i'm using the audigy sound card with the audigy2 drivers. CMSS2 enabled btw. using the creative inspire 5300 speakers! machine's an athlonxp 1800+, 512MB PC2700, etc.

the "output data format" is chosen as "16bit fixed-point" by default. any other options that i could choose which i'd benefit from?

the "dither" option is chosen "strong ATH noise shaping (recommended)" by default. any other option that'd help? i'm not clear about dithering...

what benefit would i get from resampling? if yes, what should I choose as "target sample rate" & "internal precision".

any kind of help would be appreciated!

i'd use
16bit
strong dither
resampling 48khz 32bit fast mode

The Audigy2 should sound best resampled to 88.2 KHz or perhaps 96 KHz.

Foobar does a good job of resampling, with no apparent distortion of the music. And as a result you get less noise drom the digital to analgoue conversion process (SNR for the Audigy2 is much better at 88.2 or 96, than it is at 44.1 or 48)

Can you really hera a difference between 44, 48 and 88 / 96? I thougt that resampling should help to make the sound quality better by pushing it up to the internal samplerate of the used soundcard, as i know, audigy 2 works internally with 48, so resampling to 88 or 96 is no improve to soundquality because the hardware must downsample it again to be heared. The more you don't change the sound, the more it sounds better! And in my point of view, 96 uses too much CPU time for me (1,6 Celeron, about 20% in fast mode)

I Use 16 bit output, Dither as recommended, Resampler to 48000, 32 bit perc. @ audigy 1

Let's go a little bit Offtopic (don't punish me )

I did yesterday a little "soundcheck" which quality I choose for the best, the following solutions I had:

Testfile: XXXXX - XXX XXXXX XXX.mp3 (320 kbps, HQ Lame for testing)

All Test with SBAudigy 1, Driver 324 ripped from A2
and optical out @ 48khz (Audigy standart)

Headphones plugged into a Sony MDS-JE530 DAC

WINAMP: DirectSound
FOOBAR: Kernel Streaming

1 - Windows Media Player
2 - Winamp 2.91 Directsound Frauenhoffer 16 bit, no dither, no SSRC
3 - Winamp 2.91 Directsound MAD 16 bit, no dither, no SSRC
4 - Winamp 2.91 Directsound MAD 24 bit, no dither, no SSRC
5 - Foobar 16 Bit out @ 48000 kHz resampled@32bit perc. , dither
6 - Foobar 24 Bit out @ 48000 kHz resampled@32bit perc. , dither

[*]



[*] 24 bit padded to 32 bit

End of Test, Number 2,3,4 aren't sounding good, little bit muff .
Windows Media Player sounds not bad, but a big loss of Dynamics (1)
Test Number 6 let me hear some artefacts of truncatin to 16bit, but better than all before!

Winner is: NUMBER 5! (my favorite settings)

Yes. AFAIK the Audigy2 can play at both 88.2kHz and 96KHz without doing any resampling of its own.

thanx for the help guys... i chose directsound for output, with hardware mixing enabled. that musta help.. whateva it is!

real? I read in other posts that 48 is the best for SBlive and all Audigys. Ok, i might be wrong

@ SafirXP Don't use Hardwaremixing with Audigy 2, this causes "popping" or "jumping" sound, use Softwaremixing instead! (CAN be but MUST NOT be!) An I recommend you to use Kernelstreaming as output, then you'll get the best sound performance by bypassing the windowsmixer!! (but its a little buggy, too!)

Not sure why people are recommending against resampling to higher frequencies.

Oversampling to multiples of the original material (88.2, 176.4, 352.8) is very common and used by almost all high grade CD players.

There's a lot of material on the web suggesting upsampling is a bad idea (this is resampling to numbers like 96Khx and 192KHz that aren't exact multiples). Foobar does a very good job at this though, perhaps because it uses a floating point engine (the world's first floating point resampler perhaps)

Any resampling is a bad idea: another filtering step deteriorating quality...
(a brickwall filter isn't ideal, you know)
/EDIT\Don't forget, that when using DirectSound/WaveOut it will just resample back to card's maximum./EDIT\
Foobar's resampler isn't first floating point one, but is one of the best.
But still, don't turn it on when it isn't needed

Audigy2 drivers DO support normal 24bit, so use it. Resampling is useless, turn it off,
it won't do anything good with this card. Actually, the difference isn't audible,
even if the card uses EAX effects - it uses internal resampling to 48kHz then, which sounds good,
unlike SBLive/Audigy1 one. For these cards use 48kHz resampling/16bit dithered output.

