How to Measure PCB Impedance and Component Coupling?

[js2020]

TL;DR Summary

I would like to measure a circuit boards impedance with capacitors installed

I am designing a PCB with multiple capacitors on it. I would like to measure the board impedance to compare it with results from Ansys Q3D. I can connect it to an impedance analyzer and easily get the impedance. I have recently thought about other meaningful measurements that I may be able to make, but I'm not too familiar with.

Aside from the board impedance, I would also like to measure the coupling between the two. One way that I can think of possibly doing this is applying a small known sinusoidal voltage to one set of capacitors and measuring the voltage on a second set that is not connected to the first. The induced voltage could be used to determine the coupling coefficient between the two. The voltage will be very low though so I don't think doing this with an oscilloscope would be the best method. I was also thinking about a 2 port measurement using a network analyzer but I'm not too sure how this works as I have no experience with one.

I have also been looking into what type of connector to use. BNC is pretty common so I was leaning towards that. I know that 50 Ohm is very common but there's also 75 Ohm which has lower signal attenuation. Then there's 100 Ohm and 93 Ohm in SMA. Will the connector impedance really make much of a difference here?

 

[berkeman]

Summary:: I would like to measure a circuit boards impedance with capacitors installed

I am designing a PCB with multiple capacitors on it. I would like to measure the board impedance to compare it with results from Ansys Q3D. I can connect a it to an impedance analyzer and easily get the impedance. I have recently thought about other meaningful measurements that I may be able to make, but I'm not too familiar with.

Aside from the board impedance, I would also like to measure the coupling between the two.

What "board impedance"? Do you mean the characteristic impedance $Z_0$ of controlled-impedance transmission lines on the PCB? (Probably not) Do you mean the impedance versus frequency of the power rails to ground with the decouping capacitors installed? What exactly are you trying to measure, and how are you going to use that information? What frequency range are you going to make these measurements over?

 

[berkeman]

Summary:: I would like to measure a circuit boards impedance with capacitors installed

I can connect a it to an impedance analyzer

Which impedance analyzer? Also, what accessories do you have for it?

 

[DaveE]

I would definitely use the network analyzer, unless it won't go low enough in frequency for you. It's the right tool (if it's a good one, of course) and you should learn how to use one. This sounds like a perfect project for learning new tools; I was always a lot more motivated when I had a "real" problem.

In any case you'll want to be careful with calibrations and such. Don't just do one measurement, try A-B comparisons (like maybe with/without capacitors, for example). You need to convince yourself that you are measuring the right thing.

edit: I was assuming it's a two port instrument, like S-parameters and such. That is the easy way to characterize coupling. But you can invent your own too, The problem is your setup may not be as good as the guys at Agilent, Rhode & Schwarz, etc.

 

[Baluncore]

@js2020
The impedance models will depend on the type of circuit you are building. The design of coupled transmission lines in pulse generators is not a generality that can be discussed without context.

I doubt you are talking about sinewaves. I expect you have an EMP that must be managed, such as is found in a Marx Generator. The discharge loop must be modeled differently to the charge loop as different circuits are involved. You are designing a multi-element resonant structure, a transmission line filter, and an antenna, all at the same time.

The EM energy density in the discharge current loop “antenna” is capable of breaking down nearby instruments. There are fundamental limits to the energy, proximity and topology.

 

[js2020]

What "board impedance"?

Yes, it would be look at the impedance versus frequency of the power rails to ground with the decouping capacitors installed? I am interested in determining an optimal layout for a series-parallel combination of capacitors , whether they're for decoupling or bulk capacitance. I would want from as as low frequency as possible up to a few MHz. I would like to extend this to ceramic capacitors in the future which will may have a resonant frequency of 10ish MHz so learning how to use these tools for future use is one of my goals here.

As far as what devices I have, I have a few.
Agilent 4294A impedance analyzer: 40 Hz - 110 MHz
Keysight E4990A impedance analyzer: 20 Hz - 120 MHz
Accessories: 16047E

Agilent 4395A network/spectrum/impedance analyzer: 10 Hz - 500 MHz / 10 Hz - 500 MHz / 100 kHz - 500 MHz
Accessories: none at the moment

I think I may have a Keysight E5061B vector network analyzer but it may be unavailable for a while so I can't really count on that at the moment.

