I recall seeing a definition of a filter stopband specification that had some condition in it about including only the second "bounce". Details are hazy, anyone got any pointers? My real question is how is the stopband rejection actually defined when there are ripples? TIA!
Filter stopband definition?
Started by ●December 26, 2008
Reply by ●December 26, 20082008-12-26
On 26 Des, 15:45, System Alchemist <dotter@nospam_btconnect.com> wrote:> I recall seeing a definition of a filter stopband �specification > that had some condition in it about including only the second "bounce". > > Details are hazy, anyone got any pointers? > > My real question is how is the stopband rejection actually defined when > there are ripples?The stopband is defined as the frequency band with a minimum attenuation or maximum gain, depending on terminology. So when there are ripples in the filter response, that's OK as long as the peaks don't exceed the gain specification. Rune
Reply by ●December 26, 20082008-12-26
System Alchemist wrote:> I recall seeing a definition of a filter stopband specification > that had some condition in it about including only the second "bounce". > > Details are hazy, anyone got any pointers? > > My real question is how is the stopband rejection actually defined when > there are ripples?I'm not sure there's a strict definition, at least not in the sense that you're thinking of. About the only colloquial definition would be anything beyond the first null. Usually, stopband rejection is a *requirement* that affects the filter design. In this sense, a rejection of X dB is merely saying that beyond the transition bandwidth, the filter response must be X dB below the passband at all points. -- Oli
Reply by ●December 26, 20082008-12-26
Rune Allnor wrote:> On 26 Des, 15:45, System Alchemist <dotter@nospam_btconnect.com> > wrote: >> I recall seeing a definition of a filter stopband specification >> that had some condition in it about including only the second "bounce". >> >> Details are hazy, anyone got any pointers? >> >> My real question is how is the stopband rejection actually defined when >> there are ripples? > > The stopband is defined as the frequency band with a minimum > attenuation > or maximum gain, depending on terminology. So when there are ripples > in > the filter response, that's OK as long as the peaks don't exceed the > gain specification. > > RuneYou mean maximum attenuation or minimum gain. But hey -- gain, attenuation -- what's the difference? -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" gives you just what it says. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by ●December 26, 20082008-12-26
System Alchemist wrote:> I recall seeing a definition of a filter stopband specification > that had some condition in it about including only the second "bounce". > > Details are hazy, anyone got any pointers? > > My real question is how is the stopband rejection actually defined when > there are ripples? > > TIA!One common way -- and my preferred way -- is to define a minimum gain (or attenuation) line and a maximum gain (or attenuation) line vs frequency. Usually you make these piecewise linear on lin-lin, log-lin or log-log space so you can describe them in the text, too. I like the method because it's very easy to define where the specification absolutely must be tight and where it can be loose. Most of the telecom and RF specs that I've seen specify their filters and/or spectra this way. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" gives you just what it says. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by ●December 26, 20082008-12-26
On 26 Des, 21:11, Tim Wescott <t...@seemywebsite.com> wrote:> Rune Allnor wrote: > > On 26 Des, 15:45, System Alchemist <dotter@nospam_btconnect.com> > > wrote: > >> I recall seeing a definition of a filter stopband �specification > >> that had some condition in it about including only the second "bounce". > > >> Details are hazy, anyone got any pointers? > > >> My real question is how is the stopband rejection actually defined when > >> there are ripples? > > > The stopband is defined as the frequency band with a minimum > > attenuation > > or maximum gain, depending on terminology. So when there are ripples > > in > > the filter response, that's OK as long as the peaks don't exceed the > > gain specification. > > > Rune > > You mean maximum attenuation or minimum gain.I was thinking about the stop-band spec. Maybe "maximum allowed gain" or "minimum required attenuation" is better? Rune
Reply by ●December 26, 20082008-12-26
Oli Charlesworth wrote:> System Alchemist wrote: >> I recall seeing a definition of a filter stopband specification >> that had some condition in it about including only the second "bounce". >> >> Details are hazy, anyone got any pointers? >> >> My real question is how is the stopband rejection actually defined when >> there are ripples? > > I'm not sure there's a strict definition, at least not in the sense that > you're thinking of. About the only colloquial definition would be > anything beyond the first null. > > Usually, stopband rejection is a *requirement* that affects the filter > design. In this sense, a rejection of X dB is merely saying that beyond > the transition bandwidth, the filter response must be X dB below the > passband at all points.*At least* X dB below. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●December 26, 20082008-12-26
