Advertisement

Blog

2-Port Analyzers on a Chip? Part 3: TPA Integration Challenges

A TPA consists of two or more SMUs. Two are required for basic transistor parameter measurement, and are called, for generality, the input and output ports . In general, the voltage and current of each port is measured. Consequently, a TPA has two ports with two functions (drive and sense) and two measured quantities. To bring some order to this set of combinations, the following abbreviations are used.

These abbreviations can be used to refer to either the circuit functions or their output quantities.

The second integration challenge is the number of floating supplies required. This is a major consideration in TPA design. It is a result of the need to sense both port current and voltage. Port supplies are often not grounded, and this leads to three interaction possibilities for a port.

  • Isolated: Supplies and μC interface are isolated and the drive output is isolated — with no conductive connections to elsewhere
  • Floating: Supplies are isolated and μC interface is grounded, with input level shifting to span specified voltage range.
  • Grounded: Supplies and μC interface are referenced to the external (system) ground.

If the device under test (DUT) common terminal (COM, which is usually the emitter of a BJT) is grounded, and both port supplies are also ground-referenced, then the current sense circuits must be made high-side circuits. If they are made low-side current sensors and placed in series between the output ground and source ground, they will not necessarily measure the port current correctly. The problem is illustrated by the “TPA Current Sensing” notebook page scanned below.

The basic problem is that of keeping the grounded negative-terminal (COM) currents of the two ports separated for measurement. As shown above, the common negative terminals of both ports are grounded. Through which ground connection a port high-side current finds its way back to the common grounded supply is indeterminate and depends on the relative resistance of the ground wiring.

If current-sensing circuits (ISNs) are placed in series with the port ground returns so that only the port current flows through the ISN, then the port supply cannot any longer be grounded, because it is in series with the ISN to COM and thus must have a floating ground. If the ISN were placed in series with the high-side port terminal (as in the SMU of the previous article (2-Port Analyzers on a Chip? Part 2: SMU Integration Challenges)), then the port supply could be grounded. For a grounded DUT COM terminal and low-side ISNs, the corresponding port supplies must be floating.

The disadvantage of low-side sensing is that the power supply is made more complicated, with multiple isolated outputs. Multiple windings on the converter transductor increase cost and complexity. Alternatively, high-side sensing requires floating ISN circuitry, and this can be a problem for a large OVDR range, because the ISN must eventually supply a ground-referenced output to the ADC (or DVM). If the ADC or DVM is floating, then differential input switching, more floating supplies, and isolated digital paths complicate the design.

Nobody said TPAs were simple to design. Floating circuits and high power kept the earlier curve tracers power-line-driven, where rectified sine-waves from floating 50/60 Hz transformer windings supplied the output-port drive. Even the Tek 5CT1N plug-in curve tracer used a steel-core transformer driven by a triangle-wave generator to supply the collector (output-port) drive. In the more recent Innovatia Floating Differential Source (FDS) — a specialized kind of TPA used to measure amplifier common-mode rejection and gain and to drive high-side current-sense circuits — the switching power supply has nine output windings, allowing the instrument to be free of power-line waveforms. Expect the TPA power supply to be complicated compared to measurement instruments with a common system ground.

A solution to the problem of grounding for low-side ISNs is shown by the next notebook page (below).

There might not be any good reason to ground the DUT common terminal, and it is allowed to float. A transistor as DUT is not in any other circuit when tested (the in-circuit testing feature being found only on low-priced transistor checkers) and if left floating, allows both ISNs to be low-side while requiring only one port supply to be floating. In the sketch, the output-port supply is floated. This ensures that only high-side output-port current will return through the low-side port terminal (as is required of a port, by definition). If the output-port supply were also grounded (as is the input-port supply), then we are back to the port current-mingling problem described above. Consequently, one of the two port supplies must float.

