Covox Speech Thing
The Covox Speech Thing is an external 8-bit digital-to-analog converter (DAC) designed as a simple sound output device for IBM PC-compatible computers, connecting directly to the parallel printer port to enable playback of digitized audio samples such as speech or sound effects. Released on December 18, 1987, by Covox, Inc., based in Eugene, Oregon, it was marketed as an affordable audio solution priced at around $70–$80, filling a gap in early personal computing where built-in sound capabilities were limited to basic beeps from the system timer.[1] The device operates by receiving 8-bit mono digital data through the parallel port's data lines, which are fed into a discrete resistor ladder network—a weighted array of resistors connected to each of the port's 8 data pins—to generate corresponding analog voltage levels representing the audio waveform.[2][3] This analog output is then amplified externally and played through speakers, with the sampling rate determined entirely by software timing and constrained by the parallel port's data transfer speed, typically achieving rates up to around 30 kHz but falling short of CD-quality 44.1 kHz.[3] Lacking onboard processing, mixing, or MIDI support, it relied on CPU-driven software for all audio generation and offered no input capabilities, making it a rudimentary yet innovative tool for the era.[1] Introduced during the mid-1980s surge in PC gaming and multimedia experimentation, the Covox Speech Thing found niche adoption in around 60 DOS-era titles, including adventure games like King's Quest and utilities for speech synthesis, where it provided clearer audio than PC speaker emulation and some software detected it similarly to the Disney Sound Source, enabling partial compatibility though with limitations due to hardware differences.[3] Its low cost and ease of integration—requiring no expansion slot—made it popular among budget-conscious users and hobbyists, who often built DIY clones using the published schematic.[2] By the early 1990s, it was overshadowed by full-featured ISA sound cards like the AdLib and Sound Blaster, but its legacy endures in retro computing communities, with modern reproductions and emulations preserving its role as a pioneering step toward accessible PC audio.[1]History
Invention and Patent
SRT, Inc., doing business as Covox, was founded in 1975 as a research and development firm specializing in speech technology, with headquarters in Eugene, Oregon. The company initially concentrated on developing products for speech recognition and synthesis, building on SRT's earlier work dating back to 1975 under the Covox dba. This foundation in audio processing positioned Covox to address the growing demand for accessible sound capabilities in personal computing. The Covox Speech Thing emerged from this expertise as a pioneering external audio device, invented by John L. Stewart and Bradley C. Stewart. The invention date aligns with the patent application's filing on October 2, 1987, which was granted on March 14, 1989, as U.S. Patent 4,812,847. Titled "Parallel Port Pass-Through Digital to Analog Converter," the patent describes a simple yet innovative design that connects directly to the IBM PC's parallel printer port, enabling digital-to-analog conversion without interrupting printer functionality or requiring external power. At its core, the device employs an R-2R resistor ladder network to convert 8-bit digital signals into analog audio output, supporting up to 256 amplitude levels for speech and sound reproduction.[4] The primary motivation behind the invention was to provide an economical solution for audio output on early IBM PCs, which lacked integrated sound hardware and relied on expensive add-on cards for such features. By leveraging the existing parallel port, the Speech Thing allowed users to generate speech synthesis, music, and other digitized sounds affordably, filling a critical gap in the platform's multimedia potential during the late 1980s. This approach not only minimized hardware complexity but also ensured broad compatibility with standard PC configurations.[4]Release and Market Impact
The Covox Speech Thing was released on December 18, 1987, by Covox, Inc., based in Eugene, Oregon.[1] At launch, it retailed for approximately US$70 (equivalent to about $200 in 2025 dollars after adjusting for inflation).[3][5] In the late 1980s PC market, the device provided a low-cost entry into digital audio output, contrasting sharply with pricier options like the Sound Blaster card or IBM's $1,200 Voice Communications Adapter from 1985.[1][6] It connected directly to the parallel printer port, requiring no internal installation, which appealed to budget-conscious users and hobbyists seeking to enhance the standard PC speaker's limited capabilities. The product shipped with speech synthesis demonstration software and utilities for basic digital sound playback, facilitating immediate experimentation with voice output and simple audio effects.[3] Its affordability and ease of use propelled adoption in the demoscene, where enthusiasts created music and demos well into the 1990s, even as dedicated sound cards became more common.[7] By enabling 8-bit digitized sound for early PC games and multimedia applications—supported by around 60 DOS titles—the Speech Thing marked a pivotal step toward widespread personal computing audio, laying groundwork for parallel port-based sound solutions and influencing subsequent hardware designs.[3][1]Technical Specifications
Hardware Design
