Updated:
Impact of network configuration unchanged:
1x Chord Dave - Amanero - HQPlayer - fiber
1x Totaldac - XMOS1 / XMOS2 (?) - TAS - fiber
Impact of network configuration reduced or non existent:
2x Lampizator Pacific - XMOS 2 - TAS - copper
2x Lampizator Pacific - XMOS 2 - Roon - copper
1x Pilium Elektra - XMOS 2 - TAS - copper
1x MSB Select - XMOS 2 - TAS - copper
1x Aqua Formula XHD - XMOS 2 - TAS - copper
1x T+A MP3100HV - U-hear - TAS - fiber
The goal here is to establish a reliable baseline for how to extract best performance from a particular DAC model. Even more ideal would be to be able to extend this beyond the DAC, and pinpoint it to for example the USB controller used. As
@austinpop mentioned we do have some working theories. The Amanero controller for example appears to be more sensitive then second generation XMOS and U-hear controllers. There are some very obvious differences. 5-6 years ago Amanero was the only game in town for DSD512 and PCM768 capability, so it's in a good number of DACs of that era supporting a broad range of sampling rates. Second gen XMOS can now do PCM1536 and DSD1024, in theory even PCM3072 and DSD2048. Without knowing specific design details, I can identify that USB kernel mode driver latency is 20-25 times higher for the Amanero driver then the XMOS-2 (using our own driver). It also uses larger data packet sizes, 3 times larger for 16 bits, 2 times larger for 24 bits, 1,5 times higher for 32 bits. We can very likely optimize TAS to extract maximum performance from a DAC based on compensating for these differences.
The ideal scenario imho would be something like this:
1) In TAS you select the DAC you are using and we automatically optimize the software environment for optimal performance with that DAC.
2) Based on data collected from our userbase a recommendation for a specific DAC model, which could be for example:
Chord Dave:
-Replace the internal switching power supply with a Sean Jacobs DC4
-Network environment impact, significant, invest in audiophile grade switches and cables
-Software environment, upsample to 705/768 in order to bypass the internal up-sampler with higher quality algorithms, employ a small ASIO buffer size to compensate for the large kernel mode buffer
Lampizator Pacific:
-For models predating xx-xx-xxx consider updating to current spec (replace the old Amanero interface with the new JLsounds)
-Network impact, minor, inserting a switch for copper cable lengths over ~75ft recommended
-Software environment, PCM buffer size 4096, DSD buffer size 8192
MSB Select:
-Highly recommended to purchase the MSB ProUSB / MSB ProISL interface
-Network impact, minor, inserting a switch for copper cable lengths over ~75ft recommended
-Software environment, PCM buffer size 4096, DSD buffer size 8192
DCS:
-Network impact ??
-Software environment, PCM buffer size 4096, DSD set to DoP (uses PCM frames so no separate configurable buffer)
This would take the guesswork / uncertainty out of the equation.
We can have general recommendations or good practices working in every use case, like replacing the SMPS for your router and/or switch with a LPS (or to take it a bit further, any SMPS in your entire home), this does not have to cost fortunes using the model
@marty linked to a few posts back.
