Great video! Is there any chance you could cover X and Y capacitors? I usually see them in designs as a way to reduce EMI but I don't know when they should or shouldn't be used! Is this something that isn't an issue for isolated ADCs or optocouplers?
I addressed this in another video about EMI filters on power supply inputs, specifically on isolated DC power supplies with an AC or DC input, it will come out in the next month or so. I'll give you an answer though, the use of a safety capacitor is to control high frequency return currents on the secondary side of the converter so that they can leave the system through the primary side but without coupling across the interwinding capacitance of the transformer. If you need to use isolated ground nets like this and you have a problem with radiation, this could be one solution. In a general case where you galvanic isolation, there will always be some EMI but the use of a capacitor is a matter of the scale of the EMI, if it is low enough then don't bother with a y-type cap. If you look at the dual ADC project I did, you'll notice that I did not use a cap to bridge the grounds although I did not address the reason in the video. In that particular design I don't expect a huge amount of EMI from the grounds, there is just not enough power in the board for it to be a concern. If it happens to be a problem when the design is built then the fix is quite simple.
If you're using an LDO, isn't the noise pretty minimal? That is, as long as you route your return traces for the sensor away from the high current grounds, shouldn't there be so little noise that we don't need isolation?
Yes the noise directly from the output pin of the LDO is pretty minimal. But you're ecorrect as you brought up an important layout point to withstand any induced noise from the upstream switching node, that part is important in the layout. If your sensor runs at progressively lower levels then the need for isolation increases regardless of the use of an LDO.
Yeah good point! In the past I've used instrumentation amps for analog sensors in the Microvolt range, but they also were low bandwidth, so coupled with higher order filters, we were okay without isolation, but I could see very easily how higher bandwidth requirements could lead you to consider isolation. Thanks for the follow up Zach and thanks for sharing your knowledge!
Another great explainer video, Zach. One question: Do you have any links to information about the isolation topology within isolated ADCs? All the functional diagrams I've been able to find just say "isolation barrier," which I'm translating as "then some magic happens..." I know and trust how optocouplers work.
Hello Sir, its always been great learning from you..well I have a question. To my understanding dont we need to short AGND & GND?…as a single GND is available in power ckt….kindly suggest as it will be helpful to understand ckt more clearly
In a non-isolated ADC the best way to deal with noise is to use a complete ground plane that shorts both GND and AGND. In an isolated ADC, such as what I show at 6:44, you do not short those pins because the component is intending to provide isolation between the interfaces on each side of the component. Here is a video that discusses this: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-jhphrm9tktU.html
Since this is an interface receiving a signal it was isolated ADC. If you were generating a signal into an isolated region, then maybe you want to isolate a DAC, but I am not aware of an isolated DAC that is available. There are isolated buffers for digital systems that you could use to isolate a DAC, for example the LTM2895.
In a battery powered device, you don't typically have a separated analog power source as in the reference design. So I guess agnd split is not possible without routing avdd over it.
There isn't a conflict. Here I am talking about a specific instance of splitting a plane based on a requirement for isolation. What I often see happen is people think they need isolation between analog and digital sides but they don't understand why, then they split the system ground and route over the split, that's a recipe for EMC failure. I am not advocating for that and neither does Rick (or any other EMC expert). What I'm saying is that if you have studied the situation and you can prove that the way to prevent noise into a particular portion of your system is to split a ground, or if your system absolutely requires galvanic isolation on the two sides, then there is a proper way to do the split. In a lot of cases you don't actually need the split, but most new designers will say "I have analog and digital on the same board, therefore split," this is not sufficient justification to split a plane.
so how should I split the ground? even if I used another regulator for the analog we still shared the ground. idk my previous colleague used to put the ferrite bead to split the digital gnd and analog gnd, It seems wrong but I need a solution for split that ground. @@Zachariah-Peterson
@@wisnueepis3593 I'll ask again, why do you think you need to split your ground regions? Do you have an isolated power supply, are you working on a very low SNR precision measurement application, or do you have some requirement for galvanic isolation for a type of isolated component? Are you operating at only up to about 1 kHz with only analog signals and DC? Most of the time you should not split up a ground plane at all unless you can prove you need it. If you were to split grounds for some reason, the best way to ensure you maintain DC isolation is to connect them with a y-type capacitor. The problem is that you never have true isolation unless the components you are using to cross ground splits are also isolated components that do not use any conductive coupling, they can only use optical (with an optocoupler) or magnetic (with a transformer). These are specialty cases and most of the time you should just use a complete ground unless you can prove you need to have split grounds.
@@wisnueepis3593 This is the most common question I get and I will tell you the answer: just use a uniform ground plane and keep the digital components separate from the analog components. How you layout the components above ground will always be more effective than separating ground. This will always be effective for all frequencies above audio frequencies and as long as your SNR is much larger than 1.
Now how about if the sensor has to be powered from the battery? I see that the ground split won't be an option anymore but rather noise mitigation techniques like routing and filtering.
It would be great if you could cover capacitive coupling the grounds for applications that are using a flyback or other isolated supply topology. I've had issues with audio applications using an isolated supply where WIFI Tx would cause a ground bounce that is conducted through the capacitive coupled grounds. Without the capacitive coupling the isolated side would have ground noise from the power supply causing EMC issues and audible effects on the audio signal. I understand this may be too unique of an issue but it would be greatly appreciated if you could address the approach to better understanding capacitive coupling to remove isolated power supply noise. I've tried finding application notes to address this to no avail and the approach I went with worked okay but minimizing the power supply noise through the coupled grounds would greatly improve my small signal measurements.