The Covox Speech Thing interfaces directly with the IBM PC's parallel printer port (LPT) through a standard 25-pin DB-25 connector, utilizing the eight data pins (pins 2–9) to receive 8-bit digital audio samples while incorporating pass-through wiring to maintain compatibility with attached printers.[4] This design leverages the port's TTL-level signals without requiring additional drivers or interrupts, enabling straightforward integration into existing systems.[4] The core of the device is a passive R-2R resistor ladder network that performs digital-to-analog conversion, configured with eight 200 kΩ resistors (R1–R8) connected to the data lines and seven 100 kΩ resistors (R9–R15) forming the ladder segments, terminated by a 15 kΩ load resistor (R16) to produce a linear voltage output proportional to the input byte value.[4] This network exploits the parallel port's inherent pull-up resistors (approximately 4.7 kΩ each) to source current, ensuring a high-impedance output that minimizes loading on the port while achieving 8-bit resolution without active components.[4] Complementing the DAC, the circuitry includes a 5 nF capacitor (C1) connected in parallel with the load resistor, creating a simple RC low-pass filter with a cutoff frequency of approximately 3 kHz to attenuate high-frequency artifacts and reduce aliasing from the digital sampling process.[4] The analog output emerges at microphone level, routed to a high-impedance mono 3.5 mm phone jack for connection to external speakers or an amplifier, as the device provides no onboard amplification.[4][8] Power for the Speech Thing is derived entirely from the parallel port's +5 V supply via the data lines and strobe signal, eliminating the need for an external power source and contributing to its minimalist architecture.[4] Housed in a compact external enclosure resembling a DB-25 gender changer extension, the unit emphasizes portability and ease of attachment without internal PC installation.[4]Audio Capabilities and Limitations
The Covox Speech Thing functions as an 8-bit mono digital-to-analog converter (DAC), supporting unsigned 8-bit pulse-code modulation (PCM) waveforms (values 0–255) for audio output. It employs an R-2R resistor ladder network to convert digital signals from the parallel port into analog voltages, producing 256 distinct amplitude levels.[9] This design enables straightforward playback of sampled audio, such as speech or simple sound effects, but relies entirely on software to generate and transmit data bytes sequentially via port polling. Software often detects it as compatible with the Disney Sound Source for broader application support.[3] Sampling rates are not hardware-fixed and depend on the host CPU's speed and the efficiency of the playback software, which must time data output precisely to avoid jitter. Sampling rates are determined by software timing and CPU speed, typically achieving 30–40 kHz on 80286 and faster systems but falling short of CD-quality 44.1 kHz, with no onboard buffering or direct memory access (DMA) to assist.[3] Audio quality is influenced by the precision of the resistor ladder, where 1% tolerance components are essential for linear response and minimal distortion; lower-quality clones often exhibit graininess from imprecise values. Noise can arise from parallel port interference, such as crosstalk from printer signals, though the original design minimizes this through high-impedance resistors (e.g., 200 kΩ for the primary ladder). A built-in low-pass filter, implemented with a 0.005 µF capacitor, attenuates frequencies above approximately 3 kHz to smooth output but may introduce minor phase distortion in higher-range audio.[9][8][10] Key limitations include the absence of hardware mixing, effects processing, or multichannel support, requiring the CPU to handle all waveform generation and sequential sample delivery, which burdens slower systems. Playback is strictly linear without interrupt-driven capabilities, limiting real-time applications. The device outputs at microphone level (around 0.3–1 V), necessitating external amplification for audible volume, and includes a pass-through connector to maintain printer functionality without disconnection.[9][8][3]Commercial Products
Original Covox Speech Thing
The original Covox Speech Thing, released in late 1987, was an external digital-to-analog converter (DAC) designed as a simple add-on for IBM PC and compatible computers. It consisted of a compact box that connected directly to the parallel printer port via a dedicated cable, providing 8-bit mono audio output through a mini jack for connection to speakers or headphones. Bundled with the hardware was speech synthesis software distributed on a 5.25-inch floppy disk, including the "Smooth Talker" text-to-speech synthesizer, a talking calculator, a talking calendar, a game demo, a music sampler, a sound editor, and a special-effects panel, along with multiple pre-recorded vocabulary demos for immediate voice synthesis testing.[11][1][3] Accessories provided with the device included a printer pass-through cable to maintain compatibility with existing printers without interrupting parallel port functionality, as well as an included basic audio amplifier featuring a built-in speaker and headphone jack for enhanced playback. Priced at approximately $70, the Speech Thing emphasized ease of installation, requiring no internal modifications to the computer system.[11][3] Targeted primarily at home users and developers interested in affordable audio output for applications like speech generation and basic sound effects, the device offered plug-and-play simplicity without the need for interrupt requests (IRQ) or direct memory access (DMA) configurations typically required by expansion-slot hardware. Its production was limited to the late 1980s, with sales tapering off and eventual discontinuation as dedicated internal sound cards, such as the Sound Blaster, became more accessible and feature-rich by the early 1990s. The unique selling point lay in its non-intrusive design, allowing users to add digital audio capabilities to any PC with a parallel port, bypassing the commitment of an ISA slot.[11][1][3]Licensed Variants and Clones
