Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Register Log in

FR4 and stripline very lossy

Status
Not open for further replies.
B

biff44

Advanced Member level 6
Advanced Member level 6
Joined
Dec 24, 2004
Messages
5,059
Helped
1,376
Reputation
2,748
Reaction score
1,060
Trophy points
1,393
Location
New England, USA
Activity points
37,988
Anyone try stripline on FR4? I took a chance on a really low cost job recently, and the results at 5.8 GHz were truly dismal! For instance, a 10 dB directional coupler ended up with only 3 dB of directivity. And the insertion loss is around 2.5x what I expected.

I did model everything carefully in Sonnet with an EM simulator!

I was using a Tan del=0.02, but even increasing that loss tangent much higher did not account for the poor directivity.

I have been researching prepreg materials to see if they can be super lossy, and so far have not found anything saying they were. One paper said to expect only 35% higher FR4 prepreg loss over a prepreg that Rogers makes.

So...other than not wanting to try again...any ideas why so lossy?
 

I had similar experience with homemade stripline many years ago, and wonder if this is from

(a) insufficient connection between the top and bottom grounds
(b) narrow air gap between the dielectrics.

In the Sonnet model, top and bottom are by definition ideal grounds. In the real hardware, they start as independent metal planes and you need to add many vias to connect the grounds. You should be able to test this possible error by modelling the grounds as metal polygons (instead of box boundary condition) with the actual vias from the hardware included in the model.
 

I did suspect the ground vias. They were big enough diameter holes to pass a copper wire thru, and I hand soldered around 25 wires from one side to the other...and saw no difference in performance. I was wondering if the plating did not get all the way thru from one layer to the other, but apparently that was not it.
 

Can you show a screenshot of your layout (with vias) ?
 

Biff

you should also take a look on a couple of things: first the coax to planar transition. At 5.8GHz it can be an issue if not properly designed. Second il layer registration, in low cost FR4 process it not accurate at all (especially if your coupler has stripline on different layers).

I have measured microstrips and CPWG (not stripline) up to 50GHz!! in FR4 and RO4450, and everything works well up to about 20 GHz, FR4 is lossy but it is linear with frequency.

I hope it can help.

Mazz
 
Volker, I did a quick look and did not find the file. Let me re-create it. I did not actually model the ground vias, but just made the box size the same as the via footprint.

Mazz, registration? The middle metal layer only has the two coupled lines on it. I ohmetered the thru connecitons and got << 1 ohm resistance, so the input/output vias must be making contact? yes/no? I do like the idea that it is a process problem at the vendor...as I am running out of other reasons it is working so poorly.
 

Besides insufficient ground plane connection, which can be apparently excluded, I would preferably suspect the stripline to coax transitions. Obviously an unexpected reflection factor can easily thwart the designed directivity.

To check for the different effects, one would want to measure a short versus a long test stripline with transitions.
 

If any other things works properly in the design, there is a trick how to improve the directivity in microstrip/stripline couplers, adding capacitors at the ends of the coupled line (or adding grounding areas near the bending corners).
 
Thanks, guys. You have given me some ideas to try.

I did measure the coupled port and found a pretty good (14 dB) return loss, so I initally ruled out the transistion as being suspect. But, I should measure it more closely. I have physical access to the 50 ohm termination resistor, and can replace it with a coax connector too, and make some more probing measurements.
 

Another thing to suspect is the plating. If this is a standard Nickle-based finish the conductivity is really bad and not fit for 5.8GHz.

Also the surface roughness of the copper is a major contributor to the losses in the stripline. This may be much worse than what you expect: the conductivity may be a factor 2-3 lower or more (skin effect makes most of the current flow in a thin layer where the roughness hurts you most.
 

Another thing to suspect is the plating.
Click to expand...
I don't see where nickel plating would be used on the inner layers or the inner copper side of outer layers. Every conductor surface involved with the stripline will be pure copper, I assume.

The copper roughness of prepreg/laminate copper in fact matters. But I won't expect a factor 2-3 in effective resistance.
 

Additional loss would not explain the big difference in directivity.

biff44 said:
For instance, a 10 dB directional coupler ended up with only 3 dB of directivity.
Click to expand...

It would be nice to see the layout and have the dimensions, so that we can check the simulation.
 

OK I confess I did not read the thread very thorougly. There is no Nickle plating on inner layers indeed so if the stripline is embedded that should be fine.

The figure 2 to 3 I got from very knowledgeable techical people from Rogers corporation. And I correlated this with my measurements. And they offer RF and microwave-grade substrates. FR4 will use even lower grade copper foil so it might even be worse.
 

FvM said:
The copper roughness of prepreg/laminate copper in fact matters. But I won't expect a factor 2-3 in effective resistance.
Click to expand...

It strongly depends on frequency (to be more precise: roughness vs. skin depth). Here is the paper from the Rogers guys:
https://www.sonnetsoftware.com/support/downloads/publications/5-TA1_Al_Horn.pdf

At 5.8GHz, we expect some increase in loss from surface roughness, but loss does not explain the big difference in directivity. Something else must be wrong.
 

FR4 will use even lower grade copper foil so it might even be worse.
Click to expand...
Yes, that's also my assumption. A copper roughness of 1-2 µm isn't unusual for FR4 ML, it's rather wanted to increase the peeling strength. But even with doubled losses the coupler should work.
 

Hello everybody,

I am trying to design 2.45GHz patch on ADS using FR4 substarte.But i am unable to find the substarate paramters i.e thinkness,height and dielctric constant.

I dnt know how should i proceed.
 

I think excitation of stripline from coax can cause problems. If the groundconnection from coax to both groundlayers of stripline is not enough, you may have created extra parallel plate em modes in the structure causing losses and poor directivity
 

biff44 said:
Anyone try stripline on FR4? I took a chance on a really low cost job recently, and the results at 5.8 GHz were truly dismal! For instance, a 10 dB directional coupler ended up with only 3 dB of directivity. And the insertion loss is around 2.5x what I expected.

I did model everything carefully in Sonnet with an EM simulator!

I was using a Tan del=0.02, but even increasing that loss tangent much higher did not account for the poor directivity.

I have been researching prepreg materials to see if they can be super lossy, and so far have not found anything saying they were. One paper said to expect only 35% higher FR4 prepreg loss over a prepreg that Rogers makes.

So...other than not wanting to try again...any ideas why so lossy?
Click to expand...

Any word on the S11 (input Return Loss) or the S21 (through-loss) of the main through-line?

That could give one a hint if the assumptions for Tan and dielectric constant were on the mark, or what they really are.

Jim
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top