hi, I am designing a deemphasis filter of 15/50us specification. I am using using bilinear transform method. But at frequencies near to nyquist frequency I am getting an error of 1dB. I tried pre wrapping. But I gives error reduction only at the frequency where the wrapping is applied. So I would like go for pole zero adjustment to reduce the error. What are the guidelines that I should follow while adjusting the pole and zero (I am dealing with first order filter). Is there any tutorials or pdf available, which I can follow...? --------------------------------------- Posted through http://www.DSPRelated.com
how to adjust poles and zeros for reducing error
Started by ●May 14, 2015
Reply by ●May 14, 20152015-05-14
"alpha1" <103600@DSPRelated> writes:> hi, > > I am designing a deemphasis filter of 15/50us specification. I am using > using bilinear transform method. But at frequencies near to nyquist > frequency I am getting an error of 1dB. I tried pre wrapping. But I gives > error reduction only at the frequency where the wrapping is applied. > > So I would like go for pole zero adjustment to reduce the error. What are > the guidelines that I should follow while adjusting the pole and zero (I > am dealing with first order filter). Is there any tutorials or pdf > available, which I can follow...?I would suggest using the FDLS (Frequency Domain Least Squares) filter design method instead. It is superb for simulating filters like this, i.e., analog filter magnitude and phase. Greg Berchin has written much about it - google his name and these phrases. -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
Reply by ●May 14, 20152015-05-14
On 5/14/2015 12:59 PM, alpha1 wrote:> So I would like go for pole zero adjustment to reduce the error. What are > the guidelines that I should follow while adjusting the pole and zero (I > am dealing with first order filter). Is there any tutorials or pdf > available, which I can follow...? >There is a Mathematica applet here "Transfer Function Analysis by Manipulation of Poles and Zeros" Which allows ones to see the effect of changing locations of poles and zeros. Drag a pole or a zero using the mouse and see the effect. Need a free reader to run it. The free reader is here: https://www.wolfram.com/cdf-player/
Reply by ●May 15, 20152015-05-15
On Thu, 14 May 2015 12:59:43 -0500, "alpha1" <103600@DSPRelated> wrote:>I am designing a deemphasis filter of 15/50us specification. I am using >using bilinear transform method. But at frequencies near to nyquist >frequency I am getting an error of 1dB. I tried pre wrapping. But I gives >error reduction only at the frequency where the wrapping is applied.https://groups.google.com/d/topic/comp.dsp/WSMrkyEcBBo/discussion Eight years and twelve days. Time flies. Send email directly to me if you want code for FDLS. Change "chatter" to "charter" and drop ".invalid".
Reply by ●June 9, 20152015-06-09
>On Thu, 14 May 2015 12:59:43 -0500, "alpha1" <103600@DSPRelated> wrote: > >>I am designing a deemphasis filter of 15/50us specification. I am using >>using bilinear transform method. But at frequencies near to nyquist >>frequency I am getting an error of 1dB. I tried pre wrapping. But Igives>>error reduction only at the frequency where the wrapping is applied. > >https://groups.google.com/d/topic/comp.dsp/WSMrkyEcBBo/discussion >Eight years and twelve days. Time flies. > >Send email directly to me if you want code for FDLS. Change "chatter" to >"charter" and drop ".invalid".Dear Greg Berchin, Could you please share your e-mail ID? Thanks in advance. --------------------------------------- Posted through http://www.DSPRelated.com
Reply by ●June 9, 20152015-06-09
On Tue, 09 Jun 2015 02:31:20 -0500, "alpha1" <103600@DSPRelated> wrote:>Dear Greg Berchin, > >Could you please share your e-mail ID? > >Thanks in advance."I'm sorry, (alpha1). I'm afraid I can't do that." Instead find the original comp.dsp thread at https://groups.google.com/forum/#!original/comp.dsp/wvD_EPWgnek/NDywVIz6qlMJ and locate my obfuscated address in the "From:" field there. Make the changes that I detail in the message, and you'll have my email address. Greg
Reply by ●June 16, 20152015-06-16
Hi, I went through https://groups.google.com/forum/#!original/comp.dsp/wvD_EPWgnek/NDywVIz6qlMJ I saw that Robert's poles and zeros at 48khz are really matching with the analog response. I used these poles and zeros and mapped them to poles and zeros at 44.1khz using BLT. I got error of 0.15dB only. But when I go for 32Khz or 24Khz the error is more than 0.5dB . I see Greg's and Robert's solutions are excellent. Can anyone help me with their design methods. Thanks in advance --------------------------------------- Posted through http://www.DSPRelated.com
Reply by ●June 16, 20152015-06-16
On 6/16/15 8:30 AM, alpha1 wrote:> Hi, > > I went through > https://groups.google.com/forum/#!original/comp.dsp/wvD_EPWgnek/NDywVIz6qlMJ > I saw that Robert's poles and zeros at 48khz are really matching with the > analog response. I used these poles and zeros and mapped them to poles and > zeros at 44.1khz using BLT. I got error of 0.15dB only. But when I go for > 32Khz or 24Khz the error is more than 0.5dB . I see Greg's and Robert's > solutions are excellent. Can anyone help me with their design methods.dunno fer sure who "Robert" is. i don't remember any solution i have ever done for a 15/50 microsecond deemphasis filter. -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
Reply by ●June 16, 20152015-06-16
On Tue, 16 Jun 2015 11:48:24 -0400, robert bristow-johnson <rbj@audioimagination.com> wrote:>dunno fer sure who "Robert" is. i don't remember any solution i have >ever done for a 15/50 microsecond deemphasis filter.He's referring to Robert Orban. https://groups.google.com/d/msg/comp.dsp/9kuirTaABzg/CSg4hgh0saoJ
Reply by ●June 17, 20152015-06-17
>On Tue, 16 Jun 2015 11:48:24 -0400, robert bristow-johnson ><rbj@audioimagination.com> wrote: > >>dunno fer sure who "Robert" is. i don't remember any solution i have >>ever done for a 15/50 microsecond deemphasis filter. > >He's referring to Robert Orban. > >https://groups.google.com/d/msg/comp.dsp/9kuirTaABzg/CSg4hgh0saoJyeah, Robert Orban.... I think Robert Orban's and Greg Berchin methods would realise an analog transfer function in digital domain with minimum error.... I am going through these methods.... @ Greg I have mailed u...My name is sharvin --------------------------------------- Posted through http://www.DSPRelated.com