PCB review + Analog/Digital separation?

I have built a PCB using this schematic used to charge a small wireless device.

This schematic is based on this Texas Instruments evaluation module: https://www.ti.com/lit/ug/slvu928c/slvu928c.pdf?ts=1594940372649

On top of what ti offers, I have thrown in a microcontroller to record the digital signals (LEDA, LEDB, LEDC)

This board is going to be a raspberry pi 4 shield and I am going to use the i2c bus of the atmega328p to send information about the digital signals mentioned above.

I am still learning electronics and I am debating whether I should have connected the analog and digital grounds on the board. Right now the grounds are separated, one going to GND and one to GND (picture 2 raspberry pi schematic)

The problem I am getting is that I get weird voltages on the board.

Here are some examples of voltmeter measurements:

• 5V source - GND = 5V
• 5V source - AGND = 2.6V
• TLV70033 3V3 output - GND = 1.96V
• TLV70033 3V3 output - AGND = -0.48V
• ATMEGA328p 3v3 input - AGND = -0.48V
• TLV70033 EN output - AGND = 3V

Here's a link if you want to look at the actual board. Go at REV1

Here's a list of questions I have:

1. Why is there a voltage drop on 3v3 output (was looking for 3.3V but now <2)
2. Why do I get negative voltage between ATMEGA328p 3v3 in and AGND?
3. What does BSS138 do in each case? (Q2/Q3)
4. What does DTC114 do? (Q1)
5. What is the purpose of ground separation and how is it done correctly?
• The resistors that you are using on DPWM-1A/B in the TI example schematic look like they are 10 ohm, not 10K, any reason you are using resistors 1000 times bigger? Why don't you have V33A hooked up on the BQ500212A? Your GND and AGND are shorted together, are you sure you want that? R32 looks to be a different value than the reference design... Can you go through and explain the reasoning behind the differences? Aug 17, 2020 at 12:27
• a) Pure mistake - well spotted, fine gentleman. b) I did not understand how the V33A worked so I left it out. Could you explain this to me? c) You mean shorted at raspberry pi level or before? If at raspberry pi, shouldn't these meet at some point? d) It seems to be another stupid mistake. I think I didn't believe there's such a small jump between the two resistance values. I am going to review these mistakes. Aug 17, 2020 at 13:08
• Typically you'd only have GND and AGND connected at one point to avoid a ground loop, and to reduce the noise between the two. Typically you connect them with some kind of filter, like a ferrite bead or capacitor. (Your schematic has them through 2 4.7uF caps as well). V33A is the analog 3.3V supply, which can come from the digital supply through a 10-ohm resistor and decoupling caps, but it has to be connected. R32 in your schematic is a temperature sensing setpoint, so you want to make sure you are using it correctly. Aug 17, 2020 at 13:37
• Dear @RonBeyer , do you think there are other things I should have in mind? I highly appreciate your help and review for my schematic (it's more than I could have asked), but would you like to follow the original question format (i.e the questions from 1 to 5) so I can close the topic? Thank you very much! Aug 18, 2020 at 10:31

The purpose of ground separation is to avoid disturbance and interference on the analog part from the digital signals. It is also used to minimize voltage drops on the analog part if the analog and the digital signal has the same return.

Unless you have opto or digital isolation both grounds must be connected such that they connect directly to input so that analog GND current is never shared with noisy digital or power ground.

Failing to do this can damage ESD protection diodes usually rated for low current like 5mA for high speed. Either fuse open or shorted. But AGnd=0.4V indicates it may be OK. Next that makes the uC substrate more vulnerable to latch up as and signal below AGnd can trigger large shootthru currents from SCR effect.

Alternative is a ground choke and decoupling cap to raise CM impedance at RF if single ended.

• I suspect my post was not followed. I am using decoupling caps around the board. I am interested why I get such low voltages on the 3V3 end. My microcontroller won't power. Aug 17, 2020 at 16:05
• Ohm’s Law between Analog and digital GND will explain that. They must be tied at source to not share current or get crosstalk yet be 0V difference. Aug 17, 2020 at 17:35

I was expecting an answer, but I will do this and explain what was wrong.

I solved the mysterious drop in voltage and I learned what the packages mean.

1. Because the BSS138 transistor gives 0V to the EN pin of the LDO, thus enabling shutdown mode.
2. -.4V is residual current flowing back to GND from AGND due to lack of voltage on that portion. Connect grounds together at source solves the problem.
3. It's like a electrical switch in a mosfet (or solid state switch) : Gate is the logic input (turns it ON at 0.5V), Source is the output current and Drain is the input current. When the Gate receives a current lower than 0.5V, the chip does not let the current flow fron drain to source.
4. It's like 3, but it also has resistors built in.

I learned a lot here.

Thanks everyone for contributing.