Measuring the -3dB point

[seanm]

When discussing the corner or breakoff frequency of a filter it always seems to be defined at the -3dB point. They then go on to show a nice straight line with a nice clean dropoff.

Now I have the following plot. Sorry for the size, any attempts at reducing made it less readable *and* larger in size. The x scale is frequency, the y scale is in dB.



Let's focus on the blue line (the big hump for the colour blind). If the -3dB point is taken from the peak (+7db) the frequency is roughly 170Hz. If it is taken from the flat (-3dB)  it is roughly 100Hz. This is a significant difference.

I will probably be quoting the numbers here and want to get it right 

[George Giblet]

Forthe  high-pass type responses you gave the "reference level" is the pass band region that approaches 0dB at high frequencies.

To find the -3dB you find the lowest frequency point that hits -3dB ie. the 80-100Hz zone.

Suppose your orange line dipped below -3dB, you will get two points which drop below -3dB.   In this case the meaning of -3dB is a little dubious but generally you would choose the lowest one.  The reason is that although the wobble exceeds -3dB it is still loosely considered as passband.  The steep edge is considered the transition band and usually you want the -3dB point on the edge of the transition band (since  -3dB indicates the end of the passband/start of the transition band).

[seanm]

Thanks. Using the reference makes sense and is what I have been using.

If anybody is interested, this plot is a simulation of two stage gyrators. The yellow line is with a "real" coil to use as reference. The blue line is with opamp based gyrators and the red line is with transistor based gyrators.