RF inductor on FR4 with no GDN

[Bouli]

spiral on fr4

I need to have spiral inductor on FR4 with about L=160n@80MHz. The main problem is SRF which must be over 800 MHz. To get it I need to do inductor without GDN, I thing, but I heven't seen model of that kind of inductor? Does anyone have experience with that problem?
Any advice wellcome

 

[rautio]

inductor cut gnd-plane

Your inductor MUST have a ground return. If you don't give it a path for the ground return current to flow, then it will find one on its own. Having no ground return is just like trying to light up a light bulb with only one wire. It won't work!

The Srf frequency is determined by the capacitance to ground and the value of your inductor. (It has nothing to do with inter-turn capacitance.) To decrease the ground capacitance, you do have the right idea, minimize the area of the ground that is close to the inductor. You can do this by making your ground return a strip of conductor on the same level as your inductor, right beside it. Current goes out through the inductor, it comes back through your ground return. Such a strip will have very little capacitive coupling to your inductor, and thus a very high Srf. However, it is not free. Now your inductor magnetic fields extend down below your circuit. If there is anything there it might couple to, you could have problems.

Another approach is to make a hole in your ground plane. Keep making it bigger until your Srf is high enough. Thicker substrate can also do the same thing. Again, be careful about fields extending below your inductor.

 

[Bouli]

parasitic capacitance fr4

Theory I know, maybe some practice advice?

 

[rautio]

rf lid effect

Bouli, I think my advice is practical. Let me state it a little differently.

To raise your Srf, move the ground farther from your inductor.

There is always a ground return current.

If you force your ground return current to flow far from the indutor, your inductor + ground return current make a big loop antenna. This will give you very bad results. So make sure your ground is not too far from the inductor.

Hope this helps.

 

[vfone]

tool planar inductor pcb fr4

The inductor MUST have a ground return !?!?
What about Lumped Inductors? Or Air Coil Inductors? These inductors are evaluated without ground (or you can use ground, if you want), and they are not antennas.
Have to think to a printed inductor without ground, as a lumped inductor without ground.
The printed inductor ground can be any type: solid, pattern, or no ground at all…
**broken link removed**
…the inductor will be an inductor in all cases…of course with different inductance, SRF, parasitics, Q, etc.

 

[rautio]

inductor srf pcb

Hi Vfone. Thanks for raising the issue. This is a common misunderstanding. Let's simply the problem to DC.

To make an electromagnet, take a battery and an inductor. I defy you to energize that inductor without some kind of ground return. You still have an inductor in your hand, but you can not make an electromagnetic with only one wire coming from the battery.

In the RF case, you can avoid wires by using radiation. But let's assume a strongly radiating circuit will cause undesired problems, for example, you are designing a filter or an amp, not an antenna.

Lacking radiation and DC blocking capacitors, if you can not energize an inductor at DC, then you can not energize it at RF. The ground return current must flow. The only choice you have, if you care to make the choice, is where that ground return current flows.

If you don't include a ground return in your circuit, then the current will find a ground return of its own. I have seen Si RF ICs designed with no ground return. Where was the ground return current flowing? In this case it was flowing in the substrate, substantially decreasing the Q of the inductor. The designer had no idea. He lost a lot of Q because he ignored the ground return.

Where big problems happen is in the occasional situation where the ground return current is forced to take a wide detour far from the inductor. Now you have a big loop antenna. This kind of problem is also called a "ground loop". It can be very frustrating.

So, take your designs, short out any series capacitors, and see where the DC would go if you try to energize the inductors. It will follow a complete path from + to - on your battery. If that loop has a large area, and RF will follow the same path, then you will have a very nice antenna. Any loss on that entire loop adds loss to your circuit.

 

[Borber]

+fr4 +spiral +inductor

Spiral inductors designed on FR4 can have ground layer on other side of PCB or not. Inductance of such spirall will depend on you used ground plane (lower) or not (higher). Inductance as it is it should have both connections to the circuit. Only known issue to me with inductors with no second connection (ground or not) is the case of helical resonator. There the inductance connected to ground at the one end and left open circuited at the another builds a parallel resonant circuit composed of inductance and stray capacitance of inductance to enclosure. Such circuit is a part of band pass filter, usually.

 

[Bouli]

I see that discusion goes the wrong path. I thought about spiral inductor with GND cut off only under inductor. In that case there is the return current path for the inductor and parasitic capacitance is lowered. The problem is that I haven't found any model of that inductor.
In other hand the most of model are dedicated to RFIC design, and most of papers concern that case. I thought that any of you had the same problem I have.

