Skip to main content
Topic: 48kHz --> 44.1kHz: optimal downsampling method (Read 13836 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

48kHz --> 44.1kHz: optimal downsampling method

I have recently extracted a sound file from a DVD-V using HeadAC3he. This tutorial suggests that the extraction program (HeadAC3he) be used to downsample audio if one is ultimately going to burn a std. CD from it.

However, I found that I get better fidelity if I don't downsample with HeadAC3he (i.e. rip the std 48kHz audio) and then convert to 44.1 using Nero. Not sure why this is the case?? Pehaps because the extra processing during the ripping program (HeadAC3he) -- IOW, due to both ripping and downsampling -- compromises quality.

Whatever the case, is there a consumer-level (= low-cost) solution to optimal downsampling? Is the crude, non-integer and non-multiple factored math -- 48 --> 44.1kHz -- really a limiting issue?

48kHz --> 44.1kHz: optimal downsampling method

Reply #1
I have recently extracted a sound file from a DVD-V using HeadAC3he. This tutorial suggests that the extraction program (HeadAC3he) be used to downsample audio if one is ultimately going to burn a std. CD from it.

However, I found that I get better fidelity if I don't downsample with HeadAC3he (i.e. rip the std 48kHz audio) and then convert to 44.1 using Nero. Not sure why this is the case?? Pehaps because the extra processing during the ripping program (HeadAC3he) -- IOW, due to both ripping and downsampling -- compromises quality.


HeadAC3he probably just uses a crappy resampler.  That or you're imagining things.  Did you ABX the difference?

Whatever the case, is there a consumer-level (= low-cost) solution to optimal downsampling? Is the crude, non-integer and non-multiple factored math -- 48 --> 44.1kHz -- really a limiting issue?


Its not that hard.  Its a little hard if you want to do it well and very fast, but not that big a deal.  Theres tons of free software for it.  Personally, I just use foobar to resample it (which actually uses shibatch's resampler) ,but it sounds like you were happy with Nero, so I'd just use that.

48kHz --> 44.1kHz: optimal downsampling method

Reply #2
HeadAC3he probably just uses a crappy resampler.  That or you're imagining things.  Did you ABX the difference?


No ABX (tho' I do believe in and conduct this type of test in certain cases). I (re)listened to portions of each sample enough times, on same PC, to reach a *personal*, empirical consensus.

Whatever the case, is there a consumer-level (= low-cost) solution to optimal downsampling? Is the crude, non-integer and non-multiple factored math -- 48 --> 44.1kHz -- really a limiting issue?


Its not that hard.  Its a little hard if you want to do it well and very fast, but not that big a deal.  Theres tons of free software for it.  Personally, I just use foobar to resample it (which actually uses shibatch's resampler), but it sounds like you were happy with Nero, so I'd just use that.


My version of Nero is about 3 years old, so something more recent may yield better results. Thx for Foobar suggestion.

48kHz --> 44.1kHz: optimal downsampling method

Reply #3
Quote
My version of Nero is about 3 years old, so something more recent may yield better results.

Nero switched to a more complex resampler in some 6.x version. Nero 6.6 has it but I do not recall the exact version number with what the new resampling module got introduced.

48kHz --> 44.1kHz: optimal downsampling method

Reply #4
FWIW, here are my subjective results upon comparing different re-samplers. In order of sound quality, they are:

1. Original 48khz WAV file (before re-sampling)
2. Foobar2000 (v.0.9.4.2) SSRC Ultra (48-->44.1)
3. Nero Wave Editor 2 (v.2.0.0.58) "default" (48-->44.1)
4. Foobar2000 (v.0.9.4.2)  PPPHS (48-->44.1)

I really liked the SSRC Ultra version. I wonder why it's not listed (linked to) on the foobar200.org site (i.e., in http://foobar2000.org/components/index.html)?

If anyone knows of a better conversion engine and/or tips/tweaks to better use these tools please let me know.

48kHz --> 44.1kHz: optimal downsampling method

Reply #5

HeadAC3he probably just uses a crappy resampler.  That or you're imagining things.  Did you ABX the difference?


No ABX (tho' I do believe in and conduct this type of test in certain cases). I (re)listened to portions of each sample enough times, on same PC, to reach a *personal*, empirical consensus.



Well, you  might want to save yourself some trouble and try ABXing them to see if you're really hearing a difference. 



http://www.kikeg.arrakis.es/winabx/

48kHz --> 44.1kHz: optimal downsampling method

Reply #6
SSRC is not very good. For example it is not linear-phase. Secret rabbit code I think is much better. There is a plugin for foobar2000 named foo_dsp_src9.dll v1.03 [05 June 2006] - Resampler (Secret Rabbit Code). Search it with Google...

48kHz --> 44.1kHz: optimal downsampling method

Reply #7
SSRC is not very good. For example it is not linear-phase. Secret rabbit code I think is much better. There is a plugin for foobar2000 named foo_dsp_src9.dll v1.03 [05 June 2006] - Resampler (Secret Rabbit Code). Search it with Google...


Secret Rabbit Code is here:

http://www.mega-nerd.com/SRC/fb2k.html

I would like to try it, but can't get it to install in Foobar 2000 v9.4.2.

48kHz --> 44.1kHz: optimal downsampling method

Reply #8
SSRC is not very good. For example it is not linear-phase.


Really? I thought SSRC uses windowed sinc filters (Kaiser-Bessel window).
What makes you think SSRC is not linear-phase?

Cheers!
SG


48kHz --> 44.1kHz: optimal downsampling method

Reply #10
If you want an easy way to test the SRC resampler, try SRCDrop at Rarewares, 'Others' page.

 
SimplePortal 1.0.0 RC1 © 2008-2020