High-end audio with Foobar2000 player

Discussion in 'Desktop & Laptop Computers Forum' started by Rob.Screene, Dec 8, 2003.

  1. Rob.Screene

    Rob.Screene
    Active Member

    Joined:
    Sep 29, 2001
    Messages:
    1,314
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    66
    Location:
    Berks, England
    Ratings:
    +141
    On reading about the high-end players resampling CD's to 24-bit/96KHz I decided I miss the option of highest quality SSRC resampled output that that plug-in and the MAD 24-bit decoder plug-in gave me from WinAampP under WindowsME to my m-audio Audiophile2496 and relatively high-end AV amp.

    Since adopting WindowsXP, I need to use the ASIO output to avoid the WindowsXP kmixer degrading any CD playback.

    I decided to read up on the new, free foobar2000 player, developed by a WinAMP developer who is concerned totally with quality.

    With the low cost of 200GB+ hard disks, I also took the opportunity to re-rip all my albums using Audiograbber and compress using the lossless Monkeys Audio .APE format, which requires about 600-860Kbps, so are about 3x the size of my 256Kbps MP3's but are bit-perfect copies of the originals and support ID3 like track/album/artist/year tagging which audiograbber seems to do automatically. Lots of people recommend EAC, but I found it unfriendly and audiograbber gives me repeatably bit-identical rips and seems to read at 12x off my Pioneer A03 drive.

    I have yet to do extensive listening via my HTPC in my listening room, but it sounds very, very promising via my desktop Audigy/Cambridge Soundworks 4.1 speakers in my office.

    Here's 5 reasons why I think foobar2000 is the future (apologies to Phoenix nights fans):
    (with the exception of the KS output, it's all part of the standard kit and the KS part is easily downloaded from the foobar2000 site)

    1. Decoding done internally to over 32-bit fixed precision, using 64-bit floating point accuracy.
    This is used by all processing components, such as volume, replaygain, etc. Lets say I don't think we need worry about loss of precision internally. (there are complex DSP components, like a convolution equaliser, but I won't go in to too much detail on this first post). It includes native .MP3 and .APE decoders among many others.

    2. Replaygain support
    It doesn't sound much, but it is.
    This measure your music tracks and stores a volume level adjustment gain figure, say -5.2dB as a tag on the end of the .APE file like how album and track info is stored. It doesn't mess with the digital audio at all.

    It's fantastic, as modern stuff has more and more dynamic range compression to sound louder and louder.
    e.g.
    listen to Van Morrison's Moondance after JT's Cry Me a River and Moondance sounds thin and quiet. Not any more.

    You set the volume you are in the mood for and that your system plays best at. Simple.

    This is light-years ahead of simple 16-bit peak level normalisation that I used to get audiograbber to do for me in the past.

    It actually calculates two gain figures, one per-track and the other per-album. foobar2000 then uses this figure to internally adjust the volume (at over 32-bit precision) to make each track sound the same loudness.

    You can toggle playback to use the Album mode "audiophile" gain to preserve differences in level between tracks per album, but the default "Radio" mode works brilliantly for me, even for a live album like Eric Clapton Unplugged, you don't notice any gaps or per track volume changes.

    I was surprised that the standard download included everything I needed to create this replaygain info in my .APE files. Simply select all your tracks in foobar2000, right click and choose replaygain, scan selection as multiple albums.

    3. Output bit-depth is easily configurable
    The vast internal accuracy needs reducing to output to mere mortal 24-bit or similar output for our DAC's.

    Set it to the highest depth your high-end soundcard supports well, i.e. 24-bit or 32-bit.

    It also includes dithered algorithms if you have to output at 16-bit, which apparently increases the apparent depth at 16-bit.

    4. Kernal Streaming KS output
    An easily downloaded KS output .dll can be used to output bit-perfect accuracy, avoiding the Windows KMIXER bugs.

    There are also ASIO output ones, but they seem harder to find the KS one works brilliantly for me.

    5. SSRC Resampling (at over 32-bit precision)
    You can resample to 48, 88.2 96KHz, etc again with the renowned SSRC algorithm. 24-bit 96KHz S/PDIF output, no problem. This happens at internal precision, before the final depth reduction to 24 or 32-bit, etc.

    Let your own ears decide if you prefer the best 24/96 resampling over standard CD audio from your own CD's. Oh and for free! To my ears it rids a lot of the harshness I dislike about many CD's.

