Here are the important characteristics of sound cards:
Discrete sound cards have been made in ISA and PCI versions, but ISA cards are no longer available. All recent sound adapters, embedded or standalone, use PCI. The much smaller bandwidth of ISA limited ISA cards in many respects, including generally requiring that wavetable data be stored locally, placing an upper limit of about 16 on simultaneous sound streams, and making effective 3D audio support impossible. If you encounter an ISA sound card when stripping an older system for spares, pitch the ISA card. It's not worth keeping.
FM synthesis is no longer used in current sound cards. All current midrange sound cards use wavetable synthesis, and some expensive sound cards use partial waveguide synthesis. The quality and features of wavetable synthesis vary depending on both the processor and the quality and size of the wavetable samples, with more-expensive cards producing better synthesis, as you might expect.
Each MIDI interface supports 16 channels, each corresponding to one instrument. Low-end sound cards have one MIDI interface, allowing up to 16 instruments to play simultaneously. Midrange and high-end sound cards usually have dual MIDI interfaces, allowing up to 32 simultaneous instruments. Some high-end sound cards, such as the Creative Labs Sound Blaster Live! Platinum, use a triple MIDI interface, which allows up to 48 simultaneous instruments. In general, 16-channel cards are suitable for most uses, 32-channel cards are useful for playing MIDI instrumentals realistically, and 48-channel cards are necessary only for the most complex MIDI environments. It's worth noting that Creative replaced its flagship 48-channel Sound Blaster Live! Platinum with the 32-channel Audigy 2.
Polyphony refers to the ability of a sound card to generate multiple simultaneous voices when playing MIDI. A voice corresponds to one note generated by one instrument. Do not confuse number of voices with number of channels. The 16 channels of a standard MIDI interface allow 16 instruments to play simultaneously. However, some instruments require multiple voices. For example, a piano occupies one MIDI channel, but if the musician is playing a single-note melody with one hand and three-note chords as accompaniment with the other hand, that channel requires four voices. A large number of voices is important for reproducing complex MIDI scores accurately. Voices may be hardware-based or software-based, and some sound cards use both types. A basic sound card might support 64-voice polyphony, 32 in hardware and 32 in software. High-end sound cards support 64 or more hardware voices, and may add software voices for a total of 256 to 1024 voices.
The range of human hearing is usually stated as 20 Hz to 20 kHz. All current sound cards nominally support this range or close to it, which is in fact required for PC99 compliance. However, few cards state ± dB for that range, which specifies how flat the frequency response curve is. A good card may have frequency response of 20 Hz to 20 kHz at 3 dB down. A professional-level card may have frequency response of 20 Hz to 20 kHz at 1 dB down. Inexpensive cards may claim frequency response of 20 Hz to 20 kHz, but that range may turn out to be stated at 10 dB down or some similarly absurd number, which in effect means that actual usable frequency response may be something like 100 Hz to 10 kHz.
All current sound cards support waveform audio playback at 44,100, 22,050, 11,025, and 8,000 Hz. Many also support various intermediate playback rates and the DAT-standard 48,000 Hz. Some cards record only at 44,100 Hz, although most also offer other standard rates.
Signal-to-Noise ratio (S/N ratio), stated in dB, measures the amount of signal (data) relative to noise, with higher numbers indicating better performance. A low S/N ratio translates to audible hiss. The best sound cards have 95 dB or greater S/N for analog audio; midrange cards about 90 dB; and inexpensive cards may have 85 dB or less. It's not unusual for a card to have somewhat lower S/N ratio for digital recording and digital playback. For example, an excellent consumer-grade sound card may specify an S/N ratio of 96 dB FS A-weighted for analog audio, 93 dB FS A-weighted for digital recording, and 90 dB FS A-weighted for digital playback. In a typical PC environment, noise level (both ambient external audible noise and the electrically noisy inside of the PC) and the typical use of low-quality speakers or headphones make it unlikely that anyone could differentiate between cards with S/N ratios of 80 dB or higher if that were the only difference. However, cards with higher S/N ratios are generally better shielded and use better components, which translates to better sound and less hiss.
