Says who? The filter characteristics (and hence it's order) should be derived from an understanding of the problem. It sounds like you're applying a cookbook solution that is not only narrow and simplistic, but very possibly wrong in many cases. The filter order does depend on the ADC. If a 12-bit ADC is chosen having an attenuation in stop band of less than 72dB will ensure that the unwanted frequency band does not occupy even one bit of ADC. Same way for pass-band ripple.
LPF order at DAC output
Started by ●April 23, 2009
Reply by ●April 23, 20092009-04-23
krishna_1105 wrote:> The filter characteristics (and hence it's order) should be > derived from an understanding of the problem.^^^^^^^^^^^^^^ Right.> > The filter order does depend on the ADC.^^^^^^^^^^^^^^^^^^^^^ Wrong.> If a 12-bit ADC is chosen having > an attenuation in stop band of less than 72dB will ensure that the unwanted > frequency band does not occupy even one bit of ADC. Same way for pass-band > ripple.So, if I have a 1-bit DAC/ADC, I don't need any filters, huh? Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●April 23, 20092009-04-23
Vladimir Vassilevsky wrote:> > > krishna_1105 wrote: >> The filter characteristics (and hence it's order) should be derived >> from an understanding of the problem. > ^^^^^^^^^^^^^^ > Right. > > >> >> The filter order does depend on the ADC. > ^^^^^^^^^^^^^^^^^^^^^ > > Wrong. > >> If a 12-bit ADC is chosen having >> an attenuation in stop band of less than 72dB will ensure that the >> unwanted >> frequency band does not occupy even one bit of ADC. Same way for >> pass-band >> ripple. > > So, if I have a 1-bit DAC/ADC, I don't need any filters, huh?No, I think according to this line of argument if you have a ZERO bit converter you don't need a filter. A 1-bit filter would have to be down 6db in the stopband. I guess you should also consider the spectrum of the error signal, i.e. error less than 1/2 LSB can be a problem is it is not white. But isn't it true that generally speaking, the amount of stopband attenuation depends on the resolution of the A/D as well as the spectrum of the input signal? Or am I missing somthing? (Hopefully I'm not a Lamer!) -Jeff
Reply by ●April 23, 20092009-04-23
On Apr 23, 3:32�pm, Jeff Cunningham <j...@sover.net> wrote:> Vladimir Vassilevsky wrote: > > > krishna_1105 wrote: > >> The filter characteristics (and hence it's order) should be derived > >> from an understanding of the problem. > > ^^^^^^^^^^^^^^ > > Right. > > >> The filter order does depend on the ADC. > > ^^^^^^^^^^^^^^^^^^^^^ > > > Wrong. > > >> If a 12-bit ADC is chosen having > >> an attenuation in stop band of less than 72dB will ensure that the > >> unwanted > >> frequency band does not occupy even one bit of ADC. Same way for > >> pass-band > >> ripple. > > > So, if I have a 1-bit DAC/ADC, I don't need any filters, huh? > > No, I think according to this line of argument if you have a ZERO bit > converter you don't need a filter. A 1-bit filter would have to be down > 6db in the stopband. I guess you should also consider the spectrum of > the error signal, i.e. error less than 1/2 LSB can be a problem is it is > not white. > > But isn't it true that generally speaking, the amount of stopband > attenuation depends on the resolution of the A/D as well as the spectrum > of the input signal? Or am I missing somthing? (Hopefully I'm not a Lamer!) > > -Jeff- Hide quoted text - > > - Show quoted text -I could not find the starting thread here, but anyway ... You need to consider at least 3 cases; 1) A "standard" data-acquisition-type ADC where the thermal noise is well below 1lsb of the ADC resolution 2) An ADC that is dithered to get resolution below 1LSB (or maybe it's own thermal noise does the dithering for you) 3) Oversampled A/D's (could be noise-shaped, or could just be oversampled) And you need to ask yourself 1) Do I care what the converter error spectrum looks like, or, if it's down by X dB that's all I need to know? 2) Am I trying to get resolution below the LSB (spectrally speaking) by using dither? If you answered "1" to the first group and "No" for question 1, then your needs fall into a fairly conventional category and you can specify your filter stopband according to the converter resolution; however even with this you need to be careful; for example, when you say that the attenuated signal "does not occupy even one bit of ADC", how exactly are you going to make sure that your attenuated sine-wave is nicely centered between the ADC decision points such that it does not toggle an LSB? This might not matter if you are not too picky, but the correct answer if you care about the spectrum of your signal is "use dither". If you don't care so much about the spectrum, and only care about the SNR, then life is easy. Bob Adams