The Disney Sound Source, introduced in 1990 by Disney Interactive, was an officially licensed adaptation of the Covox Speech Thing, aimed at enhancing audio in PC games through a more user-friendly design. Priced at approximately $15, it connected to the parallel port and included a 16-byte FIFO buffer to enable smoother playback by reducing timing dependencies on the CPU, supporting a maximum sample rate of 7 kHz for 8-bit mono audio. Powered by a 9V battery, the device featured a built-in speaker and integrated amplification, allowing direct audio output without external components, which distinguished it from the original Covox model by lowering CPU overhead during playback.[12] The FTL Sound Adapter, produced by Faster Than Light Games in the late 1980s, represented another licensed variant tailored for PC compatibility with Amiga-style audio in titles like Dungeon Master. Bundled with select game releases, it employed a similar FIFO buffering mechanism to the Disney Sound Source, facilitating reliable digital audio output via the parallel port while incorporating a DB9 joystick port for additional input functionality. Its design emphasized seamless integration with game software, using a resistor ladder DAC and basic filtering components to produce digitized sound effects and music, thereby extending Covox-like capabilities to specific software ecosystems.[13] Beyond these official products, several clones emerged, including the SiliconSoft SoundJr from Silicon Shack in the late 1980s, a cost-effective parallel port DAC that replicated the core resistor ladder architecture of the Covox Speech Thing for budget-conscious users.[14] Key enhancements in many clones, such as added FIFO buffers and onboard amplification, addressed the original's high CPU demands, enabling broader adoption in hobbyist and entry-level setups. DIY variants proliferated through schematics published in contemporary hobbyist publications, allowing enthusiasts to assemble parallel port-based DACs using readily available resistors and capacitors for custom audio experiments.[15]Software Compatibility
Games and Applications
The Covox Speech Thing was integrated into software applications by directly writing 8-bit audio samples to the parallel port's data register, with programs polling the port to control playback timing and achieve sample rates up to approximately 13 kHz on typical 1980s hardware. This method allowed for simple digital audio output without requiring interrupts or DMA, making it suitable as a fallback option alongside FM synthesis devices like the AdLib or MIDI interfaces.[16][2] Numerous commercial video games from the late 1980s and early 1990s provided native support for the Covox Speech Thing, primarily for digitized sound effects and voice samples, enabling enhanced audio on systems without internal sound cards. Sierra On-Line's adventure game series, including titles like Space Quest III: The Pirates of Pestulon (1989), utilized the device for environmental sounds and dialogue playback, often selectable via the game's installation program or a dedicated Covox driver. These implementations highlighted the device's role in budget gaming setups, where it supplemented PC speaker beeps or AdLib music.[17][18] Utility applications leveraged the Covox for basic audio experimentation and testing, including early WAV file playback tools and parallel port diagnostics. Covox Inc. distributed a demo speech disk with pre-recorded vocabularies and sample playback software, allowing users to stream simple digitized sounds or text-to-speech phrases via the parallel port. Diagnostic utilities, such as those bundled with hardware testing suites, used the device to verify parallel port functionality by outputting tonal patterns or noise tests, confirming data line integrity before game installations. These tools were essential for troubleshooting on entry-level PCs, where the Covox served as an accessible entry point to digital audio.[19][20] The Covox Speech Thing gained popularity from 1988 to 1992 among users of affordable IBM PC compatibles, providing a low-cost ($70–$80) solution for digitized audio in an era when dedicated sound cards like the Sound Blaster cost over $200 and were uncommon. It appealed to gamers and hobbyists on systems without expansion slots, filling a gap until cheaper ISA sound options proliferated, after which direct support waned.[6][21]Music Trackers and Demoscene
