I had been working at a new job for 2 weeks as a senior design engineer at an IC design company when my new manager asked me to help a young engineer with a problem she was having. The silicon had just come back from the manufacturer and her internal voltage regulator was oscillating. This had been an issue with the previous design, and the team was distraught over not having solved the problem.
We discussed the issue in her office. Phase margin, load capacitance, process limits etc. all seemed fine. I looked over her simulations and they also did not show any problems. I then asked, “Are you sure the silicon is oscillating?”
She took me down to the lab where the chip was mounted in an evaluation board. She put the oscilloscope probe on her chip’s internal regulator output pin and, sure enough, the oscilloscope showed a nice 100 kHz sine wave with an amplitude of 1 Vpp.
It was suspicious, though, that this sine wave had no noticeable distortion. Typically, an oscillating regulator would display something closer to a square wave or a highly distorted sinewave. I put my finger on the regulator’s output and noticed that there was no change to the frequency or wave shape. We added a 4.7 µF capacitor at the regulator’s output, but again, the frequency and shape didn’t change. Now I was really suspicious.
I told her to look at the chip’s power supply from the board. Lo and behold, the 100 kHz sine wave was riding on top of the 3.3V DC. What was going on? She said others had told her that the power supply was being affected by her regulator’s oscillating load. The oscillating load could be creating the 100 kHz sine wave on the power supply, but I couldn’t see how her regulator could produce such a low distortion sinewave whose frequency was independent of the load capacitance.
Looking over the evaluation board’s schematic, I saw that the power supply consisted of a linear regulator, a 2.2 µF ceramic capacitor, and a ferrite bead. I looked up the ferrite bead’s part number; its inductance was 1 µH. Now the problem was obvious. The 1 µH inductor and the 2.2 µF capacitor were resonating at 100 kHz and the board’s linear regulator was oscillating. We removed the ferrite bead, and everything was fine.
When I asked the evaluation board designer why he added the ferrite bead, he replied that this circuit had been used for years in previous boards. The problem was, technology had changed over those years. The original boards had used electrolytic capacitors and 100 nH ferrite beads. The electrolytic capacitors had high series resistance compared with the modern ceramic capacitor now being used, by two orders of magnitude. I also asked why they had increased the ferrite bead inductance and was told that the new ones had more inductance for the same size as the original and had greater attenuation at frequencies > 10 MHz.
Ferrite beads are extremely effective at attenuating unwanted signals > 10 MHz because they are lossy above these frequencies. At lower frequencies (100 kHz), though, they are not lossy.
In the circuit on the new evaluation board the 2.2 µF capacitor and 1 µH inductor formed a high-Q resonant circuit that the board’s linear regulator’s feedback loop couldn’t handle, and so it oscillated. A smaller inductor of 100 nH increases the resonant frequency to 1 MHz. This is in the range where the ferrite becomes lossy and the Q of the circuit is much lower, so the linear regulator was stable in the old board design.
Be aware: the evaluation board for your new IC design is also a new design.
Marshall Bell is a retired analog circuit designer who has worked for Bell Laboratories, Tektronix, National Semiconductor, SMSC, and Synaptics Corporation. He has 20 patents and has been a licensed ham radio operator since 1968.