From: "Mat Nieuwenhoven"
Reply-To: "Mat Nieuwenhoven"
X-Newsreader: PMINews 2.00.1205 For OS/2
Subject: Which opamp filter for manchester data stream?
Date: Sun, 03 Nov 2002 13:37:33 GMT
NNTP-Posting-Date: Sun, 03 Nov 2002 14:37:33 MET
Organization: chello broadband
I'm building a packet radio modem for 38.4 kbps. The radio uses FM. In order
to limit the bandwidth and concentrate the RF energy in a smaller band (thus
increasing range) I'd like to filter the manchester-encoded data stream
before it is fed to the transceiver. A manchester-encoded data stream
consists of pulses with a frequency equal to the baud rate or half of that;
thus, I have positive or negative pulses of approx 26 us or 52 us wide. The
receiver modem has a sampler that samples the received data in the middle of
each bit, so if the transmitted signal approaches a sine-wave, that is quite
accepatble. Manchester data does not have a DC component.
What kind of filtering should I use? I have available one half of a LM324
single supply opamp (2 opamps). I understand there are several active filter
topologies, like Butterworth (maximally flat response), or Bessel (maximally
flat delay), or Chebyshev (don't know what that's designed for). Which one is
best for digital pulses? I've read Bruce Carter's SLOA093, and it has simple
designs for low pass and high pass Butterworth.
Maybe I better state the things I don't know:
1. What will give a narrower RF FM spectrum after filtering, using the 2
opamps for lowpass at 38.4 or one for highpass at 19.2 and one for lowpass at
38.4? Is lowpass any use at all at the transmitting side? On AM, the highest
modulation frequency determines the bandwidth, on FM it is much more
complicated, the strength of each frequency is important. I seem to recall
that manchester data has a significant component below half the baud rate,
but I might be wrong.
2. Which of the active filter topologies is best for this, and equally
important, which one is least effected by component variations? Obtaining
single 1% resistors is not a problem, 1% capacitors certainly is, even 5%
might be difficult. However, I do have an accurate (1% or so) C meter for
small values (< 1uF). Related, are ceramic C's stable enough or should I use
polystyrene or what?
3. In an old Signetics databook (1974), with the datasheet of the NE531
opamp, they give a single opamp Butterworth lowpass design with 3 poles,
basically preceding the standard 2-pole with a R and C; the reference states
this comes from EDN magazine, 15 dec 1970. For the record: the 1KHz filter
uses 3 10K resistors in series between input and opamp, 22n C after 1st R to
gnd, 56 C after 2nd R to opamp output, 32n C after 3rd R to gnd. Is is worth
adding this (PCB space is tight)?
4. The receiver currently uses a double active lowpass filter with 2 2k2 Rs
and 2 1n Cs. Performance of the modem improved noticably with this filter
between modem input and transceiver output, however, after reading SLOA093 I
wonder if there are not component values that will yield better results. Any
As you can tell, I know next to nothing about opamp filters, so I'd
appreciate any suggestions. Does there exist free software that lets me
simulate this, inputting a user-defined data stream, simulate the filter, and
run an FFT on the output? I don't have Windows, only Linux and OS/2.
Regards, Mat Nieuwenhoven