Here's a description of the Bottlehead DAC -
Nearly four years of design and research headed by John Swenson has brought us to the release of this very special Digital to Analog Converter, the Bottlehead DAC.
For those of you who just need the highlights, here are the bullet points-
• 32 bits, up to 384kHz USB, 192kHz S/PDIF
• USB, coax digital, TOSLINK inputs
• Approx 2V output via RCA jacks
• State of the art digital filtering
The Bottlehead DAC can handle 16,24 and 32 bit audio data at sample rates of 44.1, 48, 88.2, 96, 176.4, 192kHz, and (via USB) 352.8 and 384kHz. Sample rates through 192kHz can be handled by the USB (asynchronous UA2), coax digital and TOSLINK optical inputs (assuming your TOSLINK source is capable of transmitting 192 kHz). No switches need to be thrown to select different sample rates, the sample rate is automatically sensed by the DAC and displayed on a classic 7 segment red LED display inside the 4.5†x 3†x 7.25†(approx.) cabinet.
The DAC chip used is the Texas Instruments Burr Brown PCM5142 Delta Sigma DAC with 112dB S/N, exceptionally low out of band noise (for reduced EMI and aliasing), integrated negative charge pump line driver (which means no output coupling capacitors, ICs or tubes, and a low output impedance) and intelligent muting (no pops). On its own the PCM5142 is a remarkable sounding DAC. However John has created supporting circuitry and firmware around this DAC that take its sonic signature to an even more refined state. A Field Programmable Gate Array chip and an XMOS processor are employed to achieve this.
As John puts it,
The FPGA is the "brains" of the whole thing. It is the S/PDIF decoder, buffer controller, output format generator, adjustable oscillator interface, display controller and several others I can't think of at the moment. The XMOS processor is for the USB interface.
One of the most significant aspects of the DAC is the oversampling reconstruction filter that John has designed. We actually turn off the filter that comes in the DAC and use John’s instead. Here’s John on filters:
For those few that haven't heard my thoughts on digital filters, a quick synopsis: As far as I know all digital filters inside DAC chips use special DSP "tricks" to get really good spec sheet numbers at low cost, BUT these tricks wreak havoc with the "musicality" of the sound. This is why people like NOS DACs. BUT with a NOS you still have the aliases which cause a "dirty" sound. So what I am doing is relaxing the requirements of the filter, so it doesn't have to meet those insane numbers in the DAC spec sheets, the result is I can implement a simple filter that fits in an inexpensive FPGA.
The result is something most people have never heard, the musicality of NOS without the "dirty" sound of the aliases. It's simply stunning. Listening to your favorite music with this filter is a whole new experience. There HAS to be a filter, if you just send a 44.1 or 96 or whatever, the DAC chip will use its own. The only way to turn it off is to upsample to 352 or 384.
BTW a reconstruction filter is not evil; it is a very good thing. It's just that all the DAC chip implementations have made compromises that adversely affect the sound quality. That is what my filter is all about, doing it right so you get the good things from a proper reconstruction filter without the bad side effects of the DAC chip implementations.
As to buzz words, the filter is intermediate phase. A linear phase filter has equal amounts of pre and post ringing but fairly short amount of time in each. The minimum phase has no pre ringing, but in order to get the same over all functionality it has to have a much longer amount of post ringing. The intermediate phase splits the difference and has a small amount of pre ringing and post ringing that is a little bit longer than linear phase but much less than minimum phase. This is also deliberately a short filter. That means the amount of time it spends ringing is a very short amount of time compared to other filters. The tradeoff is that the ultimate amount of alias reduction is only about 80db, whereas those in modern DAC chips are going to 130 db or more. This is the tradeoff I mentioned, by not aiming for nearly as much ultimate image rejection I get a filter that sounds much better.
There is no dither involved; it simply is not needed in the DAC.
The overall summary is that not having an oversampling reconstruction filter is NOT the best sound. It may be better than some of the horrible sounding filters used in DAC chips, but a properly done oversampling reconstruction filter sounds way better than either no filter or the ones built into the DAC chips. The secret is to find the RIGHT filter and how to implement it without degrading the sound. I have been working on this for many years and what is in the Bottlehead DAC is the culmination of all that research and listening.â€
In order to fully realize the benefits of this filter the DAC needs to be constructed in such a manner that the benefits are not masked by power supply noise, jitter, or unnecessary buffer or gain stages. And so these areas have received special care as well.
This DAC has 6 ultra low noise regulators. It has ground plane isolation with special high quality isolators on the signals. This does a very good job minimizing the ground plane noise from the digital chips from getting into the analog side of things. This gives a marvelous very "black" background. It also significantly cuts down on the outside world affects such as source component interaction and cable affects. It doesn't completely eliminate them, but it does significantly reduce these interactions.
