Designing at the Discrete Component Level—and Below

Written by Jack Carrol, Senior Engineer
image description
 integrated circuits and off-the-shelf subassemblies can save a lot of work—when they fit the application. Most of the time they do. However, not every circuit function has a large enough market to attract a chip maker's interest or to continue production after the first surge of high-volume purchases. For that matter, even if the parts makers are willing, not all electronic design problems can be solved by putting the functionality into catalog parts; sometimes the end product can only be engineered in the borderlands where the electronics collide with the physical world.

These types of design problems tend to come up in instrumentation, safety-critical products, products that are required to work flawlessly in the presence of severe external electromagnetic interference, designing for emission suppression, and operation on limited power sources. Sometimes the product may need a unique circuit that isn't in a catalog or in a textbook. That's when the design needs the engineer to work directly with the principles of electromagnetic theory, or with electrochemistry, electro-optics, or heat transfer in unusual environments.

Similarly, it can be misleading to think of all electronics as either analog or digital. Much of electronic technology fits neither neat category; for example power management, radio, and safety interlock chains, though they may incorporate analog and/or digital elements. The engineer's work encompasses all kinds of electronics.

Below are a few project examples:


Discrete power switch

The requirement was for a circuit that would switch on the power to a boost converter at the start of the treatment cycle, without causing a large current surge into the converter's input bypass capacitor that would cause the tiny battery to dip and reset the microcontroller. The solution was to charge the input capacitor at a controlled rate so that the charging current would be kept within the battery's capability. The pass element is a P-channel FET, with a capacitor connected from gate to drain. The gate is driven by a current source, using an NPN bipolar transistor and a current-setting resistor in series with the emitter. The controlled gate current, flowing into the FET's gate-to-drain capacitor, uses the Miller effect to generate a constant slew rate at the drain, charging the downstream power capacitor at a controlled dV/dt until it reaches the battery voltage.


Electrocardiograph front end

If we examine the physical measurement circuit that acquires an EKG signal from a living patient, the "electrical components" don't all appear on the parts list. We must account for the electrical properties of the patient's body and the surrounding space, as well as unrelated electrical equipment nearby.  The stray resistances and capacitances couple waveforms from the electric power system into the sense leads, at a high enough amplitude to bury the millivolt-level signal. Because medical safety rules forbid grounding the patient, some subtlety is needed to suppress the power line interference.

The EKG signal voltage is typically on the order of magnitude of 50 mV.  But the capacitive voltage divider formed by the capacitances from the body to ground and from the body to the hot side of the AC line can put the voltage of the body as a whole at 50 V or so, 1000 times larger.  Safety standards require the electrocardiograph instrument to be electrically isolated so that it doesn't ground or enable current to flow through the patient. The EKG’s internal circuits and the connecting wires also have capacitance to ground and to the AC line. The various capacitive voltage divider ratios are not only random, but they also vary dynamically as people and objects move within the room's overall electric field. The result is a closed circuit that allows line-frequency current to flow through the stray capacitances from the AC line to ground, passing on the way through the resistance of the skin and the measuring electrodes—which are typically 5 kilohms or so. That current generates an AC voltage drop, and it appears in series with the signal. The AC interference falls within the bandwidth of the signal so that it can't be removed by frequency-domain filtering, and it's often strong enough to overload the high-gain EKG amplifier.

We can hardly eliminate all AC power from the vicinity of the electrocardiograph. So how do we make the measuring circuit insensitive to its effects? By intercepting the unwanted AC electric fields on a conductive surface, completely enclosing the patient leads and the instrument front end in three dimensions. This is called passive guarding.  The guard gets its voltage from a separate electrode connected to the patient.

Because the guard is at the same common-mode voltage as the signal leads, no AC current flows in the guard-to-front-end capacitance, therefore no interfering current flows through the resistance of the signal electrodes; it flows through the guard electrode resistance instead.  Thus, the interference is effectively diverted from the measuring loop.


Photometer board

A client reported severe random drift in production lots of a photometer amplifier. It was a calibration device for precision light sources, relying on accurate measurement of a microamp-level current from a photodiode. Preliminary checks showed that the circuit was properly designed and laid out, the components were well chosen, power supplies were stable, and nothing was oscillating. Previous experience with sensitive microcurrent instrumentation suggested that degraded surface insulation resistivity was involved, the usual cause being ionic contamination on the board surface forming an electrolyte with moisture in the air. The client's manufacturing engineer confirmed that the contract manufacturer had soldered the boards with an aqueous flux. Extensive testing at other companies in previous years had demonstrated that organic acid fluxes left a cosmetically clean board, but also left a firmly bonded trace-level ionic residue that survived all cleaning attempts by a variety of processes.

The confirming test for ionic contamination was to breathe on the board. There was a large and dramatic offset shift, which recovered in a few minutes as the board surface returned to equilibrium with the room humidity. Assembly using rosin flux, followed by a saponified water cleaning process with deionized water final rinse solved the problem.

 

In a nutshell...

These are just a few examples where cookie-cutter designs do not meet the design challenge.  Sunrise Labs has experienced and creative engineers, who can do what's necessary, whether it means the direct application of fundamental physical principles, accounting for the non-ideal properties of real components, or taking into account the non-ideal electrochemical properties of real materials.

Learn more about the Electrical Engineering expertise of our team at Sunrise Labs