Partial Discharge in High-Voltage Power Electronics: 7 Brutal Lessons I Learned the Hard Way
Listen, if you’ve ever stood in a high-voltage lab and heard that faint, rhythmic "crackle" that sounds like bacon frying when there’s no breakfast in sight, you know the feeling of pure, unadulterated dread. That sound isn't just physics; it's the sound of your expensive SiC MOSFETs and insulation systems slowly committing suicide. We’re talking about Partial Discharge (PD)—the silent killer of modern power electronics. As we push for higher power densities and faster switching speeds, we’re essentially inviting this ghost into our machines. Today, I’m pulling back the curtain on how to spot it before it spots your profit margins, and the design rules that will actually save your skin. Grab a coffee; this is going to be a long, electrifying ride.
1. What is Partial Discharge? (The Monster Under the Bed)
Imagine you have a solid block of high-grade insulation. To the naked eye, it looks perfect. But inside, at a microscopic level, there’s a tiny air bubble—a "void." Because air has a much lower dielectric constant than the solid around it, the electric field bunches up inside that bubble like commuters trying to squeeze into the last train car. Eventually, the air breaks down, a tiny spark jumps across the void, and boom—you have a Partial Discharge.
It’s called "partial" because it doesn't bridge the entire gap between conductors. If it did, you’d have a full-blown short circuit and a very expensive fire. Instead, it’s a localized breakdown that eats away at your insulation, molecule by molecule, through chemical degradation and UV radiation. It’s the slow-motion car crash of the engineering world.
2. Why High-Voltage Power Electronics Are PD Magnets
Back in the "good old days" of slow-switching IGBTs, we could get away with murder. But now? With Silicon Carbide (SiC) and Gallium Nitride (GaN), we are dealing with $dv/dt$ rates that make my head spin. When you switch 800V in 10 nanoseconds, you’re creating high-frequency harmonics that love to trigger PD.
Moreover, we are shrinking everything. Power density is the new king, which means conductors are closer together, electric fields are higher, and the margin for error is effectively zero. If your clearance and creepage distances aren't perfect, or if your potting compound has a single microscopic bubble, the high-frequency stress will find it. It's not a matter of if, but when.
3. 7 Crucial Partial Discharge Detection Methods
You can't fix what you can't see. Or hear. Or sense. Here are the seven ways we hunt for this ghost in the machine.
Method 1: Electrical Detection (IEC 60270)
The gold standard. We use a coupling capacitor to catch the tiny current pulses created by the PD. It’s incredibly sensitive, measuring in picoCoulombs (pC). The catch? It’s a nightmare in a noisy environment. If your lab has a lot of EMI, your PD signal will be buried under a mountain of digital trash.
Method 2: Acoustic Sensing (The "Bacon" Test)
Using ultrasonic sensors to "listen" for the mechanical shock waves produced by the discharge. It's great because it's non-invasive and immune to electrical noise. However, by the time the sound is loud enough for a cheap sensor to hear, the damage might already be significant.
Method 3: Optical Detection
Partial discharge emits light—mostly in the UV spectrum. We use UV cameras or photomultiplier tubes to see the "corona" glow. This is fantastic for identifying surface PD on bushings or insulators, but it won't help you see inside a potted transformer.
Method 4: Ultra-High Frequency (UHF) Detection
PD pulses are lightning-fast, which means they radiate energy in the 300MHz to 3GHz range. UHF antennas can pick this up. It’s the "cool kid" of PD detection because it can filter out low-frequency grid noise, making it perfect for GIS (Gas Insulated Switchgear).
Method 5: Dissolved Gas Analysis (DGA)
Used primarily in oil-filled transformers. As PD eats the oil, it spits out gases like Hydrogen and Acetylene. By "blood testing" the oil, we can tell if the transformer is sick. It's slow but incredibly reliable for long-term health monitoring.
Method 6: High-Frequency Current Transformers (HFCT)
You clamp these around the ground lead. They catch the high-frequency transients of the PD without needing to touch the high-voltage side. It's the safest way to do "on-line" monitoring while the system is running.
Method 7: Transient Earth Voltage (TEV)
When PD happens inside a metal cabinet, it induces tiny voltage spikes on the outer metal surface. Handheld TEV sensors allow you to walk around a substation like a dowser looking for water, but instead, you're looking for impending doom.
4. Design Rules to Kill PD Before It Starts
Designing for high voltage isn't just about following a spreadsheet; it’s an art form. If you want to avoid Partial Discharge in High-Voltage Power Electronics, you need to obey these "laws of the jungle."
1. Respect the Triple Point
The "triple point" is where the conductor, the solid insulation, and air (or vacuum) meet. This is where the electric field goes absolutely berserk. You must shield these points or use stress-grading materials to smooth out the field. Ignore this, and you’re basically building a lightning rod inside your device.
2. Vacuum Degassing is Non-Negotiable
If you are potting your electronics, you cannot just pour the resin and hope for the best. You need a vacuum chamber. You need to suck every single molecule of air out of that goop. A 1mm bubble in a 10kV field is a ticking time bomb. I’ve seen prototypes last 100 hours and then fail because a single bubble finally ionized.
3. Use "Rounded" Everything
Sharp edges are the enemy. A sharp corner on a PCB trace or a busbar acts as a field concentrator. Use large radii. If it looks like it could cut you, it will definitely cause PD. In the high-voltage world, "curvy" is "reliable."
4. The 80% Rule of Paschen’s Curve
Paschen’s Law tells us the breakdown voltage of air based on pressure and distance. Never design right at the limit. I always aim for my operating voltage to be no more than 80% of the Inception Voltage (PDIV). This gives you a buffer for humidity, temperature, and manufacturing defects.
5. Common Mistakes That Blow Up Your Prototypes
I've made them all, so you don't have to. Here’s the "Wall of Shame" for high-voltage design:
- Ignoring Humidity: You test your board in a dry lab in Arizona, and it works perfectly. Then you ship it to a customer in Singapore, and it arcs over in three days. Humidity lowers the breakdown voltage significantly.
- Contamination: A fingerprint on a PCB might seem harmless. In a 5kV environment, that oil and salt create a conductive path that leads to tracking and—you guessed it—PD.
- Solder Balls: Those tiny "splatters" of solder? They are miniature mountains of field concentration. Clean your boards. Then clean them again. Then use a microscope.
6. Advanced Insights: The SiC and GaN Challenge
The industry is moving toward Wide Bandgap (WBG) semiconductors, and while they are amazing for efficiency, they are PD's best friends. The fast $dv/dt$ causes non-uniform voltage distribution across motor windings or transformer layers. This means the first few turns of a winding might see 80% of the voltage stress. If you aren't using "corona-resistant" wire, you are in for a world of hurt.
7. Infographic: The PD Survival Guide
PD Prevention Checklist
Step 1: Geometry
- Round all sharp edges
- Increase radii on traces
- Avoid "needle" points
Step 2: Materials
- Vacuum-degas resins
- Use high-CTI laminates
- Verify dielectric constants
Step 3: Testing
- Measure PDIV/PDEV
- Thermal cycle testing
- Humidity stress tests
© 2026 Kunseu Engineering Insights - High Voltage Division
8. Frequently Asked Questions (FAQ)
Q: What is the difference between PDIV and PDEV?
A: Partial Discharge Inception Voltage (PDIV) is the level where PD starts as you increase voltage. Extinction Voltage (PDEV) is the level where it finally stops as you decrease voltage. Crucially, PDEV is usually lower than PDIV. Once the "fire" is lit, it stays lit longer than it took to start!
Q: Can I ignore PD if it's below 10 picoCoulombs?
A: It depends on your insulation. For some liquid or self-healing systems, 10pC is "noise." For solid epoxies or thin films, 10pC can be a death sentence over 5-10 years. Always check your specific insulation’s endurance curve.
Q: How does frequency affect Partial Discharge?
A: PD is basically a "per-cycle" event. If you increase the frequency (like in high-speed switching), you increase the number of discharges per second. This accelerates the degradation linearly. 100Hz PD is annoying; 100kHz PD is a blowtorch.
Q: Is surface tracking the same as Partial Discharge?
A: They are cousins. Surface tracking is the progressive formation of conductive paths across an insulator. PD often acts as the "scout," damaging the surface and making it easier for tracking to occur.
Q: Why is Silicon Carbide (SiC) more prone to PD issues?
A: It's not the material itself, but how we use it. SiC allows for much faster switching ($dv/dt$), which creates high-frequency stresses and ringing in the system that can exceed the PDIV of standard insulation systems designed for 60Hz.
Q: Does temperature affect PD?
A: Absolutely. Higher temperatures usually lower the air density in voids (if they aren't hermetic) and can change the dielectric constant of solids, typically making PD more likely to occur at lower voltages.
Q: What tools do I need for a basic PD test lab?
A: At minimum, you need a high-voltage AC/DC source, a coupling capacitor, a high-frequency sensor (HFCT or antenna), and a high-bandwidth oscilloscope. For official certification, you'll need a calibrated PD meter.
Conclusion: Don't Let the Ghost Win
Partial discharge is one of those engineering challenges that reminds us that nature always finds the path of least resistance. You can have the best firmware, the fastest switches, and the most efficient cooling, but if your insulation fails, your product is just a very expensive paperweight.
The key is proactive design. Don't wait until the prototype stage to think about electric fields. Use simulation tools (FEA is your best friend here), respect the chemistry of your materials, and for the love of all that is holy, get those air bubbles out of your potting.
Would you like me to run a quick calculation on your specific creepage distances or recommend a specific potting compound for your project? Let me know!
