2-Port Analyzers on a Chip? Part 1: What Are TPAs?

In the beginning, there were curve tracers made by Tektronix. They displayed the characteristic curves of electron tubes and transistors. Then H-P expanded the concept to that of multiple stimulus-measure units (SMUs) which are drive-sense circuits. What are the challenges in integrating these kinds of instruments?

Two (or more) SMUs can be combined to make a two-port analyzer (TPA) such as a transistor parameter analyzer (also TPA) or curve tracer. Early SMUs were integrated as automatic test equipment (ATE) pin drivers for ATE testers, usually used to test ICs in production. These pin drivers did not require too much voltage or current range and emphasized precision over speed. Perhaps it is time to consider the next generation of this generic category of IC by expanding the conceptual horizons to that of a TPA.

The functional diagram of the classic curve tracer is shown below.

It displays families of curves such as the following for a NPN BJT (a) and diode (b).

The BJT vertical axis is the collector current and horizontal axis is collector-to-emitter voltage. Each BJT plot is drawn at a fixed base current as parameter. The set or family of plots results from stepping base current through a range of discrete values. For the above example, six values of IB are used. The scale factors for the per-division values are given on the instrument front-panel knob settings. (Such are the characteristics of a pre-μC analog front-panel.)

A brief perusal of curve-tracer technology of the past shows that it is expensive. While a DMM, which can also measure voltage and current over a considerable range, might cost less than $100 for 3.3 digit capability, three of them could comprise a curve tracer (VCE , IC , IB ) if they could display swept values as a v-i plot. The sweeping of one variable (vCE , for instance) while holding that of the other port constant as a parameter (iB ) to produce a parametric family of curves (of iC ) can be executed by a μC. The display is that of the instrument (whether physically one unit or separated by a comm link) and is also driven by the μC.

DMMs are low-cost but TPAs begin at several thousand dollars. Much of the reason for the high price (and cost) is a result of the ability to deliver significant amounts of power to the device under test (DUT). In the earlier curve tracers from Tektronix, the control of maximum vCE was a variac and the output-port (c-e ) waveform was a rectified power-line sine-wave.

DUT power nowadays would be pulsed to minimize heating. Pulsed capability can be implemented on a chip to minimize on-chip power-handling requirements. The picture emerges that if the required DUT voltage and current can be supplied by the SMU IC, then a much lower-cost TPA can be produced, and all of us can own a TPA and check power MOSFETs for sloppy channel or gate-channel shorts, something a DMM does not do well. To actually see the v-i curves of a transistor tells much in one display.

It all seems so feasible, yet $400 curve tracers are not to be found in the marketplace (except as old surplus 575s, not new ones). Even by using semidiscrete design, an intermediate generation of TPAs is possible with existing parts. (I am working on several of them.)

The power supply for TPAs is nontrivial in that it can require one or more isolated outputs with floating grounds. This simply means that neither terminal of a given supply output connects to the system ground node. Why? A floating supply lets a low-side (negative terminal) current sensing resistance be in series with it, connected to ground on the other side of the sense resistor. (More on this in a later article.) With switching technology, this can be readily achieved.

The Innovatia open-source Floating Differential Source (FDS), a kind of specialized TPA for testing the CMR of amplifiers and high-side current sense circuits, has three separate ground systems with supplies for the grounded source, floating source, and floating DVM. Its parts cost is about $150 US, including the circuit-board, in prototype volume. It is a single-board instrument with an AVR ATmega8515 μC. The FDS has a ±35 V grounded source (VCM) capable of outputting 100 mA. The floating source (VDM) has ±10 V output range (also 100 mA) for driving amplifier inputs. A third source is a current sink of up to 100 mA for testing high-side current sensing. (The FDS documentation is being offered to prospective suppliers or interested end-users on an open-source hardware and software basis and a few evaluation-phase prototype units are available.)

With some reconnection of the FDS functions and the addition of more range switching, the FDS becomes the Innovatia TPA204 design. It has only a 3.5 W output capability (continuous) but this is sufficient for testing most low-power transistors and even the low-power regions of higher-power devices. A serial link to a user computer provides the user interface for displaying acquired data, including semiconductor plots. The TPA202 is a non-μC version using ten-turn pots and front-panel switches for inputs.

Do you have any ideas for why the $400 TPA does not exist on the market? Or which of it you expect can be integrated?

Watch for Parts 2 (2-Port Analyzers on a Chip? Part 2: SMU Integration Challenges), 3, and 4 to come from this series.

1 comment on “2-Port Analyzers on a Chip? Part 1: What Are TPAs?

  1. Davidled
    June 9, 2014

    Building genetic oscilloscope for plotting input data (voltage vs. current) would not be a major concern. Most likely, design might consider the total design cost based on BOM. The appropriate component should be selected by designer. Hence, if engineer looks the diagram, they might check BOM list.

Leave a Reply