RMAA Tests on Audigy 2 and on Live! Player 5.1

Thikasabrik · Jun 10, 2004, 02:53 PM

I've been doing a bit of line-out to line-in testing on my Audigy 2 and on my Live 5.1 card, so here's some results, in case anyone's interested. I don't have the breakout boxes for either of them, so I can only test the line-in/out on the backplate of the cards. I have tested the A2 with both the latest CL drivers and Kx 3537. Here are some brief results...

Live! Player 5.1 Rear-out to line-in at 16 bit, 48Khz: Excellent on all counts!

I tested with the output volume at Max, with a clean DSP (default).
The AC97 was set to get line-in input and the gain was set to 1 while the line-in volume was adjusted to give RMAA 5.3 a good level (it was close to 100). The recording panel had all inputs muted except Ac97, which was at 0.

Frequency response (from 40 Hz to 15 kHz), dB: +0.00, -0.00 Excellent
Noise level, dB (A): -101.7 Excellent
Dynamic range, dB (A): 96.7 Excellent
THD, %: 0.0013 Excellent
IMD, %: 0.0042 Excellent
Stereo crosstalk, dB: -103.7 Excellent

A completely flat freq. response is most impressive! As are the other results, of course. Who needs an Audigy 2 ZS? (apart from for 24/96..) EDIT: This result is a pile of crap - I must have reset the DSP after setting the record source, since this seems to be the result of an internal loopback. See later posts for a proper result.

The Audigy 2 was a lot harder to get right. For a start, the analog output volume with the Kx drivers seems to cause very nasty distortion above a setting of about 30. At this level, line-in needs to be turned up to +10 to get a signal at the right level for the tests. Using the gain control instead seems to also distort the signal, so for this test the gain was off.

This is about the best result I can produce from the A2 at 16/48..

Frequency response (from 40 Hz to 15 kHz), dB: +0.03, -0.08 Excellent
Noise level, dB (A): -95.6 Excellent
Dynamic range, dB (A): 94.6 Very good
THD, %: 0.0081 Very good
IMD, %: 0.010 Very good
Stereo crosstalk, dB: -91.1 Excellent

Ok, it ain't bad.. but it ain't no Live 5.1, and the output distortion is evil. The Live seems to reach similar output levels distortion free.

Now. Here is a 24/48 result for the CL drivers. The output is at about 75%, Digital Output Only is Off, line-in is at 50% and the Analog Mix record source at 100% with monitoring off. All irrelevent volumes are muted...

Frequency response (from 40 Hz to 15 kHz), dB: +0.09, -0.12 Very good
Noise level, dB (A): -93.6 Very good
Dynamic range, dB (A): 92.9 Very good
THD, %: 0.0046 Very good
IMD, %: 0.0081 Very good
Stereo crosstalk, dB: -87.5 Excellent

I tested lots of combinations with these drivers, and it appears that it pipes sound through the p16v if it is 24 bit and if DSP effects are off. It also appears to mangle 44.1 khz sound no matter what (tested with the udial sample for detecting dodgy resampling - see hydrogenaudio.org), although RMAA results were not too bad at this freq.

At 24/96, the CL drivers respond thusly (with MME - DSound would not complete at 24/96 with these drivers)...

Frequency response (from 40 Hz to 15 kHz), dB: +0.01, -0.08 Excellent
Noise level, dB (A): -92.8 Very good
Dynamic range, dB (A): 92.1 Very good
THD, %: 0.0049 Very good
IMD, %: 0.032 Good
Stereo crosstalk, dB: -89.1 Excellent

Now. I'd love to be able to properly test 24/96 on the Kx drivers, but it is very difficult. This is largely because the 24 bit data format is not properly supported by the drivers yet. 24 bit padded to 32 is required, which RMAA does not support. When set to 24/96, RMAA could detect no input on the wave HQ device. At 32 bit (int) there was a signal, but I expect that only the first 24 bits contained meaningful data. The drivers were persuaded to record line-in to Wave HQ through use of the kxctrl -sac97 2 0 command, and by fiddling with the other settings until it worked. The 32 bit result...

Frequency response (from 40 Hz to 15 kHz), dB: +0.01, -0.08 Excellent
Noise level, dB (A): -97.3 Excellent
Dynamic range, dB (A): 97.2 Excellent
THD, %: 0.0082 Very good
IMD, %: 0.023 Good
Stereo crosstalk, dB: -89.7 Excellent

Not bad, if the 32 bit data is being dithered/truncated.
I also tried 16/48 through the wave HQ device, and for whatever reason, the freq. response seems to suffer...

