<weg22@drexel.edu> wrote in message
news:1192585289.299796.192650@q5g2000prf.googlegroups.com...
> Hi all,
>
> I'm looking for a good textbook that serves as an introduction to anti-
> aliasing filters, Nyquist frequency, sampling rates, SNR, low pass
> filter design, etc. Can someone please suggest a reference?
>
> Thanks in advance,
> -weg
>
weg,
It may depend on your background. If you're already well versed in linear
systems then maybe one. If not, maybe another.
The one I've used for years is:
Lathi: Signals, Systems and Communication. 1965.
It deals with signals and transforms and gives good insights. The sampled
data part is pretty short but meaningful when put in the framework of the
other stuff. I found it a good cover-to-cover read to gain understanding.
Lyons is always recommended.
If I might suggest:
Antialiasing is both an objective *and* an implementation matter.
It helps a lot to know what the objective is (thus the review of signals and
sampling).
Then, designing an adequate approach (to filter or not to filter, to sample
at a very high rate, etc.) becomes the implementation question.
One does not design an "antialiasing filter" as one would put a fuse in a
power line. Well, you can do it that way of course. Better, one designs a
system where aliasing will not be a problem. If that happens to include a
filter for the purpose of avoiding aliasing then, yes, there would be a
filter - either a lowpass filter or a bandpass filter depending on the
system configuration.
The characteristics of the filter will be as any other - based on
performance needs, "costs" of various kinds including phase and delay
characteristics perhaps.
Filters used in sample rate conversion are filters that avoid aliasing too
.... Halfband filters are an example of a nice implementation for many
situations.
I hope this isn't too convoluted and helps point you to useful
investigation.
Fred
Fred
Reply by jnarino●October 19, 20072007-10-19
On Oct 19, 3:58 pm, "Fred Marshall" <fmarshallx@remove_the_x.acm.org>
wrote:
> "robert bristow-johnson" <r...@audioimagination.com> wrote in message
>
> news:1192595769.928764.289010@t8g2000prg.googlegroups.com...
>
> > On Oct 17, 12:09 am, Jerry Avins <j...@ieee.org> wrote:
> >> we...@drexel.edu wrote:
>
> >> > I'm looking for a good textbook that serves as an introduction to anti-
> >> > aliasing filters, Nyquist frequency, sampling rates, SNR, low pass
> >> > filter design, etc. Can someone please suggest a reference?
>
> >> There's a good book on line athttp://www.dspguide/that'seven better
> >> on your desk in hard copy. "Understanding Digital Signal Processing" by
> >> Rick Lyons would be my recommendation if you're buying.
>
> > woah! i dunno if Rick even defines (or mentions) "Nyquist freqency"
> > specifically. a while back there was a dispute over the definition
> > where i was greatly surprized to find out that, of all books,
> > Oppenhiem and Schafer defined it to be the upper frequency of content
> > of what was being sampled (what i would have called the "bandlimit" or
> > "bandwidth" of the signal), whereas i previously thought that
> > EVERYBODY defined it as Fs/2. (so O&S are saying twice the Nyquist
> > frequency "must be exceeded by the sampling frequency" whereas the
> > only decent semantic i can deal with is that twice the Nyquist
> > frequency *is* the sampling frequency.)
>
> .................
>
> > r b-j
>
> This is a framework issue it seems to me.
>
> Signal centric frame:
> If the signal bandwidth is limited to B and B is known then setting a very
> high sampling rate seems to depart from saying that "twice the Nyquist
> frequency *is* the sampling frequency." Rather, the sampling frequency must
> be >2B and the Nyquist frequency is B. If the sample rate is 10B then 5B
> isn't the Nyquist frequency *for this signal*.
>
> System centric frame:
> If the signal bandwidth isn't known a priori then maybe it's OK to say that
> "twice the Nyquist frequency is the sampling frequency." Or, better still,
> half the sampling frequency is the Nyquist frequency ...... for which no
> signal frequency component should equal or exceed.
>
> Fred
If you dont mind, the real introduction is "Discrete Time Signal
Processing" by Rabiner. Its a classic, but it is not for the faint of
heart.
Juan
Reply by Fred Marshall●October 19, 20072007-10-19
"robert bristow-johnson" <rbj@audioimagination.com> wrote in message
news:1192595769.928764.289010@t8g2000prg.googlegroups.com...
> On Oct 17, 12:09 am, Jerry Avins <j...@ieee.org> wrote:
>> we...@drexel.edu wrote:
>>
>> > I'm looking for a good textbook that serves as an introduction to anti-
>> > aliasing filters, Nyquist frequency, sampling rates, SNR, low pass
>> > filter design, etc. Can someone please suggest a reference?
