PCB/Circuit Design and Modeling
I have a substantial amount of experience in PCB design, from a circuit design perspective as well as actual circuit board layout. The primary work I've done has been in designing boards from a more geometric perspective: due to the high currents and voltages achieved, the actual layout of copper shapes on the PCB is more important a consideration than the circuit design itself, which tend to be fairly simple in pulsed power. I've also designed and developed PCBs that are primarily TTL logic, with an array of reasonably complicated logic-based circuitry for use in EHT's nanosecond pulser line.
I've worked with National Instruments' Ultiboard, Eagle, and dominantly PCB123, the proprietary software linked with Sunstone Circuits. Broadly, my experience with circuit design and modeling can be summarized as such:
- Ability to design, model and create high-power circuity and logic-based control circuitry
- Familiarity with dielectric strength/breakdown voltages
- Intuitive understanding of inductive and capacitive effects, and how to minimize them
- Familiarity with permitivity/permeability of materials and the way it affects performance
- Understanding of impedance matching, and EMI broadcast reducing/shielding
High Power Circuit Design
Much of the time in pulsed power, actual circuit designs are exceedingly simple. In general, they involve a set of storage capacitors, some sort of switching component, and a load. Sometimes they include reactive elements for pulse shaping or sharpening, but by and large they are simply represented schematically when disregarding stray components. As discussed on the page regarding HV/HF Measurements, these stray components can often dominate intentional ones at high currents, voltages, and frequencies. Capacitive energy scales as V2, and inductive energy as I2, so minimizing capacitance and inductance quickly becomes critical.
Board design work then quickly transitions from circuit design to more of a theory problem to find the optimum current path to move the energy from the storage capacitors to the load, potentially through a transformer? This can be a question of just minimizing inductance, in simple cases. More frequently however, it involves a careful balance between stray L and C. This is because the impedance of a current path goes with the ratio of the two, sqrt(L/C), and at high frequency impedance matching can be as critical as anything else to eliminate dangerous reflections. Eventually, perhaps, computers will be able to immediately create the perfect plane geometries and current paths for us, but as of now, all this design work is completed through intuition and knowledge gained from experience.
Another aspect of high-power circuits is designing in a way that minimizes both broadcast of, and susceptibility to, electromagnetic interference. This involves careful consideration of how currents flow through different regions of a system, and ensuring these paths are both low inductance (to prevent voltage spikes) and separated from more sensitive drive circuity. In addition, shielding layers are often necessary to protect gate-drive circuitry, with considerations of issues of skin depth and impedance matching.
Finally, high-voltage standoffs must be considered. Even below arcing thresholds, other types of plasma discharge can develop. Often termed "corona", the very weakly ionized air can create reactive species that degrade dielectric insulation over time, leading to failure. As a result, minimizing sharp corners and points in the design is important.
I have several years of experience designing PCBs in this way, and I've loved the style of analytic and creative thinking involved in the process.
Logic-Level Circuit Design
I've also worked in designing the control logic for the nanosecond pulser line. Because the propogating signals are very short and require strong noise immunity, standard logic-based chips can be preferable to digital circuitry in the proximity of the high power switching circuitry. I've created circuitry tnvolving basic logic chips, timers, and a number of types of flip-flops to create systems that will reject noise of different frequencies and pulse widths.
Beyond the copper-based shielding included in the PCB design discussed above, additional components can be included to protect against noise. For regions with large potential fluctuations, optocouplers and optical fiber breaks can be implement to protect logic circuitry. Board-mounted shells and cages can be used to further protect against broadcast noise. Finally, in all cases, matching trace impedances is critical for preventing reflections and spurious signals in the drive traces. Avoiding trace bifurcations and ensuring an impedance-matched load at the end of each signal traces helps to protect signal fidelity.
I've worked frequently with National Instruments' Multisim software to model different types of circuits. Most of my modeling work has been exploratory, looking into new circuit topology ideas or attempting to better understand the behavior of our current designs. Much of the time these circuits are dominated by reactive elements, as is the case with pulse forming and sharpening networks, which makes modeling extremely useful in determining the dominant effects and timescales.
PCB Design Guidelines
I've included below a list of common PCB design guidelines and information, for future reference.
Also, see the files below for further "Best Practices" Documentation