boards. Next time I'll be doing this differently (holes large enough to fit over the standoffs that will be epoxied to the chassis and screws to hold down the boards on the standoffs) |
|
Assembly line work - building 4 at a time should be faster than just building one. However, putting a few components into the PCB is one of the least labor intensive parts of the job, so time savings are minimal over building one at a time. |
|
Drilling holes into the back panel of the enclosure was one of the more tedious parts of the job. The plates are made out of a pretty soft metal, so drilling gets tricky, as a small mistake can warp the whole plate. I had good results once I found a block of hardwood that would fit below the plate and let me press harder on the material without bending it. Still difficult. I may elect to have another panel machined by frontpanelexpress.com next time around. It'll easily cost $40, but the time saved is probably worth the cost, and it'll look more professional. |
|
Actually, this doesn't look too bad once it's done. |
|
Once the internal parts are connected to the rear panel, things are starting to look quite busy in the DAC. There are two DC power connectors for battery and charger, analog outs, a coax input, a toslink input with dedicated voltage regulator not installed yet in this picture), a toggle switch for those two inputs, a 12V DC input for the fan and an on/off switch for it, as well as the fan itself. |
|
Outside look once in the chassis and fully installed |
|
Here's one of those toslink voltage regulator boards being built. Not too many components, but totally hand built, even with small SMD caps on the bottom of the board to keep things within spec. This board gets screwed down against the back panel. |
|
Messy workplace while building 4 "power supply boards" at the same time. Basically, these boards are just a collection of capacitors to decouple the individual power circuits from the battery and from each other. |
|
One of the boards up close. There are mostly Nichicon Muse capacitors on the board, some small Wimas and a big power resistor that helps take the edge off the 12V that come from the battery to the DAC's LT1085 voltage regulator. Basically, this resistor keeps the regulator cooler. There are also two small molex connectors for the blue LEDs of the enclosure, powered by the DAC battery if the power switch is set to on, dark during charging. |
|
things are starting to come together - here a mock layout of where everything is about to go. Note the black material on the enclosure - that's a damping material similar to what is used in cars for noise deadening. The layout still has some flaws that will be addressed later in the build process. |
|
Different angle of the same build stage. The DAC is actually ready to play at this point, even though things need to be rigged up a little until things can be finalized. |
|
The very first power-on test. The DAC tower is held against the heat sink by a small clamp and the blue LEDs are already up and running; even the green battery checker LED is lit. Almost done? not even close! |
|
There are three more to get to that level, nor is anything done regarding the headphone output and attenuation. |
|
It does look pretty busy in the rear of the DAC once all the power hookups and coax are connected. |
|
No, that's not exactly how I envisioned the headphone output to
work. This is actually the very first rig to see if the DAC output levels are even acceptable
to bother with the headphones.
As it turned out, the volume level for the headphones was fine, but the
cheap test 100k attenuator I installed was unusable. The sound was either full volume or off. I had to
go and do some testing to determine what exactly I needed to do to have volume control over the
headphone output. I ordered some 25K noble pots hoping those would work well enough to keep the construction simple. As it turned out, the 25k Noble pots were also useless. At that point I started doing simple A/B tests with a DPDT switch as my 2-tep attenuator. The values I arrived at to get an acceptable range in volume control was so low that the only option was to build custom stepped attenuators for these units. One of the DACs was getting a nice 24 step Elma switch, while the other three would get by with a 12-step Alpha switch for a fraction of the cost. |
|
For the headphone jack I chose the "locking" Neutrik jacks, mostly because of their beefy looks. The locking part is actually pretty annoying in home use, so I took each of these jacks apart and removed the little plastic pice that turns them into locking units. Now they are non-locking and still look good. |
|
Once the attenuator values were known, I calculated the individual values for each step and ordered 2 dozen bags of resistors for the attenuators. Ready to assemble... |
|
