On 17.10.15 07:22, Sharan123 wrote:>> Feel free to reiterate your question if you need to. > > Dear Tim, > > What happens if one has access to sampled data and you have not control or > information about the system which was used to sample. All you know is, > sampling was done at sufficient rate. Probably, very hypothetical > question, but what are the implications of using incorrect sampling value > while designing a filter.The filtering measures frequency (and time) based on the sample rate only. You cannot design filters related to real-world frequencies, they are relative to the sample rate. -- -Tauno Voipio
Filter and sampling rate
Started by ●October 15, 2015
Reply by ●October 17, 20152015-10-17
Reply by ●October 17, 20152015-10-17
Sharan123 <99077@dsprelated> wrote: (snip)> What happens if one has access to sampled data and you have not control or > information about the system which was used to sample. All you know is, > sampling was done at sufficient rate. Probably, very hypothetical > question, but what are the implications of using incorrect sampling value > while designing a filter.In many cases, you want filters with specific real-world frequencies. Some old favorites are high and low blocking filters on audio amplifiers. For some audio source, there is a lot of high frequency noise, so blocking above about 6kHz helps. Also, turntable (remember those) rumble is low, and blocking below about 18Hz is good. In both cases, specific Hz values are used. But there might be plenty of cases where the filters are relative to the sampling rate. If you know a signal has high or low frequency noise, and not much useful signal at those frequencies, you could filter that out, independent of the real sampling rate. Visual signals are more likely sampling frequency independent. The pixel spacing on the sensor is the spatial frequency sampling rate, but often doesn't have an important value. More is often better, but with a zoom lens you don't really care, but you still might want to do some filtering. -- glen
Reply by ●October 17, 20152015-10-17
On Fri, 16 Oct 2015 23:22:01 -0500, Sharan123 wrote:>>Feel free to reiterate your question if you need to. > > Dear Tim, > > What happens if one has access to sampled data and you have not control > or information about the system which was used to sample. All you know > is, sampling was done at sufficient rate. Probably, very hypothetical > question, but what are the implications of using incorrect sampling > value while designing a filter.I'm not sure, at that point, that the sampled data would be of much use, so what's the point of filtering? -- www.wescottdesign.com
Reply by ●October 18, 20152015-10-18
On 10/16/15 2:10 PM, Tim Wescott wrote:> On Fri, 16 Oct 2015 00:53:29 +0000, glen herrmannsfeldt wrote: > >> Tim Wescott<seemywebsite@myfooter.really> wrote: >>> On Thu, 15 Oct 2015 09:57:37 -0500, Sharan123 wrote: >> >> >>>> I am curious as to why sampling parameter is important while designing >>>> a filter. I guess, it is sufficient that the original signal has been >>>> sampled at a sufficient rate. >> >>> I'm not sure where your confusion lies, but when you design a digital >>> filter all of it's behavior is normalized by the sampling rate. >> >> The sampling rate isn't important if your filter frequencies are all >> relative to the sample rate. >> >> It depends on how you do filter design. > > But then the sampling rate is important so that you can plug the correct > number into your filter design. >no, that's not the case, Tim. if the filter spec frequencies are all proportional to the sample rate, it only matters what those proportionalities are to determine what the correct numbers to plug into your filter.. but, for the most part, a filter design meant to have some sorta human exposure, the specs are Hz and then you need Fs. -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
Reply by ●October 18, 20152015-10-18
>no, that's not the case, Tim. if the filter spec frequencies are all >proportional to the sample rate, it only matters what those >proportionalities are to determine what the correct numbers to plug into>your filter.. > >but, for the most part, a filter design meant to have some sorta human >exposure, the specs are Hz and then you need Fs. > >-- > >r b-j rbj@audioimagination.com > >"Imagination is more important than knowledge."Robert, How can somebody know the proportions with knowing num/denum??? Knowledge is more imortant than imagination Kaz --------------------------------------- Posted through http://www.DSPRelated.com
Reply by ●October 18, 20152015-10-18
On 10/18/15 5:38 AM, kaz wrote:>> no, that's not the case, Tim. if the filter spec frequencies are all >> proportional to the sample rate, it only matters what those >> proportionalities are to determine what the correct numbers to plug into >> your filter.. >> >> but, for the most part, a filter design meant to have some sorta human >> exposure, the specs are Hz and then you need Fs. >>> > How can somebody know the proportions with knowing num/denum???well, the only example i can think of from the top of my head would be decimation filters in a sigma-delta converter. those filters would have cutoff frequency that would be specified as a fraction of the sample rate. knowledge of the ratio is what you would take to MATLAB to get your filter coefficients designed. Fs would not be in on it.> Knowledge is more imortant than imaginationtell that to Uncle Albert. :-) -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
Reply by ●October 18, 20152015-10-18
On Saturday, October 17, 2015 at 5:22:04 PM UTC+13, Sharan123 wrote:> >Feel free to reiterate your question if you need to. > > Dear Tim, > > What happens if one has access to sampled data and you have not control or > information about the system which was used to sample. All you know is, > sampling was done at sufficient rate. Probably, very hypothetical > question, but what are the implications of using incorrect sampling value > while designing a filter. > > --------------------------------------- > Posted through http://www.DSPRelated.comThis in fact is elementary. If you design a filter to have say a notch at quarter sampling and you clock the data out at 20kHz, then the real digital filter will have a notch at 5kHz. Double the rate of data passing through to 40kHz and that notch will now move to 10kHz. The sampling rate enables you to map a normalised frequency into the real-word (analogue)
Reply by ●October 18, 20152015-10-18
kaz <37480@dsprelated> wrote:>>no, that's not the case, Tim. if the filter spec frequencies are all >>proportional to the sample rate, it only matters what those >>proportionalities are to determine what the correct numbers to plug into >>your filter..(snip)> How can somebody know the proportions with knowing num/denum??? Knowledge > is more imortant than imaginationAs I noted, in audio it is likely more usual to have an actual frequency. For images, it seems to me much less likely, at least in terms of absolute spatial frequency. (On the sensor, or where?) Someone might low-pass filter an image, adjusting the filter parameters until it looks right. That is, without knowing the actual spatial frequency. And even for audio, one might have a parametric equalizer, where the knobs are adjusted for the best sound, without looking at the actual frequencies involved. Tone controls on amplifiers are rarely marked with the actual frequencies. It just isn't needed. -- glen
