Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Oversampling Better? (Read 12399 times) previous topic - next topic
0 Members and 3 Guests are viewing this topic.

Oversampling Better?

I have a emagic emi 2|6 DAC (usb, 24/96).

I have tried mad 0.14 2b for 24 bit decoding and I have put the DAC to 96khz, in order to force an oversampling at 96 khz.

The sound seems really better for me.

Is it possible or I'm dreaming?

Oversampling Better?

Reply #1 –
Oversampling adds nothing to the sound, if properly done. If not properly done, it can degrade the sound. So, unless your DAC works better at 96 KHz than at other sample rates, audible improvements are not due to the actual sound played, but other things.

Oversampling Better?

Reply #2 –
With 96, the high frequencies should be more natural because you have more points to define it.

With a sweep sound, i can hear higher with 96 and at the end of the sweep in can hear a noise (descending sweep) with 44.1.

In the other hand I have more continuous noise with 96, maybe it's because of an higher frequencie filter in 96, in this case you'r right, it's linked to my DAC.

But what about the theoric precison in high frequencies?

Oversampling Better?

Reply #3 –
Quote
With 96, the high frequencies should be more natural because you have more points to define it.

Well, that's not right, theorically, according to Fourier (or was it Nyquist? I don't remember well...) sampling theorem, there should be no difference, and in fact there's no difference, in good quality equipment and using good quality oversampling (resampling)...

Quote
With a sweep sound, i can hear higher with 96 and at the end of the sweep in can hear a noise (descending sweep) with 44.1.


...but it seems that your DAC performance at 44.1 KHz is not as good as at 96 KHz in this respect.

Oversampling Better?

Reply #4 –
The points that define the sound are in the original file, often 44100 Hz. Any converter, exept some rare crazy-audiophile prototypes, resamples the signal for the conversion. often 64 or 256 times, so that the "points that define the sound" are usually at 64x44100=2,822,400 Hz or 256x44100=11,289,600 Hz.

But they are just artificial points computed from the original 44,100 ones, they don't add any definition to the sound that was not in the original.

Also, if there are descending tones in the 44100 Hz sweep, then the converter is not running at 44,100 Hz ! Descending tones are an alias (an artifact) caused by a bad resampling.

Edit : important points : where does the sweep come from, and how is it played (program/playback settings) ?

Oversampling Better?

Reply #5 –
Quote
Also, if there are descending tones in the 44100 Hz sweep, then the converter is not running at 44,100 Hz ! Descending tones are an alias (an artifact) caused by a bad resampling.


No, the descending sweep is without resampling (44.1->44.1). With 96 (resample), i don't hear it...

Quote
Edit : important points : where does the sweep come from, and how is it played (program/playback settings) ?


http://ff123.net/samples/sweep.zip

WinXP
Winamp 2.81
Nullsoft Waveform Decorder v 2.06
Directsound 2.2.6

Oversampling Better?

Reply #6 –
i just played the sweep and found that the only way that i can render it correctly is if the sample rate in ssrc is set to 48000 Hz or greater in winamp 2.81.

Oversampling Better?

Reply #7 –
Quote
Well, that's not right, theorically, according to Fourier (or was it Nyquist? I don't remember well...) sampling theorem, there should be no difference

it's the Nyquist-Shannon theorem. when you sample a signal that has been band-limited to fs/2 hz with fs samples per second, you are storing all frequencies up to fs/2 with 100% accuracy.
A riddle is a short sword attached to the next 2000 years.

Oversampling Better?

Reply #8 –
Quote
No, the descending sweep is without resampling (44.1->44.1). With 96 (resample), i don't hear it...

Uncanny...

No problem at 96 kHz, OK...

But an alias at 44.1... The Eamgic 2/6 doesn't look like the kind of soundcard that resamples... would the master clock be on "external" ?... no, wait, it must be Windows DirectSound ! Select Winamp WaveOut instead of DirectSound, and in the WaveOut configuration, don't select the wave mapper as device, but your Emagic itself... This should solve the problem... it would be very interesting to know if DirectSound affects the sound like this.

Oversampling Better?

Reply #9 –
uhm, i *highly doubt* if selecting "wave mapper"/"primary sound driver"/whatever instead of the device itself does anything, "wave mapper" is provided just to redirect software that doesn't have user interface to select device to the "preferred" device selected in control panel.
i think you people should try toggling "allow hardware acceleration" on device tab in latest out_ds, it seems to affect a lot of weird things; when hardware acceleration is disabled, software mixing is used instead of hardware (which is default for most of directsound apps out there); software mixing seems to "emulate" waveOut (at least on win2k/xp).
Microsoft Windows: We can't script here, this is bat country.

Oversampling Better?

Reply #10 –
Quote
Quote
Well, that's not right, theorically, according to Fourier (or was it Nyquist? I don't remember well...) sampling theorem, there should be no difference

it's the Nyquist-Shannon theorem. when you sample a signal that has been band-limited to fs/2 hz with fs samples per second, you are storing all frequencies up to fs/2 with 100% accuracy.

that can't be right... if you offset by some phase (0 < phi < pi/2) the amplitude decreases. for instance, if you sample a wave at fs/2 that's pi/4 out of phase, you have zero amplitude.

Oversampling Better?

