Continue to Site

# microstrip discontinuity

#### okik

##### Newbie level 6
How can I compensate T-junction discontinuity? I need advice on how to do it easily and especially correctly. Because my designed filter not working properly due to T-junction discontinuity.

Either include a high-quality circuit model of the T-junction in your simulations, or better yet, use a 2D/3D simulator to validate and/or tune the response.

Is Operating Frequency too high ?

3.2 T-junctions
The T-junctions is perhaps the most important discontinuity in a microstrip as it is found in most circuits such as
impedance networks, stub filters and branch line couplers. A microstrip T-junction and its equivalent circuit are
shown in the Figure 3. The discontinuity capacitance for this structure has been calculated by Silvester &
Benedek (1973).

The T-junction discontinuity compensation is much more difficult than right angled bends and steps in width
discontinuity compensation techniques. The T-junctions can be compensated by adjusting the lengths of the three
microstrip lines forming the junction

Is Operating Frequency too high ?
lowpass filter with cutoff freq = 694 MHz

Q too high for your expectations? Wrong impedance ratios? not enough tuned stubs?

No specs or results?

Last edited:

Along other approaches, using tapered transmission lines at the T-junction to gradually transition the impedance from the main transmission line to the branch lines. This can help minimize reflections and impedance mismatches.

A -3dB 694 LTE LPF can be -55 dB @ 750 MHz which might be a 20 pole LPF.

Where are your specs and results?

A -3dB 694 LTE LPF can be -55 dB @ 750 MHz which might be a 20 pole LPF.

Where are your specs and results?

for calculation to microstrip form --> substrate - Er=2.33 and h=1.27 mm, Zhigh=155.9, Zlow=40.73, cutoff freq=694MHz

lowpass filter with cutoff freq = 694 MHz
T-Junctions have very little effect on the response at that frequency.
I guess there are fundamental layout errors. Can you post a picture of the layout ??
Also, have you used ideal elements or practical components from a vendor ? How about GND connections and VIAs position ?
What are ZLOW and ZHIGH

T-Junctions have very little effect on the response at that frequency.
I guess there are fundamental layout errors. Can you post a picture of the layout ??
Also, have you used ideal elements or practical components from a vendor ? How about GND connections and VIAs position ?
What are ZLOW and ZHIGH
you can find some informations in above post #9 (calculation to microstrip form -> substrate - Er=2.33 and h=1.27 mm, Zhigh=155.9, Zlow=40.73, cutoff freq=694MHz)

Why did you not give full specs for pass (PB) and stop band (SB)? (ripple and attenuation)
How did you not get -6.02 dB in your plot in the PB?

Minor but significant calc. error on breakpoint f. Show your work.

Last edited:

you can find some informations in above post #9 (calculation to microstrip form -> substrate - Er=2.33 and h=1.27 mm, Zhigh=155.9, Zlow=40.73, cutoff freq=694MHz)
You're trying to realize a Low-Pass Filter by simply transforming L-C Components onto Microstrip Lines @ 694MHz ??
Also you use a substrate with Er=2.33 and then very thin long lines, a weird pattern and bizarre result.
The path you follow is wrong. Make some Researches about Microstrip Filters and come again.

FvM

1. How did you create inductance without capacitance in your design?
2. Can you describe how microstrip inductance and capacitance is created per unit length? (important skill)

Perhaps a different topology is better suited since it relies on a ground plane.

This is just a high ripple ladder with -146dB @ ~ 1 GHz.
Then you might ask, how do you make a larger C with near null L?
When geometric ratios for L/W strips cause constraints for ratios, how can you improve it?
= Change the ground plane gap. To minimize parasitic inductance choose an impedance << 10% of fundamental.

But you can make the above into a low ripple Inverse Chebyshev 20th order or more using wide short C Stubs, i.e Z << 7.5 Ohms but watch out for crosstalk or use laser trimmed lumped caps.

.

Last edited:

The correct filter that you intend to design. You cannot transform L-C components simply placing their equivalent MS elements.

The correct filter that you intend to design. You cannot transform L-C components simply placing their equivalent MS elements.
View attachment 189362
View attachment 189364
So question is - what should be the right step between lumped and distributed form, since it can't be done directly as you wrote?

So question is - what should be the right step between lumped and distributed form, since it can't be done directly as you wrote?
If it was so, we would place every component with distributed equivalent and we would obtain desired filter... Right ?
But it's not so simple as you thought.

The correct filter that you intend to design. You cannot transform L-C components simply placing their equivalent MS elements.
View attachment 189362
View attachment 189364
Is it possible to find somewhere the exact procedure how to achieve this scheme? Or could you please briefly describe the steps how to do it? Thank you for your reply.

Use a MoM simulator package. The microstrips have much higher Q's than the lumped LC elements and are recursive every second octave.

Is it possible to find somewhere the exact procedure how to achieve this scheme? Or could you please briefly describe the steps how to do it? Thank you for your reply.
It's a Filter Design Program. I have transferred it into Sonnet and simulated.