If I put my notes here, I might be able to find them again later!
INs and OUTs of the Asus Xonar Essence STX PCIe computer sound card
Asus Xonar Essence STX, Audacity 2.1.2, VLC, Windows 10 [and DigiOnSound6 Express for 24 bit recording]
Purpose
To confirm the calibration of the sound card input and output gain. Also, to determine the relationship between the signal voltage, the recorded signal amplitude displayed in Audacity, and the signal peak and noise baseline levels in the FFT spectra.
Summary
* Setting the volume slider of the device output to 100 gives 1 V rms output for an amplitude 0.5 sine wave.
* Setting the volume slider of the device recording line input to 100 gives records a 1 V rms tone as an amplitude 0.5 sine wave, which is displayed in the frequency spectrum (FFT) as peak of magnitude 0 dB in Audacity when both channels are averaged.
* volume setting 100 needed for unity gain loopback.
* 0.5 amplitude sine wave = 0 dB FFT = 1 V rms.
* noise baseline in averaged stereo FFT is 3 dB lower than single channel measurement.
* The measured performance of the sound card is excellent.
dB scale for voltage
"dB" is 20 * log (V_measure / V_reference)
I will be consistent and always refer the absolute dB scale to V_reference = 1 V rms (1.4 V amplitude, 2.8 V peak-to-peak).
"deebees" are easy once you get used to them. Each time the voltage is doubled the signal increases 6 dB. Each time the voltage is halved, the signal decreases by 6 dB. So +18 dB is 2x2x2=6 times increase. -12 dB is 1/2 x 1/2 1/4 or 4 times smaller. A 10 times increase adds +20 dB.
Card output, with respect to signal amplitude
The thing to pin down is the relation between the signal amplitude as viewed in Audacity and the actual output voltage from the sound card. Audacity displays a waveform on a scale from -1 to 1. The "1" is not a voltage but represents the maximum value on the digital scale.
To settle this, we make a 1 kHz test tone in Audacity with an amplitude of 1 and 0.5 and check the output with a rms-capable voltmeter (Fluke 87).
At 100 volume, the 1 kHz, amplitude 1 test tone output is ~2 V rms. This is 6 dB. An amplitude 0.5 test tone is ~1 V rms. This is ~0 dB.
So if we want a 0 dB (1 V rms) test signal, we should make an amplitude 0.5 sinewave in Audacity and set the main device volume in Windows to 100. The output is 2 V rms, 6 dB, at full scale (i.e. amplitude 1).
Incidentally the volume scale is logarithmic. The measured attenuation (dB) for a 0.5 amplitude reference is as follows,
vol L (dB) R (dB)
100 0.57 0.42
90 -1.02 -1.17
80 -2.79 -2.95
70 -4.81 -4.96
60 -7.15 -7.29
50 -9.97 -10.12
40 -13.32 -13.47
30 -17.70 -17.84
20 -23.74 -23.89
10 -34.47 -34.66
(the sound card has a 0.2 dB channel imbalance, fairly common in my experience)
The default level position of the Asus driver software is 50, or -10 dB. This is probably because the default output is expected to be the built-in headphone amplifier rather than line level output to a preamp. The measurements confirm that 100 and not 50 is the correct setting for true line level (home audio standard) 2 V rms full scale output.
Reference input
Playing a 0.5 amplitude sinewave test tone in Audacity at 100 volume gives an output of 0 dB, but to record in Audacity I'd want to play the file back in another application, VLC in my case. VLC has its own volume, with a maximum of 125. At 100 the output is -6 dB. At 125 is still isn't 0 dB. I had to open the Windows Sound Mixer and turn the volume slider for VLC to 100 to get 0 dB. This is why test and measurement on Windows is no fun.
At this point, I could record the amplitude 0.5 test tone sinewave .wav file as an amplitude 0.5 .wav file in Audacity. The input volume level was set at 100. I'm not going to mess with the input level control any further since 100 gives unity gain and decreasing the line input level would only attenuate the recorded signal.
One additional concern: Audacity only records in 16 bit on Windows, even if you set the file format to 24 or 32 bit and you have a 24 bit sound card. It's a licensing thing. So to show you what actual 24 bit looks like I also recorded the test tone playback using DigiOnSound6, then opened the file in Audacity. Either way the amplitude of the recorded data is 0.5.
