Maker Pro
Maker Pro

Grounding design

Rlocus

May 17, 2016
2
Joined
May 17, 2016
Messages
2
Hi everyone,

I'm currently designing a small PCB to measure the pressure out of a wheatstone bridge. I use a conditioner from Texas Instruments (PGA309) and the ADS1251 (24bits) to convert in digital format.

I decided to separate the ground and power planes between digital and analog parts, to increase the quality of the results. I read many things about this, but I'll be pleased to have an external point of view of my design. It use 4 layers (signal-gnd-power-signal), and a picture is attached to this post ;)

Right after the power input, there is a power-on switch and a 5V regulator (low noise). Then the 5V output goes to the power plane with three vias and a decoupling capacitor. The ground "star" point is placed just above the regulator output, with a "analog return path" for the analog current. The noise from the digital part (MSP430) should not go this way (almost not) ?

On the left there is the analog part, with a voltage reference (followed by a OPA) and the the ADS1251 just under the PGA309. SPI lines and CLK lines have a 50 ohm resistor in serie and a TVS diode.
Two schottky diodes are mounted back-to-back with a resistor (1ohm) to reduce the ground potential difference (read in this article, fig2 : http://www.ti.com/lit/an/slyt499/slyt499.pdf)

I'd like to have an expert view to tell me if something look wrong or if I can improve this design in any way. I read that separating the planes could create a dipole antenna (or tripole in this case?). So maybe it is better to make only one power-plane and one ground plane for all...

Thanks in advance,
 

Attachments

  • PressureSensor.png
    PressureSensor.png
    49.9 KB · Views: 110

Arouse1973

Adam
Dec 18, 2013
5,178
Joined
Dec 18, 2013
Messages
5,178
My opinion on this Rlocus is that a split set of power 0 Volt planes is old fashioned and used to be used before people actually understood how the signal propagated. Made sense I guess at the time they had no other theory, bit like the don't use 90 degree bends in track rule, that is a myth.

For A.C signals the return path is right underneath the signal trace itself when reference to a plane. So you can see that layout and track spacing is much more important than separate 0 Volt planes to prevent crosstalk etc. And your right split planes if you are not careful can cause slot antennas which radiate and could cause possible interference. Well in fact any closed void in a plane can cause issues, even the plane clearance around a long through hole connector.

Adam
 

Rlocus

May 17, 2016
2
Joined
May 17, 2016
Messages
2
Thanks for your answer,

This is not the first time I heard about "old fashioned theory" of split planes, but both Texas Instruments and Analog devices (and others) have documents (wrote in 2012-2013) about grounding telling me to have an analog and digital plane. Here's a picture (taken from http://www.analog.com/library/analogDialogue/archives/46-06/staying_well_grounded.pdf) :

AD46-06_FIG_02.jpg


So as you say, maybe it is not necessary to split the planes, but instead to put the analog circuits separate from the current return path of digital circuits (as shown in previous picture).

Could I imagine the current return path as a straight line to the power input ?
687531path14697.png


Thanks in advance !
 

Arouse1973

Adam
Dec 18, 2013
5,178
Joined
Dec 18, 2013
Messages
5,178
I know I have seen some manufactures mention this. It might be because it's easier to explain and for the user to implement. D.C currents tend to spread out on the return plane where as A.C currents take the path of least impedance which is usually the path with the lowest loop inductance. This is because the magnetic fields produced by the send and return current is at its lowest right underneath the send trace. This is because of magnetic field cancelation from both send and return signals.

I am not saying it's wrong to split power planes, it's just not needed if the correct layout is achieved. See extract from a Douglas Brooks book.
Adam

Recall that currents— even power currents— flow in a loop. Therefore, it is important that the power loop be defined and unbroken. This typically means continuous planes for both power and ground available to all the circuits that use their currents.
Brooks, Douglas. PCB Currents: How They Flow, How They React (Prentice Hall Signal Integrity Library) (Kindle Locations 2768-2770). Pearson Education. Kindle Edition
.
 
Top