Soundcard digital pathways?, If you skip the DAC are they 'perfect'? Options

[2Bdecided]

Oversampling DACs will reduce the effect of the anti-alias filter so much that I would doubt that there is a signficant difference in performance of the filter versus a software SRC algorithm. I really don't think you are going to gain quality by resampling in software if your DAC is well designed.

I don't disagree, but there are measureable differences.

puntloos has already gone way past "audible differences"!

Cheers,
David.

I have to make one admission here.. my DAC is the 'weakest link' in the chain, in one very relevant way: It only supports 44/48Khz. I am not quite sure but I do think it does supports higher bit depths than '16'. This is why I hoped I could change bitdepth somewhere in the m-audio controls.

Hang on a moment - you have 192kHz 24-bit DACs in the M-audio card, don't you?

Whether they sound better, worse, or the same as the DAC you're currently using is something you can test.

The M-audio cards will (often) reproduce exactly what they're sent. Therefore, if you want 24-bits, you've got to send them 24-bits (e.g. from foobar2k).

Of course, if the source is 16-bits, then it doesn't matter whether you send 16-bits, or 16-bits plus eight extra zeros (i.e. 24-bits!) - IIRC (which I might not) if you look at how SPDIF works, you'll find that it's exactly the same stream either way.

If your DAC only supports 16 bits, and you send 24, then it will just ignore the bottom 8. This is truncation without dither. If the original was more than 16-bits (or has been processed into more than 16-bits - e.g. resampling, gain change etc) then this is worse than what you started with.

Cheers,
David.

[puntloos]

Oversampling DACs will reduce the effect of the anti-alias filter so much that I would doubt that there is a signficant difference in performance of the filter versus a software SRC algorithm. I really don't think you are going to gain quality by resampling in software if your DAC is well designed.

I don't disagree, but there are measureable differences.

puntloos has already gone way past "audible differences"!

-heh- well, yeah, maybe in each individual instance I am being a bit (?) picky. Still, we are talking about a long audio chain here, and differences have a tendency to amplify eachother. My goal is to optimize each step, within certain reason. Wether or not I'm entering into insanity already is up for debate, perhaps?

I have to make one admission here.. my DAC is the 'weakest link' in the chain, in one very relevant way: It only supports 44/48Khz. I am not quite sure but I do think it does supports higher bit depths than '16'. This is why I hoped I could change bitdepth somewhere in the m-audio controls.

Hang on a moment - you have 192kHz 24-bit DACs in the M-audio card, don't you?

Yes.

Whether they sound better, worse, or the same as the DAC you're currently using is something you can test.

Well see your own point (and my response) about audible differences.. anyway well my current DAC is a 'pro studio device', which was amongst the best in its class when it was built. Wether or not a 'crappy' 192k/24b dac would still sound better than this one is debatable at least. Not saying that the m-audio one is bad, I just don't know.

Additionally, my computer is (fan noise!) in another room than my stereo. I prefer to transport the sound across those 15 metres digitally instead of the old fashion way, and DA-convert at the last possible moment.

Yes, I plan to compare the two setups at some point.

The M-audio cards will (often) reproduce exactly what they're sent. Therefore, if you want 24-bits, you've got to send them 24-bits (e.g. from foobar2k).

Of course, if the source is 16-bits, then it doesn't matter whether you send 16-bits, or 16-bits plus eight extra zeros (i.e. 24-bits!) - IIRC (which I might not) if you look at how SPDIF works, you'll find that it's exactly the same stream either way.

If your DAC only supports 16 bits, and you send 24, then it will just ignore the bottom 8. This is truncation without dither. If the original was more than 16-bits (or has been processed into more than 16-bits - e.g. resampling, gain change etc) then this is worse than what you started with.

True enough.. hence my personal current goal which is to switch up to 24bits if I intend to attenuate the signal, then dither back, and send it bitdirect when I don't attenuate. (see our replaygain discussion on why Im considering attenuating in the first place )

[Patsoe]

...anyway well my current DAC is a 'pro studio device', which was amongst the best in its class when it was built.

Getting really curious here: what DAC is it?

[puntloos]

...anyway well my current DAC is a 'pro studio device', which was amongst the best in its class when it was built.

Getting really curious here: what DAC is it?

A Philips IS 5022

[puntloos]

If your DAC only supports 16 bits, and you send 24, then it will just ignore the bottom 8. This is truncation without dither. If the original was more than 16-bits (or has been processed into more than 16-bits - e.g. resampling, gain change etc) then this is worse than what you started with.

Here's an interesting piece of info: I just looked it up, and my DAC (an IS-5022 by Philips) actually supports 20bit inputs. Would sending this DAC a 24bit signal (which then gets truncated, I assume) result in worse quality than 16bit dithered? Of course, this question is only relevant if some attenuation or gain has been applied in the 24bit domain.

