Page 1 of 1

Low Pass Filter for MW and LW

Posted: 30 Nov 2016 21:54
by daveB
I have noticed interference from the 3.5 MHz - 4.0 MHz band on MW - in particular CW from the 80 m amateur band and strong AM from the 75 m band - tuning in reverse. I have hunted for a specific project for a low pass filter with a cut-off around 2 MHz - but the nearest I found was this one - which has a cut-off of 400 kHz. As I would be using standard components with 5-10% tolerance I didn't want to get too close to the top end of the MW band.

However scaling to 2 MHz is simply a matter of dividing all values by 5. And here is the resulting filter - in a pre-used small diecast box.

The filter design itself is a 9th order Chebyshev the LC design program is for one with a maximum 0.2 dB ripple and should in theory achieve around -50dB at 3.0 MHz and -80dB above 4 MHz

It's performance can be judged from the following set of screenshots. The receiver is an SDRPlay RSP1.

SDR receivers need good filtering because they will pick up signals at multiples of the sampling frequency due to under-sampling (look it up). That's how the Elad S2 gets Band 2 FM when the ADC only goes up to around 66 MHz.

Sample rate of SDR 8 mbps - MW and SW

Sample rate 8 mbps - FM Band II

Sample rate of SDR 2 mbps

Notice the mess in the 300-400 KHZ segment of the band - some of which is coming from the 75 m band and some almost certainly from the 49 / 41 m bands
These results are from the OIRT dipole fed by a 15m length of 75 ohm coax - so hardly a good match - so some ripple in the pass band is to be expected but the modelling program showed that evena 500 ohm input impedance still gave more than acceptable performance.

The actual values used - 5 uH chokes at each end (2x10uH in parallel) - 10 uH in the middle. The four capacitors were all 2n2 polyester 5%. Although the middle two capacitors should be closer to 2n42 the filter modelling program showed the results to be more than acceptable - a slight ripple above 1.6MHz before plunging off. Component tolerance will have an effect, but by modelling a 2 MHz cut-off I avoided the risk of cutting of the top of the MW band.

The modelling progam can be found here:

It will come up with an ideal set of values - but if you substitute preferred values and click on the "Do Not Press" button it re-calculates it for you.

Now I don't have to worry whether a signal is genuine or an image from elsewhere - and it is cheap to build.