Hello Ian & Tom,
I just have couple of comment to interject into the exchange...
>>>>> Tom Wrote
>
>Hi Ian,
>
>> Tom, I'm not sure I used the correct terminolgy here in this context
>> ("ringing"). Ringing, as I understand it, is repetition of the basic
>> response of the filter at harmonics of the filter's centre frequency,
so
>> if one set up a narrow-band filter at say, 300 Hz, an analogue filter
>> would show damped responses at intervals of 300 Hz. DSP filters are
not
>> supposed to do this because these responses can be suppressed in
software,
>> or are automatically suppressed by the FFT algorithm used.
>
>Ringing occurs at the exciting frequency, or near it. Any high- Q
>filter will ring, and one with broad skirts will generally ring the
least
>for a given advertised bandwidth.
>
>DSP filters are no more immune than analog filters to this problem.
Finite Impulse Response (FIR) filters don't ring in the same sense as
a traditional analog filter or its DSP cousin the Infinite Impulse
Response (IIR).
The worst thing you can get from a narrowband FIR is a short chirp in
response to an impulse input. This will persist for a couple of cycles
whereas an analog filter with equivalent bandwidth (Q) will ring all day.
>If you hear someone saying they have a 100 Hz filter that does not
>ring, you can bet the filter's skirts are gradually sloped.
The FIR example above has slopes better than 200 dB/Octave.
>A sudden change in phase shift with frequency generally indicates
>less than ideal filter design or construction.
Common FIR filters have no phase shift.
>> Group delay does not exist in DSP filters because the system is not
>> operating in realtime, and does not use frequency
dependant/propogation
>> dependant components such as inductors and capacitors; the whole
process
>> is completed before anything reaches the output.
The very fact that the system is not realtime implies substantial group
delay.
>> Therefore in DSP filters there cannot be ringing or group delay, no
>> matter what is processed - unless the dynamic range of the ADC (and
DAC
>> stage?) is exceeded..
>Not so Ian. DSP filters have or can have substantial group delay. It
>may be easier to optimize software, but the problems are all still
>there.
All digital filters have group delay even if only one clock cycle. A
more cogent parameter is whether a particular DSP filter has equal group
delay
>Exceeding the range of the D-A or A-D conversion does not cause
>ringing per se, but it certainly causes distortion products!
That just causes evil.
>The bandwidth of the filter, and slope of the skirts, affects ringing
>just as it does an analog filter. I don't write software for digital
>filters, but I do evaluate the results in communications systems
>and DSP based filters have the same problems as any other filter.
There are good, bad, and ugly DSP implementations just as in any other
facet of electronics.
>If you don't want it to ring as bandwidth approaches the minimum
>required by the signal, it needs to have gently sloped skirts.
>
>The sales pitch for DSP filters reminds me of the sales pitch for
>certain antennas, like the "lossless linear loading" that is actually
>more lossy than even a reasonably good inductor.
Sales Pitch: At a booth here, during an ARRL SW Division convention
several years ago, where one of the early DSP outboard filter makers was
demonstrating the noise reduction features of their "box". They had a
noisy taped radio signal which they alternately passed and processed with
their creation. It was an impressive demo and seemed to de-noise quite
well.
The taped source happened to be in Chinese so, I inquired whether the
system would work as well on English transmissions... and received a
somewhat pained look in response from the booth operators along with some
signs of enlightenment from the other Hams present.
73,
Marv WC6W
*
________________________________________________________________
GET INTERNET ACCESS FROM JUNO!
Juno offers FREE or PREMIUM Internet access for less!
Join Juno today! For your FREE software, visit:
http://dl.www.juno.com/get/tagj.
--
FAQ on WWW: http://www.contesting.com/FAQ/amps
Submissions: amps@contesting.com
Administrative requests: amps-REQUEST@contesting.com
Problems: owner-amps@contesting.com
|