CW Notch Filter

Homebrew CW Notch Filter Project

Why did we need an Audio Filter?

Our radio club owns four HF RIGs some new and some “gifts” that we use for public demonstrations, HAM day local schools and our yearly public special event. The RIG we wanted to use for a dedicated CW station had a 3 KHz band-pass filter and the cost for a 300 Hz CW filter was more than the radio was worth. We decided to see if an external audio filter could be added at a reasonable cost.

Audio Filters

Audio filters came in two categories DSP/FFT and analog. DSP filters are part of many HAM radio software suites used for CW and digital modes. Some of the software is free but adding a PC was too big a cost factor. Hardware DSP filters made with A/Ds and FPGAs are small but very pricey even as used equipment. Analog filters are made from a series of cascaded low and high pass filters that produce a net band-pass. Most are RC with amplifiers as the audio frequency would require very large inductors. The traditional approach of a ten stage RC network with OP amps was low in cost but the alignment was easier said than done and is likely the reason everyone doesn’t do it. One exception to the analog solution was low cost and easy to align is the National Semiconductor Switch Capacitor Filter, LMF100.

Switched Capacitor Filter

The LMF100 is a CMOS IC with precise and repeatable values of RC networks and OP amps controlled by external resistors for filter shaping and a clever external frequency selection. A better description comes from the National Semiconductor data sheet:

 “The LMF100 consists of two independent general purpose high performance switched capacitor filters. With an external clock and 2 to 4 resistors, various second-order and first-order filtering functions can be realized by each filter block. Each block has 3 outputs. One output can be configured to perform either an allpass, highpass, or notch function. The other two outputs perform bandpass and lowpass functions. The center frequency of each filter stage is tuned by using an external clock or a combination of a clock and resistor ratio. Up to a 4th-order biquadratic function can be realized with a single LMF100. Higher order filters are implemented by simply cascading additional packages, and all the classical filters (such as Butterworth, Bessel, Elliptic, and Chebyshev) can be realized. The LMF100 is fabricated on National Semiconductor’s high performance analog silicon gate CMOS process, LMCMOS™. This allows for the production of a very low offset, high frequency filter building block.

They are correct with the LMF100 and six resistors you can get the same result every time. The information provided on the LMF100 data sheet and application note 779 gave me the example circuits for cascaded Butterworth band-pass filters with a guess of 10:1 resistor ratio was a starting point. The clock and audio amplifier are directly from the LM555 and LM386 National Semiconductor data sheets.


The prototype build was constructed from hand selected components from my junk box and leftovers from other projects. The workmanship was not my Sunday best, but is good enough for testing. The LMF100 was found on eBay and I paid a little extra to get faster deliver

Filter Circuit

Cascaded Butterworth band-pass filters with a 10:1 resistor ratio on the LMF100. The input is a 1uF capacitor with a 30K resistor, right side has the networks 3K with two 30Ks (10:1). The voltage divider on the left is two 10K resistors. The output load is a 10K potentiometer.

Clock Circuit

The clock is a LM555 in astable operation, 70 KHz clock controlled by two external resistors (2K & 8.8K +/- 1K) and a 0.001 uF capacitor. The center frequency is the clock divided by 100… Example: 70 KHz Clock = 700 Hz.

Amplifier Circuit

LM386 amplifier with single power supply operating with the default internal gain of 20 configuration plus seven parts to provide the high impedance (10K) input that will drive a headset or a very small speaker with a ¼ Watt output.

Power Supply Circuit

Just a 9 V alkaline battery filtered at each IC to prevent feedback
supplies the 25 mA and is good for 30 to 40 Hours of filter operation.


The measured the band pass is shown below and tried it on the air with success so I now have a design that works and a set of curves as a reference point. After the club members get some air time with this filter over the next few months we will improve the design and packaging. Currently performance is good.

The filter band pass plot above the drop in audio.
Sound like a 300 Hz band pass or less based on a subjective listen.


The first time you plug in the filter between your radio and headset the following steps should get you working. Please note I only used a ¼ Watt amplifier so it will only power a headset for now.

1. Plug your headset into the filter and plug the fitter into your radio.

2. Select WIDE

3. Find a single CW station and tune your radio frequency for 700 Hz audio.

4. Now adjust your radio volume to comfortable level, but keep it to the low end. Making your radio volume too high will cause compression and the filter will not be as effective.

5. Select NARROW

6. Adjust the GAIN until you here the CW.

7. Again tune your radio frequency so the 700 Hz audio peak is reached. You will hear the audio increase rapidly in the last 50 Hz (650-750 Hz) and a sharpness or “tinny” sound in the last 20 Hz (680-720 Hz). I measured the peak at 700 Hz and I found could always hit in the 690-710 Hz range by ear.

8. Once you are at peak frequency adjust the GAIN to comfortable audio level, but keep it to the low end. Making the filter GAIN too high will cause compression and the filter will not be as effective.

9. Select WIDE and go find several stations operating a 100 Hz apart… If you are like me they never are around when you want them.

10. I am sure once you have done this once you will never need these instructions again.


Prototype project box and perfboard packaging.

My long term goal is to make this filter as a DIY project for TVARC provided the consensus is the filter performance is acceptable. Many questions need to be resolve before going forward.

Is the band pass too wide or too narrow? I would like to test 25:1 (2K & 51K) and 50:1 (1K & 51K) configurations. Should the frequency be adjustable or set it once during assembly? A range of 300 to 900 Hz set by an internal trimmer is the answer. Is the packaging right as a project box or should we use an Altoids tin. As the circuit is tiny it could be imbedded in a powered speaker set used on desktop PCs.

The future could be smaller or the big sound.