Reply #11 –
Quote
But an alias at 44.1... The Eamgic 2/6 doesn't look like the kind of soundcard that resamples... would the master clock be on "external" ?...


Internal clock

Quote
no, wait, it must be Windows DirectSound ! Select Winamp WaveOut instead of DirectSound, and in the WaveOut configuration, don't select the wave mapper as device, but your Emagic itself... This should solve the problem... it would be very interesting to know if DirectSound affects the sound like this.


DirectSound or WaveOut, no matter...

I have try 44.1 and 96 on several tracks and I really hear the difference...

Oversampling Better?

Reply #12 –
Quote
Quote
when you sample a signal that has been band-limited to fs/2 hz with fs samples per second, you are storing all frequencies up to fs/2 with 100% accuracy.

that can't be right... if you offset by some phase (0 < phi < pi/2) the amplitude decreases. for instance, if you sample a wave at fs/2 that's pi/4 out of phase, you have zero amplitude.

The amplitude of the samples decrease, yes, but the samples are not a wave. They still contain all the information from the original wave, including its original amplitude.
It can be seen in CoolEdit : phase-shift a high frequency wave so that the amplitude of the sample decrease. Cool Edit will still draw the original full amplitude wave between the decreased samples. It is mathematically reconstructed. I can post a picture when I'm back home.

You example at fs/2 is right. It is not possible to record fs/2. Only frequencies strictly inferior to fs/2.

Oversampling Better?

Reply #13 –
Maybe it's not the frequency itself but how to reproduce it?

Oversampling Better?

Reply #14 –
Quote
You example at fs/2 is right. It is not possible to record fs/2. Only frequencies strictly inferior to fs/2.

No, it's *sometimes* possible to get fs/2 depending on the phase - a signal alternating between +1 and -1 is definitely fs/2 - and if you amplitude modulate any signal with this +1/-1 alternating signal, you "invert" the spectrum as you'd expect (so signals at 0.1fs get shifted to 0.4fs, etc.).

Oversampling Better?

Reply #15 –
Interesting thread, I wanna add something 

I have a DMX 6fire which is capable of 24/96. When I set SSRC to resample from 44.1kHz 16bit to 96kHz 24bit the soundcard driver automatically changes its internal master clock to the resampled frequency and the sound becomes slightly higher or the highs are a little bit more present than the lows. I'm rather sure that the sound changes but I don't know why. Any explanation?

Oversampling Better?

Reply #16 –
Quote
Quote
But an alias at 44.1... The Eamgic 2/6 doesn't look like the kind of soundcard that resamples... would the master clock be on "external" ?...


Internal clock

Quote
no, wait, it must be Windows DirectSound ! Select Winamp WaveOut instead of DirectSound, and in the WaveOut configuration, don't select the wave mapper as device, but your Emagic itself... This should solve the problem... it would be very interesting to know if DirectSound affects the sound like this.


DirectSound or WaveOut, no matter...

I have try 44.1 and 96 on several tracks and I really hear the difference...

marcan: how did you set the internal clock? 44.1 or 96? and do you resample via ssrc to 96 or not?

Oversampling Better?

Reply #17 –
Quote
how did you set the internal clock? 44.1 or 96? and do you resample via ssrc to 96 or not?


I can change the clock with the Emagic Device Options.

I have tried with out_ds and out_ds_ssrc. With out_ds, the oversampling is done by Windows XP (with High Performance Option). With out_ds_ssrc, the oversampling is done by out_ds_ssrc. Personnaly, I prefer with out_ds.

Oversampling Better?

Reply #18 –
Quote
No, it's *sometimes* possible to get fs/2 depending on the phase - a signal alternating between +1 and -1 is definitely fs/2

No, the longer the signal alternates between -1 and +1, the closer to fs/2 it is.
To be at fs/2 exactly, it must alternate for an infinite amount of time. This is purely mathematical, though.
In practice, there is no point to distinguish between 22,050 Hz and 22,049.9999.

Oversampling Better?

Reply #19 –
On the Audiophile 24/96, I can measure an improvement when upsampling from 44.1kHz to 96kHz (using Cool Edit Pro) - but I'm not convinced I can hear a difference - maybe - but I've not done a blind test yet!

For example, if you play a 21kHz tone sampled at 44.1kHz, you get a quiet tone output at 44.1-21=23.1kHz (the first alias). If you resample to 96kHz (well, 88.2 is more sensible) first, then the higher quality resampling in CEP removes this alias completely. It gives poorer time-domain resolution though. Whatever - I can't hear this high!

Cheers,
David.

Oversampling Better?

Reply #20 –
Quote
It gives poorer time-domain resolution though.

What does?  88.2 kHz has better time-domain resolution than 44.1 kHz.

Oversampling Better?

Reply #21 –
Definitely, with a blind test, I can hear a difference, particularly in high frequencies!

Oversampling Better?

Reply #22 –
Quote
Definitely, with a blind test, I can hear a difference, particularly in high frequencies!

Then I'm technically correct if I call you a son of a bi....

Oversampling Better?

Reply #23 –
But Marcan said that there were low frequencies (descending sweep) at the end of the sweep file.

Oversampling Better?

Reply #24 –
I have a descending sweep (not low frequencies) with 44.1, not with 96.