Upsampling/SRC

[ack_0220]

Is it possible to upsample using software?

 

[audioengr]

Quote:

Originally Posted by ack_0220 /img/forum/go_quote.gif Is it possible to upsample using software?

Both SSRC and SRC do it on the fly with Foobar2000. Older version of SRC with Foobar 0.8.3 has the best sound upsampled to 24/96 IMO.

Steve N.
Empirical Audio

 

[tschanrm]

I have never seen any evidence proving upsampling/resampling does anything beside provide a compatible sample rate and bit depth for your sound card. Resampling arised from the incompatible sample rates between Redbook CD and DAT (Redbook being 44.1khz, DAT being 48khz.).

Resampling/upsampling may provide a different sound, but it is by no means more accurate or higher quality than the source material.

 

[audioengr]

Quote:

Originally Posted by tschanrm /img/forum/go_quote.gif I have never seen any evidence proving upsampling/resampling does anything beside provide a compatible sample rate and bit depth for your sound card. Resampling arised from the incompatible sample rates between Redbook CD and DAT (Redbook being 44.1khz, DAT being 48khz.). Resampling/upsampling may provide a different sound, but it is by no means more accurate or higher quality than the source material.

I dont agree. When transients are clipped due to insufficient sample-rate, good upsampling algorithms can add the missing samples back-in, providing more dynamics and clarity to the transients. I and my customers have done lots of listening tests that verify this. It actually gets more clear, more dynamic and less harsh.

You should try SRC and ASIO 47A SSE2 on Foobar.

Steve N.

 

[ack_0220]

Quote:

Originally Posted by tschanrm /img/forum/go_quote.gif I have never seen any evidence proving upsampling/resampling does anything beside provide a compatible sample rate and bit depth for your sound card. Resampling arised from the incompatible sample rates between Redbook CD and DAT (Redbook being 44.1khz, DAT being 48khz.). Resampling/upsampling may provide a different sound, but it is by no means more accurate or higher quality than the source material.

I have to disagree, i just learned how to upsample my musics, even with SRC on Foobar 2000 ver 9, there are noticable difference. The bass are more detailed and tight for instance.

However i am experiencing a slight problem with ASIO when i upsample to 96000. From what i know, ASIO are suppose to bypass window's KMixer and there should be no sound from other programs like MSN Messeger alert. But when i got an alert, there is actually sound coming out and Foobar stops playing any sound while the track is still running. When i down sample back to 44100 and everything works fine like how it is suppose to be. Anyone knows why?

 

[xenithon]

I have an Edirol UA-1EX USB interface which I currently use (as an interface an DAC....external DAC will come later down the line). I tried various different upsampling plugins with foobar. I always ended up preferring the PPHS plugin, and also preferred 24-bit at 48kHz....I found that 96kHz sounded more "electronic" and forced. Difficult to explain - essentially it just did not sound right (and I listen to a lot of female vocal so it was quite prominent).

 

[HFat]

Quote:

Originally Posted by audioengr /img/forum/go_quote.gif I and my customers have done lots of listening tests that verify this.

Why don't you post the protocol(s) and the results?

ack_0220, have you tried a blind test?

 

[Max F]

Quote:

Originally Posted by audioengr /img/forum/go_quote.gif I dont agree. When transients are clipped due to insufficient sample-rate, good upsampling algorithms can add the missing samples back-in, providing more dynamics and clarity to the transients. I and my customers have done lots of listening tests that verify this. It actually gets more clear, more dynamic and less harsh. You should try SRC and ASIO 47A SSE2 on Foobar. Steve N.

I thought upsamplers interpolate between samples not extrapolate beyond what was already there.

 

[audioengr]

Quote:

Originally Posted by Max F /img/forum/go_quote.gif I thought upsamplers interpolate between samples not extrapolate beyond what was already there.

No, the algorithms actually compute where the missing samples were likely to be, and some of them are excellent at this. They use several samples prior and after the missing ones in the computation.

You can set some upsamplers for straight-line interpolation (such as SRC), but is does not sound very good.

Steve N.

 

[tschanrm]

Quote:

Originally Posted by audioengr /img/forum/go_quote.gif I dont agree. When transients are clipped due to insufficient sample-rate, good upsampling algorithms can add the missing samples back-in, providing more dynamics and clarity to the transients. I and my customers have done lots of listening tests that verify this. It actually gets more clear, more dynamic and less harsh. You should try SRC and ASIO 47A SSE2 on Foobar. Steve N.

