Eng-Tips is the largest engineering community on the Internet
Intelligent Work Forums for Engineering Professionals
narrowbandpass filter 1
- Thread starter ekor
- Start date
-
1
- #2
It is the shape factor that would govern the order of the filter required, rather than the bandwidth. Decide how steep the “skirts” of the filter need to be and you can then work out how more about the response. Is ripple in the stop band acceptable? It is going to be an LC filter, but at those frequencies the LC values may need to be made using microstrip sections.
The LC values can be obtained from standard filter handbooks like Zverev. The microstrip sections would have to come from specialist software I guess, but I am not familiar with that frequency range specifically.
The LC values can be obtained from standard filter handbooks like Zverev. The microstrip sections would have to come from specialist software I guess, but I am not familiar with that frequency range specifically.
divebuddydave
Electrical
- Feb 6, 2003
- 25
- 0
- 0
Check out They have several free student versions of their very powerful design software available for download that may be of help in designing your filter.
David
David
I am an old engineer with several patents and I have learned
that if a job is difficult, it may be worth to step back
and examine if it is necessary at all.
You have to check the temperature range and the S/N
ratio, any strong noise sources in the band, modulation
mode, analog/digital, (if digital, error correction ),etc.
What about mixing it down ? ( use perhaps HF preamp
with a wider band filter and do the rest of the filtering
at IF )
<nbucska@pcperipherals.com>
that if a job is difficult, it may be worth to step back
and examine if it is necessary at all.
You have to check the temperature range and the S/N
ratio, any strong noise sources in the band, modulation
mode, analog/digital, (if digital, error correction ),etc.
What about mixing it down ? ( use perhaps HF preamp
with a wider band filter and do the rest of the filtering
at IF )
<nbucska@pcperipherals.com>
1% bandwidth is difficult to achieve, even at constant temperature, if you can tolerate the loss, look into dielectric resonator based designs, or at the extreme, the saphire loaded cavity. If not then it would be waveguide cavity based, that at that frequency can get rather large. But as nubska says, there is more than one way to skin a cat,
Here is my first attempt at the problem, based on the filter tables and methods in RF Circuit Design by Bowick.
Centre Freq= 2.45GHz
3dB bandwidth= 25MHz
using a 3 pole 0.5dB Chebyshev low-pass prototype for 50 ohm input and 50 ohm output.
This PI-section filter simulates beautifully.
Input end: parallel pair C=237pF // L=17.8pH wired across the input
Series pair C=10.36fF - L=407nH, wired input to output
Output end: parallel pair C=233pF // L=18.07pH, wired across the output
The femto-farad capacitor and the pico-henry inductor make this design look infeasible.
Since the skirts have not been specified, let’s just try a single resonant pair.
It turns out that the input parallel pair from the first try is just about right C=237pF // L=17.8pH
It could be fine tuned to give the exact bandwidth and centre frequency, but the values are pretty close and are still stupid.
My third and final attempt produces values which are more sensible and I think they might be realisable with a microstrip design. There are 5 reactive elements involved. From the input there is a 7pF capacitor to ground. The filter is symmetrical, so there is also a 7pF capacitor to ground across the output. In the middle is a 4.5nH inductor to ground and this is coupled to the input and output pins by 0.5pF capacitors. These value roughly achieve the spec required. When parasitic elements are included as well then I don’t know. That requires a proper microwave simulator rather than a simple SPICE simulation.
Centre Freq= 2.45GHz
3dB bandwidth= 25MHz
using a 3 pole 0.5dB Chebyshev low-pass prototype for 50 ohm input and 50 ohm output.
This PI-section filter simulates beautifully.
Input end: parallel pair C=237pF // L=17.8pH wired across the input
Series pair C=10.36fF - L=407nH, wired input to output
Output end: parallel pair C=233pF // L=18.07pH, wired across the output
The femto-farad capacitor and the pico-henry inductor make this design look infeasible.
Since the skirts have not been specified, let’s just try a single resonant pair.
It turns out that the input parallel pair from the first try is just about right C=237pF // L=17.8pH
It could be fine tuned to give the exact bandwidth and centre frequency, but the values are pretty close and are still stupid.
My third and final attempt produces values which are more sensible and I think they might be realisable with a microstrip design. There are 5 reactive elements involved. From the input there is a 7pF capacitor to ground. The filter is symmetrical, so there is also a 7pF capacitor to ground across the output. In the middle is a 4.5nH inductor to ground and this is coupled to the input and output pins by 0.5pF capacitors. These value roughly achieve the spec required. When parasitic elements are included as well then I don’t know. That requires a proper microwave simulator rather than a simple SPICE simulation.
Their customer support should be happy to send you a sample kit.
- Status
Similar threads
- Locked
- Question
- Locked
- Question