Objective Test Session
This section is the last one in this review so that the nuances I mentioned when listening to the card could be verified objectively. I tested the card using RightMark Audio Analyzer 6.0.6 Pro. This software reproduces a set of test signals on the given audio device and analyzes the result. The usage of the program is described in the official user manual.
To test the analog section of the sound card, a line output is connected to the line input of the same or other card (as specified for each table) with a short cable with mini-jack connectors.
Testbed configuration:
- Gigabyte GA-K8NS Ultra-939 mainboard
- AMD Opteron 165 CPU
- Corsair XMS-3200XL memory
- ATI Radeon X1950 Pro graphics card
- Hitachi T7K250 hard disk
- Rosewill Turbo Series RT550-135-BK power supply (550W)
Before measuring the analog section of a sound card it is necessary to make sure that the driver reproduces the test signal correctly, without processing it in any way. I choose the Wave sound source in the sound card’s mixer and start one copy of RightMark Audio Analyzer to record the signal with the maximum precision (32 bits). Another copy of the program is reproducing test signals with a precision of 16, 24 and 32 bits through all the available output interfaces. By the way, this test is possible thanks to C-Media’s programmers who have solved the problem of incorrect processing of Wave sources that used to provoke lots of distortions.
A 16-bit/44.1kHz signal is reproduced without distortions through any of the card’s interfaces. The results for ASIO exceed the theoretical capabilities of the format. It is the consequence of an error in the Analyzer’s reproduction procedure. When played via WinAMP with the ASIO plugin, the signal coincides with the test sample.
Every interface seems to deliver the 24-bit test signal correctly, but when I examined the noise ceiling with a high-level tone, I found that the Kernel Streaming interface was the closest to the original although DirectSound and MME hardly differed from it. ASIO introduced some distortions which were not present when the signal was reproduced with WinAMP.
Even the prehistoric Wave Out copes with the reproduction of 32-bit signal if you use the latest version of the Xonar driver. DirectSound even manages to surpass the theoretical limit of the dynamic range but the measured level of noise and the crosstalk indicate that the driver performs some rounding off beyond the 24-bit precision. Perhaps it is the consequence of the conversion into floating-point format for the volume controls and resamplers.
The ASIO result is the same as with the 24-bit signal. That’s why I can make a supposition about the error with ASIO in RightMark Audio Analyzer. C-Media allows two variants of ASIO, with 16- and 24-bit precision, whereas other developers’ implementations support 16- and 32-bit precision as more convenient for the CPU. The benchmark developer must have not provided for this.
Interestingly, the Kernel Streaming is somewhat worse here than the reference interface, too. The picture below shows the spectrum of the recorded signal in the dynamic range test to illustrate the driver’s behavior:
Dynamic Range Test Spectrum
Now I will measure the quality of the analog section of each card. First, I’ll find out what card has the best line input. For this, the line output must be measured through the card’s own line input as well as through another card’s.
The ASUS Xonar and the X-Fi Elite Pro deliver similar quality, the Xonar D2 being but slightly better in terms of noise. But I had to choose a PCI slot with lowest power noise (it was the bottommost slot of the mainboard) while the Xonar D2 had the same results irrespective of the slot.
Connecting one card to another worsens the measurement results, so the following tests will be performed using the card’s own line input. First I’ll examine the front and rear channels of the two cards. To remind you, they only differ with the opamp models.
The ASUS Xonar D2’s front output only differs from the one of the Creative X-Fi Elite Pro in terms of intermodulation distortion. The same goes for the rear outputs. Of course, the numbers make the Xonar D2 the winner but I learned in my tests of the Auzen X-Fi Prelude that a better number doesn’t necessarily give a better sound.
