This overview of the integration of electronic measurement instruments ends with a broader look at the strategy of maximum-integration instrument design. (A subsequent series on integration of a Z meter will demonstrate instrument integration issues in more detail.) We have what appears to be a goal of a four subsystem instrument for a broad range of instruments. The four subsystems are:
- Analog IC
- Microcomputer (single IC)
- Power supply (semi-discrete)
- User interface (semi-discrete)
This four-subsystem solution might be reduced further by combining the microcomputer (μC) and analog circuits on one IC. Then if the IC is not thermally limited, the supply might be included by standardizing on IC-compatible voltages. The user interface could even be eliminated by considering it a part of a larger system.
Instruments built to plug into back panels of computers use that computer interface through which the commands and acquired data move. A software interface program runs on the user computer from which the instrument is controlled. In this case, the user interface is not eliminated or integrated; the user computer is the fourth subsystem. Seeing the relationship of these subsystems differently might be all that is needed to make progress in how instruments are configured.
As it is, the four-subsystem instrument motif seems like an achievable and realistic goal. Four subsystems and not one shows recognition of the fact that integrated circuits are but a part of a larger reality, that of whole systems. What is to be accomplished by the whole system involves interfaces to human beings and power sources. Each of these expands the scope of technical considerations for integration considerably.
The highly integrated and digital (switching) nature of the μC means it is best kept in a somewhat isolated environment from analog to reduce noise. The isolation is provided by separation at the substrate level. Separate digital and analog ICs not in close proximity to each other solve problems that otherwise would require other discrete components such as shields. So in this case, semi-integration has its advantages.
Power supplies can vary considerably in what is needed, depending on the source. Battery supplies are the simplest and are not readily reducible to monolithic integration, nor would we want them to be if they are to be replaced periodically. Power-line-operated sources require converters. One way of minimizing the supply component is to standardize on various wall-mount supply voltages such as 12V.
Then in-system conversion is accomplished free from the safety compliance issues by dealing with this intermediate 12V bus input instead of 165V (or twice that in Europe). The problem with proliferating wall-mount (or power-strip-mount) and desktop converter boxes is another matter that should be considered in the overall strategy for more integration. However, they do simplify the design of what is in the integrated box.
Even with a static (DC) low voltage as input, the magnetic power components that have yet to be integrated are transductors (transformers and inductors) and capacitors. As switching frequencies increase, the possibility for a single-chip converter increase because the size of C and L decrease. We are entering a phase in converter design where multilayer ceramic capacitors are replacing electrolytic capacitors at the high end of the frequency range.
Research has been ongoing at places such as David Perreault's lab at MIT to integrate magnetic material in with the rest of the electronics. A smaller volume of magnetic material is required at higher frequencies. Suitable transformer coupling can be obtained such that the converters can transfer a given amount of energy each switching cycle from input (primary) to output (secondary) circuit. The transferred power, being the rate of energy, thus increases linearly with switching frequency. Similarly, sufficient inductance can be obtained to properly filter.
The obstacle in this trend is that as switching frequencies increase, so do power losses in magnetic materials. The best ferrites nowadays begin to have excessive losses over 500kHz, and the typical design range is in the 100kHz to 400kHz range for switchers of 100W or more. At higher switching frequencies, the per-cycle energy transfer does not have to be as large and the peak magnetic field density ripple is reduced to reduce power loss. If it is reduced too much, then more power is transferred at the same loss at a lower frequency.
All in all, integration of the supply in a monolithic form is not inconceivable, though it is not here yet. It will be best applied at first to ICs requiring a small amount of power, though most ICs cannot handle much over 5W to 10W anyway, even with a fan blowing directly down on them. With low power requirements, the inductor can be eliminated entirely by using switched-capacitor energy transfer. Therefore, the supply is conceivably integrable.
Finally, the user interface could most simply be integrated by avoiding the issue through an RF link or wired link with a fast serial protocol. This is not anything novel, and it solves the "front-panel" problem, yet it seems like a cheap way to avoid the problem of the final subsystem. If electronics technology were to trend toward a standard user interface, then any instrument (or other similar kind of device) need only communicate with it, and that can be accomplished already through the μC. This might be the final solution, though it is not really optimal for instruments intended to be highly portable, such as handheld units. Or is it? The default "standard user interface" nowadays is a desktop or laptop computer or tablet (iPad) device. These small computers are viable possibilities as front-panels for small instruments and can preserve the needed portability.
One of the ultimate limitations to integration is that as everything becomes smaller, human perception does not evolve fast enough to accommodate changes in technology. An LCD display integrated onto monolithic silicon would be hard to read, even with a magnifier, yet it might be a possibility - unless it is wafer-sized. Some full-color LCD displays selling for as little as $10 each are not much larger than some large monolithic chips. As an example, Newhaven makes a 1.8 inch, 128 × 160 dpi device.
The final integration of the four subsystems onto a single monolithic die of silicon is not preposterous, yet it is not optimal at this time. The intermediate step of integration -- that of the four-subsystem scheme -- seems to me to be the optimal goal and offers plenty of integration possibilities for present and future analog ICs.