 Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me. |
Forum Index : Microcontroller and PC projects : Proper decoupling practices, and why we should leave 100nF behind
| Author | Message | ||||
LeoNicolas Guru  Joined: 07/10/2020 Location: CanadaPosts: 503 |
I'm not a hardware expert. I found this interesting article regarding 100nF decoupling capacitors. Any thoughts? Proper decoupling practices, and why you should leave 100nF behind |
||||
| Amnesie Guru Joined: 30/06/2020 Location: GermanyPosts: 632 |
Hi, absolutly true what he wrote. The thing is: If you really want to calculate the correct value (if needed at all) there is a huge amount of things to keep in mind / to consider. What he wrote is the painful truth no one wants to hear, since there are a lot of hobbyists around which (often) don't understand the physics and math behind all this. As often in live, the correct (but unsatisfying) answer to the 100nF-question is: IT DEPENDS. Greetings Daniel Edited 2025-01-27 02:32 by Amnesie |
||||
Mixtel90 Guru Joined: 05/10/2019 Location: United KingdomPosts: 7857 |
Pah! If it doesn't burst into oscillation you must be doing something right. :) Interesting stuff, but the warning about having excessive supply rail capacitance is also important, especially on regulated supply rails. Voltage regulators can handle a maximum load capacitance before they start to lose regulation - the feedback loop becomes too slow. You start having to fit multiple regulators in order to have sufficient decoupling. The BIG problem is that decoupling is dynamic. Unless you build the PCB and then look for the problems you can't get it right. Values change with trace lengths, voltages, signal characteristics, trace proximity, PCB material (dielectric changes) and thickness and number of layers, temperature (because all components are temperature sensitive to some degree) etc. etc. It would be lovely to be able to calculate the values then build but physics doesn't work like that. This is the real reason why people have been fitting buckets of 100nF caps - it's the cheapest way to get a (mostly) working system as quickly as possible. :) Mick Zilog Inside! nascom.info for Nascom & Gemini Preliminary MMBasic docs & my PCB designs |
||||
| lizby Guru Joined: 17/05/2016 Location: United StatesPosts: 3358 |
Thanks for posting. tl;dr: because of higher speed of modern micros, etc., and significant improvements in capacitor performance along with reduction in cost of higher capacitance parts, 100nF can actually make noise worse in modern circuits, and 1uF or 2.2uF will be better and cost no more. PicoMite, Armmite F4, SensorKits, MMBasic Hardware, Games, etc. on fruitoftheshed |
||||
| Andy-g0poy Regular Member  Joined: 07/03/2023 Location: United KingdomPosts: 75 |
No, can't agree there. The subject is far too broad. "modern circuits" ??? That's the problem as Mick indicated everything interacts. What's more things like microprocessers are SLOW even if you are talking about rise times of pulses. Try talking to RF engineers who work at multiple GHz and 100's of GHz and what they have to say regarding decoupling. Andy |
||||
Grogster Admin Group  Joined: 31/12/2012 Location: New ZealandPosts: 9592 |
With SMD MLC caps on regulators(and not RF circuits), I have not bothered with 100n's in parallel with them for years. The 100n across the likes of a 100uF aluminum electrolytic seems to be where a lot of the 100n thing comes from, to help remove any HF ripple that the cheap electro can't deal with, but as far as SMD Tants or MLC's, I don't use the parallel 100n's on those. I generally stick to about 10uF on my LDO's, but then - read the datasheet, cos they recommend in there what to use with their device.  To some extent, every regulator is different and you have to follow what the datasheet says you should use - that's my golden rule, really, rather then just slapping 100n's in parallel with the output by default cos that seems to be what most people have been doing for decades now. As Amnesie said - "It depends". Usually on the regulator datasheet recommendation.  Perhaps the whole 100n thing was simply more relevant, when pretty much everything electronic was analog and there was very little digital stuff around, especially at the hobbyist level....  Smoke makes things work. When the smoke gets out, it stops! |
||||
| tgerbic Regular Member Joined: 25/07/2019 Location: United StatesPosts: 65 |
Not surprised this subject has come up again. It seems to come up every few years. Lots of engineers and hobbyists have already learned this, however many have not. Like mentioned above, it is often just a rule of thumb people seem to accept as a permanent truth. Many of us either figured this out by practice or studied datasheets/appnotes to see what the vendors recommended. Thanks for bringing up this again. |
||||
| Volhout Guru Joined: 05/03/2018 Location: NetherlandsPosts: 5054 |
When decoupling a power plane against a ground plane, there is merrit in using different decoupling values. i.e. 1u / 100n / 10n / 1n, same as when there is a separate node (like a Vdd_pll pin). This spreads out the resonance peaks. You often see this applied in datasheets. In a typical 3.3V PCB trace application, the PCB trace has far more inductance than the MLCC capacitor, being it 100n or whatever value. With varying trace length between identical capacitors, the resonance peaks also vary. And it is pretty harmless to use identical caps. The total capacitance needs to be compliant with the power source capabilities. That is clear. Although in the hobby world underestimated, a lot of thought is put into power distribution in the professional world. There are even simulation suites to help you visualize this. If you design a product for serious production (10k/month) you will spend time on this, or your desk ans a developper is littered with 100's of units ever month, for you to debug. Decoupling also impacts radiated emissions, and immunity. Certification topics. Days, and sometimes weeks spend in a non-echoic chamber finding the root cause of a spectral peak, or sensitivity to disturbance. But this forum is a whirling pool of genius idea's (same as the Electronics section where they work on power convertors and MPPT's). And as soon as you say the words "certification" and "agency" all the whirling stops. So don't worry about these things. Just splatter your 100n's and be creative. That is what counts here. Volhout PicomiteVGA PETSCII ROBOTS |
||||
| PhenixRising Guru Joined: 07/11/2023 Location: United KingdomPosts: 1360 |
 Some time ago, I had built 3 small boards, each housing a PIC18F2331. Couldn't be simpler but none of them worked at all. Then I realised that I'd forgotten the 100nf caps. They all sprang to life  |
||||
 Print this page |
 |
The Back Shed's forum code is written, and hosted, in Australia. | © JAQ Software 2025 |