>>
>> There's a good book on line athttp://www.dspguide/that's even better
>> on your desk in hard copy. "Understanding Digital Signal Processing" by
>> Rick Lyons would be my recommendation if you're buying.
>>
>
> woah! i dunno if Rick even defines (or mentions) "Nyquist freqency"
> specifically. a while back there was a dispute over the definition
> where i was greatly surprized to find out that, of all books,
> Oppenhiem and Schafer defined it to be the upper frequency of content
> of what was being sampled (what i would have called the "bandlimit" or
> "bandwidth" of the signal), whereas i previously thought that
> EVERYBODY defined it as Fs/2. (so O&S are saying twice the Nyquist
> frequency "must be exceeded by the sampling frequency" whereas the
> only decent semantic i can deal with is that twice the Nyquist
> frequency *is* the sampling frequency.)
.................
> r b-j
>
This is a framework issue it seems to me.
Signal centric frame:
If the signal bandwidth is limited to B and B is known then setting a very
high sampling rate seems to depart from saying that "twice the Nyquist
frequency *is* the sampling frequency." Rather, the sampling frequency must
be >2B and the Nyquist frequency is B. If the sample rate is 10B then 5B
isn't the Nyquist frequency *for this signal*.
System centric frame:
If the signal bandwidth isn't known a priori then maybe it's OK to say that
"twice the Nyquist frequency is the sampling frequency." Or, better still,
half the sampling frequency is the Nyquist frequency ...... for which no
signal frequency component should equal or exceed.
Fred
Reply by glen herrmannsfeldt●October 17, 20072007-10-17
robert bristow-johnson wrote:
> woah! i dunno if Rick even defines (or mentions) "Nyquist freqency"
> specifically. a while back there was a dispute over the definition
> where i was greatly surprized to find out that, of all books,
> Oppenhiem and Schafer defined it to be the upper frequency of content
> of what was being sampled (what i would have called the "bandlimit" or
> "bandwidth" of the signal), whereas i previously thought that
> EVERYBODY defined it as Fs/2. (so O&S are saying twice the Nyquist
> frequency "must be exceeded by the sampling frequency" whereas the
> only decent semantic i can deal with is that twice the Nyquist
> frequency *is* the sampling frequency.)
An interesting distinction considering what Nyquist actually did.
He was finding out how fast sampled signals could pass through
a band limited analog channel. (Specifically, telegraph pulses.)
One could also consider that any frequency between half the
sampling frequency and the upper frequency as Nyquist frequencies.
It doesn't seem, though, that those describe the band limited but
not baseband case very well.
-- glen
Reply by Rune Allnor●October 17, 20072007-10-17
On 17 Okt, 06:36, robert bristow-johnson <r...@audioimagination.com>
wrote:
> i was greatly surprized to find out that, of all books,
> Oppenhiem and Schafer defined it to be the upper frequency of content
> of what was being sampled (what i would have called the "bandlimit" or
> "bandwidth" of the signal),
On this I agree with you. If one has a signal of bandwidth B
which is AM modulated with carrier fc >> B, then one can
sample with fs >= 2B, after proper pre-processing, and still
reconstruct the original signal.
> whereas i previously thought that
> EVERYBODY defined it as Fs/2. (so O&S are saying twice the Nyquist
> frequency "must be exceeded by the sampling frequency" whereas the
> only decent semantic i can deal with is that twice the Nyquist
> frequency *is* the sampling frequency.)
I would assume O&S formulate it that way because one can
*formally* postulate that some signal exists which is
perfectly bandlimited to B, i.e.
X(w) = 0, w > B
in which case x(t) is can be perfectly reconstrued
provided ws > 2B (note the strict inequality).
Of course, we can argue till the cows come home whether
that sort of detail is interesting from a *practical*
point of view -- and no, I don't find that sort of
discussion particularly interesting.
Rune
Reply by BobF●October 17, 20072007-10-17
Two books I recommend:
"Understanding Digital Signal Processing 2e" - Richard G Lyons
Rick mentions/describes Nyquist as the 'Nyquist criterion' on pg 28.
"Digital Signal Processing - A Practical Guide for Engineers and
Scientists" - Steven W Smith
Steven describes Nyquist on pg 41.
Theses two texts together have helped me more than anything else I've found
online or in print as far as getting the concepts clearly imprinted into my
head.
