Cannot use filter beyond 30 kHz

vt.y4505 Question

Jun 14, 2016 00:00

Hi,

I have a 33 kHz sinusoidal analog waveform with Vmin 0.5V and Vmax 2.5V. The frequency spectrum of this waveform also has minor amplitudes at frequencies higher and lower than 33 kHz that affect the phase stability of this waveform, so I want to filter these components out.

I was planning to use a Butterworth low pass filter available in Sigma Studio on ADAU1761 with the maximum available sampling rate of 96 KHz. I have used these before for signals at lower frequencies, but saw that the maximum available range on these filters is 30KHz, I cannot set any cut-off value beyond 30KHz. Could anyone explain why that is the case ?

Thanks,

Vivek

Audio Signal Processors

SigmaDSP Audio Processors

adau1761

Report

ADIApproved Reply

Jun 14, 2016 00:00

Hi Vivek,

Which filter block are you using? I see that the Nth order filter does not allow frequencies above 30kHz to be input, but the Second Order General filter allows for higher frequencies and has a Butterworth LPF option.

Thanks,

Tyler

Report

vt.y4505 Reply

Jun 14, 2016 00:00

Hi Tyler, I was using the Nth order block and had to use a 4th order filter

to get rid of some unwanted frequencies. Thanks for pointing out the second

order option, I can try to see if that works.

Do you know why the Nth order option doesn't allow higher frequencies ?

Thanks,

Vivek

On Jun 14, 2016 3:11 PM, "TylerK" <analog@analog-vm.hosted.jivesoftware.com>

Report

KJBob Reply

Jun 15, 2016 00:00

Hello Vivek and Tyler,

It's also possible to design a sharp multi-section filter with analog techniques and then bring the resulting design into SigmaStudio. This was how we did it before the Nth Order Filter became available, and it's just as valid today. To see how, scroll down to the bottom of this link: The Nth Order Filter . There a highpass filter is described but the procedure works for lowpass and bandpass filters as well (you did say that your unwanted signals are both above and below the desired frequency). BTW, unless your application has tight restraints for linear phase and pass-band ripple, I suggest choosing a Chebyshev response instead of Butterworth. It can take as many as nine sections of Butterworth filter to do the work of a two-stage Chebyshev!

If your 33 KHz signal is supposed to be a sine wave (no useful info at nearby frequencies) and you just wish to "clean it up", try a second-order peaking filter with frequency of 33 KHz and Q as high as it goes (15.99). This might do the trick.

Best regards,

Bob

Report

vt.y4505 Reply

Jun 15, 2016 00:00

Hi Bob,

Thanks for your suggestions. As you ointed out, I tried using a 2nd order Chebyshev Filter yesterday, and it does as good or better than a higher order Butterworth filter. I just wish to clean the signal because I am sending it to a Lock-in Amplifier which locks on to the frequency of this waveform. Any other frequencies in this waveform adds a phase drift to this lock. Because I am using a lock-in amplifier, I also want do not want to make the digital filter response time slower than the lock-in time constant, so using a lower order Chebyshev is definitely better.

I'll check out the peaking filter idea and Nth order link you sent, and see if it helps further.

Thanks,

Vivek

Report