Rune Allnor wrote:> On 26 Des, 21:11, Tim Wescott <t...@seemywebsite.com> wrote: >> Rune Allnor wrote: >>> On 26 Des, 15:45, System Alchemist <dotter@nospam_btconnect.com> >>> wrote: >>>> I recall seeing a definition of a filter stopband specification >>>> that had some condition in it about including only the second "bounce". >>>> Details are hazy, anyone got any pointers? >>>> My real question is how is the stopband rejection actually defined when >>>> there are ripples? >>> The stopband is defined as the frequency band with a minimum >>> attenuation >>> or maximum gain, depending on terminology. So when there are ripples >>> in >>> the filter response, that's OK as long as the peaks don't exceed the >>> gain specification. >>> Rune >> You mean maximum attenuation or minimum gain. > > I was thinking about the stop-band spec. Maybe "maximum > allowed gain" or "minimum required attenuation" is better? > > RuneYes, that's better. This is why I like text backed up with pictures, or pictures backed up with text. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" gives you just what it says. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by ●December 30, 20082008-12-30
System Alchemist wrote:> I recall seeing a definition of a filter stopband specification > that had some condition in it about including only the second > "bounce". > Details are hazy, anyone got any pointers? > > My real question is how is the stopband rejection actually defined > when there are ripples? > > TIA!I think you're probably referring to an arm-waving kind of description that is more academic or analytical than practical. You've received a lot of answers addressing the practical. Consider this: The passband of a filter has often been defined as between 3dB down points - there's lots of technique from the analog world that makes this a handy definition. Choosing this point / these points has nothing to do with what the filter needs to do really. So, if one wants to, one can similarly define the stopband as beginning and ending at certain points. For equiripple filters it's usually those points where the filter gain transitions from "greater than" the stopband peak to "less than" the stopband peak - a point which is at the stopband peak itself and on the stopband edges generally. If the filter criteria are different, such as least squares or .. whatever .. then it might be meaningful in some sense to define the stopband edges differently. In the example you used, if the filter were a perfect brick wall filter then one might want to ignore the first sidelobes in defining the "stop" band. Likely someone chose this because the attenuation wasn't "good enough" otherwise. But, it remains an arbitrary definition with respect to real filter performance. Fred
Reply by ●January 8, 20092009-01-08
Fred Marshall wrote:> System Alchemist wrote: >> I recall seeing a definition of a filter stopband specification >> that had some condition in it about including only the second >> "bounce". >> Details are hazy, anyone got any pointers? >> >> My real question is how is the stopband rejection actually defined >> when there are ripples? >> >> TIA! > > I think you're probably referring to an arm-waving kind of description that > is more academic or analytical than practical. You've received a lot of > answers addressing the practical. > > Consider this: The passband of a filter has often been defined as between > 3dB down points - there's lots of technique from the analog world that makes > this a handy definition. Choosing this point / these points has nothing to > do with what the filter needs to do really. > > So, if one wants to, one can similarly define the stopband as beginning and > ending at certain points. For equiripple filters it's usually those points > where the filter gain transitions from "greater than" the stopband peak to > "less than" the stopband peak - a point which is at the stopband peak itself > and on the stopband edges generally. > > If the filter criteria are different, such as least squares or .. whatever > .. then it might be meaningful in some sense to define the stopband edges > differently. In the example you used, if the filter were a perfect brick > wall filter then one might want to ignore the first sidelobes in defining > the "stop" band. Likely someone chose this because the attenuation wasn't > "good enough" otherwise. But, it remains an arbitrary definition with > respect to real filter performance.Thanks Fred and everyone else who responded to my fuzzy question. I have a follow-on question for bonus points... In *normal usage*, when a DDC filter is specified as 80% of the output bandwidth, is the 80% point at the passband edge or the -3dB power (-6dB amplitude) point? "passband edge" in this context being the Matlab definition of where the filter response falls below the allowable passband ripple.