Frequency response (from 40 Hz to 15 kHz), dB: +0.11, -0.99 Average
Noise level, dB (A): -93.0 Very good
Dynamic range, dB (A): 90.7 Very good
THD, %: 0.0084 Very good
IMD, %: 0.011 Very good
Stereo crosstalk, dB: -91.2 Excellent

Well. After all that I think some fixing needs doing to the Kx support for Audigy 2. Without that distortion on the output (which is hopefully avoidable) the quality could probably have been even better. And the support for the p16v is, understandably, lacking. It would be nice to have support for the straight 24 bit output/input format.. could the driver not pad it itself? It would also be nice not to have to use kxctrl to get input from wave HQ for the plain Audigy 2. It's too bad Eugene is so busy these days

. Anyway, performance on the SB Live! Player 5.1 is superb - well done!

Btw, Eugene, if you're about, would you share your knowledge of the p16v chip with the ALSA driver team if they wanted it?

**dj_stick** · Jun 10, 2004, 03:16 PM

there may be an NDA about the p16v chip that prevents that

Thikasabrik · Jun 10, 2004, 04:13 PM

If there were an NDA preventing this, then Eugene must have got info on the p16v through official channels. I was under the impression that the current implementation was based on guesswork and examination of the creative open source OSS driver (or something like that). If this is so, then there doesn't seem to be any reason why the guesswork couldn't be shared, or is there?

**dj_stick** · Jun 10, 2004, 04:33 PM

if it is guesswork then no, but i read somewhere here that he signed an NDA about certain parts of the Audigy 2… this leads me to believe that the p16v chip is part of that NDA… time will tell though

Doomsayer · Jun 11, 2004, 08:11 PM

The SBLive results are unbelievably high. Are you sure you haven't messed something up? These results look like a perfect bit-to-bit transfer, which is hardly possible at all over a DAC->ADC chain.

Thikasabrik · Jun 12, 2004, 08:29 AM

Well, I don't know quite what I could have done to get a bit-to-bit transfer out of that thing.. unless it was recording the wave output directly, which I'm pretty sure it wasn't...

The sure way to test I suppose is to see if the signal gets through without the analog cable in place.. I'll try again sometime when I get a chance.

**dj_stick** · Jun 12, 2004, 09:40 AM

for perfectly accurate results, clear the DSP and only load prolog, fxbus and epilog (or 2xSRC and k1lt) and connect fxbus4/5 or 0/1 to the output on the epilog(or k1lt) and the linein input (ac97 on prolog) to the record in the epilog (k1lt)

Timme · Jun 12, 2004, 10:05 AM

the test result of SBLive! is DSP inner loop......does not go through DAC & opamp...

miguel · Jun 12, 2004, 10:15 AM

that's what it looks like, Timme

Thikasabrik · Jun 12, 2004, 11:30 AM

@dj_stick

I intend to test with the default DSP setup, because this is what I would use for recording anyway (under most circumstances), it should be perfectly possible to get a proper result with that setup if none of the effects are activated. I don't know what happened with that result, but I will look into it again tomorrow, probably.

bennetng · Jun 12, 2004, 02:32 PM

I do agree with Timme and miguel. These are obviously internal loop results. If you use RMAA 5.3 to generate a 16/48 wave file then analyse it, you can only get about -98dBA noise. RMAA can get lower than -100dBA for a 16-bit recording is because kX doesn't use dithering on the record input, the result is lower noise but higher harmonic distortion. I can demonstrate this by applying dither on the recording input:

without dither
http://www.geocities.co.jp/Hollywood...-No-Dither.htm

with dither (shaped noise)
http://www.geocities.co.jp/Hollywood.../kX-Dither.htm

DSP setup for dithering (shaped noise)
http://www.geocities.co.jp/Hollywood...Tut/Dither.png

JoshuaChang · Jun 13, 2004, 01:22 AM

several times,the newbies always forget to tune wave channel off in the record tab

**dj_stick** · Jun 13, 2004, 02:02 AM

according to Eugene, on the next release, the wave slider will be muted by default

Thikasabrik · Jun 13, 2004, 06:07 AM

Well - I suppose that tha must have been what happened - but I did check the wave volume in the record panel - honest I did! I expect I must have done something dopey, like reset the DSP after checking the record volumes..

What a shame - anyway, the A2 results are valid - I spent more time on those ones (the live is in another PC).

@JoshuaChang - I ain't no newbie, I'm just inept.

JoshuaChang · Jun 13, 2004, 07:36 PM

just use an external dac and set live digital output to 24/48,you will find that it could send perfect 16/48 signal