Could you please explain this in more detail? A sample rate of 44.1khz accurately reproduces frequencies up to 22,050 Hz. If the sound is harsh to begin with, it is from hardware incompatibility, or more than likely the source material mastering. Is the SRC you're talking about the foobar integrated v0.8.3 one? If so, they say right on the page:

Quote:

Resampling to higher sampe rate doesn't increase perceived "sound quality," only resampling to 48000Hz avoids issues with certain types of hardware

Secondy, the sample rate does not add more dynamics to the sound, the bit-depth allows for more dynamic range.

Quote:

I thought upsamplers interpolate between samples not extrapolate beyond what was already there.

Max F, you are correct in almost all cases that I know of. The only system I know that does extrapolation is mp3PRO's "spectral band replication," which can add back missing frequencies. However, to extrapolate correctly, the file needs to be coded in a special manner.

 

[tschanrm]

This may give more insight as to why, in my opinion, upsampling has been marketed as better sound quality:

Mosts DVD players use a clock of 27mhz or 54mhz, both of which are not a compatible clock frequency for 44.1khz audio. To solve this issue, some companies use a cheap PLL to allow 44.1khz sound to be played. Manufacturers, realizing that using a PLL creates inferior sound for 44.1khz audio, starting implementing upsamplers to create a compatible sample rate for the 27mhz or 54mhz clock. Hence, compared to DVD players that used a PLL for 44.1khz audio, the claim could accurately be made that an upsampler creates better sound quality.

Secondly, some upsamplers, such as AD1896, can be used as a buffer to reduce clock jitter, which can improve sound quality as well.

 

[kipman725]

right using my setup which is pretty crappy apart from my headphones upsampling sounded alot better (more detail and the sound and individual instruments were more difined and easier to position and I could hear some background noises I never could hear before) than using the actual sample rate of the actual recording. CPU usage has shot up though to about 26% from 0-1% on my 3400+ se**** so this will not be a suitible setting for listening to while gaming. If you are sceptical just download foobar and try it. I did.

*I used songs for the deaf for testing by queens of the stonage as this has lots of faint background sounds that usualy come out as mush on crappy systems (like mp3 players with bundled headphones *shudders*)

 

[tschanrm]

Quote:

Originally Posted by kipman725 /img/forum/go_quote.gif right using my setup which is pretty crappy apart from my headphones upsampling sounded alot better (more detail and the sound and individual instruments were more difined and easier to position and I could hear some background noises I never could hear before) than using the actual sample rate of the actual recording. CPU usage has shot up though to about 26% from 0-1% on my 3400+ se**** so this will not be a suitible setting for listening to while gaming. If you are sceptical just download foobar and try it. I did. *I used songs for the deaf for testing by queens of the stonage as this has lots of faint background sounds that usualy come out as mush on crappy systems (like mp3 players with bundled headphones *shudders*)

I use foobar v0.8.3 with foobar's SRC resampling plugin everyday to do active crossover/delay with an Audigy 2ZS. I have tested on my recording setup, Echo Layla 24/96, and never heard any difference between 44.1khz, 44.1khz resampled to 48khz, and 44.1khz resampled to 96khz. I've RMAA'd foobars SRC and PPHS resample alogorithms, and they provide less dynamic range and the same frequency response. I can post the results if you want to see them.

I think resampling is great compatibility tool; if you think it makes your equipment sound better, thats fine. Everybody has a personal preference to sound and what they perceive as better sound quality.

Kipman, what's your soundcard?

 

[kipman725]

well shamefully I'm using sis onboard sound here untill I have the money to buy a soundcard here so it may be somekind of hardware issue I'm having here. definatly a huge differance though, If everyone else is hearing practicaly nothing then it must be crapy hardware. Definatly not just in my head though as I keep switching back and forth and listen to specific points.

 

[tschanrm]

Quote:

Originally Posted by kipman725 /img/forum/go_quote.gif well shamefully I'm using sis onboard sound here untill I have the money to buy a soundcard here so it may be somekind of hardware issue I'm having here. definatly a huge differance though, If everyone else is hearing practicaly nothing then it must be crapy hardware. Definatly not just in my head though as I keep switching back and forth and listen to specific points.

In your case your soundcard automatically resamples to 48khz for 44.1khz source files, so having foobar resample instead of the soundcard would make a noticeable difference. Onboard soundcard resamplers are of poor quality.