One of the problems to be solved in the design of an SMU and a two-port analyzer (TPA) is to keep the voltage and current sensing from interfering with each other by causing error. This can be seen by referring again to the SMU block diagram (below).
The high-side current-sense circuit (ISN) is followed by a voltage-sense circuit (VSN) across the output terminals of the port. Any current from ISN that is lost to VSN causes an error in the output (drive) current, iDR. For drive voltage, vDR, with a small range and low maximum voltage, VSN can be implemented with JFET-input op-amps that operate from standard supplies (such as +/-15 V). However, for a TPA with a more typical range extended to hundreds of volts, resistive voltage dividers appear to be the only viable design option for measuring vDR. Historically, the divider input resistance was made high enough that the error current through the divider was negligible. In the pursuit of greater accuracy, however, this current must be compensated by some scheme.
Two possibilities exist as shown in the scanned notebook page below titled VSN Divider Error. The bottom of the divider can be returned either to the port ground (1), or else the (in this case, low-side) ISN can be bypassed and returned to the floating-ground side of the port (2). Then the divider current returns directly to the floating supply and does not cause current error in the ISN because the divider current does not flow through the ISN. However, scheme (2) causes voltage error because now the divider is not across the port terminals. The voltage drop across ISN is an additional series error voltage.
This situation gives us a design choice. We can either correct for an error current in the ISN or an error voltage across the VSN input. Both forms of compensation are shown in the two schemes given to compensate for them. For VSN compensation, the divider is returned to the floating supply ground to avoid ISN error. Then the voltage that is input to the VSN is the divided port voltage, vDR, plus the voltage across the ISN sense resistor, vS. If this voltage is subtracted from the VSN input voltage, then vSN can be corrected. The divider has a voltage attenuation of
In the reverse direction, the error voltage across ISN is vS and appears divided by (1 T) at the VSN input. Thus, if the opposite amount of voltage, or (1 T) vS were added to the VSN input, the voltage error would be corrected. That is what scheme (2) shows on the left side of the notebook page.
To sense vS without interfering with ISN, the ISN output itself, which is an amplified +vS, is a conveniently buffered voltage that can be used for compensation, as shown on the bottom of the page. The ISN gain of AiSN is removed by scaling it out. In typical implementation, however, |A| = |(1 T) AiSN| < 1 and can be a resistive divider. The divider can be formed by connecting a resistor from viSN to the output node of the port voltage divider, and the summation function in the diagram is achieved. In other words, viSN is positive and is scaled by an additional divider to effect an equal and opposite voltage to compensate for vS at the port-divider output.
The alternative compensation scheme (for connection 1) avoids voltage error by connecting the divider across the port terminals and correcting current error as ISN compensation. This scheme is shown on the right side of the page and is a dual of the voltage-correction scheme. A resistor with resistance matched to the divider input resistance is the input resistor, Rd, of an inverting op-amp. The gain of the op-amp is scaled and inverted so that it can be added directly to the uncompensated ISN output, AiSNvS, to subtract the error-current contribution, AiSN(vO /Rd). The divider Rd and op-amp Rd consist of precision resistors, and their match is not hard to achieve.
As a design choice, I prefer VSN compensation because it is easy to implement and because in the use of curve tracers, voltage precision is somewhat less important than current precision. What are hard to integrate are high-voltage resistive dividers on ICs. They need not be included because a couple of external precision resistors are not much of an imposition for compact instrument design.
This is presently the solution for DMM input-voltage range extension. The vast bulk of TPA circuitry can be on-chip while a judicious choice of what to leave off chip can result in a $400 curve tracer. With pulsed drivers, it might even be the size of a DMM or battery-powered Z meter and sell for $250. At that price it could become widespread, thereby accruing a sizeable profit for the IC maker.
Who wouldnt spend $250 for a curve tracer, even if it is low-power? A cheap TPA can also be put to an indeterminate range of other possible uses: checking diodes, avalanche (zener) diodes, shunt regulator ICs (such as the TL431), optocouplers, measuring static amplifier gains, and applications you might think of that are omitted here.
Would you buy a $400 TPA, even if it is somewhat limited in voltage and current ranges?