Then, the FFT spectrum is calculated in Audacity. 65k samples, Hanning window. The peak at 1 kHz has an amplitude of 0 dB, confirming that 0 dB = 1 V rms = 0.5 amplitude and the FFT dB scale in Audacity is correctly calibrated so that a signal in both channels gives the dB for the signal in either channel.
To confirm, I disconnected one channel. The peak in the FFT spectrum decreases to -6 dB. Uncorrelated noise adds geometrically, however, so the same math that averages the L-R signals will reduce the uncorrelated noise by 3 dB:
two signals, adds numerically
(1 + 1)/2 = 2/2 = 1 = 0 dB
two signals, adds numerically
(1 + 0)/2 = 1/2 = 0.5 = -6 dB
two noise generators, adding geometrically
(sqrt(1^2 + 1^2)/2 = sqrt(0.5) = 0.707 = -3 dB
RMAA loopback test
As a cross check, and to get some measurements for the card's performance overall, I opened Rightmark and connected the sound card inputs and outputs together (loopback). The levels are 100 input, 100 output, and this gives the correct level indicator in Rightmark (Displays -0.8, -0.9 dB).
The noise baseline close the Audacity FFT recorded by DigiOnSound. RMAA records in 24 bit. The shape and absolute values are slightly different though. Since RMAA calculates each channel separately, uncorrelated noise would be expected to be reported 3 dB higher in RMAA than in Audacity (see above) but instead RMAA shows a lower noise baseline than Audacity. RMAA apparently uses some digital filtering which may be the cause for this, I don't know.
Rightmark gives the sound card full marks. The A-weighted SNR is -117 dB, in agreement with the published specification for the input stage (-118 dB). (The output stage is -124 dB supposedly, but of course I can't measure better than the card's input stage.)
Notes
The Xonar card has socketed, user switchable op amps for the I/V conversion and line output stages. I use TI NE5532AP in all three positions.
A general comment about noise floor measurements: use the best cables you have otherwise you'll just as likely to be measuring the noise picked up by the signal on its loop-back trip as that of the electronics itself.
Purpose
To confirm the calibration of the sound card input and output gain. Also, to determine the relationship between the signal voltage, the recorded signal amplitude displayed in Audacity, and the signal peak and noise baseline levels in the FFT spectra.
Summary
* Setting the volume slider of the device output to 100 gives 1 V rms output for an amplitude 0.5 sine wave.
* Setting the volume slider of the device recording line input to 100 gives records a 1 V rms tone as an amplitude 0.5 sine wave, which is displayed in the frequency spectrum (FFT) as peak of magnitude 0 dB in Audacity when both channels are averaged.
* volume setting 100 needed for unity gain loopback.
* 0.5 amplitude sine wave = 0 dB FFT = 1 V rms.
* noise baseline in averaged stereo FFT is 3 dB lower than single channel measurement.
* The measured performance of the sound card is excellent.
dB scale for voltage
"dB" is 20 * log (V_measure / V_reference)
I will be consistent and always refer the absolute dB scale to V_reference = 1 V rms (1.4 V amplitude, 2.8 V peak-to-peak).
"deebees" are easy once you get used to them. Each time the voltage is doubled the signal increases 6 dB. Each time the voltage is halved, the signal decreases by 6 dB. So +18 dB is 2x2x2=6 times increase. -12 dB is 1/2 x 1/2 1/4 or 4 times smaller. A 10 times increase adds +20 dB.
Card output, with respect to signal amplitude
The thing to pin down is the relation between the signal amplitude as viewed in Audacity and the actual output voltage from the sound card. Audacity displays a waveform on a scale from -1 to 1. The "1" is not a voltage but represents the maximum value on the digital scale.
To settle this, we make a 1 kHz test tone in Audacity with an amplitude of 1 and 0.5 and check the output with a rms-capable voltmeter (Fluke 87).
At 100 volume, the 1 kHz, amplitude 1 test tone output is ~2 V rms. This is 6 dB. An amplitude 0.5 test tone is ~1 V rms. This is ~0 dB.
