Hi,

I designed a 6 order Butterworth LPF. Cut-off frequency is 10MHZ. My testing showed that:
At 20MHZ, the rejection is about -20dB
At 100MHz the rejection is -85dB
At 370MHz the rejection is -63dB.

So the rejection gets smaller over frequency range to 1GHz to -5dB.

My question is "is this a good filter for 10MHz cut-off frequency?"

Thank you

 

Hi,

I designed a 6 order Butterworth LPF. Cut-off frequency is 10MHZ. My testing showed that:
At 20MHZ, the rejection is about -20dB
At 100MHz the rejection is -85dB
At 370MHz the rejection is -63dB.

So the rejection gets smaller over frequency range to 1GHz to -5dB.

My question is "is this a good filter for 10MHz cut-off frequency?"

Thank you

Its not bad, but the slope looks like it starts higher than 10Mhz, but definitely usable depending on the application.

 

...and not that good if you are expecting it to attenuate frequencies above 400 MHz.

 

A filter is good if it meets your expectations which you define in your design specs.
There are many reasons for the feedthru reduction in attenuation after 100 MHz in active filters, so if you wanted to you can specify something better, you would use a passive filter design after your active filter at a controlled impedance. Yet this raises the source resistance loaded by 50 Ohms in your VNA so the loss will not be 0dB in the passband.

So in future a perfect filter design or any design is simply one that always* meets your required system specs which includes cost and availability.

1. Load = TBD [ohms min]
2. Passband (PB) BW = 100KHz (-3dB)
3. PB loss/gain = TBD
4. Stopband attenuation = TBD dB @ TBD f range

The always* depends on your environmental stresses, such as; Vcc, Temperature range, component tolerances.

There may be more specs like group delay, PB ripple, maximally flat equal-ripple, minimum group delay, raised cosine for zero data jitter , etc
Each type has names based on the math or inventor of the algorithm and are suited for specific applications.

In the old days it was hard to define specs when "pushing the envelope" so because of time or cost constraints we might have say a perfect design is 98% complete, but I prefer generally to choose realistic specs that will perform as required and then meet them. Then it's perfect if it always works. With experience you will learn how to stress your designs with environmental tests & tolerances (Monte Carlo calc) and then choose design margins for high yields or stability.

 

There is no guarantee that a particular filter response will always be monotonically decreasing. The change in phase with frequency can also create an increasing response at high frequencies, especially at very low output levels approaching the noise floor.