Filter corner frequency

Thread Starter


Joined Jan 6, 2004
Cut off / corner frequency.

Thinking of a DAC, spiting, say 2000 values/sec and the associated LPF at the output:

According to what I've read up to now, the filter should have a (-3dB) cut off frequency of 2KHz.

My question: why from the beginning I accept to design with an atenuated output since my interest is to have the signal restored with the right level, that is with 0dB attenuation.

In other words, shouldn't I design for a corner frequency where my highest signal of interest is not attenuated at all?

I do not discard misunderstand things or plainly ignoring many.
Last edited:

Ron H

Joined Apr 14, 2005
With a 2kHz sample rate, the maximum signal frequency you can reconstruct is 1kHz, so a 2kHz filter (with a high enough order) will cause minimal attenuation.


Joined Feb 24, 2006
You can put the corner anywhere you want. It does not have to be an integer multiple of the highest frequency you want to reconstruct. In fact across the passband with an analog filter there will always be some attenuation depending on the type of filter and its order. The engineering part is to specify how much you are willing to accept.

Thread Starter


Joined Jan 6, 2004
Thanks to you both for replying.

I omited to say I was oversampling with 2KHz rate a signal around 286 Hz.

My first two tries with corner frequencies of 2000 Hz (4 poles) and 2250 HZ (6 poles) had awful results.

Since I am leaving for a day of work, for the moment this is just food for thought: shouldn't have I tried for something like a corner frequency of 1000 Hz or maybe less?

What PB says, You can put the corner anywhere you want, made me think that I moved it in the wrong sense, that is, TOO high a cut off.

Will try when coming back afternoon. (Building a new filter for a new test every time IS a bit discouraging :().

MOre and more I think of this, I feel that I did wrong.

What do you think?


Joined May 11, 2009
What kind of signal do you try to reconstruct. The sample rate required to reconstruct a pure sine wave. Is very different from that is needed to reconstruct a square wave of the same frequecy. Can you post some signal snap shoot
Last edited: