What are your favorite hardware tools that you use?

Agilent or Anritsu Vector Network Analyzers for S-parameter measurements on test vehicles. My students and I use them for extracting electromagnetic parameters of materials under study over wide frequency range.

What are your favorite software tools that you use?

Best of all I like Maple software for analytical modeling. Also, I like our own “home-made” EZ-FDTD tool (developed together with IBM) for electromagnetic numerical simulations of complex structures, including those containing dispersive materials.

What is the hardest/trickiest bug you have ever fixed?

I noticed some flaws in Maple software. I was persistently getting an unexplainable from the physics point of view divergence in the modeled system response. After careful inspection, I found out that the integral expression that I was using in my analytical model was programmed incorrectly in the Maple library. The analytical expressions in the integral had to be different for different ranges of parameter, but they have given just one expression, which was incorrect for my case. I had to write this expression in Maple manually, setting aside the library one. I had the similar situation with calculating elliptical integrals as well. Sometimes there are missing solutions, and one must be very careful when using commercial software, adapting it for specific problems to be solved.

What is on your bookshelf?

I have lots of books in my bookcase – both technical and feature. As for technical literature, I will name some of my favorite textbooks and books that I always use in my work, both in English and in Russian. In English, these are D.M. Pozar “Microwave Engineering”, R.E. Collin “Field Theory of Guided Waves”, C.A. Balanis “Antenna Theory”, G. Korn and T. Korn “Mathematical Handbook”, P.S. Neelakanta “Handbook of Electromagnetic Materials”, G. Milton, “Theory of Composites”, A. Sihvola “Electromagnetic Mixing Formulas and Applications”, and B. Lax and K. Button “Microwave Ferrites and Ferrimagnetics”. In Russian, these are the books written by my Teachers: S.I. Baskakov “Radio Engineering Signals and Circuits”, D.M. Sazonov “Antennas and Microwave Devices”, Landau and Lifshitz “Electrodynamics of Continuous Media” (Course of Theoretical Physics”.

Do you have any tricks up your sleeve? (special way to analyze circuits, special process you use to make something, etc.)

Common sense. I verify solutions and results with common sense and so-called limit transitions to the known canonical solutions or other known results. When I realize that there is a contradiction to common sense or expectations, I start either looking for a mistake or start the process (measurement, derivation of formulas, calculations) again. Sometimes the very basic assumptions turn out to be wrong, or an initial hypothesis does not work. Then I have to modify the initial guess and build the model based on the new assumptions. In many cases this works. If we knew from the very beginning what we are doing, this will not be call a research!

What has been your favorite project?

I like almost all my projects on which I have ever been working, because as soon as I get deeper in any project, I get more interest of discovering or building models for new things. My recent favorite project is on separating conductor loss from dielectric loss, when taking into account conductor surface roughness. This is a project that we do in the EMC lab in Missouri S&T, and it is sponsored by Cisco.

Can you tell us a little more about your project on separating conductor loss from dielectric loss?

Correct data on dielectric properties of laminate dielectrics (dielectric constant Dk and dissipation factor, Df, which is the same as the loss tangent) used in printed circuit boards is need to high-speed digital electronics designers and PCB manufacturers. It is known that printed circuit boards have substantial levels of conductor (copper) roughness for adhesion purposes. However, if the roughness effects upon signal propagation on the printed circuit boards are neglected or underestimated, especially over the gigahertz frequency range, the dielectric properties extracted using traveling-wave methods may result in significant errors from signal integrity point of view.

We have developed the differential extrapolation method (we called it DERM – from the Greek “derma” – skin).

The method we propose to separate conductor loss from dielectric loss on actual PCBs is based on the traveling-wave method of measurements. Special pattern for test vehicles have been designed. Any test vehicle contains a test 50-Ohm 15-inch-long stripline and a number of auxiliary lines of different lengths for Through-Reflect-Line (TRL) calibration over the wide frequency range (from 10 MHz till at least 20 GHz, and potentially up to 50 GHz). Magnitudes and phases of all four complex S-parameters of the two-port test vehicles are measured over the frequency range of interest either in frequency domain using a VNA, or in time domain, using a TDR. We have developed an algorithm to extract complex permittivity data from these measurements by solving corresponding systems of equations. However, conductor loss affects the extracted results.

If the conductor were perfectly smooth, it would be easy to separate it from the dielectric loss, since conductor loss is known to behave as a square-root of frequency, while dielectric loss is proportional to frequency. However, this turns out to be incorrect in the case of a rough conductor, whose loss partially lumps into dielectric loss. We have found that in reality, loss in a PCB laminate dielectric behaves as a sum of two terms: one is proportional to frequency, and the second is proportional to the squared frequency. Rough conductor loss turns out to have three terms: a square-root of frequency, linear to frequency, and proportional to the squared frequency.

To correctly separate rough conductor loss from dielectric loss, a few (at least three) samples of test vehicles with identical dielectric and geometry, but different conductor surface roughness profiles are needed. Then the analysis of frequency behavior of loss curves on all the samples and the differences of the loss curves in pairs is done. If the roughness profiles are known, for example, from microscopic analysis of copper samples, the corresponding frequency contents of the measured total losses are analyzed as functions of a roughness parameter, e.g., average peak-to-valley amplitude. This will allow for predicting, what frequency content will be obtained, if roughness level is zero. Thus extrapolation of frequency components to zero roughness would give actual values for dielectric loss.

The analogous approach can be used if test vehicles under study are of the same dielectric, almost the same roughness, but different geometry (width of the signal trace). We called it DERM-W (W stand for “width”). Analysis of frequency contents in differential (in pairs) loss in test vehicles with different geometry will result in correct unambiguous values of dielectric loss.

By the way, the similar DERM and DERM-W techniques can be applied to extracting Dk values, too, though Dk values are not affected by the rough conductor loss as much as Df values.

Currently, we have been working on the dielectric parameter extraction tool, which will be a standard tool for the Consortium iNEMI (Intenational Electronics manufacturing Initiative). Also, we have been building a tool for roughness profile characterization using the cross-sectional analysis of PCB striplines. The images can be obtained with scanning-electron-microscopy (SEM) or optical microscopy (OM). The tool will allow for getting high-contrast profile pictures and for their statistical analysis to extract roughness amplitude parameters and correlation length, or a quasi-period of a roughness profile function.

The next stage of the project is the development of empirical design curves for high-speed digital designers so that it would be possible to estimate an effect of conductor roughness upon signal propagation characteristics (loss and frequency dispersion), and subtract rough conductor loss from the total loss, if a designer has only one type of a PCB with a known type of a conductor, e.g., standard foil, very-low profile foil, or hyper-/super-very-low-profile.

Do you have any note-worthy engineering experiences?

Many years ago I was working on my PhD dissertation, where I needed to design an 8-mm wavelength comparatively high-power generator (100 microwatt in continuous regime of operation). I used avalanche diodes in a rectangular waveguide cavity made of copper in my design. To increase the Q-factor of the cavity, every time before using the generator I cleaned the surface of the cavity with orthophosphoric acid. Since it was comparatively expensive, I used Pepsi-Cola or Coca-Cola to remove oxidation film on the copper surface. Though this was pretty successful, I burnt tens of expensive avalanche diodes. Until now I am puzzled: was it because of soda I used for cleaning?

What are you currently working on?

Currently, I have been working on three different research topics. One is related to dielectric parameter extraction on printed circuit boards. The second is developing mixing theory to predict electromagnetic properties of magneto-dielectric composites and engineer materials with desirable electromagnetic frequency responses. The third is development of a model for a ferrite filtering choke to reduce common-mode currents on cables.