Texas Instruments Speech - Integrated Circuits

Shortly after the invention of the synthesizer technology to reproduce human speech with tuned voices stored in ROM's (integrated circuits) this funny product appeared. The Speak & Spell project was started in the year 1976 and created with the TMC0280 the first one-chip LPC speech synthesizer. Later refinements to the Speak & Spell chips resulted in the TMS5100, 5110, 5200, and 5220 Voice Synthesis Processors (VSP) for use in commercial products needing synthetic speech voice output from digitally-stored words and phrases. Speech data was stored in in up to sixteen 128K ROM chips (TMC0350). Plug-in modules increased vocabulary and provided for versions in French, German, Spanish and British English. Texas Instruments continued the integration of the synthesizer technology and created dozens of VSP's and associated Voice Synthesis Memories (VSM's).

It's difficult to get information about the customized speech circuits manufactured by Texas Instruments. The following table gives an overview of the known circuits, a brief description and the products using them. 

First Voice Synthesis Processors (VSP)

Listen some words from the Speak & Spell and read the original press release of the TMC0280 to get an idea of the Integrated Circuit. 

Find here the original press release dated June 11, 1978:

First Voice Synthesis Memories (VSM)

These chips are manufactured in a metal gate P-channel MOS process and using dual-inline plastic cases. 

Later Speech Synthesizers (TSP50C0x/1x)

The TSP50C0x/1x family of speech synthesizers offer cost-effective solutions for high-volume applications. Each incorporates a built in microprocessor that allows music as well as speech capability. Texas Instruments offers five sizes of internal ROM for up to three minutes of speech. In addition, the devices can be interfaced to external speech memory.
The TSP50C0x/1x uses a revolutionary architecture to combine an 8-bit microprocessor, a speech synthesizer, ROM, RAM, and I/O in a low-cost single-chip system. The architecture uses the same ALU (Arithmetic Logic Unit) for the synthesizer and the microprocessor, thus reducing chip area and cost and enabling the microprocessor to do a multiply operation in 1.6us. Linear Predictive Coding (LPC) is used to synthesize high-quality speech at a low data rate.
The TSP50C0x/1x implements an LPC-12 speech synthesis algorithm using a 12-pole lattice filter. The internal microprocessor fetches speech data from the internal or external ROM (TSP60C18 or TSP60C81), decodes the speech data, and sends the decoded data to the synthesizer. The microprocessor also interpolates (smooths) the speech data between fetches. The output of the synthesizer can be used to drive transistor or integrated-circuit amplifiers. Some digital low-pass filtering is provided inside the TSP50C0x/1x. 
The TSP50C0x/1x  is manufactured in a 4-V to 6-V CMOS Technology for Low Power Dissipation and using 16 pin dual-inline (DIP) or 20-pin surface mountable (SO) plastic cases.

Later Speech Synthesizers (MSP50C3x)

The MSP50x3x family uses a revolutionary architecture to combine an 8-bit microprocessor, two speech synthesizers, ROM, RAM, and I/O in a low-cost single-chip system. The architecture uses the same arithmetic logic unit (ALU) for the two synthesizers and the microprocessor, thus reducing chip area and cost and enabling the microprocessor to do a multiply operation in 0.8us. The MSP50x3x family features two semi-independent channels of linear predictive coding (LPC), which synthesize high-quality speech at a low data rate. Pulse code modulation (PCM) can produce music or sound effects. LPC and PCM can be added together to produce a composite result.
The MSP50x3x implements an LPC-12 speech-synthesis algorithm using two 12-pole lattice filters. The internal microprocessor fetches speech data from the internal ROM, decodes the speech data, and sends the decoded data to the synthesizer. The microprocessor also interpolates (smoothes) the speech data between fetches. The microprocessor can calculate a PCM waveform, which can be added to the output of one of the two lattice filters to create composite PCM + LPC waveforms.

The MSP50x3x is manufactured in a 3.3V to 6.5V CMOS technology for low power dissipation and using 16 pin dual-inline (DIP) or 20-pin surface mountable (SO) plastic cases.

Later Speech Synthesizers (TSP50C4x)

The TSP50x4x family is not yet explored.

Later Speech Synthesizers Memories (TSP60Cxx)

These chips are manufactured in CMOS Technology for Low Power Dissipation and using dual-inline plastic cases. They are intended for use with the TSP50C0x/1x family of speech synthesizers.