So if we want a 0 dB (1 V rms) test signal, we should make an amplitude 0.5 sinewave in Audacity and set the main device volume in Windows to 100. The output is 2 V rms, 6 dB, at full scale (i.e. amplitude 1).
Incidentally the volume scale is logarithmic. The measured attenuation (dB) for a 0.5 amplitude reference is as follows,
vol L (dB) R (dB)
100 0.57 0.42
90 -1.02 -1.17
80 -2.79 -2.95
70 -4.81 -4.96
60 -7.15 -7.29
50 -9.97 -10.12
40 -13.32 -13.47
30 -17.70 -17.84
20 -23.74 -23.89
10 -34.47 -34.66
(the sound card has a 0.2 dB channel imbalance, fairly common in my experience)
The default level position of the Asus driver software is 50, or -10 dB. This is probably because the default output is expected to be the built-in headphone amplifier rather than line level output to a preamp. The measurements confirm that 100 and not 50 is the correct setting for true line level (home audio standard) 2 V rms full scale output.
Reference input
Playing a 0.5 amplitude sinewave test tone in Audacity at 100 volume gives an output of 0 dB, but to record in Audacity I'd want to play the file back in another application, VLC in my case. VLC has its own volume, with a maximum of 125. At 100 the output is -6 dB. At 125 is still isn't 0 dB. I had to open the Windows Sound Mixer and turn the volume slider for VLC to 100 to get 0 dB. This is why test and measurement on Windows is no fun.
At this point, I could record the amplitude 0.5 test tone sinewave .wav file as an amplitude 0.5 .wav file in Audacity. The input volume level was set at 100. I'm not going to mess with the input level control any further since 100 gives unity gain and decreasing the line input level would only attenuate the recorded signal.
One additional concern: Audacity only records in 16 bit on Windows, even if you set the file format to 24 or 32 bit and you have a 24 bit sound card. It's a licensing thing. So to show you what actual 24 bit looks like I also recorded the test tone playback using DigiOnSound6, then opened the file in Audacity. Either way the amplitude of the recorded data is 0.5.
Then, the FFT spectrum is calculated in Audacity. 65k samples, Hanning window. The peak at 1 kHz has an amplitude of 0 dB, confirming that 0 dB = 1 V rms = 0.5 amplitude and the FFT dB scale in Audacity is correctly calibrated so that a signal in both channels gives the dB for the signal in either channel.
To confirm, I disconnected one channel. The peak in the FFT spectrum decreases to -6 dB. Uncorrelated noise adds geometrically, however, so the same math that averages the L-R signals will reduce the uncorrelated noise by 3 dB:
two signals, adds numerically
(1 + 1)/2 = 2/2 = 1 = 0 dB
two signals, adds numerically
(1 + 0)/2 = 1/2 = 0.5 = -6 dB
two noise generators, adding geometrically
(sqrt(1^2 + 1^2)/2 = sqrt(0.5) = 0.707 = -3 dB
RMAA loopback test
As a cross check, and to get some measurements for the card's performance overall, I opened Rightmark and connected the sound card inputs and outputs together (loopback). The levels are 100 input, 100 output, and this gives the correct level indicator in Rightmark (Displays -0.8, -0.9 dB).
The noise baseline close the Audacity FFT recorded by DigiOnSound. RMAA records in 24 bit. The shape and absolute values are slightly different though. Since RMAA calculates each channel separately, uncorrelated noise would be expected to be reported 3 dB higher in RMAA than in Audacity (see above) but instead RMAA shows a lower noise baseline than Audacity. RMAA apparently uses some digital filtering which may be the cause for this, I don't know.
Rightmark gives the sound card full marks. The A-weighted SNR is -117 dB, in agreement with the published specification for the input stage (-118 dB). (The output stage is -124 dB supposedly, but of course I can't measure better than the card's input stage.)
Notes
The Xonar card has socketed, user switchable op amps for the I/V conversion and line output stages. I use TI NE5532AP in all three positions.
A general comment about noise floor measurements: use the best cables you have otherwise you'll just as likely to be measuring the noise picked up by the signal on its loop-back trip as that of the electronics itself.
Total Comments 0