And we have future plans to offer a battery supply in kit form that improves the black background even further.
As for jitter:
There are two different clock "topologies" for a DAC:
Source is master, DAC is slave
Source is slave, DAC is master
What matters is the jitter at the DAC chip (in the DAC). Having a fixed frequency clock in the DAC, right next to the DAC chip is the best way to implement this. You can do this when the DAC is master; it has the "master clock". USB can do this in "asynchronous mode". The BH DAC uses asynchronous mode so it can do the DAC as master. In this mode the source (usually a computer) sends the data out, but the DAC can tell it to speed up or slow down so the average data rate matches the clock in the DAC.
There is another USB mode called adaptive, in which DAC is the slave, but the BH DAC does not use this. SOME other DACs use this mode.
The S/PDIF inputs (coax and optical) just work with the source as master and the DAC as slave. Thus the DAC has to somehow synchronize its clock to the data rate from the source. This is traditionally done with a device called a PLL, which is built in to all the S/PDIF receiver chips. PLLs have much higher jitter than a good fixed frequency clock. The BH DAC does not do it this way. It cleans up the S/PDIF signal, and sends it into an FPGA (field programmable gate array) which does the S/PDIF decoding. But the special part is a digitally controlled ultra low jitter clock. This is almost as good as the best fixed frequency clocks. The FPGA tells this clock to speed up or slow down so it is synchronized to the average data rate of the source.
The result of this is that both S/PDIF and USB produce ultra low jitter to the DAC chip. This combination of ultra low jitter from BOTH S/PDIF and USB doesn't exist in any other DAC. On other DACs one or the other will be significantly worse than the other input.
I have been listening to all three inputs (S/PDIF coax and TOSLINK and USB) and can tell hardly any difference between them. There are some very slight differences, but they are very small and hard to determine. There is only a very tiny difference in sound with different source components (different computers, OS etc, at least the ones I have on hand) I've spent a lot of work trying to decrease these sensitivities and I think I've done a pretty good job. Normally I would say that a good asynchronous USB implementation will beat a good S/PDIF implementation, BUT in this case I have spent a huge amount of effort (and I really do mean huge) making an exceptional S/PDIF input. The result is that the SQ from all the inputs should be very similar. The upshot is that if you already have an S/PDIF output from whatever, you can keep on using it with the BH DAC and know that you are not loosing anything compared to one of the other inputs.
Interestingly when we started this project with John we just assumed that the DAC would have a tube output stage. We even built two early prototypes with tube outputs that used our best technology. But the introduction of the PCM5142 into the mix radically altered our perspective. The negative charge pump line driver in the chip (nope, we don’t really understand how it works either) sounded so good that anything we added after it just colored the sound. So there is NO active output stage, the output from the DAC drives a simple final analog RC filter, which is what mostly determines the output impedance of around 500 ohms.
Another great feature of this DAC is upgradeability. There are lots of interesting ideas brewing for future upgrades:
The firmware for the FPGA is stored in a special flash memory that is soldered to the board. When new firmware is needed you simply ship your DAC to Bottlehead and we will install the upgrade for you for a modest fee.
I'm designing the USB input to handle 8 channels, but the hardware and firmware to couple multiple DAC boxes together will not be available at first, it's going to take some time to get that ready.
I HAVE built a really good DSD DAC, and have been comparing the two and I can say that the BH DAC with my custom filter sounds very similar to the really good DSD DAC. I have been coming up with a scheme to add the DSD output DAC to the aux connector, so there MIGHT be a DSD add on board later on.
So why is this not a kit?
As you may gather from the description, this is not just a DAC chip slapped on a minor variant of a developer’s board made to compete with the cheap kits on ebay. There are many complex connections between many chips, regulators and other components on a multilayer board, and firmware must be installed and tested after assembly. Thus the DAC PC board needs to be manufactured in a facility that specializes in that kind of production. That board costs many hundreds of dollars. We got high blood pressure just thinking about the first phone call from an unlucky customer who blew that expensive board due to a wiring error or a slip with the meter probes, and we decided that for everyone’s sanity it made the most sense for us to assemble these DACs in house.
There will be at least one supporting kit in the future, a battery supply. We have already tested the idea and it creates an audible improvement in sonics over the stock universal wall wart supplied with the DAC.
And then there is the Nixie tube display that we have been dreaming about. That just might make it as a kit too.
OK so how so how do you get one?
Because of the high PC board production costs, and high volume parts purchases we must make to qualify for some necessary software licenses, we find it necessary to have twenty people commit to the first production run of these DACs. The buyer will commit to a $1550 order (plus $15 for shipping in the US, $25 outside the US). We are working to have a web page up to start taking pre-orders this week.