    The next step is for me or someone to create a girder setup to allow remote control of the player.

    I'd be interested what others think of this. Oh and it seems happy to coexist with WinAmp till I decide which one to keep.

    References:
    http://www.replaygain.org (good introduction, but confusingly not updated to mention foobar now supports calculating and playing these tags. Watch out for the red herring that is the MP3gain app that calculates replaygains but adjusts MP3's themselves, meaning non-replaygain aware MP3 players can benefit, but it doesn't handle .APE and is limited to 1.5dB adjustment multiples and only works in per-track "radio" mode not Album mode, if you ever decide to care!)
    http://www.foobar2000.org
    http://www.hydrogenaudio.org/index.php?act=SF&s=&f=28/

    regards,
    Rob.
     
  2. Madders

    Madders
    Active Member

    Joined:
    Jan 30, 2002
    Messages:
    2,668
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    71
    Location:
    London
    Ratings:
    +62
    Agree with you, Foobar2000 is an excellent player. I now use this instead of Winamp 2.91 - but with lossless ripped WAV files.

    Cheers,

    Steve
     
  3. Nic Rhodes

    Nic Rhodes
    Well-known Member

    Joined:
    Mar 23, 2001
    Messages:
    17,152
    Products Owned:
    0
    Products Wanted:
    1
    Trophy Points:
    133
    Location:
    Cumbria
    Ratings:
    +1,276
    good post :)
     
  4. GRees

    GRees
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    I've been using Foobar2000 to play back my .ape files for a little while now and have to agree that it seems very good. But, I can't seem to get the kernel streaming to work. I'm using a Revolution7.1 and I can only get kernel streaming working if I run Foobar immediately after installing the soundcard drivers. As soon as I go off and run ZoomPlayer or something else and then return to Foobar - "error opening device" - is all I keep getting :confused:

    Uninstall Revo - reinstall Revo - Foobar2k kernel streaming - :)
    ZoomPlayer - Foobar2k kernel streaming - :confused:

    What's going on - help please.

    I've tried using ASIO and setting playback to 32 bit fixed point but when I play back my .ape files everything is speeded up so that it sounds like there's a bit of helium in the air.

    Any suggestions?
     
  5. MuFu

    MuFu
    Active Member

    Joined:
    Jul 24, 2002
    Messages:
    1,196
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    41
    Location:
    NW2, London
    Ratings:
    +4
    Are you sure there's nothing in your system tray (or some other background task) that's using the soundcard? DigiTV, Quicktime Launcher etc...?

    I'm using foobar with a Revo too (been using it for about a year now). Resampling to 24b/96k or 24b/192k used to be worthwhile when I had an Audigy 2, but that is no longer the case. I guess it's a DAC-specific issue. The DSP chain I use is:

    Convolver=>EQ=>Advanced Limiter=>Volume Control

    The volume control is to compensate for the lowest click on my volume dial being a bit hot for night time listening when using KS (I'd rather use foobar as a preamp than the soundcard drivers anyway). Convolver gets <250Hz to +/-3dB and the EQ introduces a nice house curve. The limiter I've had to start using recently because Convolver is pushing some bass-heavy tracks into clipping. I'm not sure whether this will sort itself out once more tracks are appended with replay gain info. Maybe not. :-\

    MuFu.
     
  6. GRees

    GRees
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    MuFu, thanks for that, you were absolutely right. The culprit was DigiTV sitting in the system tray. Works a treat now. :smashin:

    Now that I've had a chance to listen to it properly, I've found that selecting 32 bit fixed point for playback and upsampling to 88.2kHz ( although I'm not totally convinced about the upsampling ) seems to be the best setting for me. I don't really understand why 32 bit sounds better than 24 bit as I thought that the Revo used 24 bit DACS or am I talking about apples and bananas?

    Rob, great post, thanks for the tip about replaygain it's made a big difference, especially when going from older to newer tracks.

    Has anyone else tried Denis Sbragions' Digital Room Correction for the Foobar Convolver plugin? I gave it a go with my old Audigy 2 ZS but it was a bit of a mixed bag. I could only use either my SPL meter line out or a borrowed mic using the mic in on the Audigy2 for calibration neither gave very good results. Has anyone tried this or something similar using a good quality mic/amp?
    Cheers
    Gary
     
  7. MuFu

    MuFu
    Active Member

    Joined:
    Jul 24, 2002
    Messages:
    1,196
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    41
    Location:
    NW2, London
    Ratings:
    +4
    Did you follow this guide?

    http://www.mooneyass.com/DRC/DRC.html

    I haven't tried it yet, but might give it a go after Xmas. My dad has a pretty nice Rode mic and preamp.