The Elma 24 position switch disassembles pretty easily, making the installation of the resistors pretty easy. It's even easier after you already soldered 48 resistors to make the thing a 25k switch, only to find out you need to desolder it and start all over again... at least it was good practice; things always look better the second time around. Note that we are in month 3 of the project at this time! |
|
The Elma decks populated with 1% resistors, all of which will never be in the signal path. All that goes between the headphone jack and the DAC output pins for left and right channels are short pieces of 5-nine silver in teflon wire and two Blackgate N-series caps. I tested 1000uF and 220uF with Sony MDR-V6 and Sennheiser HD 650 headphones and settled on the 220uF caps (MUCH cheaper, too). |
|
The way I wired the attenuators they swap resistors with each step that bleed signal to ground. The lowest volume is the smallest resistor, letting most of the signal go to ground, while at full volume there is no contact, meaning that 100% of the DAC output go to the headphones. I chose that setting because in my early tests, full volume was loud, but it wasn't distorted at all (given there's no "amplification" involved here). Any other attenuator config would have meant that I would lose peak volume. The headphones also need to be somewhat sensitive to provide a usable range in volume. A test with some AKG headphones showed that those just don't get loud enough to be acceptable with this DAC. |
|
Because it took a few hours to build it, here's a shot of the finished Elma 24 step stereo attenuator. Absolutely no popping when switching volume positions. Highly recommended as an affordable alternative to other big name products, especially if you need a custom solution like I did in this DAC. |
|
The other three attenuators were half the work - just 12 steps on each, but open and exposed to dust. I had to come up with a solution to keep dust and dirt out of the switch. There will be significant air flow in the DAC with the fan blowing, so something that would seal out the airflow had to be found. More on that later. |
|
It was time to build battery boxes. I'm skipping a full weekend of wood shop work here. 4 boxes were built out of about $40 of raw high grade poplar wood. A friend has a great shop with planer, large table saw, drill press, etc. but the boxes still were a bit off when they were done. 3 hours of power sanding got them to be "straight" and ready for the stain. Here is one of the finished boxes with black stain on the wood but no varnish yet. The varnishing (I stopped after 3 layers took another day). So much for "saving" by building things yourself. OK, these are absolute custom and fit the large batteries perfectly, but the time that went into building them would make these $200 enclosures in the real world... |
|
Here is the battery can be seen inside the finished box. At 7x6x3.5" these 18aH batteries just didn't fit into anything I could find off-the-shelf. For the power connector, I chose a simple RCA jack and some locking Dayton RCA connectors I had in the parts bin. There's only 12V and 0.5 amps flowing, so that should work ok. As battery cable I recycled some CAT5, 8 stands OFC in Teflon. |
|
Once the battery boxes were varnished, I had to tackle probably the most difficult part of the build: cutting the holes for the headphone jacks and the attenuators into the mesh steel face plates. Earlier I had already cut out the square holes for the power switches, but that was much easier than the round cuts I had to do at this stage. |
|
Finding the right tool for the job was the solution - a special cutting bit for my dremel solved the problem. But what a mess this thing made! Metal slivers in my hands for the next few days had me question why I had to use the mesh steel so badly. A custom plate machined to fit the hole would have done the job with much less pain, but obviously higher cost. |
|
With the holes finally cut into the front plates, a first test fitting of the jack and soon the attenuators could take place. |
|
And it really was getting messy on my work (kitchen) table. For my next project I'll get me a special table so I can mess up a different area of the house without feeling too guilty about it. If you are wondering what that big gray thing is on the floor - that's a DC 0-20V 10 amp Hewlett-Packard bench power supply - basically my test battery for all the builds, plus essential to set the battery checker trip levels for each LED color. |
|
But wait - even though I was almost done, I had to come up with a new tweak. For some unexplainable reason I tried building a different power supply for the clock chip in the DAC and the results were stunning. This little extra board with the fat red Blackgate stuck in the far left corner of the DAC is what I built and compared with the other DACs and it simply destroyed the unmodified DACs. So I had to order more parts, wait, the build the other three regulator boards and replace what needed replacing on the DAC PCB. Another week later we were able to move on. Bottom line - the cleaner the power you feed to the DAC clock, the beter it works. |
