Noise from oscillator on top layer to clock on bottom layer with VCC & GND layers between
What effect does an oscillator potentially have on a SPI clk signal if the oscillator is on the top layer, with GND and VCC layer in between on a four layer board. I know running it on the same layer within a certain distance will have coupling effects that could cause noise and the 3W rule. I don't see people talking about different layers as much though. The clock is on the bottom like in the picture below. I know speed of oscillator and clock will likely determine if it will actually cause issues, but I am wondering about the general physics/practical noise that could be caused.
2 answers
It's hard to answer your specific case since you didn't label the layers in your picture, or which traces are doing what.
That said, a few traces crossing each other on different layers is not much coupling. In most cases of a typical board with a typical "SPI clock signal", there won't be any problem. The SPI clock signal is driven actively both directions by some CMOS output, probably with a few 100 Ω or less impedance. Even a few pF of capacitance to another signal shouldn't have much effect.
What effect does an oscillator potentially have on a SPI clk signal if the oscillator is on the top layer, with GND and VCC layer in between on a four layer board.
Essentially none if there is at least one complete ground or power plane between the two signals. Any plane with zero impedance relative to the signals, like a ground plane, will completely block capacitive effects.
There can still be small inductive effects, but those depend on current, not voltage, and require conductors running parallel to each other for some distance. Something like an SPI clock on a typical board should be basically unaffected.
Also keep in mind that with a 4 layer board and the two inner layers being used for ground and power planes, the signal are on the two outer layers. That puts them the full thickness of the board apart, usually about 62 mils (1.6 mm). Do the math to compute the capacitance at that separation for plausible areas, and you should see how minimal it is. And, that would only apply in places where there are no planes or traces in the inner layers.
What you are proposing may be okay (depending on other specifics of the design), but your proposed 4-layer PCB stack-up looks like a compromise. Clock is a high speed signal, and so is SPI depending on the edge raise time. Clock and SPI are on different layers: top and bottom. What acts as reference layers for these signals?
A high speed signal should have a reference plane layer next to it. A GND plane layer serves as good reference plane in theory and in practice. A VCC plane layer can serve as a reference plane in theory, but there are practical hindrances. If you want to route a signal through a via from the bottom layer which uses VCC plane as a reference to the top layer which uses GND plane as a reference, then you have to connect the reference planes near the signal via. How would you connect the reference planes if they are at different potentials? You would connect them through a coupling capacitor. But then you have to have a lot of these coupling capacitors, which isn't convenient.
As a result of the above considerations, there are two non-compromised options for 4-layer PCB stack-ups.
1 - high speed signals, sensitive signals
2 - GND plane
3 - VCC plane, other power distribution
4 - low speed and non-sensitive signals
If not enough real estate for high speed signals on the top layer:
1 - high speed signals, sensitive signals
2 - GND plane
3 - GND plane
4 - high speed signals, sensitive signals
If you find yourself hard pressed to do a 4-layer PCB with two GND planes, then consider a 6-layer PCB:
1 - high speed signals, sensitive signals
2 - GND plane
3 - VCC plane, other power distribution
4 - low speed and non-sensitive signals
5 - GND plane
6 – high speed signals, sensitive signals
That's a common stack-up for a 6-layer PCB, although not the only good 6-layer stack-up.
1 comment thread
0 comment threads