    I'm not too sure whether correcting high frequencies is worth doing at all - it might well introduce more problems than it solves. If you look at Jones Rush's own results on AVS (thanks again for that link) they are astonishing, but he probably now has an uncomfortably dry listening environment.

    I'm just using Convolver rather crudely at the moment to correct frequencies under 160Hz. Works a treat though!

    MuFu.
     
  8. Rob.Screene

    Rob.Screene
    Active Member

    Joined:
    Sep 29, 2001
    Messages:
    1,314
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    66
    Location:
    Berks, England
    Ratings:
    +141
    Hi MuFu,
    Any gain, including replaygain deduced positive +dB track settings (rare, I think I only have one or two tracks with this) can push the signal in to clipping.

    Wouldn't it make more sense to drop the master gains under Preferences, Playback, Preamp rather than add any limiters?

    I currently have my 'files with RG info' setting at -1dB, even though most files have replaygain values of -6dB (i.e they are played back 7dB below the raw CD data level), some have up to +1dB.

    The reason I'm not too worried about this loss of 7dB range is because it's in 32-bit+ precision with 24-bit output. My thinking is there's plenty of range in that final 24-bit depth.

    Did you see the udial.ape download mentioned from the foobar2000 hydrogenaudio forum. It's a combination of max level 19KHz sines and some dial tones just to stop people putting the level too high and frying their tweeters.

    If the digital data isn't clipped it sounds like quiet dialling tones. If it is clipped, it sounds like someone's shooting you with a laser. It also sounds like that if the KMIXER has resampled it, so it's a good test of pure digi output and no clipping.

    cheers,
    Rob.
     
  9. GRees

    GRees
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    That is the guide I followed. It does work provided you follow the instructions exactly. The only downside is that for a novice like me, there is no explanation to exactly what it is that you're doing. I just just ended up with a sine that looked similar to the one in the instructions so assumed it was OK. Still don't know what it all means though.

    I agree with you that Jones Rush must have a very dry listening environment now. My own experience was that it made a huge improvement to the bass by getting rid of a nasty resonance low down and filling in a hole a bit higher up the spectrum. As I got a bit higher up though, things took a turn for the worse. Vocals - particularly female vocals - had a noticable hollow ring to them and the treble was a bit exaggerated. I'm assuming that this is down to the poor quality kit I used to measure my room, so I'll reserve final judgement until I get a decent mic/amp.

    How do you use convolver to just correct below 160Hz?
     
  10. MuFu

    MuFu
    Active Member

    Joined:
    Jul 24, 2002
    Messages:
    1,196
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    41
    Location:
    NW2, London
    Ratings:
    +4
    The limiter is just a stop-gap solution while I'm still experimenting with convolver. I've only noticed one track clip so far and that was a THX bass demo (with one of the more "enthusiastically" EQ'd impulse files, lol) but that scared me a bit. It's just a precaution... I'd rather use it than fry the sub!

    Thanks, I'll have a look some time after exams.

    Took SPL readings (<160Hz) using a sine wave generator and RS meter. Then messed about for ages with Cool Edit setting up various EQ parameters by a process of trial and error. The result is an impulse file that produces a reasonably flat response when used with Convolver.

    I'm sure you can adapt Jones Rush's method so that it just works for frequencies <160Hz. It's theoretically as simple as making the impulse file completely flat above 160Hz.

    MuFu.
     
  11. GRees

    GRees
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    Help - I'm being shot with lasers.

    I've downloaded udial.ape and played it back with kernel streaming, no DSPs', playback at 16 bit fixed, but I still get the lasers :confused:

    What am I doing wrong?

    Gary
     
  12. GRees

    GRees
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    Just realised what I'm doing wrong. As Rob mentioned previously - any gain, including replaygain can cause clipping - and that is what is happening to me with udial.ape. Take away the replaygain and all is well. Makes me wonder whether or not I should use replaygain on my playlist.

