Is foobar2000's native dither "good enough"?

Discussion in 'Audio Hardware' started by Nightbreed, Dec 10, 2018.

Thread Status:
Not open for further replies.
  1. Nightbreed

    Nightbreed We're only immortal for a limited time. Thread Starter

    Location:
    Riverside, CA
    A few years ago when I learned how to downsample (thanks to these forums) I began by simply using foobar's native dither (plus the SoX resampler component). I've converted many hi-res albums to redbook and have been quite pleased with how they sound.

    Then I began wondering what alternatives are out there and it seems there may be better...even better dither options for use within foobar. The technical jargon regarding the "better" ones is confusing though, and I think the consensus is that foobar's native dither is good but noisy (still don't understand that...I don't hear any unwanted noise in my conversions).

    So should I just keep rolling with foobar's dither or would it be worth it to switch to another type of dither? And should I bother re-converting all my hi-res stuff again using a different dither?
     
  2. Eleventh Earl of Mar

    Eleventh Earl of Mar Somehow got them all this far.

    Location:
    New York
    I use foobar for all my MP3 converts (the FLAC stays at home... lol) and no issues myself.

    If you have to ask why it should sound better, it probably doesn't matter but I'll be killed for mentioning it here.
     
    Max Florian and Nightbreed like this.
  3. Plan9

    Plan9 Mastering Engineer

    Location:
    Toulouse, France
    I have compared different sorts of dither but never to the Foobar one.
    Usually the difference is very slight and is more akin to feeling and perception than to pure audibility.
    Sometimes they can lower the S/N ratio quite a bit but it's usually not a problem as only the higher frequencies are altered.

    My advice is if you don't hear anything wrong with your downsamples, keep them that way. The possible gain will be very slight anyway. Life is too short to bother re-converting all your stuff.
     
    Last edited: Dec 10, 2018
    Max Florian and Nightbreed like this.
  4. Plan9

    Plan9 Mastering Engineer

    Location:
    Toulouse, France
    A bit of advice if you don't know: don't use dither prior to converting to MP3. Do the encode from the 24bit file directly.
     
    Last edited: Dec 10, 2018
    vwestlife and albertop like this.
  5. Nightbreed

    Nightbreed We're only immortal for a limited time. Thread Starter

    Location:
    Riverside, CA
    Ain't that the truth! Thanks, I won't bother changing anything then. :righton:
     
  6. Plan9

    Plan9 Mastering Engineer

    Location:
    Toulouse, France
    If however you get the courage to compare different ditherings in Foobar, I'd be interested to know what are your findings. :)
     
    Nightbreed likes this.
  7. Eleventh Earl of Mar

    Eleventh Earl of Mar Somehow got them all this far.

    Location:
    New York
    Yeah, as it goes I should mention I don't have too many 24-bit files though, all of them were converted straight.
     
  8. carlwm

    carlwm Forum Resident

    Location:
    wales
    To my luddite brain, that is the single most incomprehensible post I've ever read! May as well have been written in Magyar! :D
     
    Nightbreed likes this.
  9. Nightbreed

    Nightbreed We're only immortal for a limited time. Thread Starter

    Location:
    Riverside, CA
    LOL! I hear ya, I felt the same way not long ago about this topic. That's still how I feel when reading the technical posts of people that really know their stuff haha. :laugh:
     
    carlwm likes this.
  10. jbmcb

    jbmcb Forum Resident

    Location:
    Troy, MI, USA
    I did a bunch of research on this a few years ago for the same reasons. At the time, the best dithering algorithms were in Weiss Saracon and whatever came with Cubase - I think the software version of Apogee's UV22. The best free dithering algorithms were SRC (Secret Rabbit Code) and SoX.

    Keep in mind that the difference between the better resamplers is *very* slight. I think the overall delta was a tenth of a dB difference in noise floor. I don't think you can go wrong with SoX.

    You can check out the technical comparisons here:
    SRC Comparisons
     
    Tommy SB and Nightbreed like this.
  11. Nightbreed

    Nightbreed We're only immortal for a limited time. Thread Starter

    Location:
    Riverside, CA
    Thanks for this! So at worst, an album that's been downsampled using a dithering tool that's not the "best", the outcome will only be a little bit of hiss? And probably not even that since the noise that's there is generally outside the scope of human hearing. Do I have that right? I still don't plan on re-doing any of my old downsamples, I just want to see if I'm understanding the general concept.
     
  12. jbmcb

    jbmcb Forum Resident

    Location:
    Troy, MI, USA
    It's probably lower than the noise floor of most DACs. If you are at all concerned you could dither a couple of tracks using different methods and use the ABX tool in foobar2000 to see if you can tell the difference. If you can't, then there's your answer. I couldn't.

    By the way, at the time I was using Cubase as I had a copy that came with some audio gear I bought. It isn't supported on newer OSes, so I just use foobar2000 to do dithering now, and it sounds fine to me.
     
    Tommy SB and Nightbreed like this.
  13. Plan9

    Plan9 Mastering Engineer

    Location:
    Toulouse, France
    Dithering =/= resampling
     
    albertop and Nightbreed like this.
  14. Plan9

    Plan9 Mastering Engineer

    Location:
    Toulouse, France
    You got that right except it's not hiss per se but digital noise.
     