|
3 more supplies built. These three supplies use Blackgate Nx capacitors and sound even better than the original build, while they are more compact. |
|
Time to wire things together up front in the DACs. This unit has a 12-step attenuator and the smaller 220uF Blackates right on the headphone jack terminals. |
|
Here the "dust jacket" I came up with to keep the switch contacts protected from dust. Simply a piece of 2" heat shrink left over from some custom power cords I built in the past. These turned out to be a little tight on the switch in some positions, but once I put some cuts into it with a sharp knife things got more flexible and acceptable. At $45 less than the Elma switch, you have to compromise somewhere. Later I wrapped up the top row of resistors with Teflon tape (just like the entire Elma switch, just to make sure there's no metal to metal contact ever inside the box). I also used Teflon tape on the output capacitors (2.0uF Auricaps) which I wrapped in Stillpoint ERS paper, then in the Teflon tape, before holding everything together with a piece of heat shrink. |
|
Back to testing, now almost complete and with the headphone attenuator installed. |
|
Rub-on lettering to tell users what is what on the back of the unit. This stuff is difficult to apply, and even more difficult to get straight. It's easy to rub off if you don't like it. I figured I have it on there for the other users who didn't build it and need to know what plug is what. Once they learn it they can remove the lettering. |
|
The front of a nearly finished DAC, just before screwing down the top lid. |
|
Back to testing, this time with the big battery hooked up and still with the clamp holding the DAC tower. I was waiting with the epoxy to the very end, even though the DAC tower can be disconnected from the PCB. I just wanted to be absolutely certain that I didn't have to remove the tower for any reason and then be faced with the glue. |
|
Varnished battery box with finished battery cable sitting next to the rack in final testing. All four DACs were tested for at least 4 hours each before I proceeded to wrap things up. |
|
Here the inside of the DAc with the Teflon-wrapped Elma attenuator just before the final closing of the lid. ERS paper on the ICs on the boards will be complemented with a big pice covering the DAC tower. The heat sinks are all bonded to the chassis at this stage, plus an extra finned heat sink has been attached to the DAC tower facing the inside of the enclosure. The cooling is so effective, the chassis fan is really not needed, even though it's not a bad idea to force a little fresh air through the DAC to extend the life expectancy of the electrolytic capacitors that are next to the hot regulator on the main PCB. |
|
A closeup view of the heat sinks showing the epoxy at the contact areas. |
|
Now some ERS paper on top of the chips and the unit is complete. Parts cost per unit varies from $560 to $610, depending on the volume attenuator chosen. This is not a cheap DAC, but I can't wait to compare it to $2K+ units (knowing that those clearly can't drive headphones nor be used as passive preamps). |
|
|
It's DONE. The project too easily 3 times as long as initially estimated. Much of that extra time went to figuring out how to wire up the attenuator and to solve the layout issues inside the DAC. However, it was well worth it, since now the DAC not only can play through my preamp, I can also listen to headphones, or simply plug the power amps into the DAC and use the very same attenuator to control the output volume of the DAC to the amps, turning it into a "passive preamp DAC" - the volume range needs some tweaking for this use but I already have some ideas of how to deal with that in the next build that's already underway. |
|
Here's a "stack-o-DACs" - 6 channel DAC anyone? They are all slightly different, but then that's what you get with hand-built components. Each weighed in at 30 pounds including battery and all other accessories (cables, charger, fan power supply) when I shipped them to their patient new owners. |
|
Another angle of the hand-cut face plates. About as straight as you can do
it by hand, plus they'll never be seen again together :).
Stay tuned for my 5th unit to be completed. In that one I am doing things differently with many of the components, plus all the 5V supplies move off the DAC PCB and are custom built. I'll try to include a USB interface instead of the toslink as well, while using Blackgates and Auricaps throughout this time. I'll also try to calibrate the attenuator to work with my power amps better than the current builds that are calibrated for headphones. A toggle switch with a single high-grade series resistor may be the solution. |