 

[rautio]

Actually, extracting models like that is an area I am presently working on with a high degree of success. You can not model this circuit in SonnetLite, you would exceed the layer limit (three layers required, 1) partial ground, 2) spiral, 3) bridge out).

If you have access to one or the other of higher levels of Sonnet, put your spiral inductor in and do a Pi Net model synthesis (works best for lossless case). You have to use some care there (as described in the documentation) to be sure you have good results, but that should get you your lumped model. If all you need is inductance, Q, and Srf, just get that from the data. In about 10 seconds you can plot inductance and Q and see what you have. Move things around until it works the way you want. Full disclosure: I work for Sonnet. However, if you have easier access to any other planar tool, they should work too. Try to stay away from the volume meshers for this problem, they will take way too much time.

As for my present research, if you send me two port S parameters of a spiral in Touchstone format, I will reply with a model extracted using my latest techniques. Can not discuss what those techniques are just yet.

I think it was good that vfone raised the issue about ground. It is very widely misunderstood, and that causes a lot of grief and missed deadlines. Kind of scary actually. It needs discussion.

Not sure I fully understand Borber's post, but the stray capacitance to the enclosure comment is a clue. That stray capacitance goes to ground. The ground return current flows through that stray capacitance to the ground and back to the source. The ground return path must be complete. Perfectly OK at RF if the complete RF path includes stray capacitance. If you take away that stray capacitance (impossible to do completely) and there is no radiation, the circuit will not work.

Actually, there is a very easy way to almost completely eliminate the ground return current path. Somewhere in your test setup, you have a signal source. Probably has a coax cable coming from it. Take a small knife and completely cut the braid of the coax, leave the center conductor alone. OK, you still have a very small capacitance across the gap in the braid, but I'll bet your circuit quits working!

 

[Bouli]

There is nothing better than experimen. I did few tens of inductcors (diferent size each), measured it, then use s2p files. There is no better way.

 

[Borber]

I am not shure are we talking about the same thing. As I understand the problem it is to have ground plane or not on the opposite side of PCB and the inductor may be connected to ground or it is a part of circuitry and has no ground connection.
It is clear that inductor with ground plane on other side of PCB has smaller inductance as with no ground plane. Interturn and stray capacitances will define SRF and it will be higher for inductor without ground plane. Cutting of one connection of the inductor or not is not the question here.

 

[rautio]

Let's say a lumped equivalent circuit for your inductor exists. It has resistors, inductors, and capacitors. These lumped elements are hooked up in a network. You want to excite the inductor, so you put a voltage source (at some frequency) on port 1. Terminate port 2 with a resistor, or some arbitrary impedance. If you really have a complicated circuit of some kind on port 2, the entire complicated circuit will have a value of impedance terminating port 2 of the inductor. Just terminate the inductor with that impedance.

In your schematic of the equivalent circuit, some of the components will be connected to ground. For example, your voltage source is connected to ground. The port 2 terminating impedance is connected to ground. If nothing is connected to ground, no current will flow. You MUST have a complete circuit for current to flow. Look at your schematic. Where are the ground symbols? That is where your ground current flows. Look at your circuit. Where do the schematic ground symbols go in your actual circuit?

Some of the components in your equivalent circuit that are connected to ground will be capacitors. For example, the spiral inductor has a capacitance to ground. That capacitance is what resonates with the inductor to set Srf. If you move ground farther away, the capacitance will be less and Srf will go up.

If you move ground too far away, you make a big loop antenna and you will get crazy results, this is called a "ground loop".

If you don't know where ground is, then you don't know what your circuit will do. Maybe it will work OK, maybe not. If you want to be sure your circuit works OK, you must know where ground is and make sure it is OK. If you can not find the ground, the electric current flowing in your circuit will. Good luck.

 

[nandopg]

Just to reinforce the nice posts made by rautio, who is absolute right, which pcb thickness would you pick if you are interested to get a minimum radiation from the traces? The answer is: that would be the thinnest possible, what means the one with the ground plane as close as possible to the pcb traces. Of course to implement a coil with a certain Q using microstrip line in this situation, the geometry of line would have to be properly chosen and a certain maximum Q would be reach.
I remember in late times when field simulators were not available, I had to do a planar power combiner using microstrip line. In order to withstand with the power level I had to use a pretty thick Teflon substrate (ground far from the traces). What happened was that the height of the lid had to be set to 5 inches to eliminate some odd coupling modes that took place inside the structure, the well known lid effect. This is because in order to have just one return to the traces the top lid had to be really high. If I could use a thinner substrate, for sure the lid's height would be smaller. Taking this fact to the limit where no ground plane there exists, the pcb traces will turn a radiator.

This fact is very pertinent to this nice discussion.

NandoPG