    Perhaps it's a case of - no gain no pain :D

    Whilst I'm still in a Foobar kind of mood - I was messing around with resampler and tried 176400 Hz which worked OK with my Revo. But when I looked at the Revo Input it showed 174600 Hz. Surely this is a error by M-Audio as 44100 Hz x 4 =176400 Hz.

    Anyone else noticed this?
     
  13. Rob.Screene

    Rob.Screene
    Active Member

    Joined:
    Sep 29, 2001
    Messages:
    1,314
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    66
    Location:
    Berks, England
    Ratings:
    +141
    udial.ape didn't have any replaygain information on it. Did you add it by accident?

    cheers,
    Rob.
     
  14. GRees

    GRees
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    Hi Rob

    I got the "lasers" the first time that I played udial.ape. Then I realised that whilst I had been fiddling with the settings, I had set "files without RG info" to a little above zero. After resetting this back to zero, I had a nice clean playback.

    It was only then that I tried applying replay gain to udial.ape and, lo and behold, got the "lasers" back.

    Forgive me for being a bit slow with this, but does it mean that by applying replay gain to my playlist, I am going to cause clipping?

    Cheers
    Gary
     
  15. Rob.Screene

    Rob.Screene
    Active Member

    Joined:
    Sep 29, 2001
    Messages:
    1,314
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    66
    Location:
    Berks, England
    Ratings:
    +141
    Hi Gary,

    In short,
    Set Preferences, Playback, Replaygain, Preamp
    Replaygain mode: Album mode
    Files with ReplayGain info: -1.2 dB
    Files without ReplayGain info: -0.3dB

    I don't get clipping, well except if I replaygain wierd SOUNDING test tones....

    In detail,
    Q:does it mean that by applying replay gain to my playlist, I am going to cause clipping?
    A: Not usually.
    As replay gain aims for 89dB average level, it usually applies a gain between -2 dB and -7 dB to the music i.e. it's played back quieter.

    The modern, dynamically compressed stuff like pop gets reduced more, say -6.5dB, the other stuff less.

    The most extreme I've noticed is Mozart Piano Concerto No 19 in F, KV 459 - Allegro vivace,
    Right click and choose Show File Info...
    replaygain_track_gain = +1.99 dB
    replaygain_track_peak = 0.576110
    replaygain_album_gain = +1.11 dB
    replaygain_album_peak = 0.999481

    So, in this case, it's a very quiet track and ReplayGain has calculated this track needs +1.99dB gain, but see the peak for the track is 0.576110. Think of that as 57.5% of the max available volume on a CD, so it won't clip. It's on average a quiet track without big peaks.

    Also, see later two values for the the whole classical album album (i.e. if playback is in 'Audiophile' per-album mode, as opposed to 'radio' per-track mode).

    Here the replaygain required is less, at +1.11 dB, and the recorded peak is right up at 99.95%. In this case setting playback mode to per-album and pre-amp to = -1.2% so playback will be +1.11 -1.2 = -0.09 dB adjusted = no clipping. This means other tracks on the album are louder on average, with huge peaks somwhere, well that's classical for you!

    The conclusion is anything with a plus replaygain may clip if it used the original gain available and the pre-amp settings are left at the 0 dB defaults.

    In getting replaygain to evaluate the set tone as a human ear would pervieve it, well it does sound very quiet doesn't it?
    The phone tones are quiet and the 19KHz sine is inaudiable, hence replaygain giving it a +2.0 dB gain and pushing the 99.9999% signal 2dB in to clipping.

    Two solutions:
    a) Check the 'use peak info to scale down to avoid clipping' checkbox, although this would have ment mozart would not get the +1.11 gain and would sound on average 1.11dB lower than Macy Gray! Kind of dilutes the advantage of replaygain, eh?

    b) (my preference) with a 24-bit/32-bit soundcard output, dropping the pre-amp setting to -1.2 dB and upping my amplifier's volume a similar amount is a nicer solution. I also now use per-album mode, not per-track as track volume differences are sometimes intentional and it means replay gain doesn't get over excited by one or two quiet interlude type tracks on an album.

    I also check the Show Clipping warnings checkbox in that page's output section. Just so I know.

    I have noticed the clip warning is sensitive and may also be because of the 16-bit to 64-bit to 24-bit conversions that I sometimes see clip warnings for non-replaygain files with the
    If the original recording was within 99.9%, it still shows a clip warning unless it give it a relatively tiny -0.3dB pre-amp setting.

