Hi in a second Channel video which I'll link it down below and at the end if you haven't seen it, seen it and I highly recommend you subscribe to Eevblog too. There's tons of videos that I put over there. hundreds and hundreds of videos that sort of don't make it to the main channel. Anyway, we took a look at this super low-cost Jlc PCB it was $72 delivered for a four layer board.

Five of them. Absolutely insane. Anyway, Um, somebody asked the very good question: hey, what about the internal ground planes inside this thing and we'll take a look at this in a minute. Up close under the microscope, there's actually four layers on this board.

I've got an internal ground and power plane covering this entire board and hey, how effective is the capacitance between those? Because the capacitor remember is just two metal plates separated by a dielectric material. The dielectric material in this case is the epoxy resin fiberglass in side this thing. but you know I'm sure I've mentioned this in previous videos on you know, controlled impedance, traces, and all sorts of other stuff. I'll link it.

mean if I can find him in that if you put ground and power for example, in this case, ground and five volts inside those inner layers which go all over the board like this, that can actually be a decoupling capacitor that goes across effectively your entire board. But how effective is it? Can we actually measure the capacitance? Is this power plane? Yes, we can measure the capacitance. Let's do that right now, but we can do more than that I've got the probes connected into just one of the bypass capacitors here. Doesn't really matter which one and we'll test that, will actually verify that later on.

But let's have a look. we've got a nominal capacitance here of oh, let's call that to her. Bang on to Nano Farad's and if we actually change, that's a 1 kilohertz. If we change the test frequency here, 10 kilohertz, it's basically the same thing.

Doesn't change much. So at a hundred kilohertz, Well, you know, at one point nine, One Nano Farad. so you can see that. You know, even at a hundred Hertz There we go to Nanofarad, so it's quite even across the board.

And the dissipation factor. Look at that point, double O five. The lower value, the better at the different frequencies. That's actually pretty.

That's a pretty good capacitor, isn't it? It's not too shabby at all. So there you go. Somebody asked to Nano Farad capacitance on this particular board, but it's going to depend upon the stack up of the board and as it so happens, so I've actually ground down this board and we can see the internal layers. Let's go to the mantas microscope have a look.

You can see that there's an internal layer up there, but you can see the top copper layer in there now, internal copper and the bottom internal copper along there. And of course, then you'll have the outer copper layers top and bottom. You can see that the internal copper layers are very close to the outside our copper layers of the board and these are very thin prepregs they're called. They get these met.

They get these from the raw PCB supplies. They don't make them themselves, they order them from our various companies in manufacture. These are prepreg and core layers, so they actually get to double-sided boards that are incredibly thin and then they glue together what's called a core in there and the core in this case because this is a 1.6 millimeter outer thickness board and then the internal core will be typically about 40,000 millimeter thickness and then they glue the two layer boards top and bottom to those as part of the manufacturing process. So the problem with this is that having a one millimeter core in there, there's actually quite a big gap between the internal copper layers.

So the bigger the distance between two plates of a capacitor, the lower the capacitor and the less effective it actually is. So in this particular arm, stack up that you get and you don't have any choice in this because when you order a like a really cheap you know, four layer board like this, you get whatever stack up the manufacturer gives you, but you can't actually order your own stack up. It's just like a more custom job and they'll charge you more fruit, but you can't actually do that you can specify. so instead of those copper layers being like so far apart like that in near the top and bottom services, you can say hey I want a two layer board in the middle please are very thin prepreg in the middle so that my power planes are as close as physically possible together so that you get increased arc capacitance.

You get a greater effective capacitance that way. But the trade-off is is well, the good thing about the one that we've got here is that having the copper internal copper as close as possible to the top and bottom layers, it's easier and more effective to do a two route controlled impedance traces on the border. And I'm sure I've done a video somewhere on controlled impedance traces, but the thing with this is that well, you can do it the other way. You can say hey I want those two layers together in the middle to give me a greater plane capacitance because that is a more important parameter to me.

but then because the internal distance between the inter and the thickness between the internal plane and the top and bottle signal layers is greater, you have a harder time doing those are controlled impedance traces. for a given control you like and say you want to do a 50 ohm transmission line, you need a much wider trace the greater the thickness of your material between the signal layer and the ground plane to do a micro strip for example. So it's all the trade-off if you wanted the best of both worlds. in this case, you would have to go to a six layer board.

Now let's actually try a different board. Let's get this out of Nano Board Two, which is an eight layer board. So once again I've ground away the layers in here. Let's just pick one of the ones which I believe is the 3.3 volt rail.

So it's going to have a rather large ground plane on here. We dedicated the ground planes to the various rails but 13.5 Nanofarads because they're closer together and check it out. This one's actually much more interesting than we had before. You can actually see the eight layers in there.

count them and you can see how they actually pull back from the edges. and you can actually see the plating on the through-hole I Did you know the main mounting hole there? So that really is quite fascinating. But you can see how they're actually made up of little individual pairs. As I said, these are they.

These are bought bare as like two layer boards of very incredibly thin ones. Then they just glue them together with these are like various thickness cause to make up your stack. that's why it's called a stack up to stack all the layers together to give you your final 1.6 millimeter board. And the exact thickness of the actual prepreg and cause used depends upon a what outer diameter board you want.

You know one point six nominal, but the one point six may include the copper or may not and also the copper thickness. the copper weighting Whether you want one ounce, half ounce to ounce, whatever it is that increases the thickness of the copper so that all adds up so they've got to adjust for that to give you a particular stack up and you can. As I said, you can actually specify this with the manufacturer. you can go hey, I want this specific? You know thickness with these prepregs and you can order specifically Material You might order like a Rogers brand and material for example, because you've got the datasheet for it has the proper controlled dielectric you need and everything else.

so you can you know specify all this sort of stuff until the cows come home. but you can see that the internal calls are much closer together. So using you know a basic capacitor manufacturing formula, you know it's basically the capacitance is the square area times the distance of between those particular plate. So I don't remember which actual wire pair inside.

There is the rail that were actually measuring, but it's one of those and they're going to be very close together to get that huge capacitance. But hey, we can do more than just measure the capacitance of this thing. We can get the full impedance plot versus frequency because we have the Bode 100 analyzer we you've seen in previous videos. We'll just use the jig and we'll just hook some wires over to the ground plane and we can calibrate out the Y's and we'll get a full impedance versus frequency plot and get the capacitance as well versus frequency.

Let's do it. Okay, so let's use the analyzer software. When we want the impedance measurement here, we're using the test jig that we use for you know, with the little slot so we can actually measure some real capacitors as well and we'll compare the ground plane of the two PCBs with some real bypass capacitors. So let's do it.

So what we want is hundred Hertz to fifty Meg No worries, we want. let's get a lot of data points of us. get 813 DB It's lower receiver bandwidth just because we can and we have to do the calibration first. So let's go full.

perform your calibration now. Open Short Load. We've got You've seen this. We have a short and a load thing.

It's got a 50 ohm resistor on it so we'll do the open compensation you have to do this to get rid of to compensate for the test leads and the test jig and the contacts and everything else will do the capacitors first. So open, short and load done. Sweet! And now we can measure some real capacitors. So first up we'll get a 2.2 nano farad ceramic which is close to the nominal 2 nano farad's that we measured.

and so let's whack that in and let's give it a ball. What we want though, is admittance here. and we want the parallel capacitance there. So let's run that single sweep and it'll scale properly.

It we can. Well, we can scale it now. Now of course, the impedance. It's of course it's linear with frequency like this and then there'll be a resonant point and it'll go back up.

It'll go up like that, but in this particular case, because it's such a small value, it'll probably be beyond of the 50 mega Hertz so we probably won't see it go back up. No and just sort of see it start the tail off there. so it would have gone down and then back up like that. but we can get different values to show that so we can just optimize that.

There we go, And then we can get our cursor here. and what's the capacitance? You can see it there up the top there. Those values change in 2.3 nanofarads. There you go, and of course that it drops off like it.

at higher. at this frequency, it's really starts to change. You can start to see that the lead inductance and the other parasitics of the capacitor are starting to change. that and the capacitance is going up.

burn up, burn up and up and up and up and up and up. Up and up. Well, it's not been going up by much 2.9 but you can see that it varies with frequency. All right.

I'll save that one. We've got a hundred in film capacitor. Let's whack that in. Give it a burl single shot.

Here we go. Of course it's a different impedance. You'd expect that, but the bot will turn off the well. we can leave.

We can leave the other one there and now we'll turn it off. And here we go. The scale will optimize it in a second, but we should see. this one should be within the range.

Will it come back up? Boom. Optimized. There, We go. There we go.

There's our nice response. There's our resonant peak right down there at about what 8.5 megahertz or there abouts and you can see that the capacitance is near enough to 100 in 96 In like that, it's a bit of a bit of a peak there, where it jumps up to 140 and then it goes negative. but don't worry about that, that's just a negative in quote marks. That's a quirk of the calculation in this case.

But yeah, there isn't a peak all right. Now we've got a ceramic hundred in instead of the film. measure that one just so that we have some baselines to compare baseline frequency response, impedance curves to compare against the ground plane. You can see that it's very similar, slightly higher in capacitance, but it shouldn't Maybe Will it have the same resonant point? We just don't know.

It's not based on the capacitance, based on the lead frame and everything else. See that one's a bit. that one's a bit peaky. Er.

so there you go. It's got a it's got a sharper dip down here. that's probably due to the longer or maybe not you the longer leads on that new hard to tell anyway, the parasitics of that capacitor that that has a sharper peak like that and it's a slightly different frequency. You know, like seven megahertz or something like that.

We've got another one. Just another leader die won't bother doing like a surface mount one. It's no point now. we just want to get a couple of caps.

Twiddle your thumbs. Of course you know you can get more data points so you can lower your receiver bandwidth and take get more resolution and take more time and stuff like this. But there you know this is good enough for our purposes. Similar.

Whoa. A different response there again. So there you go. you can see the different written.

This is the yellow ceramic. highly technical description. the yellow white versus the orange. These come from my junk bean film cap and the 2n2 ceramic.

So there you go. We've got a nice little collection of responses. now. let's go to the ground plane.

now. What I'm going to have to do? Because I'm going to use these leads to jump over from the test cheek over to the board. They're got inductance in them. They're inductors.

Remember every piece of wire, every trace on a PCB everything. Every conductor has inductance in some way, shape or form. Even the ground plane has spread inductance. So yeah, I'll recalibrate open short load at the end of these test leads.

Just so we you know, take those out because you know these are equivalent to having long leads on your ceramic like this and really like sticking out of your board. So you know we, we just want to get a bit better. Okay, so what I'm going to do? Let's put it in one of the bypass capacitor holes down here down near the edge of the board and we will do another test which is not near the edge of the board just to show that there there will be a little difference. You'd expect a little difference, but not much.

The ground plane is generally the ground plane. It's just like one big capacitor. there'll be you know subtleties in there, but like not much at all. So you remember this measured about two nano farad's So here we go.

We've got it in there and I've done the open short short low compensation with the leads in place. Let's go. Here we go and the impedance is going to be different. Oh yeah, Here it comes here.

It comes here. it comes. Will we get it within the frequency range? Will it will the resonant point be within there? Let's give it a ball. Come on.

Hey there we go, it's going back up. Look at that has a nice response. Wow that's a real that is really good. that is a superb capacitor.

It because you ideally you want you don't want a really sharp resonant point. You want a you know a broad range like that. So that is that is superb. And it was.

Where's our cursor? There we go. So sorry. I'll turn these off because that's a bit. that's a bit messy be comparison and will optimize.

Yeah yeah, here we go. So what's our capacitance? There we go. 1.8 nano, Farad's 1.8 1.8 v and then it drops off 1.7 and then it starts to really taper off down there and boom she's gone. Ski at, you know, 50 mega Hertz but that's really interested.

Frequencies about 14 point C and a 15 Meg's something like that. but it's a really broad response especially compared to let's well compared to the 2 n - lets you know there's the - n - ceramic we had before. Unfortunately we don't have the frequency range but it's going to have a dip similar to the other ones we're seen here. But let's look at 100 in film cap.

Well let's look at that one. it's nice dark color there. We go there we go and look. it's a really sharp resonant peak there.