This post has been edited by puntloos: Aug 23 2006, 18:23

[Patsoe]

Here's an interesting piece of info: I just looked it up, and my DAC (an IS-5022 by Philips) actually supports 20bit inputs. Would sending this DAC a 24bit signal (which then gets truncated, I assume) result in worse quality than 16bit dithered? Of course, this question is only relevant if some attenuation or gain has been applied in the 24bit domain.

My guess is that the electronic noise in the DAC will sufficiently dither the signal, since for equipment of that time noise will probably be somewhere around the 18-19th bit. Should be good.

[puntloos]

Here's an interesting piece of info: I just looked it up, and my DAC (an IS-5022 by Philips) actually supports 20bit inputs. Would sending this DAC a 24bit signal (which then gets truncated, I assume) result in worse quality than 16bit dithered? Of course, this question is only relevant if some attenuation or gain has been applied in the 24bit domain.

My guess is that the electronic noise in the DAC will sufficiently dither the signal, since for equipment of that time noise will probably be somewhere around the 18-19th bit. Should be good.

Yeah from the specs the DAC's SNR is around 103dB a-weighted.. that sounds to be around 18-19..

[2Bdecided]

Here's an interesting piece of info: I just looked it up, and my DAC (an IS-5022 by Philips) actually supports 20bit inputs. Would sending this DAC a 24bit signal (which then gets truncated, I assume) result in worse quality than 16bit dithered? Of course, this question is only relevant if some attenuation or gain has been applied in the 24bit domain.

My guess is that the electronic noise in the DAC will sufficiently dither the signal, since for equipment of that time noise will probably be somewhere around the 18-19th bit. Should be good.

Yeah from the specs the DAC's SNR is around 103dB a-weighted.. that sounds to be around 18-19..

If it's A-weighted, I'm not sure that's even 16-bits! (I could calculate - it probably is, but little more). Is that anymore than the M-audio card?

The analogue noise on the output can't dither a truncated digital signal on the input. It might hide the truncation distortion, but that's a different issue - and not guaranteed either.

Cheers,
David.

[odyssey]

Creative X-Fi cards let you select the master clock frequency, including 44.1 kHz, and also have a bit-perfect mode in "audio creation" mode.

I just tried to enable bit-matched playback on my X-fi (on x64), however i think theres a fault in the driver because I get 48khz output on every material i try to playback. I tried contacting Creative about this, but have not yet got an informative answer.

Okay there wasn't a fault in the driver, it does just not work as I expected. You can only get bitmatched output on the X-fi though the WAV ADPCM driver (whatever that is - maybe someone can enlighten me). Appearently any WAV played back sends the signal with correct frequency to my reciever - What I don't get, is that I would assume that all output (mp3's etc.) that are decoded would be a PCM signal, hence I don't get why they are not routed the same way as plain WAV files.

Does anyone know if kernel streaming or anything else does this?

[coburn_c]

On my X-fi, in "audio creation mode" it allows you to set the clock to 44.1Khz. Now I'm not sure what effect this has, but it sure doesn't change the output. The card can only output 48Khz and 96Khz, and it resamples every bit of it. This does have the plus side that analog content is sent out over stereo pcm.. but has the downside of turning 44.1KHz DTS content into.. crap.

[odyssey]

On my X-fi, in "audio creation mode" it allows you to set the clock to 44.1Khz. Now I'm not sure what effect this has, but it sure doesn't change the output. The card can only output 48Khz and 96Khz, and it resamples every bit of it. This does have the plus side that analog content is sent out over stereo pcm.. but has the downside of turning 44.1KHz DTS content into.. crap.

It's still resampled. You need to check "Bit matched playback" which greys out the field you can specify samplerate.

I figured that foobar sends this correctly and should be bit identical.

[coburn_c]

On my X-fi, in "audio creation mode" it allows you to set the clock to 44.1Khz. Now I'm not sure what effect this has, but it sure doesn't change the output. The card can only output 48Khz and 96Khz, and it resamples every bit of it. This does have the plus side that analog content is sent out over stereo pcm.. but has the downside of turning 44.1KHz DTS content into.. crap.

It's still resampled. You need to check "Bit matched playback" which greys out the field you can specify samplerate.

I figured that foobar sends this correctly and should be bit identical.

Well, for one, enabling bit matched playback does NOT grey out the samplerate field, nor does it enable bit matched playback. My DTS audio CDs still come out resampled to 48KHz and sounding like crap on a stick(terrible distortion). Creative's website even states the it only supports recording of 44.1KHz, not output.

Not that I'm shocked or anything, Creative Sucks.

Edit: OK.. with bit matched playback and the sample rate set to 44.1KHz it does indeed put out a 44.1KHz signal... after it converts it to 48KHz and back..

Edit#2: Kernel streaming works.. *clap*

This post has been edited by coburn_c: Oct 7 2006, 04:35

[CSMR]

You will not get bit perfect playback unless there is a dedicated connection between the application and the sound card; for example ASIO or kernel streaming. Otherwise the sound has to be mixed which entails volume reduction.

