**Career Technical Writing High-Lights**

In academia we have to maintain a *vita* for Tenure and Promotion
purposes and also for seeking financial support from various agencies.
Now that I am retired it does not appear that I need that vita anymore
but it is interesting that in looking at the different publications
down through the years it portrays a trail of evolving interests
and experiences. The following list of publications, along with
anecdotal notes, provide a view of what I see as some of the best
experiences in my professional career. - FHI

Irons, FH, "Magnetometer for Small Specimens," IBM Technical Report,
Endicott, NY, Jan 31, 1955, 18pp, Code:TR105.096.359

This was my first engineering job and took me out of state for the
summer of 1954. It convinced my Dad that I was not going to return to
the farm. I designed the magnetometer from a schematic in RM Bozorth's
physics book. It was used to measure the magnetic properties of thin
films in support of the development of the magnetic drum memory device.
Later the magnetometer was built into a servo-controlled system to
measure samples automatically.

------, "Digital Storage of Statistical Data," *IRE Trans. on Nuclear
Science*, Vol.NS-7, No.1, Mar 1960, pp 43-48.

This graduate research work used a 14-track tape recorder to store digital
samples from a photo-multiplier tube in the study of low level radiation
effects for the Radio-Activity Center at MIT. It was fully transistorized
and stored digital samples in parallel on a short loop of tape.
The concept centered around the random nature of pulse arrival times to
not have to preserve time information. This system eliminated the need for
lab technicians to manually create histograms from thousands of samples
taken from strip chart recorder outputs. The storage system would record
for several days and then produce the statistics of the experiment in seconds.
A beginning in automation!

------, "Active Filters: Properties and Applications,"
*Frequency*, Mar-Apr 1964

This tutorial paper was in connection with the Guillemin Networks (GNI)
adventure over the period 1962-67. The publisher was another new firm,
like us, and so we published a paper there in cooperation with helping them
get started. GNI pioneered the use of active filters in the development
of many systems, such as: vocoders; seismic array receivers; off-shore
drilling platform stabilizers; gated ranging radar circuits; underwater
sonic analyzers; and many others. The paper talked about active filters
and their advantages and touched on the basic problem of stability.
At that time, there was a fad for negative impedance converters and other
positive feedback approaches, but Dr. Guillemin was never convinced.
The paper hinted at, but did not really explain, the approach favored by
Guillemin. He did not trust the patent system (he really admired Armstrong)
and so we went for a Trade Secret approach to protect our product position.
The successful development of the opamp as a component made the whole
problem a moot point and made it possible for most companies to design
their own active filtering right into their own electronics.

Bryant,TG & ------, "An Analysis of Coupled-Pair Microstrip
Transmission Lines," Univ. of Maine, D.I.C. Report No. 855(S-196)
to M.I.T. Lincoln Laboratory, 1968.

Tom was my first graduate student. This work stemmed from a summer job
that Tom had at Lincoln Lab in the summer of 1967 under the guidance of
Jerry Weiss. Their work resulted in the publication of a famous and oft
referenced paper on coupled-microstrip lines. I was the local
supervisor for the work but did make the contribution to use strip-charge
finite elements rather than line charges. This removed a troublesome
singularity in the formation of the problem and led to
highly convergent results for the algorithm.

------, *An Analysis of Scattering from Dielectric Obstacles in
Rectangular Guides*, PhD Thesis, Lehigh Univ., 1971.

This work was performed at Lehigh University, from 1969-71, and involved
the development of algorithms to compute equivalent circuit effects for
resonant obstacles in wave-guiding structures. The purpose was to automate
CAD tools for the design of UHF filter structures. This was an
interesting problem that really pushed the finite element methods of
Roger Harrington and was limited by computer speed and memory space.
Two-port reactance versus frequency dependence was estimated for each obstacle,
through resonance, in a non-homogeneous field space. It was a
challenging exercise and it would be interesting to apply today's super fast
and large memory computers to see how they would perform on the problem.
We came within a few percent with extremely limited computing power,
by today's standards. The key to the problem was in trying
to determine the set of eigenvalues that yielded the waves that contributed
the most stored energy in the determination of reactance values.
This turned out to be a frequency dependent phenomena.

------ & Naimpally,S, "Iterative Active All-pass Structures for
Arbitrary Attenuation Requirements," IEEE 19th Midwest Symp on Ckts &
Sys, Aug 1976, pp 399-404.