    This will of course ruin any dolby digital or DTS tracks music tracks you play with foobar2000 because they rely on bit perfection, which of course get changes by the -0.3dB digital volume drop. I think a better solution is to replaygain all your ripped music, but not DTS/DD stuff and set the non-replaygain pre-amp to 0.0dB, but remember you'll still have to reset output to a lowly 16-bit and remove any resampling to hear them properly.

    cheers,
    Rob.
     
  16. GRees

    GRees
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    Hi Rob

    Thanks for taking the time to explain that, it's starting to make a lot more sense now :smashin:

    Now I'm beginning to doubt the wisdom of having my Revolutions' analogue output hooked up to a 5 channel power amp. Apart from the potential destruction of my speakers, I don't have the added flexibility of increasing the gain on a separate pre-amp. Any reduction in gain in Foobar is going to lower output even more and I've already got everything at maximum. This problem is particularly noticable with films - I just can't get the damn thing to go loud enough - but that's another story.

    Thanks again

    Gary
     
  17. Rob.Screene

    Rob.Screene
    Active Member

    Joined:
    Sep 29, 2001
    Messages:
    1,314
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    66
    Location:
    Berks, England
    Ratings:
    +141
    No input gain switch or adjustment on the power amp then?

    HTPC direct to power amp is not something I've any practical experience of. Seems quite common on AVS though.

    cheers,
    Rob.
     
  18. CrispyXUK

    CrispyXUK
    Active Member

    Joined:
    Feb 5, 2003
    Messages:
    3,751
    Products Owned:
    0
    Products Wanted:
    1
    Trophy Points:
    71
    Location:
    Land of the Living, Essex
    Ratings:
    +45
    Wish I had the balls to try that
     
  19. GRees

    GRees
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    Nope, unfortunately not. It's just a simple 5x120W THX Ultra power amp with an on/off button and a 12V trigger - that's it.

    I know it sounds suicidal, but the only problem I have is with the power on/off popping noise and that just means turning my amp and sub on/off before the PC on/off. This wasn't an issue when I was using the Audigy 2 ZS because it only produced a very quiet pop - even when we had a power cut. That was great as it meant I could use the 12V trigger - taken off my mobo fan header - to power the amp on/off. Unfortunately the same can't be said for the Revolution or NVidia Soundstorm - ouch.

    Some rogue application suddenly cranking the volume up isn't really a concern because as I said earlier, even with everything at maximum settings I still can't get it to go loud enough. It's nowhere near the volume - with the same power amp - as my recently departed AV32.

    I think the only solution is to do some research into alternative soundcards to find out whether or not they offer some sort of adjustable output gain. I notice that the new RME9632 seems to offer just that and it also claims to have built-in speaker protection. No bass management though - and it's a bit pricey.

    Oh well...

    Dear Santa...............
     
  20. sapgem

    sapgem
    Standard Member

    Joined:
    Dec 11, 2002
    Messages:
    83
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    6
    Ratings:
    +0
    Sorry to say it again.. but missing out the pre-amp between the sound card and power amp is not recommended. Digital attenuation not only reduces the volume, it also reduces the SNR of the the sound card output (by exactly the same amount expressed in dB).
     
  21. CrispyXUK

    CrispyXUK
    Active Member

    Joined:
    Feb 5, 2003
    Messages:
    3,751
    Products Owned:
    0
    Products Wanted:
    1
    Trophy Points:
    71
    Location:
    Land of the Living, Essex
    Ratings:
    +45
    yeah me too???????????????
     
  22. Rob.Screene

    Rob.Screene
    Active Member

    Joined:
    Sep 29, 2001
    Messages:
    1,314
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    66
    Location:
    Berks, England
    Ratings:
    +141
    Hi sapgem.
    I understand what you are saying. The noise floor stays the same irrespective of the dB digital attenuation.

    In the case of these m-audio 2496 cards, we're talking of 104-106dB S/N, so losing 6dB from that will still leave 98 dB S/N.

    For me the lack of harshness using this playback source is very very enjoyable, although I'm only feeding to a THX Ultra Denon via digital 24/96 and on to some Linn Index II's and a sub, because otherwise if I fed analog the Denon would ADC it first to to do Bass management and time alignment. Even in this perhaps humble set-up, with very little room treatment, I'm finding this one of the best music replay sources I've ever heard. Period. I did find the Meridian system had potential to sound as clean if not better at the Event II, but I do find most other digital systems harsh.

