Post subject: LC Low Pass Filter refuses to work properly
Unread postPosted: Thu Apr 07, 2005 2:47 pm
very perplexed. I designed a 5th order LC lowpass (Chebyshev) filter
to have a cut-off frequency of 108 MHz (the upper frequency of the FM
radio band) and have a minimum attenuation of 40 dB at 216 MHz. I went
through all the usual procedures like normalising the design and getting
the frequency scaling factor etc. The component values I obtained and
the circuit layout is as follows: The input is connected between C1
(38 pF) and L1 (115 nH), the other leg of C1 goes the ground. Then the
second leg of L1 is connected between C2 (66 pF) and L2 (115 nH), the
second leg of C2 also goes to ground. Now, the second leg of L2 is connected
between the output and C3 (38 pF), the second leg of C3 also goes to
ground. Sorry I can't draw a picture here but its just a basic layout.
I chose to build an LC lowpass circuit as I thought there would be too
much power dissipation across the resistor in an RC lowpass design?
The trouble is that when I built my lowpass filter it behaves the same
as a Notch filter. That is, the signal gets attenuated at (approx.)
the designed cut-off frequency, but slightly beyond this frequency the
signal level shoots up to its original (unattenuated) level. I have
used proper components (trimmer capacitors and tuneable coils) and their
values are very close to the ideal values listed above. I kept my leads
as short as possible, I screened the two inductors in case they couple
with each other, I etched some copper away from the ground plate under
the inductors in case this might affect the coil Q. None of these steps
seem to make any improvement - my LC lowpass filter is still behaving
like a notch filter! I have simulated my design on ADS and it works
almost exactly as designed. It should be so simple but I just don't
know why the thing isn't working in real life! You can probably tell
that I'm quite frustrated by my filter (or lack of filter!) so any helpful
suggestions will be most welcome.
Post subject: Filter respons
Unread postPosted: Fri Apr 08, 2005
There are two parts to the operation of a filter:
1. The filter circuit itself, including parasitics, and
2. The "context"
- the circuitry surrounding the filter. Since most filters work by reflecting
the power that is to be rejected, in the stopband the input VSWR gets
That means that you generally don't want lots of
transmission line between filters and whatever drives them - cable length
can affect filter response in odd ways. So: How are you testing this
Unread postPosted: Fri Apr 08, 2005 2:46 pm
Thank you for your reply. I am testing the filter on a Network Analyser.
I have also looked at it (my filter) on the spectrum analyser. That
is, I input a known signal from the signal generator and sweep through
the frequency range and no matter what way I look at it or no matter
what I do, it still behaves as a Notch filter. I have built it and re-built
it, I have tried different components, I have reduced the length of
my leads etc. Now I am going to try an build it one more time. I'll
keep you informed.
Unread postPosted: Fri Apr 08, 2005 4:26 pm
really don't underatand why you use tuneable coils? Why not use fixed
value wire wound coils, like those manufactured by Coilcraft.
From the description of your problem sounds like you have some sort
of resonance frequency in your filter. Could be that one of your coils
is working at its SRF (Self-Resonance Frequency). Use fixed value components
with high Q as possible for the inductors (as their Q is the dominant).
What is the BW of your filter is it 20MHz (88-108MHz)?
your filter on a board with good ground plane. Use vias to connect between
the GND nodes to the ground plane. You should also take into consideration
the Er and thickness of the substrate (even tough this is a relatively
low frequency, this might affect the filter's performance), the Er of
the filter will also dictate the 50 ohm trace width.
Unread postPosted: Sat Apr 09,
2005 9:57 am
I agree with the gentlemen above. At this low frequency
the simulator should be "dead nuts" on the response. Most likely, the
tuned coil is not behaving as you think it is, as suggetsed. I bet if
you replace the tunable components with fixed components, like coilcraft
coils and ATC caps, your filter will behave as expected.
Unread postPosted: Tue
Apr 19, 2005 12:09 pm
I'm happy to report that I eventually
managed to get my LC lowpass filters working, thanks in no small part
to the suggestions made in response to my question. I took the advice
of using wire wound ceramic chip inductors rather than tuneable coils.
This change, along with keeping copper/wire lengths to a minimum and
establishing very good grounds, helped to make the filter function properly.
Once again, many thanks to those who contributed.
Unread postPosted: Wed Apr 20, 2005 6:20 am
You welcome. Nothing
compares to experience!