Catto / Topic Map · Electronic Measurements II Level 5
Electronic Measurements II · 120 h · Topic 10 of 10

Advanced measurement automation systems

With a Cpk of 1, three out of every thousand power supplies fall out of specification. With 1.33, thirty per million. The test system not only decides which unit passes: it measures how good the process is.

PXI and LXI IVI drivers Test sequencer Cp and Cpk Yield

01From the bench instrument to the test system

The measurement automation of Measurements I connected bench instruments to a computer with SCPI commands. In production, where thousands of devices are tested per day, that is not enough: compact, fast and easy-to-maintain systems are needed. There are three main modular platforms:

PlatformIdeaAdvantage
PXIInstruments as modules inside a chassis with a PCI Express bus, shared synchronization and triggersSpeed and density: dozens of instruments in one chassis
LXIStandalone instruments over Ethernet, with a web interface and synchronization through IEEE 1588Distributed instruments, no chassis
AXIeHigh-performance modules in a larger chassisHigh-end instruments, such as many-channel oscilloscopes

On top of the hardware a layer of interchangeable drivers is used, such as IVI, which present a common interface per instrument class: a program written for one multimeter works with another from a different manufacturer without changes.

02Sequencers and parallel testing

A test sequencer organizes the test of a product as a list of steps, each with its measurement, its limits and its action on failure, and stores the results with the serial number. The same measurement code is reused across different products. To increase the number of units tested per hour, testing is done in parallel: several units at once, sharing the expensive instruments and using the cheap ones in parallel. The bottleneck is usually the shared instrument, and the design of the system consists of balancing the times.

03Process capability

The test system decides whether each unit passes or not, but the accumulated data say something more: how well the manufacturing process meets the specification. If the measured quantity has mean \( \mu \) and standard deviation \( \sigma \), and the specification runs from \( LSL \) to \( USL \):

\[ C_p = \frac{USL - LSL}{6\sigma} \] \[ C_{pk} = \frac{\min(USL - \mu,\ \mu - LSL)}{3\sigma} \] \( C_p \) compares the width of the specification with the spread; \( C_{pk} \) also takes into account the shift of the mean. With \( C_{pk} = 1 \), 0.27% falls outside the specification; with 1.33, about 30 parts per million; with 1.67, fewer than one. The automotive industry usually demands 1.33 or more.
Lab · process capability and yield

The output voltage of a power supply, specified as 5.00 ± 0.10 V. The bell curve is the distribution of production; the red zones, the units out of specification. The histogram shows 2000 simulated units.

The uncertainty of the measurement itself has to be added: with a narrow specification and an instrument that scatters, good units are rejected and bad ones accepted. The guard bands from testing and standards also apply here.

04In the laboratory

Activity 1 · A sequence

Program in Python a test sequence for a laboratory power supply with a multimeter and an electronic load over LXI or USB: no-load voltage, load regulation and ripple, with limits and logging of results.

Activity 2 · Real Cpk

Measure the resistance of 50 1% resistors from the same bag and calculate Cp and Cpk with respect to their tolerance.

Activity 3 · Test time

Measure the time of each step of the sequence and propose how to reduce it: fewer averages, range changes, parallel testing.

05Common mistakes

  • Looking only at Cp. A narrow but shifted process has a good Cp and a bad Cpk.
  • Calculating the capability with few samples. The estimated standard deviation varies a lot; tens or hundreds of units are needed.
  • Forgetting the test uncertainty. It adds to the spread of the process.
  • Writing the test tied to an instrument model. Interchangeable drivers avoid rewriting it when a device is changed.

06Self-assessment

Specification of 10 ± 0.3 mA, mean 10.1 mA and σ = 0.05 mA. What are Cp and Cpk?

Cp = 0.6/0.3 = 2; Cpk = 0.2/0.15 = 1.33.

What fraction falls outside the specification with Cpk = 1 and the process centered?

0.27%.

What advantage does LXI have over PXI?

The instruments can be distributed over a network, with no chassis, and are synchronized through IEEE 1588.

What are IVI drivers for?

So that the test program does not depend on the instrument model within a given class.

07Further reading

  • Douglas C. Montgomery. Introduction to Statistical Quality Control. 8th ed., Wiley, 2019. Process capability, the Cp and Cpk indices and statistical control.
  • PXI Systems Alliance and LXI Consortium. PXI and LXI Specifications. Available on each consortium’s website. The definitions of the modular platforms and their synchronization.
  • IVI Foundation. IVI Specifications. IVI Foundation. The instrument classes and the interchangeable drivers.
Development of the topic “Advanced measurement automation systems” of Electronic Measurements II (Level 5), based on the curriculum of the UTN Electronic Engineering program, 2023 curriculum — Ordinance No. 1849 of the UTN Higher Council. Back to the Topic Map · catto.ar