b0bb
Headphoneus Supremus
- Joined
- Oct 15, 2014
- Posts
- 1,663
- Likes
- 888
The elephant in the room here is the high powered switched mode regulator (VRM) driving the CPU of the computer host supplying the audio data via usb.I'm definitely tempted to try a nice linear power supply for my SU-6 at some point, but I also don't want to spend a whole pile of money on something that may not even make a discernable difference. I think I would more so do it in an attempt to eliminate all single switch mode power supplies from being either on the same circuit as my gear, or even sitting close by.
The discussion below puts in some actual numbers
I use a NUC8i3, it consumes 7W while delivering audio to the DAC.
CPU operates at 1V, this means there is 7A of pulse current flowing into ground while running.
Assuming ground plane impedance of 1 milliohm (0.001 ohm), there will be 7mV at the ground connection closest to the CPU.
This 49dB below full scale output on a typically unbalanced RCA connection of 2V max output, effective SNR here is 49dB
The level of noise at the metal shell of the USB connector on the PC is the critical one here.
Intel certifies the NUC8 as compliant with FCC Class B (Title 47, Part 15 Clause B)
https://learnemc.com/emc-regulations-and-standards
The conducted interference allowed by the FCC varies across frequency with an average of 60dBμV
This translates to 1mV noise at the USB metal shell.
1mV ground borne noise to the DAC.
SNR is 66dB a far cry from the 130dB SNR the ES9038Pro converter on my DAC is capable of.
Once the connection between the metal shells of the USB and HDMI connector on the SU6 is broken, a big chunk of the lost SNR comes back.
(The PC and DAC will require the earth gound connection aka 3pin vs 2pin power to wall socket to make both sides are properly tied to ground to prevent uwanted EMC radiated inteference and potentially to meet local electrical code requirements)
The 9038 uses a DS modulator, in simple terms it counts and sums up energy pulses.
Spurious EMI adds unwanted energy pulses affecting performance.
Meeting the FCC ClassB noise requirements is a non-trivial task requiring careful construction of the motherboard and PC enclosure.
It buys up to 17dB improvement in ground borne noise in simple endpoint duties.
This is something the average DIY PC will not be able to achieve, this leaves a lot of potential performance on the table when it comes to DS DACs.
The numbers are from a simple endpoint.
On a full blown PC running HQPlayer and GPU offloading, CPU power consumption goes up to 110W, 110A peak current vs 7A, resulting in roughly 15x or 24dB more noise, around the CPU.
(What noise levels exist at the USB connector is an open question in this DIY case as EMC testing is quite specialized and expensive)
A proper FCC certified EMC testing setup starts north of 50K USD. The good ones are above the 500K range.
(The NUC8 has wifi and bluetooth, this meant Intel had to do extra work to keep motherboard noise down in the 0-60MHz baseband and 2.4xGHz transmission range, this has the flow on benefits at the 5-30MHz range where my DS DAC operates)
Last edited: