can ffdshow audio do this dream resample to 192000hz ?

[vola]

if ffdshow can't do that
There are other software?
with 192000hz
:goodpost::goodpost::goodpost::goodpost::goodpost::goodpost::goodpost::goodpost::goodpost::goodpost:
hi

i love ffdshow audio

This improves sound quality
For example, from 48 000 HZ
TO 96 000 HZ

http://img826.imageshack.us/img826/1584/52018169.png

Many sound cards today offer 192000HZ
Why not give the possibility to make UPSCALE TO
192000HZ with audio ?
Through ffdshow audio

ALC 892 supports- 192 000 HZ also
all asus card and more

[Gser]

Hahahahahaaa it doesn't improve quality unless the source is 192000Hz.

[vola]

Hahahahahaaa it doesn't improve quality unless the source is 192000Hz.


How do you say to me it's not good?
Have you tried 192 000 with FFDSHOW?
have you tried 96000 HZ ?

You know how to tune this at all ?

Do you have any home theater system?
Or you use headphones

[Midzuki]

If you do have an audio source @ 192kHz,
then simply DON'T USE ffdshow. :)

[TinTime]

If you've got a burning desire to resample then you could give Reclock a go.

[vola]

If you do have an audio source @ 192kHz,
then simply DON'T USE ffdshow. :)

Are you serious?
The whole idea is sound analysis on 192 000 ON LINE
It does not transform as
WAV TO MP3

This deep sound analysis

[pandy]

simply put 192000 instead 96000 - ffdshow support various sampling rates...

i don't know why You pushing for 192000 - this make no sense for me but it is up to You...

My fault - ffdshow will flag output as 192ksps but not resample the data - this mean that audio will be played twice slower - should be raised as an issue for ffdshow developers

or i was wrong with this and issue is in driver that can't work well with 192 (output from ffdshow is flagged correctly)

[Ghitulescu]

As pandy said, you won't get any quality improvement, quite the opposite (depending on your hearing and the used algorithm).
Candy camera, Crazy animals and so on won't look better just because they are sent HDTV - the same is for audio.

[ramicio]

I have ffdshow resample to 96khz, but I did this with the intention of better quality when it downmixes 5.1 stuff to stereo. I just wish it supported 192khz.

[pandy]

it can be better if Your DAC is better with 96 or 192ksps - but most of the DAC's are bit worse in fact - they are (usually) multibit Delta Sigma converters - they internally resample audio by some constant factor to achieve final conversion frequency - usually few maybe up to few tens MHz - shifting resampling from final stage of DAC to source usually means that real performance can be slightly worse - but of course everything depend from DAC - but don't expect miracles - no data means no data - You can put funny filters, even try to replicate bandwidth - still there is no good data at the input of DAC - maybe if You are "golden ear", you have source with 24 bits, extremely clean power supply after all this maybe You can hear difference but... in real life (ie PC case) You may expect more problems form various noise sources, more problems from jitter or even more problems from software than You can gain for pushing from 48 to 96 or even 192ksps.

24 bits, 48 ksps are more than in enough in normal life - for extremely situation 24 bit, 96 ksps is more than You can hear (especially if a good noise shaping is used) - 192 ksps is interesting but as a substitute to software defined radio or some lab signal generation than for normal listening from PC.

[ramicio]

I find it better to mix audio with a higher sample rate. If ffdshow was nicer one could resample a 5.1 track, mix it to stereo, and then sample it back down. This is going to turn into a damn hydrogenaudio thread i think. 24/48 is way better than CD, and I told them that, and they lashed out at me, so screw that site. Since when is it best to just have the bare minimum, what is so wrong with going better?

[Ghitulescu]

it can be better if Your DAC is better with 96 or 192ksps - but most of the DAC's are bit worse in fact - they are (usually) multibit Delta Sigma converters - they internally resample audio by some constant factor to achieve final conversion frequency - usually few maybe up to few tens MHz - shifting resampling from final stage of DAC to source usually means that real performance can be slightly worse - but of course everything depend from DAC - but don't expect miracles - no data means no data - You can put funny filters, even try to replicate bandwidth - still there is no good data at the input of DAC - maybe if You are "golden ear", you have source with 24 bits, extremely clean power supply after all this maybe You can hear difference but... in real life (ie PC case) You may expect more problems form various noise sources, more problems from jitter or even more problems from software than You can gain for pushing from 48 to 96 or even 192ksps.
Could you direct me to a site, article, spec sheet or similar that describe this negative effect? I was convinced about the opposite.
I find it better to mix audio with a higher sample rate. If ffdshow was nicer one could resample a 5.1 track, mix it to stereo, and then sample it back down. This is going to turn into a damn hydrogenaudio thread i think. 24/48 is way better than CD, and I told them that, and they lashed out at me, so screw that site. Since when is it best to just have the bare minimum, what is so wrong with going better?
One cannot have a better quality by using a better format when the, in this case, audio definition doesn't allow it? A telephone conversation won't gain anything from being converted to 192/24.
Oh, yes, when the material already lies in 192/24 or 284/48 or whatever, well, now we're talking something, but resampling it from say 48 to 192, apart from dithering noise, is useless.
Most audio HW that work internally with 24b actually take the analog input, not the digital one. If they take a digitla one, say 16b, usually they use 20b, 4+4, to avoid the clipping top and bottom during sound manipulation. Or 24b (6+6). In the end the signal is still 16b, and even if the output is 20 or 24b, the resolution of the sound is still 16b.

[ramicio]

You really think music is made entirely in 16/44.1? I doubt it. It's mixed in a higher samplerate and bit depth and THEN converted to RedBook. It is well known that higher sample rates (24/192) are beneficial for mixing sounds and music, even if each channel being mixed is not as high of quality (24/48) as the project.

[MatLz]

Don't kick my *ss if I say nonsensicalhugemistake.
Upsampling the audio samplerate can be done by copying each (or interleaving) audio frame ?
That won't give more "quality", right ?
But by interpoling, let say the avg between 2 consecutive frames....can that give better "quality" ?
Is that a kind "audio dithering" ?
I repeat, don't kick the newbie ! :D

[Midzuki]

Upsampling the audio samplerate can be done by copying each (or interleaving) audio frame ?

Do you mean, let's say, going from 48kHz to 96kHz by simply duplicating every sample? Yes, it's possible, and it's lossless too. :devil:

That won't give more "quality", right ?

Of course it will not.

But by interpoling, let say the avg between 2 consecutive frames....can that give better "quality" ?

"Better" than what ??? :)

Is that a kind "audio dithering" ?

No.

[ramicio]

For instance, if you have a wav file ripped from a CD and upconvert it to 24/192, there will be no benefit whatsoever, and probably even loss because when you dither the bits there is minor quantization noise. I use minor because you will never hear it. The point I am trying to make is, say you have a DTS-HD Master extraced to wavs, or a 5.1 DVD-A track. What I do is open the 6 mono tracks (24/48) in Goldwave, and mix them into a 24/192 file, and then downsample to 96 khz. I suppose even 24/48 after all is mixed is fine, but I have the hard drive space, so I can afford the extra room it takes up and I don't care about ABX.

[Ghitulescu]

You really think music is made entirely in 16/44.1? I doubt it. It's mixed in a higher samplerate and bit depth and THEN converted to RedBook. It is well known that higher sample rates (24/192) are beneficial for mixing sounds and music, even if each channel being mixed is not as high of quality (24/48) as the project.
The music is analogue captured. It may be analogue processed or digitally. It also might be analogue connected to digitally processing units. Finally it's the final mix that is captured/downconverted to 44.1/16 for CD distribution. Entirelly in the digital realm is possible only with Techno/house/etc. music.

Do you mean, let's say, going from 48kHz to 96kHz by simply duplicating every sample? Yes, it's possible, and it's lossless too. :devil:
"Line doubling" is extremely bad for music. Interpolation would be better.

[Mug Funky]

- mixing at a higher samplerate makes no sense. the discrete channels are all sample-aligned, so why would there be an improvement to adding them together in a higher samplerate?

- processing at a higher samplerate often has advantages when distortion is involved - the distortion is typically broadband, meaning it is spread more-or-less evenly throughout the passband. when you upsample, filter, then downsample, the portion of the distortion that is above the output nyquist is filtered out, giving up to half of the noise from distortion you'd otherwise get.

but bit-depth is what you really want if all you're doing is a fold-down from 6 channels to 2 or whatever.

the only gotcha here is if the DAC for whatever reason works better at a given sample-rate. this is usually because of cheap brickwall filters. in this case upsampling with a good digital filter and relying less on a bad analog filter in the DAC makes sense.

[Blue_MiSfit]

^Nailed it

Derek

[vola]

Why not try to build a resample to 192000hz?

WHY NOT BE Practice ?

Is this a technical problem?
Build In FFDSHOW

[MatLz]

"Better" than what ??? :)hum....well....more samples means closer to an analog wave form, right ?
So if we interleave an interpolated sample....
No ?

But I'm pretty sure 192KHz is overkillplacebo.
96KHz should be enough.

[vola]

hum....well....more samples means closer to an analog wave form, right ?
So if we interleave an interpolated sample....
No ?

But I'm pretty sure 192KHz is overkillplacebo.
96KHz should be enough.

With all due respect

Why not try?

Who knows maybe might add?

Is it a large economic investment to try?

Can be on a hit ? OR NOT



I PLAY WITH

Amplifier +7.1+AMD Processor
I hear this series and movies


http://www.fastup.co.il/v.php?file=21197462.png

http://www.fastup.co.il/v.php?file=36635222.png

http://www.fastup.co.il/v.php?file=52606778.png

http://www.fastup.co.il/v.php?file=74139141.png

http://www.fastup.co.il/v.php?file=25586837.png

http://www.fastup.co.il/v.php?file=77572.png

http://www.fastup.co.il/v.php?file=5083901.png

http://www.fastup.co.il/v.php?file=71323725.png


and Sound card 889 or 892 set to 5.1

http://img153.imageshack.us/img153/5660/50758175.png

[Groucho2004]

- mixing at a higher samplerate makes no sense. the discrete channels are all sample-aligned, so why would there be an improvement to adding them together in a higher samplerate?

- processing at a higher samplerate often has advantages when distortion is involved - the distortion is typically broadband, meaning it is spread more-or-less evenly throughout the passband. when you upsample, filter, then downsample, the portion of the distortion that is above the output nyquist is filtered out, giving up to half of the noise from distortion you'd otherwise get.

but bit-depth is what you really want if all you're doing is a fold-down from 6 channels to 2 or whatever.

the only gotcha here is if the DAC for whatever reason works better at a given sample-rate. this is usually because of cheap brickwall filters. in this case upsampling with a good digital filter and relying less on a bad analog filter in the DAC makes sense.

Finally someone with a clue who doesn't recite collected nonsense from various forums.

[Ghitulescu]

No consumer is able to obtain 96/24 sound that is really 96/24, not to mention 192/24.
The consumer gear is limited to some 80dB of dynamics (the now defunct DAT recorders), today even less. Using 24b to store what can be stored in 10b is generally overkill.
The same comes to bandwidth: 96kHz allows up to 30kHz (I know what the theory says :)), which no consumer gear is able to capture.

So why 192/24 and stuff? Because in the studios the signal is several times processed and reprocessed and transferred and so on. Both the bandwidth and the dynamic assures that the signal has enough space of manoeuvre against eg clipping or noise floor. Even in the digital realm, which is not yet THE standard. It's only after the last mixing, when the signal finally becomes a distribution format (CD, DVD, SACD, DVD-A, BD).

Besides, even the reproduction of 192/24 is more a placebo as a reality. 120 dB is the entire spectrum a human ear can feel, from the sound of the "growing grass" to the painful noise of a turbojet. Anyway, 24b should suffice to describe the auditive nature that surrounds us. The placebo consists in each of the following 3 items, separately and in combination:
listening a concert of 144 dB dynamic (the max of 24b) needs the written aproval of the townhall :)
144dB means a voltage difference of 1000000000000000000000000x from the maximal value to the lowest one. Take the SACD which is analogue driven over -10dB wires (500mV): how low must be the lowest voltage :)?
now, can anyone imagine how strong are the intrinsec noise currents compared to the smallest ones (see above)?

So no fully 24b capturing range, no 24b reproduction range. And of course an almost useless "finesse" in capturing and reproducing the 10-16000Hz hearing range.

[pandy]

Could you direct me to a site, article, spec sheet or similar that describe this negative effect? I was convinced about the opposite.


I need to search my datasheet archive - from some time IC manufacturers like Analog Devices or TI limit amount useful data in their datasheets...

but rule is quite simple - cheap DAC mean Delta Sigma - Delta Sigma is limited in bandwidth - internally they use something like oversampling from 32x up to the 256 (maybe higher) ratios - more oversampled data - better processing gain it is clear - so if
feed more data (higher sampling) i reduce oversampling and thus processing gain - restriction is quite obvious - of course new technology still push parameters up - frequency of work is higher thus gain can remain on same figure but as always there is balance between technology and results that can be achieved.


One cannot have a better quality by using a better format when the, in this case, audio definition doesn't allow it? A telephone conversation won't gain anything from being converted to 192/24.
Oh, yes, when the material already lies in 192/24 or 284/48 or whatever, well, now we're talking something, but resampling it from say 48 to 192, apart from dithering noise, is useless.
Most audio HW that work internally with 24b actually take the analog input, not the digital one. If they take a digitla one, say 16b, usually they use 20b, 4+4, to avoid the clipping top and bottom during sound manipulation. Or 24b (6+6). In the end the signal is still 16b, and even if the output is 20 or 24b, the resolution of the sound is still 16b.

Yes, so there is no sense to push lossy compressed and limited in bandwidth signal by upsampler only to artificially increase sampling ratio.

[pandy]

You really think music is made entirely in 16/44.1? I doubt it. It's mixed in a higher samplerate and bit depth and THEN converted to RedBook. It is well known that higher sample rates (24/192) are beneficial for mixing sounds and music, even if each channel being mixed is not as high of quality (24/48) as the project.

but of course but during conversion from such high sampling always low pass filtering and noise shaping is used - after such processing there is no data over stop band of the filter.

[pandy]

No consumer is able to obtain 96/24 sound that is really 96/24, not to mention 192/24.
The consumer gear is limited to some 80dB of dynamics (the now defunct DAT recorders), today even less. Using 24b to store what can be stored in 10b is generally overkill.
The same comes to bandwidth: 96kHz allows up to 30kHz (I know what the theory says :)), which no consumer gear is able to capture.


10bits is only 60dB dynamics, for 80dB 14 bits is required
In real life natural spectrum of many instruments quickly going dow - energy over 25khz can be really low and unhearable.


So why 192/24 and stuff? Because in the studios the signal is several times processed and reprocessed and transferred and so on. Both the bandwidth and the dynamic assures that the signal has enough space of manoeuvre against eg clipping or noise floor. Even in the digital realm, which is not yet THE standard. It's only after the last mixing, when the signal finally becomes a distribution format (CD, DVD, SACD, DVD-A, BD).

Besides, even the reproduction of 192/24 is more a placebo as a reality. 120 dB is the entire spectrum a human ear can feel, from the sound of the "growing grass" to the painful noise of a turbojet. Anyway, 24b should suffice to describe the auditive nature that surrounds us. The placebo consists in each of the following 3 items, separately and in combination:
listening a concert of 144 dB dynamic (the max of 24b) needs the written aproval of the townhall :)
144dB means a voltage difference of 1000000000000000000000000x from the maximal value to the lowest one. Take the SACD which is analogue driven over -10dB wires (500mV): how low must be the lowest voltage :)?
now, can anyone imagine how strong are the intrinsec noise currents compared to the smallest ones (see above)?

So no fully 24b capturing range, no 24b reproduction range. And of course an almost useless "finesse" in capturing and reproducing the 10-16000Hz hearing range.

Even if usual max audio level is around 2.5VRMS this not change anything, in many cases it is very difficult to create something better than 128dB unless we not go to cryogenics electronics - everything what have temperature higher than 0 K make noise - resistor connected to sensitive AC voltmeter (highly hypothetical) will create some small voltage related to its temperature - it is noise. this is valid for all conductors, joints, soldering point, each contact produce noise... noise is everywhere

https://secure.wikimedia.org/wikipedia/en/wiki/Johnson%E2%80%93Nyquist_noise

48ksps/24bit is more than sufficient in normal life, 96ksps/24bits it way beyond human ear abilities, 192ksps is good for measurements lab ;)

[Ghitulescu]

10bits is only 60dB dynamics, for 80dB 14 bits is requiredI know :), the technique allows 16b but the sources are not :)
Even if usual max audio level is around 2.5VRMS this not change anything, in many cases it is very difficult to create something better than 128dB unless we not go to cryogenics electronics - everything what have temperature higher than 0 K make noise - resistor connected to sensitive AC voltmeter (highly hypothetical) will create some small voltage related to its temperature - it is noise. this is valid for all conductors, joints, soldering point, each contact produce noise... noise is everywhere
Exactly, even with +4dB connections (studio standard) 128dB are hardly achievable (capturing and transporting).48ksps/24bit is more than sufficient in normal life, 96ksps/24bits it way beyond human ear abilities, 192ksps is good for measurements lab ;)
96/24 is probably the best compromise between achievable quality and space requirements.

[pandy]

I know :), the technique allows 16b but the sources are not :)

Exactly, even with +4dB connections (studio standard) 128dB are hardly achievable (capturing and transporting).
96/24 is probably the best compromise between achievable quality and space requirements.

Accordingly to the equation form https://secure.wikimedia.org/wikipedia/en/wiki/Johnson%E2%80%93Nyquist_noise it is impossible to achieve 128dB in 30 - 40kHz bandwidth for nominal 600 ohm studio at the room temperature.

128dB should be possible (barely) with low impedance equipment for 20kHz bandwidth

[Midzuki]

"Line doubling" is extremely bad for music. Interpolation would be better.
:confused: :confused: :confused:

Would you mind explaining why is it so? :)

I've already done that (manual audio upsampling by sample duplication), and my "non-audiophile" auditory cortex didn't detect any "critical injury" :p in the final result. :o

[pandy]

http://www.dsplog.com/2007/03/25/zero-order-hold-and-first-order-hold-based-interpolation/

there are two ways for "Line doubling" and many ways for interpolation...

[Midzuki]

Thanks for the URL.

Honestly, I didn't know that a lossless transformation could ever be worse than a lossy one. :eek:

[CpT]

This...
- mixing at a higher samplerate makes no sense. the discrete channels are all sample-aligned, so why would there be an improvement to adding them together in a higher samplerate?



and Sound card 889 or 892 set to 5.1
http://img153.imageshack.us/img153/5660/50758175.png

After all this you're running onboard sound?

[ramicio]

I can see I'm just banging my head against the wall here trying to make any sense to anyone else about mixing waveforms. High frequencies suffer when mixed together at a low sample rate. If you take a 16/44.1 and make it 24/192 you will have a more coherent waveform and when mixed you will have a more coherent waveform, then you can downsample. No one's nerdy BS talk is going to change my mind on this. Science is not absolute and I see how things go for waveforms with my own eyes. My ears are another thing and I don't care if I can barely tell a difference, I like to have the best quality one can get. I don't see why people need to bring DACs and dB into discussion here. Bits always turns into a damn dB discussion and sample rate turns into a damn hz discussion. So sick of it. I say it only applies to a simple sine waveform. I think about more resolution with either bits or sample rate. You start mixing sounds together at low quality and things will suffer. I get why the OP wants this feature, and I wish it was there too. I wish one could stack filters and use them twice so I could resample, downmix, and resample.

[Ghitulescu]

@ramicio

You're right. However, bandwidth is Hz and bits are dB. Nobody can change this. And, most important, all are numbers.
What you say makes sense in the analogue world, like sinewave and stuff.

Any processing in the digital world (numbers) needs a lot of algorithms, none is perfect.
The very same way one cannot let the PC judge the artistic quality of a painting or a movie, one cannot let the PC judge the music quality. A man can go beyond the numbers and see the logic behind (abstractization) while a PC does only implement pre-created algorithms on data. Same data can be seen as a JPEG or as a WAV or as an AVI. Remember the old CD-players that played noise from Data-CDs? For you it's noise, for them just data that needs to be played. This is the whole point.

[Ghitulescu]

Not very long time ago, some digital mixers worked internally with 20b and output to DATs (for CD factories) only in 16b. How to fit 20b into 16b? Simply, thought the engineers of that time: since the master fader summed all the useful signal in the "top bits", the lower ones are simply floor noise or unused (zeros) data. So they cut the lower 4 bits (in PC language they simply shifted the data 4 positions downwards).
The result was not always acceptable, today most downconverters use noise shaping based dithering algorithms, as they yield the best (for now) results. Which algorithm to use, well that's another issue.

What is importance of this example? Well, the main fader was not graded into bits or Hz, but in dB. A sound engineer uses dB. The whole chain is measured in dB, the specs are in dB. Everything that surrounds an engineer is in dB. Because this is how our senses work, not because someone thought that the job of an A/V technician is so easy that it needs to be complicated with mathematical formulas.

[pandy]

I can see I'm just banging my head against the wall here trying to make any sense to anyone else about mixing waveforms. High frequencies suffer when mixed together at a low sample rate. If you take a 16/44.1 and make it 24/192 you will have a more coherent waveform and when mixed you will have a more coherent waveform, then you can downsample.

this is simple not true...

if You add two samples 44.1/16 together, You will receive 44.1/17 sample - they are completely coherent in time domain and complete coherent from quantization point of view as coherent adding of the 2 bits together can be - there is no magic in adding two 16 bit digits - they give You perfect and fully coherent 17 bit value. What is worst as 192 is not dividable by 44.1 You will receive "incoherent" to source samples after resampling.
High frequencies not suffer at all...


No one's nerdy BS talk is going to change my mind on this. Science is not absolute and I see how things go for waveforms with my own eyes. My ears are another thing and I don't care if I can barely tell a difference, I like to have the best quality one can get. I don't see why people need to bring DACs and dB into discussion here. Bits always turns into a damn dB discussion and sample rate turns into a damn hz discussion. So sick of it. I say it only applies to a simple sine waveform. I think about more resolution with either bits or sample rate. You start mixing sounds together at low quality and things will suffer. I get why the OP wants this feature, and I wish it was there too. I wish one could stack filters and use them twice so I could resample, downmix, and resample.

Do some experiment - 2+3=5 - multiply both by 10 i.e. 20+30=50
divide by 10 and You will receive 5 - nothing nerdy here...
this is simple not even math but arithmetic - definitely You over estimate binary arithmetic if You believe that shifting bits from right to left and add them together can give You better quality.

And believe me - if something apply to the sine waveform it also apply to series of sine (or cosine) waveforms.

If You insist to upsample Your source take this good advice and made this not on 192ksps from 44.1 but on 176.4 or 352.8ksps - after decimation to 44.1 You will receive more "coherent" samples than after up-conversion from 44.1 up to 192ksps.
Why? - search for differences between synchronous and asynchronous sample rate conversion.

btw search also for "golden ears" debate about non-oversampled DAC's in audio ;)

"Non oversampling DACs are by many people (including me) regarded as better sounding than conventional oversampling systems."

http://www.justblair.co.uk/creating-a-diy-non-oversampling-dac-with-pcm1704.html


http://www.sakurasystems.com/articles/Kusunoki.html (http://www.sakurasystems.com/articles/Kusunoki.html)

[vola]

This...




After all this you're running onboard sound?
What a good sound card should be purchased?

I hear only through SPIDF

[nurbs]

@pandy
If you want to understand where ramicios concerns stem from look here (http://www.hydrogenaudio.org/forums/index.php?showtopic=84208&view=findpost&p=726646). Basically his experimentation has lead him to believe that the Nyquist-Shannon sampling theorem is wrong if there is more than one frequency present in an audio file, therefore he oversamples so that at least the downmix isn't affected by that perceived problem.

[ramicio]

If you think you can digitize an analog signal and have it come out exactly the same as it was before it was recorded you are insane. If you think only frequencies that you can audibly hear are the only ones that make a sound sound realistic you are insane. I prefer to have an analog signal as unmolested as possible, so I use high sample rates and a 24 bit depth. Sorry I do what I do, and I feel I should be able to get music is something higher than Redbook. 24/48 would suffice, and no loudness. Do you really think if they would have had say the DVD medium back when CD was being invented that they would have chosen 16/44.1? I think they chose what they chose to be DECENT enough for the storage medium. The other reason being recording off PAL tapes and the sample rate matching BS. I believe if they had more storage they would have chosen better quality. You all argue dynamic range and peak frequency, I argue resolution.

[CpT]

What a good sound card should be purchased?
I hear only through SPIDF

I've used the "Plus" version of this one - no complaints.
http://www.newegg.com/Product/Product.aspx?Item=N82E16829271002

Currently I run m-audio and luv it. But tbh m-audio is a bit overkill for a media center pc.

->Creative sux...

[pandy]

@pandy
If you want to understand where ramicios concerns stem from look here (http://www.hydrogenaudio.org/forums/index.php?showtopic=84208&view=findpost&p=726646). Basically his experimentation has lead him to believe that the Nyquist-Shannon sampling theorem is wrong if there is more than one frequency present in an audio file, therefore he oversamples so that at least the downmix isn't affected by that perceived problem.

hmmmmm i see the source of his faith - maybe this help a bit http://src.infinitewave.ca/

[Groucho2004]

"Non oversampling DACs are by many people (including me) regarded as better sounding than conventional oversampling systems."

Same here. I have a Audio Note DAC1 fed by the SPDIF OUT of a LynxOne card. Sounds great!

[vola]

I've used the "Plus" version of this one - no complaints.
http://www.newegg.com/Product/Product.aspx?Item=N82E16829271002

Currently I run m-audio and luv it. But tbh m-audio is a bit overkill for a media center pc.

->Creative sux...
Thank you
What will give me this card through SPIDF hearing???

How it will manifest itself ???

the card from 2007
not support hd ONLY (AC97)
not support win 7 64 ( Out of a bad experience with ASUS)

Looking for something new

[pandy]

What with older Creative cards with good driver like KX Drivers?

http://kxproject.lugosoft.com/

IMO Creative make quite good audio cards.

[Ghitulescu]

@pandy
If you want to understand where ramicios concerns stem from look here (http://www.hydrogenaudio.org/forums/index.php?showtopic=84208&view=findpost&p=726646). Basically his experimentation has lead him to believe that the Nyquist-Shannon sampling theorem is wrong if there is more than one frequency present in an audio file, therefore he oversamples so that at least the downmix isn't affected by that perceived problem.

Nobody here said ramicio is wrong in his sayings. The problem is that 2 aspects are discussed here: the "feeling" (which is human only) and the numerical representation (which is computational). Music is not dB and Hz and bits, but since this is how the digitised music is all about, we have to restrict ourselves at what we had: bits, Hz and dB.

The only way both aspects can be unified is to use the same parameters from acquisition to reproduction. Any resampling & Co would inherently involve one of the lossy algorithms, so criticised.

In short, upsampling 44k1/48 to 192 or 16b to 24b would only make the sound worse, because of the imperfection of algorithms. Should the source already be present in 192/24 then the manipulation algorithms would work better than on the same source but in 48/16.

[nurbs]

How is he not wrong then? There is stuff that helps with what he does, like doing the downmix at a higher bitdepth, and stuff that doesn't help, like upsampling before downmixing. He does only the latter, because that's what makes him feel better. I get the implications of having a higher quality source, but he doesn't have one in this case.

[Groucho2004]

In short, upsampling 44k1/48 to 192 or 16b to 24b would only make the sound worse, because of the imperfection of algorithms.

Generally, you're probably right. However, there are exceptions like the DACs from Wadia and the DCS upsamplers which employ very sophisticated interpolation algorithms. I have heard a top of the line Wadia DAC which is very impressive.

Still, the only reason I can think of to go this route is to shift the sampling frequency to much higher values (192 KHz or even 384 KHz) so only a simple low order anti-aliasing filter is needed.

My personal choice is however a non-oversampling DAC which sounds very natural to me.
The Audio Note DACs only have a simple passive 3rd order low pass after the DAC and you can certainly see the "steps" when you put a 10 KHz signal on the oscilloscope.
Most electronics engineers would tell you that this concept would lead to horrible sound and one would need an extremely high order low pass to filter out the digital noise. And yet, these DACs sound fantastic which would imply that the human ear appears not to be that sensitive to the digital noise.

[pandy]

"very sophisticated interpolation algorithms" which is always only interpolation - it can be good for some type of signal and wrong for other - not optimal from general point of view... i like Wadia design, i like philosophy behind Wadia but interpolation is only interpolation - filling the gaps between real data samples by "fake" ones - how good are "fake" data - maybe it shall be completely different set of algorithms or some adaptive algorithm, self tuning, different for Miles Davies or Pink Floyd or Beethoven - who knows?

issue with "non-oversampling DAC" is always issue related to the clock recovery - but what in case when we don't recover clock, when we use one master clock for whole system? when all processing is properly synchronized - never ending disputes of "golden ears" - "vinyls are better than CD" - ok, we can assume that this is ok but how to explain that source for vinyl and CD is exactly the same - digital one? on some point such debate turns to faith, believes and religion - there is no place for science on this area.

[Groucho2004]

"very sophisticated interpolation algorithms" which is always only interpolation - it can be good for some type of signal and wrong for other - not optimal from general point of view... i like Wadia design, i like philosophy behind Wadia but interpolation is only interpolation - filling the gaps between real data samples by "fake" ones - how good are "fake" data - maybe it shall be completely different set of algorithms or some adaptive algorithm, self tuning, different for Miles Davies or Pink Floyd or Beethoven - who knows?

issue with "non-oversampling DAC" is always issue related to the clock recovery - but what in case when we don't recover clock, when we use one master clock for whole system? when all processing is properly synchronized - never ending disputes of "golden ears" - "vinyls are better than CD" - ok, we can assume that this is ok but how to explain that source for vinyl and CD is exactly the same - digital one? on some point such debate turns to faith, believes and religion - there is no place for science on this area.

All good points and I agree that filling in missing samples is better left to our brain than to electronics. But that's just my opinion.

I have several albums on CD and also as vinyl and it's not always the vinyl that sounds better.

I believe that nowadays the most damage is done in the mastering, or even worse the "re-mastering" process. The trend is to make music sound louder which mostly means that dynamic compression is applied and the signal is driven into clipping. Not difficult to imagine how that sounds.