Half-duplex sound cards can either play sound or record sound, but not both at the same time. Full-duplex sound cards do both simultaneously. For simple tasks?listening to CDs or playing games?a half-duplex card is adequate. More advanced audio functions, such as Internet telephony and voice recognition, require a full-duplex card. Most midrange and all high-end sound cards are full-duplex.
In the past, software wrote directly to the sound card. That meant that compatibility with proprietary standards?initially AdLib and later Sound Blaster?was important because if your game or application didn't explicitly support your sound card, you simply couldn't use sound with that software. Microsoft took the initiative away from sound card manufacturers by incorporating standard sound APIs into Windows. Here are the standards you should be aware of:
Sound Blaster compatibility, formerly a sine qua non for any sound card, is now largely immaterial except to those who still use DOS software, including DOS games. True Sound Blaster compatibility requires fixed IRQ, I/O port, and DMA assignments, whereas PCI cards are assigned resources dynamically. Within those constraints, all Creative Labs sound cards and most competing cards boast (nearly) full Sound Blaster compatibility. If you still use DOS applications, though, it's worth verifying whether real-mode drivers are available for a sound card before you purchase it.
Microsoft DirectSound (DS) is a component of DirectX. Developers can write to the DS API, rather than to the underlying hardware, with the assurance that their software will function with any DS-compatible sound card. DS compatibility has replaced Sound Blaster compatibility as an absolute requirement for any sound card.
Microsoft DirectSound3D (DS3D) is an extension to DS that supports 3D positional audio, which is a technology that manipulates sound information to extend stereo imaging to full surround sound, allowing sounds to appear to come from any position around you. For example, when you're playing an air combat game and your missile hits a bandit in front of you, the sound of that explosion comes from the front. But if you didn't notice his wingman on your six, the sound of his missile blowing off your tail comes from behind. The realism of DS3D imaging in any given situation depends on the means used to reproduce the sound (two speakers, four speakers, or headphones) and the hardware capabilities of the sound card. But whatever the physical environment, DS3D provides noticeably better imaging than older 2D technologies. If you intend to use DS3D-enabled software, it's important to have hardware support for DS3D in your sound card because DS3D positional effects that cannot be processed in hardware are processed by the main CPU, which can bog down system performance.
Although Aureal went bankrupt in spring 2000, many cards with Aureal chipsets remain in use, and such cards were still available new as recently as late 2001. Aureal A3D is a proprietary 3D positional audio standard that is available only on sound cards based on the Aureal Vortex and Vortex2 chipsets, which have been made by Voyetra/Turtle Beach, Diamond Multimedia, Aureal itself, and others. A3D is available in two versions. A3D2.0 is supported only by the Vortex2 chipset, whereas the earlier and less-capable A3D1.0 is supported by both the Vortex and Vortex2 chipsets. A3D1.0 provides realistic 3D imaging even on dual-speaker systems or headphones. A3D2.0 provides extraordinary 3D effects, particularly on quad-speaker systems. A3D achieved broad support from game software manufacturers. For software without A3D support, A3D hardware drops back to using DS3D.
Creative Labs EAX (Environmental Audio Extensions) is basically a proprietary Creative Labs extension to DirectSound3D. EAX 1.0 is technically less ambitious than A3D2.0, but provides reasonable 3D imaging. EAX 2.0 and EAX Advanced HD Multi-Environment are significant enhancements that match A3D2.0 in most respects and exceed it in many. Given the dominance of Creative Labs, the various flavors of EAX are widely supported by game software.
Midrange and high-end sound cards have an onboard DSP, which is a general-purpose CPU optimized for processing digital signals, such as audio. In 2D mode, the DSP provides enhanced audio effects such as chorus, reverb, and distortion. In 3D mode, it processes 3D-positional audio (e.g., DirectSound3D or EAX) algorithms locally, removing that burden from the main CPU. Inexpensive sound cards use the host CPU, which reduces performance significantly, particularly during complex operations such as 3D rendering. Any accelerated sound card should accelerate 32 or more DS and DS3D sound streams in hardware.