Xonar HDAV1.3 Deluxe: Asus HTPC sound card does Everything

Viewing page 6 of 7 pages. Previous 1 2 3 4 5 6 7 Next

TEST RESULTS

Power Consumption

Due to the dearth of sound card "loading" tools available, power consumption was measured only at idle. While we can make guesses about which tasks are most power intensive, there is no easy way of isolating those tasks from the stress they put on the other parts of the system. Spot checks during simple stereo playback did not reveal a significant boost in power consumption under load.

Asus Xonar HDAV1.3 Deluxe Estimated power consumption
System State	System Power
	AC	DC (Est.)
Base System	42W	27W
Base System + HDAV1.3	54W	38W
HDAV1.3 Only (Calculated)	12W	11W

Even at idle, the power gobbled up by the HDAV1.3 is significant. While not in the same league as power hungry graphics cards, it did eat up about 11W at idle, increasing the total power consumption of our modest test bed by about a third.

Objective Test Results

It is tempting to simply state that the acoustic characteristics of the Xonar HDAV1.3 are excellent and leave it at that. So called "objective" testing is difficult to do well, and SPCR has no existing body of results to use as a baseline. While Rightmark's Audio Analyzer (RMAA) is a useful, standardized tool, it is difficult to use correctly and interpret the results in a meaningful way.

A proper lab test using RMAA would test each input and output in isolation using reference equipment with known audio characteristics that are significantly better than the card being tested. Ideally, the test suite would include multiple reference levels and usage patterns. Unfortunately, advanced though the SPCR lab is, we are not set up to do this kind of advanced electronic testing.

So, we offer a compromise. RMAA has a test mode called external loopback testing in which the output of the card is looped back into the input via an external cable. The test suite is then run, and the results include the performance characteristics of both the input and the output sections of the card. Whichever section has the worst performance will limit the results, but, barring distortion in one section that is corrected by the other, both the input and the output should be guaranteed to have performance at least as good at the test results.

Some other problems with the test:

• Ideally, tests should be done either at the maximum output the card can provide (to maximize the possible signal-to-noise ratio and thus test all devices to their maximum potential), or at a predetermined set level (to provide a standardized test across all models — perhaps at the Redbook CD Audio "line" level of +6 dBV). The RMAA loopback test does not allow output levels to be set properly.
• The RMAA tests do not test typical usage scenarios, in which the card's mixer, DSP, and levels can all affect audio performance.

So, with those caveats, we present the results of our loopback test for 96 kHz, tested at bit depths of 16, 24, and 32 (int). Only a portion of our RMAA test data is reproduced here. Those who are interested may download our complete set of results, including every combination of sample rates and bit depths from 44.1 to 192 kHz and 16 to 32 bits. The SAV files included in the results may be opened with RMAA, downloadable from Rightmark's web site.

Frequency response is flat to ±0.05 dB from 20~20,000 Hz.

The noise floor is about -117 dB for 24 and 32 bit depths, and -100 at 16 bits.

Test @ 96 kHz	16 bit samples	24 bit samples	32 bit samples
Frequency response (from 40 Hz to 15 kHz), dB:	+0.05, -0.05	+0.05, -0.05	+0.05, -0.05
Noise level, dB (A):	-100.0	-117.1	-116.9
Dynamic range, dB (A):	100.0	117.0	116.7
THD, %:	0.0007	0.0006	0.0006
IMD + Noise, %:	0.0029	0.0009	0.0009
Stereo crosstalk, dB:	-98.3	-115.3	-114.1

The results at 24 bit / 96 kHz are representative of the card's performance overall. Tests at 44.1 kHz or with 16 bit samples showed worse performance, but this reflects on the limitations of these formats, not on the card itself. We are still scratching our head trying to figure out how our 16 bit results managed to exceed the maximum theoretical noise level for 16 bit audio (96 dB). This wikipedia article suggests that the 96 dB number is based on assumptions about the digital-to-analogue conversion, so perhaps the results are a reflection of higher quality DACs ... or perhaps they indicate a problem with RMAA's tests.

At 48 kHz or 24 bits and above, performance was essentially the same for all audio formats, suggesting that these results are the true representation of the card's abilities (most likely limited by the performance of the inputs). It also suggests that the performance (at least as far as noise is concerned) exceeds what is required to play back the vast majority of audio sources (which are 16 bit) "perfectly".

Obviously, noise is not the only concern, but the frequency response and various distortion tests were also universally excellent. Frequency response was essentially flat through the entire audible range, and THD and IMD were barely measurable.

It should be noted that our results more or less duplicate those obtained by Asus in their own RMAA tests. We know of two other sites that have done RMAA tests on the HDAV1.3, both of which produced different (and worse) results from our own. Elite Bastards examined the card here, while Atomic PC obtained results so bad as to throw their methodology into question in this review. In fairness to Atomic PC, we obtained similar results in our first attempt by routing the signal internally by recording the "Wave" input. Most likely, the poor results are a result of audio mangling within the operating system and not a reflection on the hardware, but it does illustrate that there are many factors that affect audio performance beyond the hardware itself.