"robert bristow-johnson" <rbj@audioimagination.com> wrote in message
news:1192595769.928764.289010@t8g2000prg.googlegroups.com...
> On Oct 17, 12:09 am, Jerry Avins <j...@ieee.org> wrote:
>> we...@drexel.edu wrote:
>>
>> > I'm looking for a good textbook that serves as an introduction to anti-
>> > aliasing filters, Nyquist frequency, sampling rates, SNR, low pass
>> > filter design, etc. Can someone please suggest a reference?
>>
>> There's a good book on line athttp://www.dspguide/that's even better
>> on your desk in hard copy. "Understanding Digital Signal Processing" by
>> Rick Lyons would be my recommendation if you're buying.
>>
>
> woah! i dunno if Rick even defines (or mentions) "Nyquist freqency"
> specifically. a while back there was a dispute over the definition
> where i was greatly surprized to find out that, of all books,
> Oppenhiem and Schafer defined it to be the upper frequency of content
> of what was being sampled (what i would have called the "bandlimit" or
> "bandwidth" of the signal), whereas i previously thought that
> EVERYBODY defined it as Fs/2. (so O&S are saying twice the Nyquist
> frequency "must be exceeded by the sampling frequency" whereas the
> only decent semantic i can deal with is that twice the Nyquist
> frequency *is* the sampling frequency.)
>
> anyway, i wanted to see where Rick weighed in on this semantic
> dispute, and, much to my surprize, Rick side-stepped the issue and
> didn't mention the term at all! i ribbed him a little about that on
> this forum (a little too hard, for which i was sorry). anyway Rick
> just calls it "fs/2" (what i would call the "Nyquist frequency") and
> "B" (what O&S call the "Nyquist frequency"). actually, Rick does
> label fs/2 (and its negative) the "folding frequencies".
>
> but actually, for what the OP wants, i would say Rick's book is a very
> good choice. just don't look into it for a definition, or even a
> mention, of the otherwise ubiquitous term "Nyquist frequency".
>
> r b-j
>
Reply by robert bristow-johnson●October 17, 20072007-10-17
On Oct 17, 12:09 am, Jerry Avins <j...@ieee.org> wrote:
> we...@drexel.edu wrote:
>
> > I'm looking for a good textbook that serves as an introduction to anti-
> > aliasing filters, Nyquist frequency, sampling rates, SNR, low pass
> > filter design, etc. Can someone please suggest a reference?
>
> There's a good book on line athttp://www.dspguide/that's even better
> on your desk in hard copy. "Understanding Digital Signal Processing" by
> Rick Lyons would be my recommendation if you're buying.
>
woah! i dunno if Rick even defines (or mentions) "Nyquist freqency"
specifically. a while back there was a dispute over the definition
where i was greatly surprized to find out that, of all books,
Oppenhiem and Schafer defined it to be the upper frequency of content
of what was being sampled (what i would have called the "bandlimit" or
"bandwidth" of the signal), whereas i previously thought that
EVERYBODY defined it as Fs/2. (so O&S are saying twice the Nyquist
frequency "must be exceeded by the sampling frequency" whereas the
only decent semantic i can deal with is that twice the Nyquist
frequency *is* the sampling frequency.)
anyway, i wanted to see where Rick weighed in on this semantic
dispute, and, much to my surprize, Rick side-stepped the issue and
didn't mention the term at all! i ribbed him a little about that on
this forum (a little too hard, for which i was sorry). anyway Rick
just calls it "fs/2" (what i would call the "Nyquist frequency") and
"B" (what O&S call the "Nyquist frequency"). actually, Rick does
label fs/2 (and its negative) the "folding frequencies".
but actually, for what the OP wants, i would say Rick's book is a very
good choice. just don't look into it for a definition, or even a
mention, of the otherwise ubiquitous term "Nyquist frequency".
r b-j
Reply by Jerry Avins●October 17, 20072007-10-17
weg22@drexel.edu wrote:
> Hi all,
>
> I'm looking for a good textbook that serves as an introduction to anti-
> aliasing filters, Nyquist frequency, sampling rates, SNR, low pass
> filter design, etc. Can someone please suggest a reference?
There's a good book on line at http://www.dspguide/ that's even better
on your desk in hard copy. "Understanding Digital Signal Processing" by
Rick Lyons would be my recommendation if you're buying.
Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
Reply by ●October 16, 20072007-10-16
Hi all,
I'm looking for a good textbook that serves as an introduction to anti-
aliasing filters, Nyquist frequency, sampling rates, SNR, low pass
filter design, etc. Can someone please suggest a reference?
Thanks in advance,
-weg