 

[js2020]

Don't just do one measurement, try A-B comparisons (like maybe with/without capacitors, for example). You need to convince yourself that you are measuring the right thing.

edit: I was assuming it's a two port instrument, like S-parameters and such. That is the easy way to characterize coupling. But you can invent your own too, The problem is your setup may not be as good as the guys at Agilent, Rhode & Schwarz, etc.

I definitely plan to do more than one measurement. I have at least 4 different tests planned that I would like to compare with Q3D.

A two port measurement would be good for measuring the coupling. That's new to me though so I plan to read up on it and experiment a little.

 

[js2020]

@js2020
The impedance models will depend on the type of circuit you are building. The design of coupled transmission lines in pulse generators is not a generality that can be discussed without context.

I doubt you are talking about sinewaves. I expect you have an EMP that must be managed, such as is found in a Marx Generator. The discharge loop must be modeled differently to the charge loop as different circuits are involved. You are designing a multi-element resonant structure, a transmission line filter, and an antenna, all at the same time.

The EM energy density in the discharge current loop “antenna” is capable of breaking down nearby instruments. There are fundamental limits to the energy, proximity and topology.

This project doesn't involve EMPs, but I would like to build a foundation that can be expanded to any area. This application is just for decoupling on a DC bus. I am building a high voltage high power converter. This capacitor will be used for decoupling between the source and however many phase legs. I found different results for my Q3D simulations depending on what I simulated, so I want to verify which is accurate. Determining the coupling between capacitors is an added bonus as it will give a more accurate insight into what design approach works best for a given application. I would like to find an optimal capacitor arrangement based on whatever the end application is.

 

[Baluncore]

A power supply decoupling capacitor will need low series inductance at high frequency. The power supply will have the bigger reservoir caps some distance away. To bypass or decouple a DC supply will need fast small ceramic capacitors in parallel with slower film capacitors. Using two or more different types in parallel should damp any “mono-cultural” resonance, and they can share the resistive divider.

Bolted flat straps must be used, not wires and impedance matched connectors.

Have you seen the range of pulse capacitors manufactured by HK Film Capacitor Limited?
https://www.filmcapacitor-st.com/
I have trouble matching their spec's with lower cost fabricated modules.

 

[js2020]

Yes the power supply is pretty "stiff", which allows me to use relatively low capacitance for decoupling. Actually, this decoupling is more for decoupling between two phase legs. Multiple ceramics will be useful for providing the lower inductance. Over designing based on voltage is really important for ceramics, in general, since they are prone to fail short due to stress fractures. Kemet has a line that they've designed to increase the probability of failing open. I think a few parallel branches are easy enough to implement with these and can provide an extra buffer for reliability.

The HK Film Capacitor Limited caps are really nice. They have really good specs. Matching their specs will be hard and maybe not even possible. Considering the cost and modularity, it would be nice to have something close though. Another cap I like is Electronic Concepts. They have a wide range of products.
https://www.ecicaps.com/

 

[js2020]

So with this, if I have 4 large overlapping planes...2 MID layers in the center and +DC, GND on the outsides, is it important for me to worry about the thickness between layers for impedance matching? I think capacitance and inductance calculations will be pretty complex and probably iterative if I consider the actual loop inductance and capacitance. Any suggestions here are appreciated. Otherwise, I was going to just place a 50 Ohm BNC near the middle of the series capacitors that are lined up in a row.

 

[Baluncore]

is it important for me to worry about the thickness between layers for impedance matching?

Impedance matching should be the least of your worries. Why do you think you must match the impedance of the tracks on the PCB?

A PCB needs wide tracks to lower the series inductance and to carry the peak current. On multi-layer PCBs many vias will be needed to reduce the inductance and carry the current.

The BNC connector is the weak point because it forms a 50 ohm = 50 volt/amp bottleneck. You might do better with a couple of 6 mm spade lugs.

 

[js2020]

For impedance matching, I wanted to make sure to get as accurate results as possible which I thought would be possible by impedance matching. This is just for a test board that I will be using for analysis to determine how I will layout my final design. If I wanted to use a coax for actual power, I would look for something with even lower impedance.

The PCB will have the widest tracks possible; they will be copper planes that cover the full PCB (minus cutouts to meet clearance requirements, board edge, etc.).

The BNC would be for measuring impedance and coupling between other capacitors. The final design will have 90 degree lugs like these.
https://www.steinerelectric.com/Pro...Copper-Conductor1-4-in-Stud1-Bolt-Holes-19490