dougi
Addicted to Fun and Learning
Intro
Not sure which forum to put this one in as it straddles a few. Given that most of my music is now via a PC, either UPNP using Foobar or Spotify, I thought I would attempt to work out the sample rate that gave the best performance with my digital preamp, being a Lyngdorf DPA-1. Since it has DSP for room correction and crossover, it resamples everything to 96kHz using an ASRC chip. Not sure which one, but another forum said even later Lyngdorf products use an AKM AK4122, circa 2004 so that also fits with this circa 2010 preamp.
TLDR? Best not to feed it multiples of 44.1kHz. See how windows audio resampling goes for Spotify and resample in Foobar too.
Setup
Measurement setup was PC using REW->iFI nanoIDSD as SPDIF bridge->DPA-1 digital in, XLR out to XLR in of RME ADI-2 pro ADC. Java used in REW and sample rate for generator set there as I was connecting two different devices. RME ADC using 96kHz sample rate. Note that the DPA-1 seems to give itself some headroom as -1dFS from REW, no input or output trim set in DPA-1, gives about -3dBFS from the RME ADC. RME was set to +13dBu ref level and DPA-1 balanced out is 8.8V (+21dBu) for full scale so about 10dB headroom it used with RoomPerfect bypassed. This headroom, and that REW Java is constrained to 16 bit, explains the slightly higher than expected SNR figures.
Note that for listening I normally use the digital out of the DPA-1, bypassing it's WM8740 DAC and output, using the RME as a DAC.
Measurements
REW RTA plots are presented in pairs. One for 1kHz THD and the other CCIF intermodulation for each input sample rate. They are at the bottom.
Obervations
44.1kHz multiples have lower harmonic distortion than 48kHz ones but there are other products in the THD plots and much higher IMD levels in the CCIF plots.
Conclusions
It looks like the best approach on balance is to do software resampling to 96kHz.
Questions
IMD for 44.1 kHz multiples seem quite bad. Is this real, or some REW or other measurement artifact? The AK4122 datasheet is sparse, only giving selected 1kHz THD numbers and no other measurements.
Should I try and measure Windows and Foobar software resampling and how? Record test tones of REW at different rates and use Foobar with different output settings?
44.1
48
88.1
96
176
192
Not sure which forum to put this one in as it straddles a few. Given that most of my music is now via a PC, either UPNP using Foobar or Spotify, I thought I would attempt to work out the sample rate that gave the best performance with my digital preamp, being a Lyngdorf DPA-1. Since it has DSP for room correction and crossover, it resamples everything to 96kHz using an ASRC chip. Not sure which one, but another forum said even later Lyngdorf products use an AKM AK4122, circa 2004 so that also fits with this circa 2010 preamp.
TLDR? Best not to feed it multiples of 44.1kHz. See how windows audio resampling goes for Spotify and resample in Foobar too.
Setup
Measurement setup was PC using REW->iFI nanoIDSD as SPDIF bridge->DPA-1 digital in, XLR out to XLR in of RME ADI-2 pro ADC. Java used in REW and sample rate for generator set there as I was connecting two different devices. RME ADC using 96kHz sample rate. Note that the DPA-1 seems to give itself some headroom as -1dFS from REW, no input or output trim set in DPA-1, gives about -3dBFS from the RME ADC. RME was set to +13dBu ref level and DPA-1 balanced out is 8.8V (+21dBu) for full scale so about 10dB headroom it used with RoomPerfect bypassed. This headroom, and that REW Java is constrained to 16 bit, explains the slightly higher than expected SNR figures.
Note that for listening I normally use the digital out of the DPA-1, bypassing it's WM8740 DAC and output, using the RME as a DAC.
Measurements
REW RTA plots are presented in pairs. One for 1kHz THD and the other CCIF intermodulation for each input sample rate. They are at the bottom.
Obervations
44.1kHz multiples have lower harmonic distortion than 48kHz ones but there are other products in the THD plots and much higher IMD levels in the CCIF plots.
Conclusions
It looks like the best approach on balance is to do software resampling to 96kHz.
Questions
IMD for 44.1 kHz multiples seem quite bad. Is this real, or some REW or other measurement artifact? The AK4122 datasheet is sparse, only giving selected 1kHz THD numbers and no other measurements.
Should I try and measure Windows and Foobar software resampling and how? Record test tones of REW at different rates and use Foobar with different output settings?
44.1
48
88.1
96
176
192