Investigating Distortion

All of our RMAA tests were done with the line-in level at 100 and the output level at 40. Why 40? This was the only level we could get which was high enough for RMAA to test but low enough not to cause serious distortion at the input. This caught our attention, because we also noticed serious distortion at the outputs when doing our speaker tests during setup. So, we decided to investigate by pulling out our multimeter and actually measuring the output levels electronically. And this is what we found...

At 100 (maximum) volume, our 1 kHz test tone (recorded at -3.2 dBFS, or 3.2 dB below the maximum possible level) measured 2.26 V_RMS. That's a lot. In fact, the standard line level for RCA outputs is a mere 0.316 V_RMS. Expressed in decibels, the standard line level is -10 dBV, while the Xonar was outputting +7dBV. Adding in the potential 3.2 dB that could be gained by maximizing the test tone brings the difference between the Xonar's output and the expected line level to a whopping 20 dB. No wonder the output was clipped. Have pity on the poor pre-amps that have to accept such a hot signal.

Now, to be fair, the line level is just a guideline; many products do not meet it exactly, and all designs allow for some headroom. Unofficially, many devices follow the Redbook audio standard such that the loudest possible signal is output at +6 dBV (2.0 V_RMS), a headroom of 16 dB. At +10 dBV, the Xonar's maximum output isn't that much higher than the unofficial standard, but it was high enough to cause distortion in situations where there shouldn't have been any. We did notice that the Xonar's driver software correctly reports the peak of the graphic equalizer as +20 dB. We also verified that a 0 dB peak on the equalizer correctly output a line level signal of roughly 0.322 V_RMS. But folks, we have to warn you, blasting the HDAV1.3 at full volume into a substandard amplifier or receiver could permanently damage your equipment.

Further investigation revealed that the "40" volume setting that we used during the RMAA tests yielded an output level of 1.427V, or +3 dBV. This suggests that the line-in (which clipped above this level) is not quite up to the task of accepting a full strength signal from a CD player or any other device with peaks above +3dBV.

We also noticed that, while boosting the volume from 40 to 100 increased the output voltage by almost a volt, the level in decibels increased only by 4. This behavior is indicative of the fact that the volume control in the Xonar control panel sets the volume linearly, not logarithmically as it should. The practical effect of this is that it is extremely difficult to set low volumes reliably. For example, consumer line level (the level at which you'd be guaranteed not to clip most consumer electronics) can be set at about ~13.5 on the volume meter, but it is difficult to set with any precision.

Xonar HDAV1.3 Output Levels
	Volume Setting	Output Voltage	Output Level
Maximum Output (Calculated)	100	3.266V	+10 dBV
Maximum Output with Test Tone	100	2.260V	+7 dBV
Maximum Level of Unofficial "Redbook" Standard	100	2.000V	+6 dBV
RMAA Test Level	40	1.427V	+3 dBV
"Consumer" Line Level	~13.5	0.316V	-10 dBV

With this new knowledge in hand we can revisit the results from the RMAA test and see why they aren't as reliable as they might seem. Tested with the volume set to 100, the HDAV1.3 could potentially show a 7 dB greater dynamic range and signal to noise ratio (never mind that a recording with 124 dB of dynamic range could blow your eardrums if you listened to it). On the other hand, even if it could output the signal cleanly without distortion (and distortion increases as output goes up), its high output level is likely to cause distortion in whatever is plugged in to it.

By the same token, if it was tested at consumer line level (when the equalizer peak is 0dB), the incredible 117 dB SNR would magically drop to 104 dB. In theory, this is the level that should provide the best possible analogue performance, allowing plenty of room for transient peaks. Not only will distortion drop at lower levels, but keeping the levels closer to where other devices expect them to be should help avoid distortion further down the signal chain.

The vulnerability of RMAA's test results to variable line levels makes us speculate whether this is the reason for the HDAV1.3's high output levels. Boosting levels is an easy way to artificially inflate signal to noise ratios (it's one of the reasons why "pro" line is 12 dB higher than "consumer" line), and it would not surprise this reviewer to learn that the practice is widespread among all high-end audio cards. And, so long as the rest of the equipment in the signal chain is also designed for levels with 20 dB of headroom above "standard" line, this kind of design can offer a measurable improvement in SNR. However, we repeat our caution that high line levels can damage equipment that is not designed for this kind of signal.

Previous 1 2 3 4 5 6 7 Next

Audio|Video|Misc - Article Index

Help support this site, buy the Asus Xonar HDAV1.3 Deluxe from one of our affiliate retailers!