    The ability to bring your albums from different years all to the proper listening level it was aimed at, plus it's done with superb depth of precision really makes all material sound great. Even Annie Lenox Diva, Deacon Blue Raintown, etc and all those early not-quite 16-bit really CD's of that era, which when combined with bright Linn's really need a refined source. I can't see me changing from this as a CD replay source. Definitely power-amps, active crossovers and high-end speakers are next.

    For Gary, he's also removing any degredation an external analog pre-amp might have added, let alone what most AV processors/integrateds might do. It's just for some reason the line-level outputs aren't hot enough for his chosen power amp.

    regards,
    Rob.
     
  23. Madders

    Madders
    Active Member

    Joined:
    Jan 30, 2002
    Messages:
    2,668
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    71
    Location:
    London
    Ratings:
    +62
    I tried the HTPC to amp thing for a while. I didn't have a problem with the "popping" sound, you just make sure you turn things off in the correct sequence. I am using an RME-Audio card. The sound quality never sounded inferior when turning the sound down on the RME-Audio control panel. The main reasons for getting a processor again was for processing of Dolby Digital when using Sky+ and ease of turning sound down/up without selecting control panels!!

    Cheers,

    Steve
     
  24. MuFu

    MuFu
    Active Member

    Joined:
    Jul 24, 2002
    Messages:
    1,196
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    41
    Location:
    NW2, London
    Ratings:
    +4
    Rob, sapgem,

    I deleted my original post on this because I wanted to double-check M-Audio's specs.

    The Revo and the 24/96 digitally attenuate with 36-bit precision internally; I assume they resolution-enhance and filter bitstreams before mixing them. Subsequently, even if you reduce output by 36dB (i.e. by a factor of 64), you're still getting a 30-bit signal sent to the DAC (which will discard the 6 redundant LSBs). It's therefore pretty much impossible to noticeably reduce the source resolution of a 24-bit recording, let alone most CDs. Reduced SNR isn't an issue because at the last stage in the digital domain the signals are all full DAC res (or greater).

    How is this unsatisfactory? Am I missing something?

    Edit - wouldn't be the first time. :D

    MuFu.
     
  25. Rob.Screene

    Rob.Screene
    Active Member

    Joined:
    Sep 29, 2001
    Messages:
    1,314
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    66
    Location:
    Berks, England
    Ratings:
    +141
    Hi MuFu,
    I think you are wrong assuming digital volume attenuation and DSP is done using only the extra, least-significant bits.

    With an upsampled 16-bit signal, the result is very close to the 16-bit original, give or take a few bits caused by DSP, such as volume (pre-amp/replaygain), resampling, dithered noise shaping, etc.

    Sample values (Hex)
    CD 16-bit original = 7F FF
    foobar (let's pretend it's only 32-bit! as the actual 64-bit floating point is hassle to read/write)
    = 7F FF 00 00
    replaygain -5.9dB = 78 50 6F EE
    resample to 96KHz= 78 12 EE D0
    (now many samples, above is just one)

    Round/optionally apply dither filter to 24-bit output:
    result = 78 12 EF

    See each time that the left most-significant bits stay very close to the original, it's only the added precision bits to the right that get maintained then discarded, so the extra bits are only giving us depth and precision so that perceivable noise isn't added at each DSP stage.

    IMHO, I think it's these precise calculations and wide dispersion of rounding errors is why Foobar2000 sounds so good.

    cheers,
    Rob.
     
  26. sapgem

    sapgem
    Standard Member

    Joined:
    Dec 11, 2002
    Messages:
    83
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    6
    Ratings:
    +0
    My way of thinking about digital attenuation is to think about how many different possible output voltages the Revo has to generate to avoid losing any information.

    To avoid losing any information in a 16bit audio file the Revo has to be capable of generating just over 65,000 different levels of output signal. Every bit you add increases this number by x2. So if you have -36db of digital attenuation the Revo has to be able to distinguish 24 bits (8 extra bits), thats over 16 million different levels of output signal.

    It may be that its easy to generate 16 million different levels of output signal from a PC sound card. But it seems to me that this may be the limiting factor. After all if a soundcard could generate that number of outputs then it should approach a SNR of 132db for 24 bit audio (the Revo gets 108db i think, and THD may affect the output ability as well - not sure quite how, seem to remember someone on HTPC explaining).
     

Share This Page

Loading...
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice