Quick Settings:
Winamp
Use Shibatch mpg123 input plugin, enable ReplayGain
Sound Blaster Audigy 2 ZX - mpg123 at 24-bit, ASIO output, 24-bit resampled to 96KHz, no dithering
Sound Blaster Audigy 2/NX/LS - mpg123 at 24-bit, DirectSound w/ SSRC output, 24-bit resampled to 96KHz, no dithering
all other SB cards - mpg123 at 16-bit, DirectSound w/ SSRC output, 16-bit resampled to 48KHz, triangular dithering/noise shaping
most 24-bit cards - mpg123 at 24-bit, ASIO output, 24-bit, no resampling, no dithering
Foobar 2000
Sound Blaster Audigy 2 series - 24-bit padded to 32, no dithering, resampler DSP at 48KHz, Kernel Streaming output
all other SB cards - 16-bit padded to 32, dither: strong ATH, resampler DSP at 48KHz, DirectSound or WaveOut
most 24-bit cards - 24-bit padded to 32, no dithering, ASIO or Kernel Streaming output
Step 1: Picking the Right Hardware
Sound Card
When looking for a sound card, first consider what you are going to be using it for and how you are going to use it. If you plan to use it for games, you may want a card with DirectSound hardware acceleration so that it does not affect your framerate. If you intend to use the analog outputs of your card for headphones or speakers, you’ll want a card with a good Digital to Analog Converter (DAC), whereas the DAC does not matter if you are using the digital output.
The most important factor in the quality of your sound card is the chipset on which it is based. The most popular chipset, by far, is the EMU 10k series, found in Creative Lab’s Sound Blaster cards. While Sound Blaster cards are widely despised among audiophiles, it is largely the fault of their poor drivers and DSP, not the chipset itself. In fact, EMU Systems produces their own line of sound cards that far exceed the Audigy in sound quality. However, since they are based on a different chipset and the Audigy line are the only sound cards to use the EMU 10k1 and 10k2, it is recommended that you avoid those two chipsets. The reason is that Sound Blaster cards are not capable of bit-perfect output. That means they cannot take a sound source and send it straight to your speakers without altering it, in this case my resampling. An audio CD is recorded at 16-bits with a sampling rate of 44.1kHz. If you play that CD through an Audigy 2 (for example), the output is actually 16/48 instead of 16/44. The Digital Signal Processing (DSP) chip on the Audigy 2 (and all SB cards) is upsampling the sound to 48kHz, a process to reduces sound quality. Fortunately, there are ways to reduce the effects of Sound Blaster resampling, but if you are considering a new sound card and want the best sound quality for your music, avoid them.
So what is a good chipset? The most common answer is the Envy24 series, produced by a division of VIA formerly known as ICEnsemble. There are variations of this chipset such as the onboard Envy24PT and the budget Envy24HT-S. All produce high-quality sound rivaling home theater equipment costing over a thousand dollars, but the original Envy24 is the best of the series. You can find this chipset in many cards produced by M-Audio, such as the highly popular Revolution 7.1 and the Audiophile 24/96. Terratec also makes Envy24-based cards, and there some cheaper budget cards have emerged with it recently such as the Chaintech 7.1 and the Mad Dog Entertainer. The Envy24 chipset is by no means the absolute best on the market, but it has served as a great entry point to many PC users exploring higher quality PC sound. Many of these cards are still marketed at the average consumer, and so they maintain a pretense of game support. The fact is that none of them are as good for gaming as an Audigy 2 or an nForce2 board with SoundStorm, though some of you may be willing to make that tradeoff for higher-quality music and DVD playback. Alternatively, you could just have two sound cards. As you get into higher-end audio cards, you’ll find that they’re marketed more towards recording artists and sound engineers. These cards have the best sound quality, but some may not have any game support at all.
Here is a list of some high-quality sound cards:
Chaintech AV-710 - This card is based on the Envy24PT. It has comparable sound quality to the M-Audio Revolution, but costs far less. However, expect to spend some effort getting ASIO and Kernel Streaming to work, and finding the best drivers. The ones that come with it are not very good. cost: $30
ESI Juli@ - This card is based on the Envy24HT-S. Interestingly, the PCB is white. Current consensus seems to be that it performs on par with the Audiophile 24/96, but may have superior drivers. Being based on an Envy24HT, you may see a performance hit in games due to the lack of hardware acceleration, but music and DVDs should perform admirably. cost: $179
ESI Waveterminal 192x - Older reviews seem to indicate that this card had some issues that have since been resolved in later revisions. Currently considered a good card for the money. cost: $229
E-MU 1212m - This card has been gaining a lot of popularity lately. It uses the same DACs found in Digidesign’s Pro Tools HD I/O module (pretty much the standard in professional audio recording), and thus has great analog output. Despite being made by the same company that produces the chips for the Sound Blaster cards, the 1212m is quite capable of 44.1kHz output. However, it currently lacks 24/96 support in its WDM drivers. You’ll only be able to get 24-bit playback through ASIO. cost: $200
RME Digi96/8 Pad - One of the best-sounding cards on the market within a reasonable budget. It also has optional balanced XLR outputs, if your equipment supports that. One interesting thing about this card is that ASIO or Kernel Streaming output is not necessary because the card’s drivers intercept any DirectSound or WaveOut stream and send it straight to the card, bypassing Windows. Note that Dolby Digital passthrough is possible on this card, but hasn’t been confirmed with the current drivers, so do some research before buying for DVD purposes. cost: $300
M-Audio Audiophile 24/96 - This is one of the most popular cards for HTPC use due to its high-quality DACs and attractive price. Envy24-based. cost: $159
M-Audio Revolution 7.1 - A cheaper card that sounds almost as good as the Audiophile and has slightly better (though still not very good) game support. However, unless your budget is under $100, you’re probably better off buying the Audiophile. cost: $98
Terratec DMX 6Fire LT - This card is also based on the Envy24. Unlike the other cards, it has DirectSound hardware acceleration for up to 16 channels. The same is true of most other Terratec cards. I haven’t included a link because their server appears to be down. cost: $179
I recommend researching these cards, as well as others by the same manufacturers, before making your final purchase.
Speakers/Headphones
I won’t go too in-depth in this section but rather just provide some tips for speaker or headphone shopping, and some good brands to look at. The most important thing to keep in mind when shopping for PC speakers is that no matter how much you spend, PC speakers will always be inferior in quality to similarly-priced home theater speakers. Depending on who you ask, the best PC speakers on the market are either the Klipsch ProMedia 5.1 Ultras or the Logitech Z-680s, but neither can hold a candle home theater speakers in that same price range. There are also some companies that manufacture high-end PC speaker systems costing $600 or more, but information on them is sparse, probably because everyone just buys real speakers instead.
However, there is a downside to using home theater speakers with your PC. With PC speakers, the analog outputs from your sound card are usually directed through your subwoofer or a decoder box. These devices examine the front channels and remove all of the sound below a certain frequency, called the crossover, then redirect it to your subwoofer (a process called bass management). They do this because games do not separate the bass from the other channels; they just send it all to the sound card, which does not have enough power to properly manage the bass by itself. Home theater speakers don’t do this because they are connected to a receiver that handles this job. Unfortunately, most receivers can only apply bass management to digital or stereo analog inputs. If you’re using 5.1 analog connections from your sound card, your receiver will not manage the bass. That means you either let the sound card do it or you have no bass. Some sound cards, such as the M-Audio Revolution, have decent bass management, but the most popular gaming cards made by Creative do not. If you do decide to purchase home theater speakers for your PC and intend to use 5.1 analog connections, you have a few options. You can buy a 3rd-party bass management solution, such as the Outlaw ICBM, you can use a card that is less effective at gaming but has decent bass management like the Revo, or you can use the SoundStorm APU found on nForce2 motherboards. The nForce2 has pretty horrible analog output, but its digital output is fairly high quality and — more importantly — has the ability to encode your game sound into a Dolby Digital signal in real-time. That means you can use a digital connection to your receiver, which will then apply as much bass management as you want.
With that out of the way, there are a few things to keep in mind when speaker shopping. First, this is not an area where the Japanese dominate. American companies simply have superior speaker design. As such, you should stay way from Sony, Pioneer, Kenwood, or any other Japanese-made speaker. That’s not to say they don’t make good receivers; just don’t buy their speakers. Second, don’t buy Bose. Some people love them and some people hate them, but they are overpriced regardless. The reason they are so popular is simply that they have the best marketing team of any speaker company bar-none. Also, don’t follow their ideaology that tiny cube speakers can produce sound equal to giant floorstanding speakers. There are a lot of great satellite speakers out there (I use some myself), but keep in mind that larger speakers will always have a better frequency response. Lastly, avoid home-theater-in-a-box packages.
As far as headphones go, the best sound quality is going to come from a circumaural open headphone, circumaural meaning it encloses your ear and open meaning it allows air to pass through the ear pieces. However, while open headphones have a wider soundstage, they also leak a lot more sound. It will be easier for people around you to hear what you’re listening to, and harder to ignore outside noise while you’re using your headphones. You may prefer to use closed headphones, which can still provide excellent sound quality with a bit more isolation, but they are not as common as closed headphones. Two companies that consistantly produce high-quality headphones are Grado and Sennheiser. While other companies have a few good headphones in their lineup (like the Sony MDR-V6), you can be assured that pretty much any Sennheiser model is going to be worth your money. Personally, I use the Sennheiser HD-212 Pros, a closed headphone that you can find on Amazon.com for around $45. For more extensive headphone advice, check out Head-Fi.
Step 2: Setting up Your PC
Analog or Digital?
A digital connection is not inherently better than an analog one. While sound transmitted over an optical cable is immune to RF interference, that shouldn’t be an issue if you route your audio cables away from power cables and other electronics. It is also important to remember that a single cable, digital or analog, can only transmit 2 channels of uncompressed sound. To achieve surround sound, the signal must be compressed into a Dolby Digital or DTS stream, or multiple cables must be used. The one exception to this rule is that Creative sound cards and Creative speakers have a special interface that allows them to send uncompressed surround sound over a digital cable. This will not work with speakers or sound cards not made by Creative; you must have both. So if you’ve got your speakers hooked up digitally and you’re getting Dolby Digital from your DVD movies but only stereo sound in games, that’s why.
Of course, there can be other advantages to going digital. Your PC (as well as most audio equipment) processes audio digitally. It reads the source (a CD, MP3, etc.) digitally, applies whatever processing is necessary, then either send it digitally or converts it to analog. Either way, the sound will be converted to analog at some point because that is the only way that speakers can play it. Like everything else, converting from digital to analog or analog to digital will reduce sound quality, so you want it to happen as little as possible. If you send an analog signal to your receiver or decoder box and then tell the receiver to apply some DSP effect like stereo surround or Pro-Logic, it’s going to convert it to digital again to apply that effect, then back to analog to be played by your speakers. As such, you’d be better off just sending it a digital signal in the first place so that your sound card never has to convert it to analog and your receiver never has to convert it to digital.
On that note, you want to make sure that your CD-ROM drive is sending its signal digitally. If you built your own system or tinkered with it any, you probably remember the gray cable running from your CD-ROM drive to your sound card. This is an analog cable, and your CD-ROM drive is reading the CD digitally, converting it to analog, and sending it to your sound. While digital versions of this cable exist, there’s no reason to use them because the sound can be sent digitally over the IDE cable connecting the drive. I recommend just taking the cable out completely unless your CD-ROM drive is so old that it does not support digital audio extraction (DAE). To enable DAE in Windows, go to Device Manager and open up the properties for your CD-ROM drive. Go to the Properties tab and ensure that “Enable digital CD audio for this CD-ROM device” is checked.
Windows Settings
There are also a couple things you can do in Windows to make sure you’re getting the best possible sound quality. First, make sure that the hardware acceleration and sample rate conversion quality are on their max settings. To do this, go to the sound properties in Control Panel and click the Advanced button under Speaker settings. Click the Performance tab and make sure that both sliders are all the way to the right. Another thing you can do is set your master volume in Windows to 75 or 80% and then adjust the volume on your speakers to the desired level. Keeping the master volume below 100% helps prevent clipping, a form of distortion that occurs when the sound you are playing is louder than your system can handle. There are other ways clipping can occur as well, but we’ll get to that later.
Step 3: Checking your Music
I could write a lot in this section, but there’s already a guide by Moguta over in the Behind the Music forum that handles it nicely. What I will cover is ReplayGain. If you have read Moguta’s guide, you know about MP3Gain, a program that normalizes your music so that you don’t have to keep reaching for the volume control everytime the track changes. Used correctly, it also helps prevent your music from clipping. It is a good idea to use this program on any music you intend to burn to CD, transfer to a portable player, or play in any form outside of Winamp or Foobar. Inside of Winamp and Foobar, it is unnecessary if you use the recommended settings. Foobar and Winamp with the mpg123 plugin have support for ReplayGain, a standard made by the developers of MP3Gain which accomplishes the same objective by adding meta tags to your MP3s instead of changing the audio itself. The downside is that the music is only normalized on ReplayGain-compatible players, but the sound itself is never touched, making the process easily reversible. In Winamp, you need only follow the instructions in the following paragraphs to take advantage of ReplayGain. Foobar users will need to run a scan on each file the first time it is played. You can simply select the entire playlist, right-click, and click “scan selection as album” under the ReplayGain sub-menu. Of course, if the files in question are not part of the same album, do a per-file track gain scan instead.
ReplayGain will work on any format. MP3Gain is obviously for MP3s only, but there is also VorbisGain for Ogg Vorbis music. Note that if you decide to use MP3Gain or VorbisGain after scanning a file with ReplayGain, you should erase the ReplayGain data first, as it can cause volume issues. You can safely rescan it afterwards if you wish.
Step 4: Setting up your Software
What Media Player to Use
The most important feature of a music player is that it must give you as much control as possible. You need to have extensive input and output options for multiple formats. That narrows it down quite rapidly. The two most popular players are Winamp (obviously) and Foobar2000. Winamp has the larger community — and thus more plugins — as well as a more user-friendly interface that can be readily changed with skins. You probably already know this. If you don’t, that may mean you’re using Windows Media Player for your MP3s, in which case you really need to pay attention to this part. Generally speaking, you will get better sound quality from Foobar 2000. I will go into why in the next section. However, while FB2k’s interface is very efficient, it doesn’t have built-in skinning like Winamp. Skinning support is available via plugins, but there aren’t nearly as many available as there are for Winamp. For the purposes of this thread, I recommend FB2k, but I will list optimization settings for both players.
Which Decoder to Use
There two components of a MP3 player that affects its playback quality: the input component that decodes the file and the output component that sends it to your speakers. With less-popular or lossless formats, you won’t have much of a choice as to which decoder to use. There will probably only be one. However, there are quite a few MP3 decoders available. Winamp utilizes code by Fraunhoffer, the company responsible for the MP3 format. While you might think that the people who created the format would have the best decoder, we already know that they do not have the best encoder (that would be LAME). By contrast, Foobar2000 uses an open-source library called mpg123. There is also a popular 3rd-party decoder for both players known as MAD. Of course, if you’re not playing an MP3, it won’t make a bit of difference.
Foobar2000 users should stick with the default decoder. MAD is available if you prefer it, but any quality differences that might exist are indiscernable. Winamp users should take a hint from the Foobar users and download the mpg123 decoder for Winamp. The two most important features of this plugin are its support for ReplayGain and 24-bit output, both of which the default decoder lack. ReplayGain is a version of MP3Gain that losslessly adjusts the volume of your MP3s to prevent them from becoming louder than your system can handle, which clips off some of the sound. The difference is that ReplayGain stores the adjustments in metatags similar to ID3, as opposed to altering the sound itself. 24-bit decoding improves sound quality by providing a larger dynamic range over which to spread distortion caused by rounding up an MP3 with a floating-point bit value to 16-bits. That distortion can be pushed into inaudible frequencies, providing a cleaner sound.
Output Methods and KMixer
All output plugins — excepting burning or transcoding plugins — use one of 4 output methods: WaveOut, DirectSound, ASIO, or Kernel Streaming. Ideally, you should stick with one of the latter two. ASIO is an API for bypassing Windows and communicating directly with your sound card. Kernel Streaming does the same, but has far less support and might be considered more of a hack. The reason it is important to bypass Windows is because of KMixer, a system process that handles all audio output in Windows. KMixer — much like Creative sound cards — has the nasty habit of resampling all audio to 48kHz. This is supposedly because DSP effects are easier to apply at 48kHz, and most game sound effects and movies are recorded at 48kHz. However, you don’t want your 44.1kHz music being resampled, so use ASIO or Kernel Streaming unless your sound card resamples anyway, in which case the difference between ASIO/KS and DirectSound/WaveOut may not be noticeable to you. If ASIO is supported by your sound card (most modern ones do), you’ll need the appropriate output plugin for your player. Winamp users should download the Winamp5 ASIO output plugin and Foobar users should download the latest ASIO plugin for Foobar2000. If your card does not support ASIO or the drivers suck, you can try Kernel-Streaming. One of the downsides to Kernel Streaming is that it will tie up your sound card. No other applications will play sounds while you use it. For Winamp, download the Kernel Streaming Output Plugin. For Foobar 2000, just be sure to check it during the installation process, as it is included. Note that you will receive errors if you set the output bit-depth to something your card does not support.
Note: In the Sound Blaster Audigy 2 series, only the Audigy 2 ZX and ZX Platinum support 24-bit ASIO output. For the original Audigy 2, Audigy 2 NX, or Audigy 2 LS, you should use Kernel Streaming output in Foobar. The Winamp KS plugin lacks resampling, so stick with DirectSound w/ SSRC if you use Winamp.
If you cannot get either of these output methods to work, WaveOut should be used for Win95/98 systems while DirectSound will provide better compatibility on Me/2k/XP systems. There is no difference in sound quality between the two. However, if you own a sound card that resamples internally (including all Creative sound cards), you can benefit from using an updated WaveOut or DirectSound output plugin that can do a better job of resampling than your hardware. WinMe/2k/XP users should download DirectSound w/ SSRC while Win95/98 users should grab WaveOut w/ SSRC.
Winamp Settings
You should already have the Shibatch mpg123 plug-in, so open up its properties. Under the decoder tab, set the output format to 24bits if you have a 24-bit sound card. Otherwise, leave it at 16. Go to the Replay gain tab and ensure that “Enable” and “Hard Limiter” are checked, and that the drop-down box is set to Album gain. If you find that your music is playing too softly, you can boost it a little using the Preamp slider on this tab. However, if the music is going to clip, the hard limiter will kick in and prevent the volume from going too high.
Now you need to adjust the settings of your output plugin. In most cases, all you will have to do is enable and set resampling to 48KHz if you have a card that resamples. In the case of DirectSound or WaveOut w/ SSRC, set the bits-per-sample to 24 if you have a 24-bit card (again, only if your card resamples). Uncheck fast mode unless you find the increased CPU usage not to your liking. Set dithering to triangular spectral shape and leave noise distribution on its default setting of triangular.
If your card does not resample, you don’t need to adjust any of these settings. The other options such as hardware acceleration should only be experimented with if you are having issues such as stuttering sound. They are usually better left disabled.
Foobar 2000 Settings
For Foobar 2000, I recommend kernel-streaming output for the same reason I recommend ASIO: it bypasses KMixer. You don’t have to download any plugins; it comes with the main installation. Just make sure you select it, or it won’t install. If you use 2 sound cards, be sure to select your music card as the output device under the kernel-streaming settings. Under playback, you may need to fool around a little depending on what card you have. Personally, I use 24-bit fixed-point decoding padded to 32-bit with dithering disabled on an M-Audio Revolution 7.1. The Revolution 7.1 is a 24-bit card, but padding the output to 32-bits makes the job easier on the CPU (which processes data in 32-bit chunks) without degrading sound quality. For an Audigy or Audigy 2, I recommend 16-bit with dithering enabled. Dithering allows you to apply the amplifier and other DSPs if you so choose without noticeably affecting sound quality. However, 24 padded to 32 will generate kernel-streaming errors on some PCs, so you can also try plain 24-bit to see if it works for you. All other settings are best at their defaults, so leave them as they are.
Notes, FAQ and Resources
Most of the tips here are based on established consensus in the PC audiophile community. The utmost precaution is taken to ensure that sound quality testing is as unbiased and scientific as possible. The “placebo effect” (as commonly known in the medical field) also applies here, and a quality test is not objective if the participant has prior knowledge of variables used. If you intend to do your own tests on these or other adjustments to your audio setup, be sure to perform them “blind.” Have someone else play the variable music against the original music for you, so that you do not know which is being played. You can also perform an ABX test, which will analyze your new file against a reference file and determine what changes have been made in the sound. These are standard methods of trying out new sound improvements, and you should never claim something improves audio quality without at least doing a blind listening test, if not an ABX test. Foobar2000 now includes a built-in ABX component, making this easier than ever. Simply queue up the two files you would like to compare, do a ReplayGain scan on both of them, then select them, right-click, and click “ABX two tracks.”
FAQ
Q: Does my (insert onboard sound here) resample?
A: Most likely, yes. There are a few motherboards that do not (those based on the Envy24PT chipset). I would compile a list, but then you could just look on your motherboard for a chip with the Envy24 logo. Of course, this is only on motherboards with VIA chipsets.
Q: I thought (insert Creative card) didn’t resample?
A: Despite what you may have heard, all Creative sound cards resample in some form or another. Supposedly the Audigy 2 NX does not resample in hardware, but does so instead through the drivers that send the audio over USB. All other cards will resample in hardware, even if you disable all EAX effects, CMSS, etc. Even if it appears the card is giving you 44.1KHz output (such as the digital out on an Audigy 1), it is in fact upsampling to 48KHz and then back down to 44.1. The easiest way to test this is to try to play a DTS-CD, as the resampling destroys the DTS encoding and you’ll only hear static. Bit-perfect cards with ASIO or kernel streaming output can play them.
Q: Does my sound card support 24-bit playback?
A: All of the cards listed previously in this guide support 24-bit playback, in addition to all PCI M-Audio cards, most Terratec cards, and many others. The Audigy 1, despite its advertising, does not support 24-bit playback. The Audigy 2 does. The nForce SoundStorm APU does not support 24-bit playback. For any other cards, ask in this thread or check the specs on the manufacturer’s website.
Q: What’s the big deal with SoundStorm? I have it and it sounds like shit.
A: You’re probably using the analog outputs. The SoundStorm specification calls for quality parts, but motherboard manufacturers — ever looking to cut costs — use the all-too-common and all-too-crappy Realtek ALC650 to handle the analog output on their nForce2 motherboards. If you use the digital output, the sound comes straight from the APU and doesn’t go through the Realtek codec.
Q: Does resampling to 96KHz improve sound quality?
A: No. Resampling never improves sound quality for the same reason that reencoding a 128kbps MP3 to LAME APX doesn’t improve sound quality. The original analog waveform has already been sampled a certain number of times in its conversion to digital form. Resampling does not magically add more samples from the original wave, it just guesses. The only reason to ever resample to 96KHz is if you have an Audigy 2 card and cannot use Kernel Streaming output. In the other output modes, the Audigy 2 only supports 24-bit playback at 96KHz.
More questions as I think of them.
Resources
Here are some sites where you can acquire information and tips on high-quality computer audio:
http://www.hydrogenaudio.org
http://www.3dsoundsurge.com
http://www.3dss-forums.com
http://www.hardforums.com/forumdisp…p?s=&forumid=33
http://www.avsforum.com/avs-vb/foru…p?s=&forumid=26
http://www4.head-fi.org/forums/forumdisplay.php?f=59
http://www.nforcershq.com/forum/vie…32876f61d08d479 - for nForce SoundStorm users
Hope this works!!
