Why this filter has such poor insertion loss characteristic

[yorkyong]

My design is a hairpin microstrip bandpass filter. fc=3G BW=300M
I think it can be considered as a lossless passive circuit
BUT...
where out ot pass band, S11 is about 0db and S21 is about -30db or more
where in the pass band, S11 is -10db and S21 is -13db!!!
what a great insertion loss!

the shape of db(S21) curve is nice, and insertion loss is the only problem. Network Analyzer was well calibrated.

How can I explain that |S11|^2+|S21|^2 much less than 1 within passband? And where the energy has gone?

 

[goxy]

Re: Why this filter has such poor insertion loss characteris

What substrate do you use? Maybe problem is in the delectric losses.

Please post here your filter design.

 

[VSWR]

Re: Why this filter has such poor insertion loss characteris

A description of your design would be helpful. Maybe you can post a drawing of the filter with dimensions? Don't forget to provide substrate data as well (type, thickness, epsilon-r).

 

[borich_03]

Re: Why this filter has such poor insertion loss characteris

You can try to simulate the filter by any 3D or 2.5D EM-Solver. And you'll see your EM-field pattern. So you can find where you lose HF energy and mismatching. I think your design is wrong.

 

[yorkyong]

Re: Why this filter has such poor insertion loss characteris

to borich:
I have run the simulations both in @ds and @nsoft ensmble. They agree well whether conductor resistance is considered or not. When consider metal loss, there is -3 ~ -4db insertion loss.

to goxy and GaAs
I'm not sure the name of the substrate. It is a common and cheap material because the filter is not design for serious purpose.
Er=2.65, thickness=0.3mm

filter dimension:
all traces are 0.2mm width. Input and output traces are 0.8mm for 50ohm
the spaces from left to center are 0.2, 1.0, 0.35, 1.0, 0.4, 1.1mm
their length is 17.2

I had doubted with dielectric loss. But some other designes using the same substrate have good performance.

 

[AL0117]

Re: Why this filter has such poor insertion loss characteris

Hi,

I will try to simulate your filter later, I will let you know my results.
Till then you could checl the following article:

"A 2.6GHz Microstrip Hairpin Filter Design Using CAD and EDA Tools"
you can d/l it from"

hxxp://img.cmpnet.com/edtn/europe/mwee/pdf/Nov1.pdf

A similar article appeared in Applied Microwave and Wireless.

Also there was an older appnote for serenade (Ansoft / Comsoft)

"Using Serenade Design Environment to Design a Microstrip Hairpin Filter
for the 1900MHz Wireless PCS Band."

I don't find anymore the article on the web, but if you need it I will post it. Also I have the project files too.

I'll be back with the simulation results.

rgds, Al

 

[AL0117]

Re: Why this filter has such poor insertion loss characteris

Hi,

I used Er = 2.65, H=0.3mm, t=0.035mm for the simulation, and I used tapered input for matching the in/out lines. I simulated in the linear simulator, optimized, and after that I run the EM simulator. The results are not so bad. Take a look to the below pictures. Check your design, for the mesh at the U folded arms. Also, the discontinuity at the i/o and 50 ohm line could be a problem.

rgds, Al


lila+blue trace= linear simulation, red and green EM

 

[yorkyong]

Is it possible that the loss is caused by the connection of SMA to microstrip? I only solder them at the tip of SMA.

And papers about the above issue is expected. Private message please for not enough downlod points.

 

[goxy]

Re: Why this filter has such poor insertion loss characteris

S11 parameter of fillter is about 10dB and can be improved, but S12 is ok. Look into posted graphics below. So problem is not in the design.
Pin of SMA conector should be short.
Also check connectors with 50ohm microstrip line instead your filter.

BTW. Here you 30points


**broken link removed**

 

[yorkyong]

to goxy
thx for the points
The reason why I cannot improve more is the manufacture limit of the trace width and spacing. Both minimum are 0.2mm

 

[yingyang]

Re: Why this filter has such poor insertion loss characteris

Hi,
You can improve it by staggering the inner hairpin component.

 

[Debeli]

Re: Why this filter has such poor insertion loss characteris

What have you ment by "I only solder them at the tip of SMA." ? If you don't have propper ground connection for the SMA connector your signal will experience great loss so the connector body must be good connected to the ground. At 3 GHz you can be quite sure that length of the connector tip does not make any larger problem. It is the best to use microstrip SMAs with flat tip and spacing to insert board in so you can easilly solder them to board an get good ground.

Regards

 

[VSWR]

Re: Why this filter has such poor insertion loss characteris

This is the SMA connector to use for interconnection to the board:

https://www.conec.com/section24/s1/drw/009.pdf

The tab width is 1.27 mm, somewhat too wide to perfectly match the 0.8 mm 50 ohm track width of your design. But shouldn't give any significant mismatch at 3 Gigs. The SMA connector body shall be soldered directly to the ground Cu plane of your microstrip filter board. I have sucessfully used this kind of board test connectors for microstrip circuits up to about 10 Gigs.

 

[filterman]

Re: Why this filter has such poor insertion loss characteris

[Sure you can use this connector to Xband, but you have to match connector to 50 ohm microstrip line. Going from 50 ohm coaxial to 50 ohm microstrip might cause at 3ghz bad return loss (-16-14db). You have discontinuity of coaxial pin to microstrip width+discontinuity of ground plane+fields i coax are quite symetric, while in microstrip most of the field is in substrate, at low frequency you can get resonable RL , but the higher you go in frequency the worse RL will be. (try multiplying VSWR of such connector with a VSWR of the designed filter.
Now the main loss is in your case conductor loss+dielectric if you don't RF
substrate material. To minimize loss use low dielectric material, and low impedance level in resonators. You can also use thicker material for substrate.

 

[cross coupling]

Re: Why this filter has such poor insertion loss characteris

The mesured S11 in you design is not so good.
But the S21 should not be so bad even at this provided S11 value
The dielectric loss of the filter at this frequency range should not be so large if you use suitable substrate for RF/microwave.

 

[borich_03]

Re: Why this filter has such poor insertion loss characteris

I think your problem is wrong input matching. Try to use tapered input.

 

[microstrip_line]

Re: Why this filter has such poor insertion loss characteris

I suggest you refer to the book "microstrip filters for RF/Microwave application".
And the performance you have to take account the SMA connectors, fabrication error, and so on.