    Nightbreed likes this.
  15. jbmcb

    jbmcb Forum Resident

    Location:
    Troy, MI, USA
    You're right, it's just that typing out "dithering and re-sampling" over and over again is tiring :)

    Usually when you're talking about going from any kind of hi-res to redbook, you're implying both, anyways.
     
    Nightbreed likes this.
  16. Randy Robinson

    Randy Robinson Well-Known Member

    Location:
    Manhattan, Ks
    If you want to experiment, Tascam has a free "Hi-res Editor" that can down sample from DSD or PCM to as low as 16/44. I've down sampled DSD to 24/96. Sounded ok to me.

    TASCAM Hi-Res Editor | DOWNLOADS | TASCAM
     
    Nightbreed likes this.
  17. testikoff

    testikoff Seasoned n00b

    Below are spectra of null-test deltas between 24-bit original & 24-to-16 downrezzed versions without dither (NO DITHER) & with various dithers: Audacity TPDF [rectangular] (AURA), Audacity sloped TPDF [triangular] (AUTA), Audacity noise-shaped (AUNS), foobar2000 noise-shaped (FB2K) & MDA plug-in noise-shaped (MDA):

    [​IMG]
     
  18. Nightbreed

    Nightbreed We're only immortal for a limited time. Thread Starter

    Location:
    Riverside, CA
    Oh hey, I was hoping to get your feedback on this. It was your posts that made me want to investigate other options but I couldn't reply to the old threads. Thanks for this graph, but I can't tell exactly which one is supposed to be the "best". FB2K is pretty extreme compared to the rest, so I assume it's either the worst or the best lol.

    EDIT: If I'm reading it correctly it appears as if foobar is one of the quietest until 20khz and then it's the noisiest (which I assume is why you say you don't use it). And if that's the correct way to read it then I suppose MDA would be the best.
     
  19. testikoff

    testikoff Seasoned n00b

    ^^^ I use MDA dither for my 24-to-16 bit downconversions (it has -79dBFS amplitude & I find it more precise, least intrusive & substantially more quiet that, say, TPDF). FB2K dither is pretty extreme (its amplitude is close to -60dBFS & it also tends to modify original 24-bit recording's frequency response above 15kHz). AUNS is quite good, but still a tad high in noise amplitude for my taste... X axis range on my graphs is 20Hz...20kHz, BTW.
     
  20. Nightbreed

    Nightbreed We're only immortal for a limited time. Thread Starter

    Location:
    Riverside, CA
    Can anyone tell me what foobar2000's "Smart Dither DSP" is? I thought it was the same as native foobar dither, but since I don't see this listed in "Components" it leads me to believe this one is better than native.

    foobar2000: Components Repository - Smart Dither DSP

    EDIT: I think I ran across an MDA dither plugin I can use in foobar. Is this the same MDA you use, testikoff? I don't plan on redoing any old conversions, but going forward I may consider using MDA via foobar.

    mda Dither DSP (foo_dsp_mdadither) – Component for foobar2000
     
  21. testikoff

    testikoff Seasoned n00b

    ^^^ Looks like a TPDF (rectangular) dither that leaves digital silence untouched. Still it will likely have less resolution in low & mid frequency areas than a moderate noise-shaped dither a la MDA, methinks...
     
    Nightbreed likes this.
  22. Nightbreed

    Nightbreed We're only immortal for a limited time. Thread Starter

    Location:
    Riverside, CA
    I just installed the MDA plugin in foobar to take it for a spin. Since I just did a new needle drop of Iron Maiden's "Live After Death" I converted it to redbook using both foobar native dither and MDA to compare. In the MDA plugin settings I chose "second-order noise shaped" which is the default (I hope that's the correct setting...I left all other settings default as well). Unless it's the ultimate placebo effect I'm pretty sure I can hear a difference between the two downsamples. The one I did using MDA has noticeably crisper highs; the cymbals sound more crisp especially and it sounds closest to the hi res version to my ear. Even the guitar crunch seems to have a little more bite. The foobar version in comparison sounds a little warmer and not quite as crisp (although still really good). I'll have to listen using good headphones later and report back. For now I think I like MDA the best, unless as I said, it's placebo.
     
  23. Nightbreed

    Nightbreed We're only immortal for a limited time. Thread Starter

    Location:
    Riverside, CA
    Never mind what I said about MDA sounding better. I tried the ABX test in Foobar and all samples sounded more or less identical. Then, to top if if, I don't even think I had the MDA plugin configured right because I downsampled to 8 bit so I could hear the difference and the foobar native dither was the only one that affected the digital noise (and the only one to show up in the Spectrum). The MDA sample sounded exactly like the one with no dither (and no evidence of it in the Spectrum) so I don't even think it did any dithering at all lol. MDA may very well be better but this is making my head hurt and I think I'll just leave well enough alone for now.
     
  24. albertop

    albertop Forum Resident

    Great piece of information, thank you!
    Have you tried dBPowerAmp? I generally use its dither (the triangular one) to convert 24bit files into 16bit files.
     
    Nightbreed likes this.
  25. Nightbreed

    Nightbreed We're only immortal for a limited time. Thread Starter

    Location:
    Riverside, CA
    I've never used dbPowerAmp for dither but it's what I use to rip my CDs to FLAC. I'll have to play around with it and compare to foobar. Thanks!
     
    albertop likes this.
Thread Status:
Not open for further replies.

Share This Page

molar-endocrine