Platform forum

(Super Simple) Layout Feedback - considerations I might have missed

SockThief , 02-21-2023, 04:50 AM

inspired by the recent posts on EasyEDA I thought I would take a crack at it myself - and I have to say, for most of my projects it's actually making life much easier than KiCAD, so really enjoying it

one of my first projects that I was attempting was a really, super simple Feather adapter board for connecting 3x DS18B20 1-wire temp sensors

I think my schematic is rather neat, so I am pretty ok with that, but when i was doing layout, I realised that in this case I have LOADS of board space and so there is very little external factors influencing my component placement and routing choices. With more complex designs I can find myself saying "this is basically the only way to route this, will it work?" if it won't work I need to go and move components etc. However in this simple case, it's all up to me.

Although the minimum with EasyEDA is two layers, for some reason on such a layout that feels like cheating, and you should be able to do it on one layer... on the other hand, if you have two layers "for free" then you should make use of them to make for a neater layout, right?

so my first attempt (single layer) was like this:

I have the ground trace snaking around and through the screw-terminals. Due to the data lines and the Vcc, i think this is the only way to do it on a single layer, and i have no doubt it will work. However, realising that the headers are through hole, and i have a free layer, i did this

and then what I would do, is make a power polygon (plenty of space) and the entire bottom layer can be ground. i might also make the 1-wire pull-up 0805, since there is no reason not to.

what do people think? any feedback, suggestions, improvements?
robertferanec , 02-21-2023, 05:57 AM
I would go with something similar what you have on the second picture.

I would use 2 layer PCB
- GND on bottom
- 3V3 plane or wide tracks on the top
- signal on the top

I would maybe add 100nF capacitors close to the connectors (between power and gnd - you don't have to fit them, but if cables are long, maybe you will need to experiment with it, so it may be good to have at least the footprints there)
SockThief , 02-21-2023, 06:12 AM
thanks hadn't thought about a bulk capacitor for that, but a good call.

(i find the nomenclature around capacitors to be painful... in this case I am using "bulk" as being a long cable, the voltage might drop and so a cap to stabilise that could be helpful... i guess you can call that filtering, but filtering is something different to me - hope that makes sense!)
qdrives , 02-21-2023, 01:09 PM
Remark on the layer count: 2 layer is free in the software and perhaps in low quantity production too, but in most cases a single layer is a (tiny) bit cheaper.

On the capacitors - the 100nF capacitor @robertferanec mentioned are NOT bulk capacitor, but decoupling capacitors. Bulk capacitors are for power supply stabilization and energy storage. You could state that > 100uF is needed to be considered bulk, but that also depends on current consumption and voltage.
The decoupling capacitors are there to overcome voltage drop over the inductance due to current consumption fluctuations.
Another reason to place them here is for filtering, as high frequency noise is 'shorted' to Gnd when it comes in or goes out of the board.
SockThief , 02-21-2023, 01:37 PM
hey @qdrives - long time hope all is well!

re: layer count; yeah, that was my (badly explained) thought; normally you would find single layer boards being a bit cheaper, hence an advantage to doing everything on a single layer, but in this case you are unable to do anything less that double sided, so may as well use it right?

re: Bulk/Filter/Decoupling/Coupling/Bypass... we have had many a conversation on this! the way I have always thought/pictured/interpreted (and this is maybe where I can get myself into trouble!)
- filter: this is about attenuating/removing frequencies from a signal
- bulk: this is about keeping a voltage rail stable (though bulk is a terrible name; as you correctly point out, bulk would be specifically large values back towards the power supply that handle the bulk of the work)
- coupling is about linking sections of a circuit
- decoupling is about isolating sections of a circuit
- bypass is going around a circuit element or section

and yes, there will be plenty of overlap here! However, you will see bypass and decoupling used interchangeably as a capacitor close to the voltage supply pins of a chip in order to stabilize the rail (and no doubt it will be doing filtering as well)... so I think (finally) I have an ok grasp on the concepts and function... but I still have some problem with the naming conventions which are never consistent and don't always describe the full function of the component the compulsive part of my brain that desires everything to be neat and unambiguous is not happy for the record i also have yet to fully reconcile with the statement "shunts high frequency noise to ground" - it makes it feel like something is "directing traffic" so to speak... but I am still white-boarding this one... if you analyse it from the perspective of complex-impedance / reactance (i.e.: mathematically) it makes perfect sense, however if you imagine electron flow I haven't got it to add up yet....
qdrives , 02-22-2023, 03:20 PM
Decoupling and bypass are two different names for the same function. And yes, filtering is almost the same.
The difference between decoupling and filtering is that the first is often higher capacitance and not so much on high frequency, whereas the filtering capacitor often needs to filter high frequencies and do not need to have such high capacitance for it to function correctly.

In this case the suggestion for the capacitors near the sensor connection is to both 'stabilize' the voltage and, especially my suggestion, to filter outgoing and incoming noise. For an EMC/EMI perspective that would be good for the signal too, but 'big' capacitance (>= 1nF) would make it functionally fail, so other filters techniques are required (if needed!)
In the picture below I tried to draw how noise (AC!) will prefer to go through the (filter) capacitor and not the (parasitic) inductance. However, as a capacitor, and the traces to it, have inductance as well, not all the noise will go that route.

Coupling capacitors are for making an AC coupling. So it is a series connection and not a parallel one. The capacitor eliminates the DC component.

And yes, bulk capacitors are mostly near power supplies. Bulky in physical size (compared to decoupling) and large capacitance.

A simple though is also that the mains supply is 50/60 Hz. That gives 100/120Hz rectified. That times needs to be bridged by the bulk capacitors so that the circuit behind it, keeps sufficient power. A switch mode regulator switches at say 100kHz. That is 1000x as often. The inductance of the wires and traces makes that the power supply (or mains) cannot instantly 'correct' the voltage at the load. That is why the IC needs decoupling capacitor to fill that local gap.
To make this a little more complicated: https://www.youtube.com/watch?v=oL6qjhJH_m4
qdrives , 02-24-2023, 03:02 PM
Robert Feranec listened to you. He made a video on capacitors together with Eric Bogatin.

SockThief , 02-27-2023, 12:05 PM
qdrives I know! i was away for a few days with the family when it came out, but put a note in my notebook. robertferanec has made a great number of videos coincidental to what I am (re)-learning... I am actually preparing a response to your post above - be back soon
Use our interactive Discord forum to reply or ask new questions.
Discord invite
Discord forum link (after invitation)

Didn't find what you were looking for?