Any filter in fact processes several bits at a time. Left-over
energy from previous bits affects the filter state of the current bit. It's
a fundamental physical limit, you can blame Mr. Fourier.
A filter with finite time response (limited ringing duration) has
infinite frequency response (that is, it is a poor filter), and any filter
with finite frequency response has infinite time response (ringing). It's a
byproduct of Fourier's transform. Practically however, we only care about
ringing so-many dB below the main response, and it's the same with frequency
response, so we can get rid of the infinite nonsense.
What it means is that any filter that processes just one bit would
have pretty bad frequency properties. - it wouldn't reject noise or qrm that
well. A better filter will have a ringing response that covers multiple bit
times. The secret is in controlling exactly how the ringing impacts bits in
the past, and those to arrive in the future, and trading that off with good
frequency response properties.
The repetitive pattern does not excite many of the problematic
filter states. The more random the pattern the more likely a deleterious
filter state will impact the results.
A rather large amount of information exists on this subject, that of
data modem filter design.
-- Tom, N5EG
From: email@example.com [mailto:firstname.lastname@example.org] On
Behalf Of Bill Turner
Sent: Thursday, April 22, 2004 10:19 AM
Subject: Re: [RTTY] Multipath, testing, soundcards
On Thu, 22 Apr 2004 09:57:32 -0500, Tom McDermott wrote:
>pattern (specifically: RYRYRYRY...) is the worst possible pattern to
>utilize for such test. The pattern should have the most random-like
>pattern of bits. This is because badly-designed modem filters can pass
>the repeating pattern test with good results, yet yield poor copy with
>more random characters.
Could you please explain this further? It would seem to me that a filter
which only processes one bit at a time would have no *knowledge* of what
came before or what follows afterwards, and therefore would not be capable
of recognizing whether a pattern was repeating or not. In other words,
would MARK MARK MARK MARK be processed differently than MARK SPACE MARK
SPACE? Perhaps you are saying that a poorly designed filter would not
handle the transition from MARK to SPACE correctly?
Either way, this could be proven empirically pretty easily I would think.
QSLs via LoTW
RTTY mailing list
RTTY mailing list