This was a surprisingly neat little solution to the approximation problem
using a cascade structure of unit-gain one-pole all-pass networks.
Fourier theory is used to obtain a set of real coefficients
to fit an arbitrary attenuation function. The technique is unique, and
coupled with Lanczos smoothing to weight the truncated Fourier series
coefficients, the resultant responses come arbitrarily close to the objective
function. The circuit realization easily lends itself to IC fabrication
technology and is a very practical solution for obtaining a simple and
general purpose filter synthesizing architecture.

------ & Gilbert,MG, "A New Formulation of the Approximation Problem,"
*IEEE Trans on Ckts & Sys*,Vol. CAS-24, No.5, May 1977,
pp 231-241.

This paper solves the filter approximation problem to fit a ratio of equal
order polynomials to arbitrary attenuation requirements as specified on a
dB vs. frequency space by means of a straight-line connected function.
This is a problem that Dr. Guillemin envisioned the computer being able
to solve but he never did get to see that happen. The algorithm uses curve
fitting techniques as described in his *Synthesis of Passive Networks*
for the Hilbert transform relationship between phase and loss functions.
The minimum phase thus derived is split into two monotonically increasing
parts, one for the numerator and the other for the denominator.
These phase functions yield unique polynomials that are a close fit to the
objective function and are very close to a global minimum
for optimizing the selection of the polynomial parameters.
Thus the solution to the approximation problem is reduced to invoking
physical conditions for the initial choice of polynomials in an
optimizing procedure. The solution is very practical for active filters
where the numerator can easily be obtained from the pole-generating circuits
in a state-variable type of structure. In addition, since the method uses
zeros as effectively as poles, the procedure yields a minimum order
in its solution to the approximation problem.
This is my favorite paper for my whole career.

**Special Projects**~~~Somewhere along the way, in the mid-1970's,
I realized it was important to enable the graduate student's to work on what they wanted
to work on and not let my own personal research interests dominate their choice.
This resulted in some really fun projects, e.g.:

__Bill Jeffrey__ ('74) - Designed, built, and evaluated a variable
resolution PCM system for speech signals using synchronized pseudo-random
dither at the transmitter and receiver. The system was able to work at a
3-bit resolution using this technique. Typical telephone systems use 8-bits.
Bill used recorded tapes and carefully designed test methods to determine
the reliability of untrained listeners for evaluating this system at different
resolution levels.

__Andy McClellan__ ('75) - Built a spread spectrum receiver/transmitter
to tag and track aquatic animals in a noisy environment. This was in
cooperation with the Zoology department and we actually tracked fish in the
Stillwater River.

__Jim Lester__ ('76) - Jim was a radio ham and wanted to build a narrow-beam
antenna array structure but the problem was how to measure the pattern.
He built a skirt array, ala Kraus, on the (flat) roof of Barrows Hall and
used radio astronomy methods to estimate the pattern as a selected star
passed through the antenna aperture. That was one real exciting project
in freezing mid-winter but Jim succeeded in his goal. The antenna remained
on the roof until sometime in the 90's before it was removed.
It was impressive to "see" the sun for several hours after it had set below
or before it rose above the horizon.

These students proved to me that I did not have to be expert in the subject
to help them learn about and work effectively in a subject area of interest
to them. I ended up learning a lot too!

-----, "Standoff Single-Station Command Guidance Technology:
Experimental Results"(U), Tech Report 744, M.I.T. Lincoln Laboratory,
Aug 1986. (Secret)

This is a classified report, but it should be declassified around 2011!
There is Unclassified information included in the Appendices that document
how the system was designed and implemented. It is the results that are
classified. We implemented 8-ary codes with programmable
SAW convolvers operating over 160 to 320 MHz. There was a lot of high-speed
circuitry to synchronize acquisition and to decode received data.
The generation of modulating and demodulating codes involved a lot of digital
and analog design and overall presented a very capable
agile-beam state-of-the-art spread-spectrum modem link.
This flying system was able to demonstrate reliable performance when
subjected to substantial main-beam jamming. This was probably the best
technology project that I had the privilege to work on during my career.

Rebold, TA, & ------, "A Phase-Plane Approach to the Compensation
of High-Speed Analog-to-Digital Converters," IEEE Intl Symp on Ckts &
Sys, May 1987, Philadelphia, Pa. pp 455-458.

This study established the beginning of work to characterize ADC behavior
so that errors could be removed by post-processing methods for spectral
analysis applications. Here we made the premise that flash converter error
could be modeled using the ADC state and input signal slope at the time
the sample was acquired. This proved to be fundamental to most of
the successful dynamic compensation methods for several years until the
Pipeline and Sigma-Delta devices arrived on the scene and became technically
feasible. It is interesting that in ADC design, as in other fields as
well, if you start measuring a device for certain performance and specify
the procedures and errors, designers will "design out" the errors in
subsequent designs. That is in the case where a product
has enough volume to justify the design effort and cost.

Hummels, DM, -----, Cook, R, & Papantonopoulos, Y,
"Characterization of ADCs Using a Non-Iterative Procedure",
IEEE Intl Symp on Ckts & Sys, May 1994, London, England pp 5- 8, Vol.2.

This paper spells out the FFT procedures used to obtain ADC error functions
from a set of sinewave calibration tests. The method is by direct inversion
of a single matrix and obtains a result that provides error correction over
most of a given Nyquist band.

------, & Hummels,DM, "The Modulo Time Plot - A Useful Data Diagnostic
Tool", *IEEE Trans. on Instrumentation & Measurement*,
Vol.45 No.5, pp 734-738, Jun 1996

During the 1990's we wrote a lot of sophisticated papers dealing with
obtaining ADC error functions, but this one simple performance evaluation
procedure was always being used and taught to new students.
It is a simple procedure to determine if your data acquisition system is
really working correctly __before__ you invest time to collect calibration
data for an ADC. It is simple to use for any acquisition system and anyone
that ever used it was always impressed by how much you
could tell about what was going on. So we wrote it to document the procedure
for others to use. Basically, with sinewave testing, you align the data set
over one full period of the test signal and you can see whether you have noise
in the system, missing states (due to bad connections, etc),
over-driving the ADC (clipping), the statistical independence (or lack of)
in residual errors, etc. This is a practically useful procedure and a popular
procedure with our lab assistants.

------, Riley, KJ, Hummels, DM, & Friel, GA, "The Noise Power Ratio -
Theory and ADC Testing," *IEEE Trans on Instrumentation &
Measurement*, Vol 49, No.3, Jun 2000, pp 659-665 (ISSN 0018-9456)

Back in the mid-1980's when I started work on characterizing ADCs,
the noise power ratio (NPR) test was thought to be the best test for ADCs for
broadband signal applications. Initial work indicated that this was not true
since the test allowed excessive harmonic distortion to go
undetected. We tested ADCs that fit ideal 12-bit performance using existing
NPR theory but actually had enough distortion to exhibit 3 or more lost bits
of resolution when tested with sinewaves. So this paper resolves that issue
by finding out that using a random signal with uniform
pdf yields results that agree with sinewave testing whereas the signal with
a Gaussian pdf does not yield corresponding results. It was a relief to
finally demonstrate the difference through using a programmed DAC to generate
specific signals and then perform the NPR test with these
synthesized test signals. The paper also presents a full development of
the theoretical relationships for each type of signal and it provides the
basis for the results used in the IEEE Standard 1241 on ADC testing.

Irons, Fred H, *Introduction to Electrical Circuits and Signals*,

Vol.1, University of Maine Printing Services,

1st Edition, 1992, 2nd Edition, Dec, 2000, 414 pp

Irons, Fred H, *Introduction to Electrical Circuits and Signals*,

Vol.2, University of Maine Printing Services,

1st Edition, 1993, 2nd Edition, Apr, 2001, 453 pp

I wrote and used these texts for 5 years to introduce a design oriented and modern
circuits course to the 2nd year Electrical and Computer engineering students at the
University of Maine. It was a successful venture as the course, the text, and myself
all received high evaluations. The text is based on my experiences over the
years and it was a fun project to assemble and teach introductory circuit theory
to take advantage of computer math tools in a fashion that other popular texts do not do.
It was a pleasure to see the number of students that were inspired to go on for PhDs and
to become committed to an academic career as a consequence of their experiences in
learning the material in these texts. I have always felt that my receiving the
Maine Distinguished Professor Award was a direct result of teaching this course
and writing these texts.

Irons, FH, *Theory and Design of Active Filters for Integrated-Circuit Applications*,

University of Maine Printing Services Sep, 2003, 410 pp

This text is derived from a technical elective course that I taught in the '90's
on the subject of active filters. The text was written as a retirement project
to document the many techniques that are not used in other texts and to provide
a modern teaching reference for the subject. Unique approaches are included for
both the state variable approach and for switched capacitor techniques. The text
concludes with a chapter where the above paper on the approximation problem is
redeveloped using Matlab. In addition it is demonstrated on several examples and
shown to be a very useful approach to solving the filter design problem. The course
was always fun to teach as the students liked it very much and it always received high
ratings from the students. It was one of my favorite courses to teach.

You are visitor number
at this site.

Last update: Nov 23, 2011