How to reduce metal losses
I am looking at some filters in the 30+ GHz range based on microstrip. I find that metal losses are becoming critical. (Insertion loss of -1.5 dB with lossy substrate and PEC strip, -7dB insertion loss with lossy substrate and silver strip). I am using metal thickness which is 10 times the skin depth at my frequency of interest. Are there any guidelines in general to minimize metal losses in planar microwave circuits ?
Thanks a lot.
-svarun
Surface resistance (Ohms/square) sees essentially no change (actually it increases very slightly) once you get above about 3 skin depths thickness of metal. Going to larger thicknesses can decrease loss but only because you have more surface area (on the sides of the lines).
To further decrease loss, you can select a substrate with lower dielectric constant, select a thicker substrate (provided it is not so thick that radiation clobbers your filter), switch to suspended substrate microstrip, switch to CPW, or, if possible, select a wider bandwidth for your filter.
Make absolutely certain that you have no magnetic nickel anywhere in the circuit, including any connectors you might have.
If you want to go on the wild side, switch to high temperature superconductors.
Thanks Dr. Rautio. I read your paper on Microstrip Conductor Loss Models " published in MTT. I have a fair idea now. I want to add this piece of info. Just let me know if I am correct.
At very high frequencies, it looks like you will have 4 line currents in a microstrip conductor. (Firstly the skin depth effect causes volume currentts become surface currents and next, the current crowds at the conductor edges and you will end up with 4 line currents corresponding to the four corners. Is this correct ? Thanks.
-svarun
Hi Svarun -- Of course, there will be some current everywhere, but the strongest current will be on any and all sharp edges. For a thick microstrip line, that will be the corners as you describe. Because of this, no matter what EM analysis you use, it should always have fine meshing on all sharp edges, no exceptions, when you want high accruacy.
Whether or not a pure surface current model is good enough depends on the situation. For example, if skin depth is on the order of the gap between lines, and you need high accuracy, a pure surface current model for the side current is definitely not good enough. Our multiple-sheet model works really well in situations like that. You can use a multiple-sheet model in any planar EM tool, but Sonnet has the generation of multple-sheets fully automated.
One really crazy thing I have seen (from one major EM vendor in particular) is the idea that the high edge current emerges when the skin depth and the width of the line are about equal. That is totally completely wrong. I have no idea how they can go around saying that with a straight face! I have even seen a very well known EM researcher make the same statement in a recent publication! As frequency increases from zero, the high edge current emerges when R per unit length is about equal to ωL per unit length. It is a major point in my paper on loss. Heck, you are a practicing engineer and you are reading research papers, seems like researchers and EM vendors should read research papers too!
dear Arun,
As far as I can think, the only way to reduce ohmic losses is to somehow increase the effective electron density in your design. Though, the method to increase effective electron density depends on the particular design