The Covox Speech Thing found significant adoption in the demoscene and among users of music tracking software during the early 1990s, where its ability to output 8-bit sampled audio via the parallel port enabled affordable digital sound production on budget PC systems. Popular DOS-based trackers integrated direct support for the device, allowing composers to playback modules with digitized samples. For instance, Modplay, released in 1990 by Mark J. Cox, included Covox compatibility and was even customized for distribution with Covox sound cards, facilitating playback of Amiga-style MOD files on PC hardware.[22] Similarly, Inertia Player (1995) supported mono and stereo-on-1 Covox output modes, targeting demoscene musicians with optimized assembly code for 386+ systems.[23] Later trackers expanded this integration for more advanced composition. FastTracker 2 (1994) and Scream Tracker 2.3 (1994) provided native 8-bit sample playback through the Covox, with Scream Tracker's early versions specifically accommodating the device's DAC for monophonic output alongside PC speaker support. Impulse Tracker (1995) further refined this with configurable Covox modes, including dithering options to enhance audio quality during module rendering. These tools democratized sampled music creation, as the Covox required no dedicated sound card, making it ideal for hobbyists and demo groups constrained by hardware costs. In the demoscene, the Covox played a key role in producing chiptune and sampled audio for size-limited competitions, such as 4k and 64k demos, where its low expense and parallel port simplicity outperformed the internal PC speaker for fidelity. Early PC demos by groups like Future Crew leveraged such peripherals for immersive soundtracks, contributing to the scene's growth on MS-DOS platforms in the late 1980s and early 1990s.[24][25] Composers adapted techniques like sample rate conversion to the device's practical 12 kHz limit—achieved by timing byte outputs at approximately one per 83 microseconds—and applied dithering to reduce quantization noise in MOD files, preserving dynamic range in low-bit-depth playback.[8] The demoscene community amplified the Covox's longevity through shared resources, including TSR drivers and utilities that enabled seamless compatibility with Amiga MOD formats, allowing PC users to remix and playback tracker modules originally designed for the Amiga's Paula chip. This grassroots support sustained the device's relevance into the mid-1990s, even as dedicated sound cards like the Sound Blaster became more accessible. Despite these advantages, limitations persisted: real-time playback demanded intensive CPU cycles for bit-banging data to the parallel port, often consuming nearly 100% of processing time on 386-era systems and necessitating combinations with the PC speaker for multichannel effects or beeps.[22][8]Emulation and Operating System Support
Emulation of the Covox Speech Thing is available in several DOS emulators, enabling compatibility with legacy games and applications that originally supported the device. DOSBox, first released in 2002, includes support for Covox Speech Thing and Disney Sound Source modes through its parallel port emulation, allowing users to configure the device via settings likedisney=true for audio output in compatible software.[26] Similarly, DOSBox-X provides detailed emulation of the Disney Sound Source, which maintains backward compatibility with the Covox Speech Thing by simulating the parallel port's data transfer for 8-bit digital audio. DOSBox Staging further enhances this with explicit Covox Speech Thing and Disney Sound Source support, alongside related devices like the Stereo-on-1 DAC, to improve accuracy in sound reproduction.[27]
Driver support for the Covox Speech Thing primarily exists in MS-DOS environments through custom terminate-and-stay-resident (TSR) programs and utilities that handle ADPCM decoding and parallel port output. For instance, STDRIVER.SYS can be loaded in CONFIG.SYS to initialize the device for games requiring direct hardware access, while utilities like those in the Covox Speech Thing software package provide TSR-based playback for sampled audio.[28] These drivers facilitate integration with DOS applications by managing the low-level port I/O necessary for real-time sound generation.[29]
Under Windows 3.1, partial support is achieved via dedicated drivers that leverage parallel port APIs to route audio output, though performance is limited by the era's resource constraints and lacks full DMA capabilities.[30] The driver package for Windows 3.1 and later multimedia versions enables basic playback of WAV files and simple sound effects through the LPT port, but requires manual configuration and is not optimized for multitasking environments.[29]
Support in other operating systems is more limited, with Linux offering experimental drivers like pcsndrv for kernel 2.x that treat the device as a mono DAC via the /dev/lp interface, allowing basic audio output but without advanced features like compression handling. Modern Unix-like systems generally lack native drivers, relying instead on emulation layers for compatibility. No built-in support exists in contemporary operating systems without additional software bridges or virtual parallel ports.
Software such as virtual sound emulators can configure the Covox Speech Thing to mimic outputs from other cards, including Sound Blaster for PCM playback or AdLib for FM synthesis, by intercepting and redirecting audio streams through the emulated DAC. This approach, seen in tools like OPL2LPT adaptations, extends compatibility to software expecting those interfaces while using the Covox's simpler hardware simulation.[31]
Emulation challenges include maintaining precise timing for high sample rates, as the original device's direct port writes demand consistent CPU cycles to avoid distortion, a issue exacerbated in virtual environments with variable host performance. For Disney Sound Source variants, accurately simulating the 16-byte FIFO buffer is critical for flow control and autodetection, preventing buffer underruns that degrade audio quality in real-time applications.[32]