Electrical Engineering Stack Exchange is a question and answer site for electronics and electrical engineering professionals, students, and enthusiasts. It only takes a minute to sign up.
Sign up to join this community
I'm connecting CdS photoresitor and ADC (MCP3204) to the RPi. And MCP has DGND and AGND.
I am about to connect both GNDs to the GND on RPi's GPIO, naively.
Is that problematic if I want to read the value from ADC at maximum rate? Say... 100kSps.
MUST I use separate GND for each of them? If it can't be, putting capacitor, or diode can mitigate some potential risks?
I think there is no definitive answer here and you should consider every one and pick the best of all.
Why? This is common, when one talks about ground loops, return currents, or just proper grounding to get tons of replies that come up with completely different conclusions. They all have good technical arguments. All will explain that they have done so or so thousands of time and that it solved the problem all the time... Well, probably they are right and this is true.
BUT
Proper grounding is dependent of the application. Good practices in analog design are disastrous in high speed digital. RF grounind has nothing to do with high power DC grounding. Then comes the argument of safety ground (Earth) and ESD/EMC/EMI issues. For instance, there is a good book about the subject "Electromagnetic Compatibility Engineering" by Henry Ott, which is very good. But good at what? The book is oriented towards EMC/EMI. It doesn't address (or at least not thoroughly) proper grounding for audio for instance.
This doesn't answer your problem alone, but combined with the other answer you will get, it may help a little bit.
In order to contribute more to the final answer I would say that: What defines if your solution is correct or not? In your case it's if the amount of noise you get at the input of your ADC, relative to AGND at the ADC pin, is lower than the resolution of that ADC. Thus we need to know the amplitude of your input signal and the requirement in term of noise.
There has to be a current path from GND on the converter to GND on the controller. However, the AGND and DGND pins on the converter must be connected to each other (see http://www.analog.com/en/content/raq_groundingadcs/fca.html). If noise may be an issue, then you could put a ferrite bead or LC filter in. Definitely do not use a diode.
Usually ICs recommend using seperate GNDs for analog and digital because of ground loops mostly caused by high frequency transients. When digital signals switch from a 0 to 1, there is a current spike as the power supply reacts to this greater current demand. This isn't too big a problem for digital circuits because the logic levels for digit signals have a large tolerance built in (+0.5V is still a digital 0 for example). But for analog devices it is a significant problems, especially with an ADC where +0.5V difference in an analog signal is HUGE. What is then recommending is to use 'Star Grounding'. This is a pretty good article explaining star grounding. Essentially, you connect all your grounds at 1 centrallized location. And This is a good article about grounding in general, which also touches on star grounding.
Just connect the two separate ground pins (AGND and DGND) at a single point, like the analog ground pin on the GPIO of the RPi. The document linked by Alex Forencich, states that if doing star grounding, make it as close to the supplies as possible to improve noise performance on the analog side.
A capacitor will only limit DC but the noise from the digital side will be high frequency. If needed you can throw on a ferrite bead, but make sure you know the affects first. I think I have referred to this document a few times.
