Mivera
Major Contributor
Hi Guys,
Lately the popularity has grown with DAC's that have discrete 1 bit DSD sections. The reason for this is due to the discovery that DSD simply sounds better when it doesn't have to go through the SDM/SRC/multibit conversion process that it must go through in most chips that were designed primarily for processing PCM. Reading around on forums there seems to be quite a bit of confusion on the different ways this can be accomplished. There's no 1 cut and dry superior way to do this. There's several variables at play that can make any method superior based on the implementation and care given to the design. So don't get hung up on the method due to buzz words and marketing hype.
Now let me get into some different implementations.
Method #1, use a DAC chip that has a DSD direct bypass mode:
Although this may seem like a boring way to do it, with proper care and implementation it can be one of the best ways. The main drawback of using off the shelf DAC chips is simply because they are resource constrained. For PCM processing they must do a lot of work. This is why the best DAC's out there use FPGA's or DSP chips to handle this workload, then send the signal to the chip already filtered/modulated. However when you're simply using a chip that has a DSD direct mode as a conduit to pass the DSD through, they are extremely transparent. They also do a great job of handling the traffic flow. The best designs that use this direct DSD path, use extremely well implemented discrete low pass filters after the chip as well as ultra low jitter clocks, and superior board layouts. A great example of a DAC that uses this implementation is the Phison PD2:
Method #2: Use an FPGA:
It's possible to handle everything inside an FPGA. It just needs to be programmed with all of the filters required to filter the noise out of the DSD. Although the PS Audio Directstream DAC's and the EMM Labs DAC's use an FPGA, they aren't pure DSD DAC's like some of the other designs. This is because they have an SDM/SRC section programmed into the FPGA that up samples the DSD/PCM. So the signal passes through that section before it makes it to the DSD filtering. However it's not necessary to do it this way. A pure DSD DAC can also be made. Acko from AckoDAC has one in the works called the Directdrive DSD:
Method #3: Use completely discrete components:
This method uses no master clock and relies on a device such as an USB interface to generate a clean jitter free DSD signal. Then it's simply passed through discrete analog low pass filter components. Example's of this is the Lampizator DSD DAC's, AAVIK C-300, and Signalyst DSC-1.
Method #4: 1 bit DAC chips followed by lowpass filter:
This method is like what the Mola Mola Makua and the new T+A DAC 8 DSD uses. However the Mola Mola isn't a pure DSD DAC. It uses DSP chips to upsample everything to a 1 bit format, then uses 32 of these 1 bit chips to filter the DSD. The T+A is more of a pure DSD DAC. It works more like the Signalyst DAC, only uses 6-1 bit chips like the Mola Mola instead of 32.
So there we go. Lots of ways to accomplish the same thing. Just remember a 1 bit DSD signal is very sensitive to jitter. So the clocks used, and board layout is one of the most important things. With proper care with jitter and filtering, any of these methods has the potential to provide better DSD sound than can be had by running DSD through the SDM/SRC/Multibit conversion of a modern SDM DAC chip. However, it doesn't automatically mean it will sound better, it just means that with a very good implementation it can sound better.
Lately the popularity has grown with DAC's that have discrete 1 bit DSD sections. The reason for this is due to the discovery that DSD simply sounds better when it doesn't have to go through the SDM/SRC/multibit conversion process that it must go through in most chips that were designed primarily for processing PCM. Reading around on forums there seems to be quite a bit of confusion on the different ways this can be accomplished. There's no 1 cut and dry superior way to do this. There's several variables at play that can make any method superior based on the implementation and care given to the design. So don't get hung up on the method due to buzz words and marketing hype.
Now let me get into some different implementations.
Method #1, use a DAC chip that has a DSD direct bypass mode:
Although this may seem like a boring way to do it, with proper care and implementation it can be one of the best ways. The main drawback of using off the shelf DAC chips is simply because they are resource constrained. For PCM processing they must do a lot of work. This is why the best DAC's out there use FPGA's or DSP chips to handle this workload, then send the signal to the chip already filtered/modulated. However when you're simply using a chip that has a DSD direct mode as a conduit to pass the DSD through, they are extremely transparent. They also do a great job of handling the traffic flow. The best designs that use this direct DSD path, use extremely well implemented discrete low pass filters after the chip as well as ultra low jitter clocks, and superior board layouts. A great example of a DAC that uses this implementation is the Phison PD2:
Method #2: Use an FPGA:
It's possible to handle everything inside an FPGA. It just needs to be programmed with all of the filters required to filter the noise out of the DSD. Although the PS Audio Directstream DAC's and the EMM Labs DAC's use an FPGA, they aren't pure DSD DAC's like some of the other designs. This is because they have an SDM/SRC section programmed into the FPGA that up samples the DSD/PCM. So the signal passes through that section before it makes it to the DSD filtering. However it's not necessary to do it this way. A pure DSD DAC can also be made. Acko from AckoDAC has one in the works called the Directdrive DSD:
Method #3: Use completely discrete components:
This method uses no master clock and relies on a device such as an USB interface to generate a clean jitter free DSD signal. Then it's simply passed through discrete analog low pass filter components. Example's of this is the Lampizator DSD DAC's, AAVIK C-300, and Signalyst DSC-1.
Method #4: 1 bit DAC chips followed by lowpass filter:
This method is like what the Mola Mola Makua and the new T+A DAC 8 DSD uses. However the Mola Mola isn't a pure DSD DAC. It uses DSP chips to upsample everything to a 1 bit format, then uses 32 of these 1 bit chips to filter the DSD. The T+A is more of a pure DSD DAC. It works more like the Signalyst DAC, only uses 6-1 bit chips like the Mola Mola instead of 32.
So there we go. Lots of ways to accomplish the same thing. Just remember a 1 bit DSD signal is very sensitive to jitter. So the clocks used, and board layout is one of the most important things. With proper care with jitter and filtering, any of these methods has the potential to provide better DSD sound than can be had by running DSD through the SDM/SRC/Multibit conversion of a modern SDM DAC chip. However, it doesn't automatically mean it will sound better, it just means that with a very good implementation it can sound better.
Last edited: