Hi I did a recent video on months in which I'll link in down below and at the end of this video if you haven't seen it and that's the process of in the particular case of the video I did removing bypass capacitors from a circuit to see if it still worked I removed them one by one and ultimately yes, the circuit did still work, but it really it was just for fun. It wasn't really a demonstration that you shouldn't use bypass capacitors in your circuit because I've done which are linking up here. it'll be a YouTube card up here to a very popular our 30-minute whiteboard tutorial on what our bypass capacitors, how they work, why you use different values of bypass capacities in certain situations, and I also do some practical demonstrations using a poor-man's network analyzer to see the frequency response of that. but I thought I would follow on from this or really part two of the more technical tutorial side of things of why you want to use bypass capacitors and why you actually want to put them right near the components and loop area which I've talked about in the previous months in video and many other videos as well.

So I thought we'd do a practical demonstration rather than the network analyzer one before with frequency responses and all that sort of thing. Actually get some bypass capacitors and actually put them in different places in a circuit and actually see the effect. They have not only high frequency bypass capacitors, but lower frequency bulk supply decoupling capacitors as well, so let's get into it now. a practical demonstration of bypass capacitors.

Really quite difficult to do on a regular, like, complex and proper product. PCB It's much easy to do if you actually set up an experiment for it. So what I've got here is a just a single-sided our copper clad. PCB I haven't done anything to it I Could have used a double sided board, round it out and things like that, but it's much easier.

I Just used some marked copper tape here to simulate traces and have one big ground plane on the bottom. So we're talking about transmission lines here, but also talked about Loup area and it just allows us to put the bypass capacitors in different locations along here and see how it makes a difference. Five volts DC Coming in here, the green is the ground that just goes to the big copper ground plane there and the +5 volts goes to this copper tape which goes all the way up here to the top pin of a 1 megahertz crystal oscillator. This is a standard crystal oscillator you should be familiar with here.

This pin is uh, soldered directly down to the ground plane that's ground pin. The positive pin is soldered directly onto that input tape here. this is just a not connected pin and the output is connected to this copper tape up here. and as I've shown in the previous video, I won't go over it again.

but I'm using a proper low inductance probing solution here with my one gig bandwidth Tektronix probes you might be able to see that I've just got a tiny little loop of wire in there so that I can actually connect my probe directly into there so we're going to have and the ground is a very short ground using the little probe wire attachment here. very important if you want to get the best signal integrity possible which is what we want to do here and this can direct it directly to the ground plane. So we're getting an excellent high signal integrity probe in solution and then the output from that goes along this copper tape which is five millimeters wide and it's actually around about one point six millimeters thickness. so standard PCB and that gives you roughly 50 ohms impedance and we got that terminated in two 100 ohm resistors in parallel.

so it's terminated in 50 ohms. and then we've got the same low inductance start probing solution there as well. So the entire point of bypass capacitors of course, is for digital systems which switch from 0 to 5 volts, which we can see up here on the screen. Here's the output of the oscillator on channel 1.

Here 0 to 5 volts, it's only 1 megahertz. The frequency actually doesn't matter what we worry about is the transition time here and it's reasonably fast. It's a H CMOS oscillator, so we've got like 2 nanoseconds our for time in the rise time is going to be similarly. so we just wants something with a fast edge so that we can see the transitions on the power supply.

So with what we've got is a basic circuit here which you can imagine is a product PCB which you would design. You'd have your power input here and it'd go along some power traces to your digital chip, which you're particularly interested. in. this case, we've only got one, but it might be multiple chips in multiple systems and then it drives an output load.

In this case is driving a 50 ohm transmission line 50 ohm load. so we need that to actually get a decent pulse currents in this case, 5 volts on 50 ohms so that we can actually get some large current transitions in the signal trace and more importantly, flowing through the ground plane so that we can actually see the effect of bypass capacitors. Because bypass capacitors matter more for things that take large amounts of current when they're transitioning and it doesn't have to be a resistive load either, This trace down here will also have capacitance. and if this was dry, if this was just a regular signal wire driving another, you know.

CMOS TTL Digital Gate. Over here that digital Gates got input capacitance. The trace has input capacitance to ground and when your signal transitions like this, you remember capacitive Impedance our formula that it actually acts for a brief period of time, acts as a low impedance, or effectively, if it's an infinite transition time like that, it's effectively the capacitor operates as a short circuit. So even if you have no resistive loads unlike what we've got in this circuit, if you've just got traces and capacitive input gates, all input gates have capacitance, even if it is only a couple of Pico Farad's couple of puffs.

Every time you transition in your circuit, it takes a little gulp of current from your power supply and that's what bypass capacitors are designed to help with. So I'll briefly go over bypass capacitors again, but you really have to watch my art 30 minute tutorial video to really understand what's happening there. So I recommend you watch that first. This is more a practical demonstration, but you have basically two different types of bypass capacitors in a circuit.

You have your bulk power supply capacitance which generally goes right at the power supply input or at the output of a voltage regulator or whatever it is, and generally that will serve all the chips on the board. so it basically stores charge and delivers it for the lower frequency R type events in your circuit. like you know, the 50 hundred Hertz mains input ripple. for example, on a traditional linear AC bridge rectified power supply, for example, it's smooth that out.

whereas bypass capacitors like a hundred Nano, Farad or Point One microphones that you typically put right next to each IC Just as sort of like an industry rule of thumb, these store charge which actually provide the energy for the higher frequency switching transitions which we get in here. So what we're going to take a look at here is we've got some low frequency stuff happening here, and we've also got some high frequency stuff happening in here. So we'll be able to use the different bypass capacitors and we'll see how these handle the different types of our scenarios. So let's get to it.

So we've got absolutely no bypass capacitance on this circuit at the moment. it's just switching at 1 megahertz with those fast transitions. It's not recommended. Don't not have any bypass capacitors in your design And Channel 1 here.

the yellow waveform as I said, is the output of the oscillator down here, and that's the one we're triggering off, whereas channel 2 the blue one here is actually the power supply pin directly on this chip. because when you're looking at bypassing, you're concerned about in this particular case, concerned about the actual component which is in this case transmitting. or it could be the receiver chip over here, for example that's actually receiving the signal or both of them. Anyway, we're concerned with that power supply rail.

How stable is that rail relative to the switching currents that this thing is taking? In this case, every time the output goes higher like this, it's got a drive that the 50 ohm load so we get. it's basically drawing a big the current like that. So if you have a look here, we've actually got 200 millivolts per division here for the power supply and that's a lot. Look, we've got maybe like 300 millivolts peak to peak of this low frequency ripple we'll call it even though it's like 1 mega Hertz like that.

Ok, it is still in this particular case, the lower frequency switching stuff and that's quite a lot to have your 5 volt rail vary by. you know, 300 millivolts peak to peak. That's a that's a lot of ripple on your power supply. That's horrible.

That's because we've got no bypass capacitors on there and in this case it's actually taking due to various our parasitics in the circuit because we've got no capacitance whatsoever. It's actually taking what turns into or what looks like a sinusoidal our waveform here. and also you can see the droop in there. and if we actually change the scale on our channel 2 here and we move that up, we can see that that power supply corresponds directly with the droop in the output waveform.

So that's duded. No capacitance and various parasitic capacitances and other things in the circuit which we won't particularly worry about. And if we zoom right in at a hundred millivolts per division on our power supply, this is the high frequency ripple there that we want to get rid of with our point 1 micro farad high frequency bypass capacitor near the chip. and it's the worst on the negative transition here.

This Oh will concentrate on that. So let's look at the effect of a 330 micro farad cap, a bulk decoupling capacitor on the circuit. So I'll put it down here right at the input where you'd normally have it. So we expect this to affect the low frequency ripple stuff.

Get the polarity correct and Bingo. look at that. It goes away. Magic.

That's the effect of bulk. Look, there's virtually none of that ripple and crap that we saw before. Yeah, there's high frequency noise there, but that's not the job of this capacitor, so it's doing it Excellent up there of getting rid of that low-frequency stuff. That's what your bulk D couplings for.

But check it out. Even though our low-frequency stuff has gone, our high frequency stuff is still in there. It doesn't get rid of that. but AHA Let's put this near the chip up here, which is a good design practice and see what happens.

Here we go: I'll put it directly on the probe and directly on the pin and the ground plane of this chip. It doesn't get any better. There we go, It reduced it a little bit. It has some effect of course because it is working as a high frequency bypass capacitor, but this electrolytic due to its various parasitic inductances and whatnot inside.

and the ESR inside this thing. It's just not good enough. As a high frequency bypass capacitor, it's really only good for bulk decoupling. Watch my previous video to see what's actually happening inside this capacitor.

But let's do exactly the same thing with a hundred Nano Farad film capacitor, which they work quite well as a bypass capacitor. So let's whack it in here in exactly the same location as before. That one is a bit more effective, but let's try your more traditional our ceramic capacitor like this. There we go.

That one's done a reasonable job and bit not much better than the film cap. Really, probably about the same. Let's show the effect of that bypass capacitance again. the Point One: Notice the height of the spikes up there.

They're just off our screen there, but if you lower that down, look it gets rid of those effectively. But the point 1 micro farad is on its own is not enough to get rid of the lower frequency ripple inside there. You need both capacitors in this particular case, so I'll clean that up again. Here's the power supply ripple without the cap and with the cap, there you go.

You can see there's still a bit of high-frequency stuff in there that's going to have to do with the type of cap and the inductance of the leads and other traces. And you know parasitics like that? You can see that it got rid of a good bunch of that? high frequency switching stuff. The reason why this little naught Point 1 Microfarad 1 doesn't get rid of the low frequency stuff, and the Big 330 mic does is because this can store a lot more charge so it can deliver that charge to smooth out that high current stuff that we've got in there. If we didn't have a very low impedance load like we've got here and it wasn't our drawing much current, then we wouldn't actually get that low frequency stuff.

And I can show you that by lifting the legs of those resistors there. and all we get is the high frequency our switching. So that's what would happen in a circuit if you were just driving another digital gate that just had a switching capacitance. It's just because it's driving a capacitive line.

It's actually or and/or a transmission line in this case. But effectively every trace is a transmission line. But we won't get into that. That's what's causing this ring in here because there's not sufficient bypass so that once again, we're on 100 millivolts per division.

That's an awful lot of ripple happening on your five our power supply. It's horrible. It's got all sorts of ramifications in terms of our signal integrity, glitches in your circuit transitions, and ground bounce, and all sorts of, you know, a weird and wonderful stuff which we won't get into. but if we connect the load, bingo, We've got that lower frequency switching ripple as well due to the high pulse currents actually or high transmission driving currents going into that load.

Now watch the size of these high frequency are switching transitions on the power supply rail as we move our bypass capacitor closer and further away from our device being decoupled and probed. So if I put it fairly close up there, look at that. there's that. there's our signal level.

You can see where they are and if so, as I slide it towards there, hopefully you'll be able to see that. There you go. As we get closer and closer to the chip, it laws in amplitude. Don't forget as close as we possibly can.

Bingo! That's as low as we can get with this particular bypass capacitor because it's got the particular type and the leads on there remember. leads like this always have inductance. That's why surface mount bypass capacitors close to the chip are gonna be better than through-hole ones, whereas the Bob decoupling capacitor it's not going to matter where on there we actually put it. It's gonna do the same job at the top as it does down the bottom because it's due to the higher frequencies.

It doesn't matter about the lead length or the trace length here, but there is a limit to that if we actually I go and put this I even use a bigger one I'll use a 2200 microfarad one if I put that here, it's going to do exactly the same thing. It's going to get rid of that ripple. but if we go, put it right over here. there is a limit to the effectiveness of this thing.

There you go. It changed it a little bit, but really doesn't do a huge amount because we've got all the extra inductance of the leads here and everything else. And it's closer to the lower impedance source over here, so it's got to be placed reasonably close to the low impedance ground over here copy at the other end of the cable right over here. It's not nearly as effective, and by the way, we don't need 330 microphones to get rid of that either.

We can use a in this case, a half a micro farad here, another film cap, and we can put that there and it's going to do a quite respectable job of getting rid of that as well. You can still see there's a little bit of low-frequency stuff in there, but not much, so you know even that does a reasonable job. You don't need to overdo it on your bulk decoupling capacitor. It all depends on the amount of bulk current actually being taken in your circuit and at what frequency.

So let's now try the best possible bypass if we can get for this particular scenario which is a basically a leadless and that's what they are. a leadless capacitor surface mount capacitor 1206 soldered directly to the ground plane and the pin. Let's give that a bill, that's probably the best we can do. It's still going to go through the ground plane is still going to go up the lead into the package and the ground lead on the other side is are quite tall on the package.

But anyway, this should be the lowest amount of high frequency switching noise that we get. Check it out, that's absolutely a Mae Look at that. We got not much there at all. You remember what we had last time we had it was maybe the same height there, but there was some more undershoot there that is really good, so that's obviously get rid of most of it.

You can't really eliminate it entirely, because ultimately there are going to be package limitations. even surface mount. Even leadless surface mount packages like those capacitors, there's still got some inductance in them, the ground plane still has some inductance, the bond. Why? If you're using a surface mount chip, the lead of the chip has some inductance in it as tiny as it is, and then the bond wire going over into the chip internally, that's got some inductance in it, etc.

etc. And it's And also the probing solutions got a little bit there, so a little bit here, a little bit there, but that's still pretty good for that sort of a leader package there. I like it, so if we combine that with our bulk decoupling here, we've gotten rid of almost all of our switching stuff. Nice.

So I know you're thinking: Dave What if we actually change the value of the capacitor? Does that make a difference in the high frequency content? If you use a lower value cap, will that do it? Because I mentioned in a previous video why you want to use. You know you might want to use different value capacitors in parallel for different frequency components. Well, let's try our naught point 1 Microfarad one again. There we go.

Reduces it like that. Ok, in this case, I've got the white reference waveform there I Stored the Hundred Nano Farad cap and now we'll put in the 2.2 nano Farad cap in exactly the same location as you can see. There are some differences there, but basically it's it's not really going to change the peak. The peak, which is around about there is basically the same with both of them, but the hundred n had more undershoot like that.

whereas if you put both of them there, she'll be able to combine them so that you know having the two bypass capacitors on there can make a difference. Different values: hundred nano ferrets and 2.2 n. That's the combination because the smaller capacitor, the 2.2 Nano Farad will take care of some of the more higher frequency components, but it all interacts as I explained in the previous video with the lead inductance. Like all the packaged inductance and the parasitics in the circuit and everything else.

So what do we talk about when we talk about Loop Area in terms of our current flowing like a complete path like this: Well, we have our power supply input over here, we have our driving chip, we have our load, and we have our return ground path. So let's assume that we have our bulk decoupling capacitor are right at the input here. Well, when you talk about in this case, switching currents and the high frequency is involved and this is how transmission lines work Well, currents in the circuit will always take the lowest impedance or lowest resistance path from the source through the circuit and then back to the ground terminal like this. So if we have our bulk decoupling capacitor over here, for example, then our current will flow up here into our chip.

It'll flow along here like this, and then it will actually return from this ground point here. and I'll take the lowest impedance or lowest resistance path. And for low frequency stuff, lowest DC resistance paths will basically be straight through there. I Know it, it distributes through the PCB and everything like that, but it's basically going to take a direct path.

So all that right around there. that is our loop area and that's where the current has to flow. And here's the trick. The larger the loop area, the larger the physical distance and circle like that, and the higher frequency you go, the more it's going to act like an antenna and it's going to radiate electromagnetic or EMI electromagnetic interference.

It's going to just generate all that and your device may not pass your A C E FC C compliance which I've done a separate video on. So you always want to minimize this loop area. Now for low frequency bulb decoupling, it doesn't matter, that's why it doesn't matter where you put it effectively, it still works even if you put it right over here to the input and effectively. that's weird.

In a ground plane, that's where it's going to flow, start and end it. So that's okay. But high-frequency stuff. and it's a different ballgame for high-frequency stuff.

We're shown that the bypass capacitor is more effective over here, right on the chip itself. So effectively, this capacitor becomes the source for all those high frequency transitions we've seen and it'll do the same thing. It'll flow out the your high frequency currents will flow out here like this, but your return path won't be back over to your large decoupling capacitor over here because it's a lower impedance at that higher frequency to actually travel under that. I've shown it sort of like next to the transmission line, but it's actually under the transmission line and you can prove this.

You know mathematically in field equations, you know all sorts of weird and wonderful advanced theory to what show that this is the case. but the current will actually flow back under that transmission line so that becomes your loop area. So here's where good high frequency design comes in and why: You put your high frequency bypass capacitor right next to the chip because you're minimizing that loop area for generating electromagnetic interference. If you put this bypass capacitor well away from the chip over here, then it has no choice but to follow that as the lower impedance ground.

And if you do that, bingo, You've got this large area again at high frequencies. And when you have that large loop area what What? What? wha you're probably going to be are starting to fail your EMC compliance. This thing's going to be radiating to buggery and it can also pick up things as well. The larger the loop area now I'd love to actually show you that on the board and I actually was hoping that I'd be I would actually measure you the current and the mapping flowing through our the board under here like this using my I Am I Proba 520 Cut positional current probe which has a magnetic car head on it.

but really, you can't pick up the currents if you put it on. here. There we go directly on the trace. You can actually see the switching currents in the trace, but unfortunately it's down in the noise floor for the actual our current path.

But all is not lost yet. Look at this if I take my bypass capacitor and put it over here or anywhere and probe right on the leg. Look at that. Bingo! You can see that all that current is flowing through the lead of that poor little bypass capacitor.

So all that so that shows that it has to be flowing across the ground plane like this and all the way back to that cap. So the closer we put it over here, then this smaller the loop area we're going to get now. Unfortunately, this doesn't have the bandwidth to really show the detail in the high frequency switching stuff, which we've been used to. but as I showed in the Munson video, there's an IBM research paper which I'll link down below where they've actually are visually mapped the currents in the ground planes like this and you can check that in, link down below.

but here's a screenshot of that and it's very cool. Unfortunately, we don't have the tools to do that, so it really does matter where you put your bypass capacitor in the circuit and why it should be near the chip, but there's a whole lot more involved in this. It's not always as simple as that, but that's why it's a good like you know rule of thumb. just to you know, have a bypass capacitor.

It's values not that critical in most cases, next to each or you know in nearby groups of chips in your particular digital layout, but that can vary depending on whether you've got a full ground plane like this one or whether and you're not. You've got a double sided board and it's all filled in and higgledy-piggledy and grounds running everywhere. That's a different kettle of fish. and I hope to show that better in a follow-up video to this.

I Hope it works alright. So what we're going to do now is take a very crude and look at what this thing is radiating. I've got my hell you're doing antenna hooked up to here. Yes, it is a piece of solder.

No worries. I've got my Riga Wire spectrum analyzer here. Got a 500 megahertz span on here. I got it switched off.

so that is like our baseline at Minus 77. DBM There, don't worry about the setup, this is not. You know, an absolute first-class type measurement system. We just want to see if we can see a difference by putting the bypass capacitors on here.

Okay, that's our baseline around Minus 77. DBM Okay, it's just come on. This sweep here will show us our spectrum. There we go.