Where is the ground plane is really and quite an awesome. But that's what you'd expect because a ground plane is just two sheets of copper separated by a dielectric. Just like capacitor, it is a capacitor. Okay, so let's see if it makes a difference going towards the middle of the PCB sorry if you can't see this I Don't expect to see subtle differences but not big ones because it shouldn't matter because the interconnect, you just look at it like a edge effects of the plane and stuff like that.

So, but the actual connection is the same so it shouldn't actually matter. So here we go. Boom. It's almost look at that.

It's almost identical, but we could see a difference in the resonant peak down here. perhaps. But I doubt it not with that broad response. If we had a a nice sharp response, a really sharp peaky response, not know it's it's identical.

There you go, makes no difference. Where you actually connect the ground plane, it's exactly the same. Alright, we're going to do the LTM ground plane now and I've got to hold this with my fingers. but trust me, I'm not touching it and it doesn't make a difference.

So let's run that you remember this one was what? 13 nano farad, something like that was it. There we go, this one. once again, it's going to be an excellent, sorry, screwed that up. Let's do it again.

Here we go and of course it's going to be a higher capacitance. You can see it up there yet. 13 points Something right? 13 our dinner ferrets And where will our resident point be? Will it be here? Well I'd have a Sim. Oh yeah, there we go.

Look at that. it's even broader. It's a broader response. Wow that's really great.

Look at that compared to the light blue the with the other one so well I don't know. They're very silly if you move it and you know similar. Anyway, no broad shape. they're beautiful capacitors.

ground planes fantastic. We don't win a chicken dinner so whilst a ground plane does have that beautiful response like that, it doesn't You know it's not a real PT response which can get you in trouble sometimes. with you know if you have an LC resonance thing you don't near one you can really come a gut. So and it's not just ground planes by the way, it's you know, like just regular bypass capacitors.

I Mentioned this before I'm sure is that you know if your thing is switching at the correct frequency, if your circuit switching at just the right resonant frequency and you combine that with the inductance your traces with the impedance of the resonant point of your bypass capacitor your you can really screw things up. Okay so actually so that's one of the dangers are putting too many bypass capacitors on there of different values as you might potentially hit a resonant point at some switching frequency and that can just cause your circuit to go completely screw it. Anyway, the impedance of let have a look at the the Marone colored one here and it's impedance is down. In the you know, it's like tens of millions Stuff like that.

where is the impedance of the ground plane is in the order of like you know, six ohms for the something like that. What is it? Yeah, four point, six ohms there for that, and maybe like 11 ohms or something for the other gigatron board. So you know they're not as lower impedance. Whilst they have an excellent capacitive response, their impedance just isn't as low at the certain frequencies.

so you know and they're not as good, not as good there. And there's the film. Cap that down, it goes down to 100 milliohms I impedance so you know. and there's the two end to ceramic.

Yeah, that one's lower as well. So the ground planes aren't particularly low impedance, but they do have a proper capacitive response as you'd expect. now. I've just spent about 20 minutes showing you that the internal layers of a PCB actually work as a capacitor.

They are actually a capacitor. and in theory you might think that they are low inductance because they're a ground plane, right? They don't have like little leads coming out of them and stuff like that, and that's kind of sort of true. But when you put a chip on it that you actually want to bypass, that's where the problems start coming in. So you might have this massive ground plane that's like this big like this and it's just solid copper all over it.

and it's one big capacitor. Yeah, it might be tens of nanofarads that we saw before, but can that actually replace an actual 10 nano Farad bypass capacitor? The answer is no, and it's obvious why. You actually if you actually think about it when you, if you've got a large plane like this, the capacitance is spread over the whole surface of it. So that's the whole you know, 10 20 nano farad's if you're But you're looking at a little chip which wants to bypass between the but right at this point between the positive and the negative, rail power and ground, it needs to do it.

At that point, it doesn't would have to go right across the board to get the capacitance. So if you're trying to bypass one pin which is one little point like this on an otherwise big board, then the amount of actual capacitance in that little area is actually rather small. It drops away to bugger off, so it doesn't actually retain much charge, which is what a bypass capacitor doesn't mean. It holds charge so that when a digital circuit switches suddenly really fast, it can take that gulp of current from directly from that little bypass.

that little source of energy, which is the bypass capacitor right there. So having a capacitor that's this big doesn't really help you because that one point source in effect which might be low impedance technically from that, you know just at that one point. if you want to take the whole capacitance right over the whole thing, then it's got to spread out across there. And therefore the inductance spreads out.

And it's actually a phenomenon called inductance spreading that spreads across the ground plane. Which may a ground plane or power planes just rather ineffective as bypass capacitors because that inductance spreading across the ground plane. jazz hands, um, spreads across. then it that tends to dominate over the small amount of capacitance that you're talking about at that one little point source in your board.

Anyway, if you go to like application notes for FPGA is like this are Xilinx One I'll link it in down below. It's very comprehensive. You know it goes into all sorts of stuff. You know the equivalent circuit of a capacitor and the impedance response curve that we had and how buyers can make a huge difference on.

You know, the inductance of Vias makes a massive difference and all that sort of stuff and good and bad techniques. And they actually talk about plane inductance. here. they talk about power plane inductance spreading in particular, inductance per you know, Picot Henry's per square for example.

and you know capacitance per square and it just spreads across. And whilst it does something, it's really not a replacement for like a one nano farad for example and these are application notes will often show you that you put a 1 micro farad, 100 n, a 10 in and a 1 in all in parallel across there and you know they give various reasons for that. And I've done a video and we can go down here. It's the whole idea is that you are spread.

you create as I said, create a larger frequency band of lower impedance. so when you put the multiple capacitors in parallel. but unfortunately the ground plane is not a replacement for one of those smaller capacitor values due to inductance spreading and the big nature of the power plane, the big nature of the plate of the capacitor. But having said that, power planes aren't really effective as bypass capacitors.

that's not a reason not to keep them closely spaced and one on top of the other just like we saw before with a mentor thicknesses minimum as possible because if you when you keep them close together like that, you get a small little maybe benefit from the capacitance. but that's not the reason you do it. You keep them like that so that the effective inductance, loop, inductance and plain inductance and everything else is the lowest possible so you don't want to just go power planes of uselessness bypass capacitors. So therefore I'll put one on the top layer and one right on the bottom layer and just separate and by the biggest distance.

No, you're creating more problems. So you do actually get benefits but in other ways by actually keeping them as close together as possible and actually having them on the same prepreg inside. so you have ground and power on a small little prairie. that's where you get the most benefit.

and then of course you've got the inductance. dropping the via down from the top and bottom layers through to the ground and power planes. but having the planes together is a greater overall benefit than not. And there's actually a good paper on this by Steve where it was an older design con paper all out link it in down below some terror speed and it actually talks about and actually models and demonstrates like built practical examples of why you know spreading inductance doesn't work and he actually you know sets up our experiments and and that sort of stuff.

I won't go into huge amount of detail in it but he basically you know builds up two different you know boards and things like that and and probe some and all that sort of stuff and he also makes the claim that it and provides started to back it up that it's not necessarily the best thing to do to add the multiple values of capacitor in parallel like the hundred and the one Mike the hundred end the ten end You can actually get away with the same value in parallel just in multiple locations and stuff like that and you know that's going to be true and not true depending on circumstances. There's so many different things going on here that it really is difficult to model and actually measure practically measure this sort of stuff. So a lot of bypass some engineering is like just like follow the best practice and cross your fingers and nothing hits some resonant peak or something like that because it's not that you can go oh my. output is going to switch it precisely twenty Point Two megahertz.

Therefore, I Have to select this capacitor to avoid this and do this. Maybe in theory, you know if you go to Pluto you might have the time and effort to be able to you know, analyze and model and and build and simulate and measure that kind of stuff. but in most cases you know you just whack multiple different values in parallel. Follow the application.

note from my you know Xilinx or Altaira or whoever it is and just you know she'll be right. Oh look, I'm big anyway I hope you found that interesting. It is a fascinating subject and you can go down the rabbit hole on this one if you really want to. It really is quite a complex subject.

Just by passing start, you know a bypass capacitor. It stores energy. At a point in about now, There's a lot more to it than that if you really want to get into it. But anyway, I hope you did find that interesting.

If you did, please give it a big thumbs up. As always, you can discuss in YouTube comments down below or over on Eevblog a forum and I've mentioned it before but I mentioned again my patrons often get to see videos including this one early, so if you want to do that might want to support me on patreon link is down below. Catch you next time.