One of the challenges in integrating all the electronics of measurement instruments is brought on by the versatility of instruments.
Imagine having to buy a separate instrument for each different front-panel setting -- hardly a feasible situation. Instruments have multiple functions with multiple parameters that can be set over wide ranges. This is such a basic feature of instruments that it would barely deserve a mention were it not for its prominence in the list of instrument design considerations. Consequently, one of the major aspects of instruments is configurability, with each configuration referred to as an instrument setting.
Reconfiguration, or changing settings, is mostly implemented with switches or electromechanical relays. In older equipment, these were switches that could be operated from the front-panel (whether the switch was physically at the front or offset into the instrument with a coupling shaft.) These switches either directly controlled the signal paths (and maybe switched some control lines); or they operated relays that controlled the signal paths.
Electromechanical relays are being integrated as microelectromechanical systems (MEMS) develop, and it would be desirable to integrate them with transistor circuitry without too many performance sacrifices. Even so, research prototypes of MEMS relays have power, voltage, and current limitations caused by size. Additional discussion of the use and development of MEMS devices can be found here. As crude as they seem, discrete electromechanical relays are high-performance parts that exhibit features that help meet instrument criteria.
CMOS/DMOS switches are accurate for lower voltage and current switching and are well-suited for many of the configuration requirements in instruments. The long-defunct Signetics made low-capacitance, V-groove, DMOS SD5000 quad Mosfet switches with high dynamic isolation, useful for switching high-bandwidth waveforms. They were used in an early microprocessor-based instrument, Northwest Instrument Systems Apple II-card function generator.
This technology has migrated into present-day analog integrated switches. As the distinction between high-speed switches and solid state relays blurs, the possibility is emerging of replacing electromechanical relays with small, fast power Mosfets. IXYS and Maxim have, in different ways, made the Mosfet switch more robust. IXYS (with acquisition, C. P. Clare) has been extending downward in application-space from power electronics and Maxim has improved the ESD protection and increased the voltage range of CMOS switches.
In many instances in instrument design, a 4066 quad CMOS switch is quite adequate. Where that part was not quite good enough, the next generation of the device, such as the MAX4066, might suffice. Switches with less than an Ohm of on-resistance have all but accomplished the goal of integrated configurability.
For high-bandwidth switching and isolation, however, linear BJT differential-pair switches as found in scope channel switch circuits and in PLL phase detectors and transconductance multipliers provide higher performance, though with more support circuitry. All of these schemes are integrable with a biCMOS process. So it would seem that some companies have a semiconductor process that could integrate many but not all instrument configuration circuits.
In a previous article, An Instrument on a Chip? A Look Back, the prominence of the rotary switch of DMMs illustrated the importance of instrument configuration. An alternative to the big switch in non-microcomputer-based instruments is a programmable ROM or PLD containing the configuration logic, plus solid-state or electromechanical switches.
An example is the ESI 253 RLC meter, shown below. The front-panel switches do not directly switch internal waveforms but are cold-switched in that they only encode digital logic input to the PROM which outputs control lines to the actual switches. The switches can be optimally placed to eliminate mechanical linkages or cabling noise and their interconnections carry no noise- or bandwidth-sensitive waveforms, only near-static logic levels.
The big switch, while inelegant in some ways, is an elegant cost solution to the configuration problem for inexpensive, hand-held instruments. For maintenance and fault diagnosis, it can be a troubling hindrance, especially if the circuitry you need to probe cannot be accessed because the switch must be in contact with the circuit-board for the circuit to work. Cold-switching allows the front-panel to be separated from the analog real-time circuits (usually through a microcomputer). Consequently, the analog circuitry can be designed and tested separately as a modular block of the system.
Nowadays microcomputer-based instruments integrate the configuration and front-panel interface into the microcomputer programming. Of course, the microcomputer itself is a separate IC, but at this point in time, this does not seem like a bad tradeoff because it separates most of the noisy digital activity from sensitive analog circuits. A two-chip (plus front-panel and power supply) instrument solution is a goal yet to be achieved and seems like a hopeful next step in the quest for greater integration of instruments.