Connecting VDD to VDDA rail

[ghamille]

Hi all,

I have seen a lot of designs around STM32 microcontrollers (blogs and videos) where they mention doing “additional filtering” for the VDDA pin. I think I understand why
Usually though, they will also mention the power comes from the already filtered VDD rail. As a newbie, I am wondering how to connect the VDD rail to the VDDA one.

In the attached diagram I am working on, would I connect 1 --> 3 or rather 2 --> 3. I would love to also understand why it has to be connected a certain way. Any pointer to learning resources would be appreciated.

Thanks a lot for your help!

[selcuk]

You may connect circle 1 to circle 3. Since circle 2 is GND net, you cannot connect to it. I guess circle 3 continues to VDDA through the ferrite bead.

The purpose is to prevent the noise on VDD net from propagating to the VDDA net. But if you don't place the ferrite bead intentionally between the VDD and VDDA, it is better to not use it (i.e. short VDD to VDDA). If you know the noise frequency and characteristics and can calculate a bead value then you can use it. Otherwise you cannot be sure about the effectiveness or signal degradation.

You may read a book about EMC or signal integrity to have better knowledge about filters. This is an open access one.

Filters:
https://link.springer.com/chapter/10.1007/978-3-031-14186-7_15

Ferrite beads:
https://link.springer.com/chapter/10.1007/978-3-031-14186-7_11

[ghamille]

Thank you for the resources.

For the decoupling, I was going by the datasheet for my specific microcontroller (STM32L052K6T6: https://www.st.com/resource/en/datasheet/stm32l052k6.pdf).
So what you are saying that if I don’t calculate the exact value for the ferrite bead, might as well connect all of those capacitors (both for VDD and VDDA) in parallel, and use that to power both pins?

[wasedadoc]

Are you using the any of the ADC or DAC?  If not the benefit of a filter is irrelevant and you can connect VDDA and VDD directly.

[ataradov]

So what you are saying that if I don’t calculate the exact value for the ferrite bead, might as well connect all of those capacitors (both for VDD and VDDA) in parallel, and use that to power both pins?

If you know that your application will potentially have lots of switching noise and you care about performance of the analog blocks, then place the ferrite to have options to experiment later.

You may not  be able to always calculate the exact value you need, sometimes you just build the board and try different values. And in the end you may end up with just a 0 Ohm resistor instead of a bead.

But generally, if you are not optimizing the cost to the cents, having some generic bead like 470 Ohm @ 100 MHz will at least not hurt anything.

[selcuk]

As suggested, you may use a generic value ferrite, or 0 ohm to calculate later. You may short VDD to VDDA and will be okay for most of the time. This is even better than a wrong ferrite.

But you shouldn't parallel the capacitors of different pins of the MCU in the layout. Even if you short the VDD to VDDA, you need to use recommended capacitors of the datasheet for each supply pin. There is one 100nF and 1uF for VDDA and you need to place them very close to VDDA pin and preferably without vias between pin and capacitor. There is another one 100nF and 1uF for Vref and you need to place them very close to Vref pin. There is also one 10uF and N x 100nF for VDD and you need to place 100nFs to each VDD pin one by one. Then place 10uF to a common point to feed 100nFs.

[ghamille]

Are you using the any of the ADC or DAC?  If not the benefit of a filter is irrelevant and you can connect VDDA and VDD directly.

That’s a good point, I’m only using a few I2C components together and I guess these are considered digital and not using ADC or DAC?

[ghamille]

As suggested, you may use a generic value ferrite, or 0 ohm to calculate later. You may short VDD to VDDA and will be okay for most of the time. This is even better than a wrong ferrite.

But you shouldn't parallel the capacitors of different pins of the MCU in the layout. Even if you short the VDD to VDDA, you need to use recommended capacitors of the datasheet for each supply pin. There is one 100nF and 1uF for VDDA and you need to place them very close to VDDA pin and preferably without vias between pin and capacitor. There is another one 100nF and 1uF for Vref and you need to place them very close to Vref pin. There is also one 10uF and N x 100nF for VDD and you need to place 100nFs to each VDD pin one by one. Then place 10uF to a common point to feed 100nFs.

Got it. The reason I thought I should put them in parallel is this video, where he goes on to explain how this (slightly different) STM32 MCU is wired:


I think I may not understand what he’s doing or the concept of “parallel capacitors” fully then

[selcuk]

Ah yes. They are parallel in your schematics. That is not wrong. The Net connections are correct. But while doing the layout, you need to pay attention to connect one 100nF to pin 1, one to pin 24, one to pin 36 and one to pin 48. So that, there are 4 VDD connections and 4 100nF capacitors. You will route 3.3V to the capacitor then to the pin for each of them. Each pin will draw the current from its related 100nF capacitor. Those capacitors will draw from 10uF capacitor and goes like that.

If you don't use 120ohm ferrite, then you still need to connect that 1uF and 10nF to pin 9.