A few years ago, I worked on digital radio for a customer. It was a very complex design, but the reward for an electronic engineer such as me was enormous. It was a pleasure to work on every single part of the design. As I recall, I bought a few books to ensure that I was doing a good job, and the best book that I purchased at the time was titled "Grounds for Grounding" by Wiley, which I highly recommend, in particular chapter 9, which was the most useful chapter for me. The book is extremely well written, and it goes into great detail about how to connect a noisy digital ground with a clean analog ground of a device such as an ADC or DAC using a few different methods. It explains why they are called AGND and DGND on a datasheet and why they should be placed on the same plane on a datasheet. The guide also explains how to minimize the amount of noise between them and so on. In spite of that, that book did not explain to me what a "noisy" ground is, and what a "clear" ground looks like. As a result, I am sharing this with you at this time. There is no problem in getting 50 - 100 mV of 'ground noise' when we design a pure digital board. From the perspective of most designs, it is tolerable. What if we have an 8-bit DAC (or ADC) instead of a 16-bit DAC (or ADC)? Is it possible to tolerate a ground bounce of 50 mV? Assume for the time being that we will be using 1V as a voltage reference for our DAC. In that case, with an 8 bit DAC the resolution is 1V/256 = 3.9 mV. Similarly, we can build a table like this 8 bit -> 3.9 mV -> 48 dB 16 bit -> 15 uV -> 96 dB 18 bit -> 4 uV -> 108 dB etc. So in conclusion, we can say that despite the fact that we have an 8-bit DAC, we still need to be very careful. In the case of the above design, after spending a sleepless night studying the above book on how to best connect the "AGND" to the "DGND", we made the decision to use a single GND plane as the best way to connect the two. However, even though I was terrified of being able to use microstrips on a PCB, there was no choice since the PCB stack had been decided by others months earlier. As a result, I had to ask myself the following question: Can I know how much of an impact I will have on the "AGND" if I have a microstrip near my DAC? Due to the fact that I was looking for a noise insulation of about 80 decibels, this is a very important question to ask. Having said that, it is now time to return to the theory, and I made another table this time, with the dimensions of the strip lines in the vicinity of the DAC in mind. w/h = 1 -> 24 dB current reduction at a distance of 5w w/h = 5 -> 78 dB current reduction at a distance of 5w etc. It became clear to me that with the stack-up I had, I would never be able to achieve the dynamic insulation I wanted, and since I was not able to use strip lines and asymmetric strip lines, I was going to be in trouble. I decided to add a low pass filter to the distance of the DAC as well as a digital buffer so that the noise between the two GNDs could be minimized. As you can see in the drawing above, the resistor is the only component you will see; however, you can imagine an 18pF capacitor close to the DAC to complete the circuit. As far as I am concerned, it worked perfectly. If you do not wish to buy the book mentioned above, I recommend you take a look at AN from Analogue Devices. It will make you grow as an engineer, at least I did. There are a lot of good suggestions to be found in them. I hope you enjoyed it!
test