Pretty filthy look at this size. So this is a hundred megahertz, 50 megahertz per division. So at around about like a hundred and twenty-five megahertz is a big broad being content in there at around about what 230 megahertz or something. We've got a some content in there and some higher up our stuff there so that's with no bypassing at all.

All right now I'm going to put on the 100 micro farad bulk decoupling cap. We're still going to get a lot of that high-frequency content. It's changed it a little bit. Look, we've still got some content here at 250 odd megahertz and we're still got all this broadband content down there around the hundred megahertz mark.

Now let's just put out on our point 1 micro farad bypass cap. We'll gotta wait for the cycle to start again. Here it goes. And Bingo quite well know our content around there has narrowed, but we still have some content up at 250 megahertz.

Why is it so well? let's take a look at the scope screen. it'll tell us Alright, this is with our point 1 micro farad bypass cap. If you have a look, we're at 2 nanoseconds per division. What is the period there? Well, it's about 4 down a seconds.

What's that So roundabout? That 250 megahertz mark that we saw? So that small amount of content there at the 250 megahertz mark is going to be due to that high frequency stuff there. And if we put both bypass caps on there, it's going to be not nearly as high around that 250. We've basically neutered that out now, but you can see how if we remove the bypass caps, it's actually shifted frequency somewhat because the parasitics are all different in there. so it's going to ring at a different frequency so that also can cause a problem if you try and mix your bypass capacitors.

I've explained this in the previous video and due to the parasitics inside these capacitors and the parasitics in the trace and the lead links and everything else, you could potentially get these to resonate at a frequency that you don't want them to resonate at So it's not always. you know, 100% guaranteed. The best idea to put multiple caps in parallel or even choosing the wrong value Bypass cap could shoot, could form a resonant tank circuit at a particular frequency and you could end up getting the spike on your spectrum. and well, that comes down to Murphy Usually open to the Friday afternoon, so let's go for broke and put on our bulk decoupling cap here and our two smaller ones reasonably close.

There we go and let's see how this spectrum changes. We're going to wait for it to see how it just knocks it all down at the between. like you know, 200 megahertz and 500 megahertz. You can see how it's changed drastically by adding those bypass caps and if I take them off warm, the crap starts coming back.

So all this horrible broadband content here and here is caused by all this ringing in here. Look at it. this is just just horrible. And the amplitude is, you know, incredibly high, so it's just radiating like buggery.

and well, that kind of stuff. Yeah, you're probably not gonna pass your C II FCC emissions compliance. So I've put that touch 12:06 ceramic capacitor back and you can see that our high frequency switching noise there is like bugger all really. But we're still getting this content right up here at like a hundred and twenty-five odd megahertz, but everything else is reasonably low.

and that's with our hundred. that's with our bulk decoupling cap on there if we remove the bulk decoupling cap. Yeah, it doesn't really affect any of the that bulk high-frequency content at 125. and if we switch it off, of course, you can see that it all buggers off.

so all that content is being radiated by our circuit under test. And of course, we've got the one megahertz fundamental oscillator as well, spewing out the stuff. So it's not just the high frequency ringing on there, but you could definitely correlate the high frequency ringing to that. What was it? You know, 250 odd megahertz peak on there and all this stuff matters.

I Mean it's you know these maximum Peaks matter when you're testing armed. EMC Compliance. And if we have a look at our one megahertz fundamental here on a 10 megahertz, our span. so we're one megahertz per division there, there's our fundamental at one megahertz, then the harmonic at three megahertz, five seven nine, and so forth.

So I hope you enjoyed that video is a bit longer than I expected, but hopefully it shows a difference between bulk decoupling capacitors and the higher frequency ones and having multiple ones in parallel in terms of not only our signal fidelity over here, but also in terms of loop area and how that actually generates electromagnetic interference. So I hope you enjoyed that. If you did, please give it a big thumbs up. And as always, discuss down below and the other videos will be linked in at the end somewhere here.

Catch you next time.