1980s – Multiple-Channel Data Acqui

1980s – Multiple-Channel Data Acquisition
The major innovations in transient testing methodology were
10 SOUND & VIBRATION/JANUARY 2011 www.SandV.com
accomplished in the 1960s and ’70s. In the 1980s, developments
in data acquisition and parameter estimation led to the significant
advancements in multiple-reference impact testing (MRIT) of the
’90s.
Data Acquisition Developments. It became clear in the ’70s that
to improve modal parameter estimation, it would be necessary to
develop multiple-reference parameter estimation algorithms and
affordable multiple-channel data acquisition systems. One of the
major problems with parameter estimation was the consistency
of the measurement database. Measurements taken at difference
times from different reference points were inconsistent. As a result,
estimated modal parameters were inconsistent even when the fit
to individual measurements appeared to be excellent. To address
this problem, a large multiple-channel affordable data acquisition
system would be required. In the late ’70s, the groundwork
for the application of a multiple-channel acquisition system was
developed with the formulation of the multiple-input/multipleoutput
(MIMO) FRF technique. Initially, a four-channel system
with two inputs and two outputs was used, followed by testing
with an eight-channel system with two inputs and six roving
response channels. Several vehicles were tested in this manner
with encouraging results.
The dream was a system with two to four inputs and hundreds
of responses that could be measure simultaneously and multiple
reference parameter estimation algorithms that could extract modal
data from this set of measurements.
Parameter Estimation Developments. In the early 1980s, with
the breakthrough development of the polyreference time domain
algorithm (PTD)9 the parameter estimation part of the dream came
true. This was followed by the development of the eigenvalue realization
algorithm (ERA)10 a few years later. The PTD algorithm
was a multiple-reference version of the least-square complex
exponential algorithm, and the ERA algorithm was effectively a
multiple-reference version of the ITD method. The PTD method
could run effectively in a small minicomputer system; however,
in the early ’80s, the ERA required a larger mainframe computer.
As a result, two different groups of users were employing the two
methods: NASA was the primary user of ERA; industrial users
(machine tool, auto companies, etc) used the PTD method. By the
mid ’80s, a more general unifying approach to parameter estimation
was being developed. The unified matrix polynomial approach
(UMPA) concept was being developed in the late ’80s and early
’90s. Using the UMPA concept, all important parameter algorithms
could be rederived from a common starting point.
Important mathematical techniques like singular value decomposition
(SVD) became a significant part of these developments. In
fact, a parameter estimation procedure based on SVD, the complex
mode indication function (CMIF), was developed in the late ’80s
and perfected in the early ’90s. It became a standard parameter
estimation tool used with multiple-reference impact testing (MRIT)
in the ’90s
The second part of the dream, the ability to measure hundreds of
channels simultaneously, took a little bit longer to develop. In 1981,
Boeing Aircraft Company conducted the first large-scale modal test
of the Boeing 767, where up to 128 channels of responses could
be measured simultaneously. The raw time data was recorded to
a large disk file and was post-processed into FRF measurements.
This was the ideal case; by recording the raw data, it was available
for post-processing after the test object was released. Different
signal processing could be used to enhance certain aspects of the
analysis, such as zooming into a certain frequency band to enhance
an important target mode.
Unfortunately, this type of data acquisition was too expensive in
the early ’80s for most users. The development of the Struccel system11
in the mid ’80s significantly reduced the cost of the multiplechannel
sensor systems. The widespread demand for CD players
and further developments in digital music led to mass production
of delta sigma ADCs that were available and very inexpensive by
the early ’90s. This made it possible to design a very inexpensive
multiple-channel data acquisition system. The possibility of conducting
a test with hundreds of channels became practical in the
90s for groups outside the aerospace and auto industries.