@meischder: Comparing A1 SSRC to A1 w/o SSRC is pointless. Apples. Oranges.
Try to find Peter's Winamp SSRC output plugins, then repeat the test.
The difference will be hard to notice...

I'm afraid your completely mixing up apples and pears. Over here it's custom not to post when you are familiar with the subject, or atleast it's quite common.

Sigma-delta approaches and similar does not apply. The only time you want to resample, if you'd even want to bother is when your onboard D/A has a worse antialiasfilter at given resolution than SSRC gives you, if you'd even hear the difference

so it seems

try the following

Oversampling reference

Second oversampling reference

A quote from the second of there:



Yes there are disadvantages to resampling to higher frequencies (it is signal processing), but there are also advantages (the Audigy2, Revolution, even the Lynx 2 do a better job filtering at 88.2Khz than they do at 48Khz). IMHO foobar does a fabulous job of signal processing making the disadvantages insignificant.

hehe.
I'd guide you to some more explanatory sources if I had more links at hand, I'm sure someone else does, if not I'll shout some in a jiffy, I just need to have a break
your statement is still irrelevant in sense of usage and purpose of effect.

from your sources I'm also granted to read:


edit: typo

OK so i think I see your point. The DAC in cards like the Audigy2 and Revolution will be oversampling DACs (is that right?). In which case the sound card is doing oversampling to a higher frequency regardless of input frequency (mathematically equivalent to resampling).

In this case it seems to me that Foobar2000 is doing a better job at resampling than the sound card DAC is. The whole basis of my argument is that to my ears Foobar2000 sounds unambiguously better resampled to 96KHz (or 88.2). I did extensive A-B testing and the differences were not subtle (APE files, Kernel Streaming, Revo w/th high quality power supply). As an aside, 88.2 KHz was not better when doing A-B testing with Winamp + SSRC plugin (the music in this case showed more detail and air, but lost some of its musicality, with the timbre of instruments subtely changing).

... Getting slightly of topic but still a nice read. I have gotten quite a few questions answered myself.

Thank you guys!

By doing loopback tests testing 88.2 kHz playback I must conclude that the A2 can not playback that samplerate without resampling (severe IMD). 48/96/192 samplerates seem to be the only ones natively supported.

Disclaimer: It is possible that given the hybrid nature of the A2 the fact that I'm recording and playing back at the same time might cause this behaviour.

So what is the conclusion for an A2 card?

meischder tests gives that resampling to 48KHz and using 16-bit output is best for the A1. Without any resampling, like in Winamp, it sounds worse in his test. Also, he says that the A2 works internally with 48, so resampling to higher would be bad because the card needs to resample back again. This should also mean that running without resampling should produce worse sound, as would resampling to higher than the card work with internally?

What about the 16/24 bit setting then? Is a 24-bit and resample to 48KHz a good choice for the A2?

Now the A2 does spec 24/192, but is that only for digital passthrough? I use 24/96 (5.1) for WinDVD and S/PDIF passthrough to my 24/96 surround amplifier, and it sounds very good. As the A2 does play DVD-audio, it would be strange if it only does 48KHz internally (as DVD-audio isn't allowed digital passthrough). I have listened to DVD-audio and the quality was amazingly good. I didn't have the same music on CD though, so couldn't compare.

The A2 DSP only handles 16/48 signals but there is external (to the DSP) hardware that properly handles 24/96 and 24/192. Using the Nyquist theorem it can easily be proved that the A2 is capable of >48 KS/s samplerates.

I suggest you upsample to 24/96 using Kernel Streaming. All objective sound parameters are best in this mode (24/96) for the A2 (FR, SNR,DR,IMD,THD etc). You can also use Hardware Mixed DirectSound (Windows will downsample to 16/48 if you don't enable Hardware Mixing on this card) or WaveOut in this mode (24/96).