This post has been edited by CSMR: Oct 7 2006, 03:03

[odyssey]

On my X-fi, in "audio creation mode" it allows you to set the clock to 44.1Khz. Now I'm not sure what effect this has, but it sure doesn't change the output. The card can only output 48Khz and 96Khz, and it resamples every bit of it. This does have the plus side that analog content is sent out over stereo pcm.. but has the downside of turning 44.1KHz DTS content into.. crap.

It's still resampled. You need to check "Bit matched playback" which greys out the field you can specify samplerate.

I figured that foobar sends this correctly and should be bit identical.

Well, for one, enabling bit matched playback does NOT grey out the samplerate field, nor does it enable bit matched playback. My DTS audio CDs still come out resampled to 48KHz and sounding like crap on a stick(terrible distortion). Creative's website even states the it only supports recording of 44.1KHz, not output.

Not that I'm shocked or anything, Creative Sucks.

Edit: OK.. with bit matched playback and the sample rate set to 44.1KHz it does indeed put out a 44.1KHz signal... after it converts it to 48KHz and back..

Edit#2: Kernel streaming works.. *clap*

Bitmatching playback will not disallow other streams - Like I said, i had problems activating it with Winamp, but foobar2000 works out of the box (no kernel streaming)

You will not get bit perfect playback unless there is a dedicated connection between the application and the sound card; for example ASIO or kernel streaming. Otherwise the sound has to be mixed which entails volume reduction.

Everything sent to ADPCM (well that's what Creative said, but I thought that was a CODEC) will be bitmatched.

[CSMR]

What you have just said would imply that bitmatched playback is not bit perfect.

[odyssey]

What you have just said would imply that bitmatched playback is not bit perfect.

No

[CSMR]

Then what happens if you have a bit-perfect stream which uses the full range and another stream starts to play at the same time?

[coburn_c]

Then what happens if you have a bit-perfect stream which uses the full range and another stream starts to play at the same time?

Well, for me, it kills all playback.

P.S. whats with the freaky religious sig?

[CSMR]

(Which is that is not compatible with what odyssey said.)

This post has been edited by CSMR: Aug 26 2007, 01:53

[odyssey]

Then what happens if you have a bit-perfect stream which uses the full range and another stream starts to play at the same time?

Just checked - My X-fi/Windows seems to be stuck at 44Khz no matter what I play now in "Bit-matched Playback"-mode Even with Kernel streaming plugin.

(Which is that is not compatible with what odyssey said.)

I put the sig up at easter and haven't got round to changing it.

I did not say that you can play multiple streams with different sample rates at the same time, I said (like I'm having some trouble with now) that Bitmatched Playback does not disallow other streams like DTS/DD etc to be sent to external reciever.

[coburn_c]

Just checked - My X-fi/Windows seems to be stuck at 44Khz no matter what I play now in "Bit-matched Playback"-mode Even with Kernel streaming plugin.

Yeah, mine screws up a lot when I actually, you know.. use it. Usually a reboot fixes it.

I did not say that you can play multiple streams with different sample rates at the same time, I said (like I'm having some trouble with now) that Bitmatched Playback does not disallow other streams like DTS/DD etc to be sent to external reciever.

I'm pretty sure I can't output an ac3/DD/DTS stream and a pcm stream at the same time. And what are these 'other' streams any way? The whole point of bitmatching is to get a bit identicle output, why would it disallow DD/DTS? I needed bit matched playback to play DTS CDs.. You're very confusing...

[odyssey]

I'm pretty sure I can't output an ac3/DD/DTS stream and a pcm stream at the same time. And what are these 'other' streams any way? The whole point of bitmatching is to get a bit identicle output, why would it disallow DD/DTS? I needed bit matched playback to play DTS CDs.. You're very confusing...

Umm maybe, but without Bitmatched playback, DD/DTS are still sent to the reciever - One could wonder, why it would not be bit-identical without that setting then. Actually by "other streams" I meant streams with different sample rate, but while the stream does not break apart when playing content with different samplerates in Bit-matched playback, it makes the feature quite untrusty.

I'm on deep water now Need more testing... or maybe just better drivers.

[coburn_c]

What about this card?

[Marvin77]

Hmm, interesting.

I found this review
http://www.guru3d.com/article/sound/363/

The review compares the X-plosion to the X-Fi for Games, Movies (DVD) and Music. As far as I understood, the verdict is the following:

Games: X-Fi better because the reviewer could not get EAX to work on the X-plosion. The X-plosion was still not that bad though.

Movies: X-plosion better for DVDs since it allows encoding of Dolby Digital 5.1 and DTS on the fly (means transfer to your receiver/speakers with a single cable).

Music: the RMAA numbers of the X-Fi are clearly better. Subjective listening tests found X-Fi slightly better in half of the tests and the X-plosion slightly better in the other half. Note that the author claims that he can hear the difference between the cards and proves it by making ABX tests with Foobar2000.

This post has been edited by Marvin77: Oct 10 2006, 18:53

[coburn_c]

That review was kind of crappy.. but those cpu utilization numbers were scary. 10% to play a movie? Lost 6fps to dts encoding? I thought it was done in hardware? oh well..