Page 301 - Semiconductor Manufacturing Handbook

Geng(SMH)_CH19.qxd 04/04/2005 20:00 Page 19.28

INSPECTION, MEASUREMENT, AND TEST

19.28 FINAL MANUFACTURING

Clock distribution in a system is necessary because different resources running concurrently will
often run at different frequencies, so they must be capable of dividing a master clock rate down as
needed. In fact, sometimes frequencies are not multiples of each other and it may be necessary to
support multiple unrelated (frequency) clocks within the system. This is also desirable when imple-
menting the undersampling technique mentioned previously.
Power supplies deliver the necessary voltage and current to power up the device and keep it run-
ning. Usually, some relays are used on the DUT board to facilitate signal multiplexing and other
uses; these require drivers to turn on the relay coils.
Basic Test Setup for a Mixed-Signal Device. Load board layout is often overlooked, yet is a critical
factor in overall test performance. Poor design or layout can lead to significant noise and cross talk and
undesired attenuation issues. These issues can be compounded when dealing with multisite designs.
Digital, analog and utility signals and their associated return paths should be as isolated as possible.
Utility supplies, such as those used for relay control, are generally the noisiest supplies on the tester.
There are two basic parameters in the selection of the analog instrument—resolution and band-
width. Resolution is the number of discrete values the module can source and measure, usually spec-
ified in bits. For example, a 16-bit digitizer can resolve 2^16 discrete values. With a 2-Vpp input
range, the resolution would be the full scale of the range/ the total number of discrete values,
2V/2^16 or 30.52 µV. A parameter that quantifies the impact of nonlinearity and noise within the
bandwidth of interest is the effective number of bits (ENOB). The ENOB calculation is
.
ENOB = SNR meas −176
.
602
It is critical that this value be associated with a specified frequency range. Another method that
is often used to specify the performance of an analog instrument is the noise power or spectral den-
sity. In this case the noise power is spread out over the frequency spectrum and the units are
expressed as volts per root hertz. It is a useful way to express the noise floor of a resource in a stan-
dard terminology.
The basic test setup includes powering up the device, setting up the logic pins to achieve the testing
mode desired, and then sourcing a stimulus and measuring a response. The test list will be defined
according to the test specifications found in the data sheet. Since a converter is a very common
mixed-signal device to test, it will be briefly examined. There are two fundamental tests that are used
to characterize the performance of the devices—transfer curve and distortion tests.

Transfer Curve Tests. DAC transfer curve testing is accomplished by measuring the discrete ana-
log output voltage for each code and comparing these measured values with the ideal. This compar-
ison derives the fundamental measurements of differential nonlinearity (DNL), integral nonlinearity
(INL), offset error, and gain error. DNL measures the least significant bit (LSB) step size of the
device relative to an ideal step size. INL measures the deviation of each point, in LSBs, relative to
an ideal curve. Offset error is the offset of the output, at code “0,” relative to the ideal curve and the
gain error is the difference from the ideal at full scale.
Although the same parameters are used to measure ADC accuracy, the calculation methodology
is different due to the inherent differences between DAC and ADC transfer curves. While there is a
one to one correlation of input code to output voltage for a DAC, an ADC has a range of input val-
ues that correspond to a single output code. Therefore, a histogram method is utilized in the calcu-
lation of the transfer curve characteristics. The histogram plots the number of occurrences of each of
the converter codes and through some statistical mathematics the transfer curve for each code can be
plotted. This would be the deviation from an ideal transfer curve for each code.

Distortion Tests. Distortion testing is a method of dynamically testing a converter and the approach
is the same for both ADCs and DACs. The primary goal is to generate (or stimulate) a single
frequency to the DUT input and then measure or sample the output, convert the time sampled values to
the frequency domain, utilizing DSP with a Fourier transform, and quantify the undesired, distortion,

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)
Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.