Piezoelectric Effect in CPU power circuit ?

[fzabkar]

Does anyone understand this?

https://www.dell.com/support/article/en-au/sln80281/the-solid-state-drive-ssd-on-my-laptop-makes-a-strange-squeal-buzzing-noise?lang=en

Some portable systems equipped with solid state drives (SSD) emit an audible squeal when the system is idle or under light use. This is normal behavior.

In order to save energy when the Central Processing Unit (CPU) is idle, the CPU can be set to enter a lower-power mode, also known as a "C-State". The intent of lower power modes is to cut power use inside the CPU. In certain situations, a high-frequency buzzing noise can be heard from the system. This buzzing noise appears with a change in processor activity, SSD load, or if Universal Serial Bus (USB) devices are attached.

The noise comes from the processor’s power circuit and can only be heard when the processor is in C3 (clock-stopped or sleep) power state. The noise is generated from the processor's power circuit, which is caused by a phenomenon referred to as the Piezoelectric Effect. When a specific voltage is applied to these solid state components, they begin to resonate producing sounds that fall within the range of human hearing (15 – 20 KHz).

[wraper]

What to not understand. It's known for a long time that class II/III ceramic capacitors may sing under certain circumstances. They basically act as piezo buzzer when ripple current flows through them. The most common occurrence is in PC graphics cards under 3D load. When there is ripple current of audible frequency flowing through MLCC, you may hear it.

[wraper]

https://product.tdk.com/en/contact/faq/31_singing_capacitors_piezoelectric_effect.pdf

[fzabkar]

Doh, I forgot about capacitors. In fact one of my first power supply builds had a "singing ceramic capacitor", and it was a linear, dual tracking topology which I copied from the application circuit in an NS datasheet.

Thanks.

[EEEnthusiast]

Even power inductors make this sound due to magneto-striction and other effects. I had run into a problem on one of my designs that required me to change the inductor and the switching frequency to get the sound below reasonable levels.
An inductor is nothing but a speaker without the diaphragm...

[magic]

The noise comes from the processor’s power circuit and can only be heard when the processor is in C3 (clock-stopped or sleep) power state. The noise is generated from the processor's power circuit, which is caused by a phenomenon referred to as the Piezoelectric Effect. When a specific voltage is applied to these solid state components, they begin to resonate producing sounds that fall within the range of human hearing (15 – 20 KHz).

Misleading gibberish but roughly true. The noise is caused by periodically varying voltage across ceramic caps. It doesn't happen when the CPU is in C3 state, it happens when the CPU periodically changes C states or P states (possibly including turbo boost). There is also some inter-core dependence, in that usually all cores run from the same voltage and it has to be the highest voltage requested by any core.

You can play with it using tools like c2ctl or k10ctl. The noise goes away if you configure equal voltage on all P states and perhaps disable some of the deep C states (or keep one core at 100%).

It's been awhile, I don't remember details.

edit
I also think that no resonance is required, it will work at any frequency. That being said, equal excursion at lower frequency yields lower SPL from any acoustic transducer, so you typically only hear the highest frequencies. As for the real content of the noise, it's likely 1kHz and tons of its harmonics because many things in software are scheduled with millisecond granularity.

[fzabkar]

I was focusing on the statement that "a specific voltage is applied to these solid state components", which implies that the semis are singing. Even more confusing is that the symptom is a noisy SSD.