Which port supply should be chosen to float? This design consideration involves the types of amplifiers that are the drivers of the SMUs. An amplifier capable of outputting both bipolar currents and voltages is a four-quadrant (4Q) amplifier that requires both positive and negative supplies and corresponding driver stages of both polarities. (A graph of v versus i will have four quadrants with the four polarity combinations of +/-v and +/-i .) A way around this is to design a single-supply 1Q amplifier with only +v , +i out. Then if the amplifier is floated (requiring a floating supply and input translator circuitry) the output terminals can be swapped with a DPDT relay or manual switch (on an analog front-panel) to select the polarity. In this case, the output-port supply must be floating. This conveniently allows the input-port supply to be grounded, as shown in the top sketch of the above notebook page.

Low-side ISN design is not terribly challenging, in that the circuit nodes are all near, if not at (system) ground. In the lower sketch of the above notebook page, OISN is grounded, but the sense resistor of IISN is not connected to ground on either side, and floats. On the port low-side terminal, it is grounded through the OISN. Typically, the voltage dropped across ISN sense resistors is a fraction of a volt, and the IISN is floating not far from ground, though it is required to be floating. Therefore, to output a ground-referenced IISN voltage, it must have a differential amplifier with inputs across the sense resistor, and its output reference node must be ground.

Related posts:

12 comments on “2-Port Analyzers on a Chip? Part 3: TPA Integration Challenges

  1. etnapowers
    July 2, 2014

    The separation of the grounds of the power supplies for the forcing and sensing is really important because the current flowing in the common ground might modify the ground potential and it might affect the measurement operation.

  2. etnapowers
    July 2, 2014

    It's very interesting the link to the description of FDS features, by Innovatia.

    It can work both in Grounded (CM) Source and in Floating (DM) Source .

  3. D Feucht
    July 2, 2014

    The Floating Differential Source is not exactly a TPA but it is similar. It was designed for manufacturing test of boards having floating analog inputs that need to be tested for CMRR.

    It also has a current source for testing floating or high-side current-sense circuits. Although it is somewhat application-specific, the general capabilities of the three sources on it, and the isolated 4-digit DVM make it useful for a wider range of applications as a somewhat general workbench tool.

  4. fasmicro
    July 6, 2014

    >> The separation of the grounds of the power supplies for the forcing and sensing is really important because the current flowing in the common ground might modify the ground potential and it might affect the measurement operation.

    That is usually the hallmark of good designs. Ignore what your ground is doing, burn the chip efficiency. When it comes to measurement system, that becomes more important.

  5. etnapowers
    July 7, 2014

    @Dennis, it's interesting to me to know what are the main differences between FDS and TPA, thank you for your clarification. The test of boards is really important especially in the debug phase, and it's very important to have an application board that incorporates many features to speed up the debugging procedure.

  6. etnapowers
    July 7, 2014

    @fasmicro, that's right. A good design of a test board has to take into consideration the need of the presence of a power ground pin, to absorb the pulsed currents, and a analog ground pin as well, to have a reference potential that allows to take high precision voltage measurements.

  7. fasmicro
    August 3, 2014

    >> A good design of a test board has to take into consideration the need of the presence of a power ground pin,

    It seems that is coming from experience. Newbies experience that point firsthand when the board/chip returns and nothing works. Then they will know that GND is not just as the professor says it is. There is more to GND in the practical design world. Good piece.

  8. etnapowers
    August 4, 2014

    @fasmicro: you're absolutely correct, it comes from experience. The difference between a college course and an industry environment is just this: In the real world the theory is useful to model some effects and to start a design but experience teaches what else has to be taken into account to make the designed circuit working effectively

  9. goafrit2
    August 9, 2014

    >> he difference between a college course and an industry environment is just this:

    College prepares one to have the capacity to learn when in the industry. A good background is important to succeed in the industry.

  10. etnapowers
    September 2, 2014

    Agreed. The background is very important, not only from a technical point of view: a good engineer for example has to carefully evaluate the costs and the gain margin of a device , to be able to propose a profitable product.

  11. fasmicro
    September 2, 2014

    >> a good engineer for example has to carefully evaluate the costs and the gain margin of a device , to be able to propose a profitable product.

    But that is one of the things few universities teach in engineering programs. The economics of design is never a huge subject.

  12. etnapowers
    September 5, 2014

    @fasmicro: you're right, a good university course should include the economic aspect of an engineering project, to teach to the students that one key factor of the industry is not only the effectiveness of an IC or a electronic device, but also its profitability.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.