Crystal Ladder Filter: Build a Simple SSB or CW Filter
The heart of a good superheterodyne radio is its narrow filter, which separates the station you want from its neighbours. In expensive equipment it is a factory-made block, but a home builder can make a very good one from a handful of ordinary crystals. This is the crystal ladder filter, a design known for decades that gives a bandwidth of a few hundred hertz for CW or about 2.4 kHz for single sideband, from parts that cost very little. This guide explains how it is built, how the bandwidth is set, why the crystals must be matched, how to terminate it and how to measure the result with a simple analyser.
Why a Crystal Is a Good Filter Element
A quartz crystal behaves like a resonant circuit with an extremely high Q, often tens of thousands, which is far beyond what a coil and capacitor can achieve. The equivalent circuit has a series resonance at a frequency fs, where the impedance is very low, and a parallel resonance a little above it, where the impedance is very high. The two are only a few kilohertz apart at the frequencies of interest. The narrow gap between them is what makes the crystal ideal for building filters with a bandwidth of a few kilohertz or less. The general background is covered in the articles on the crystal filter and on the crystal oscillator.
A single crystal gives a response that is too narrow and uneven for speech. By connecting several crystals in a ladder, with small capacitors between them, the individual resonances spread into a flat passband with steep sides. The more crystals, called poles, the steeper the sides and the better the rejection of neighbouring signals.
The Ladder Structure
In the usual arrangement the crystals are connected in series along the signal path, and between each pair of crystals a capacitor goes to ground, as a shunt. The input and the output are terminated with a resistance, typically a few hundred ohms. The crystals all have almost the same frequency, and the shunt capacitors determine how far apart the individual resonances are spread. A larger capacitance pulls the poles closer, giving a narrower passband, and a smaller one spreads them wider. The bandwidth is therefore set mostly by the value of the shunt capacitors, and the number of crystals decides the steepness.
The design procedure was described by Milton Dishal, and many programs use his method to calculate the capacitor values for a given bandwidth, number of poles and crystal parameters. Popular programs for amateurs take the measured values of your crystals and give the capacitors and the termination. You can also begin with a known, published design and adjust it. The table lists typical choices.
| Purpose | Poles | Typical bandwidth | Remarks |
|---|---|---|---|
| CW filter | 4 to 6 | 300 to 600 Hz | Larger shunt capacitors, higher insertion loss |
| SSB filter | 5 to 8 | About 2.2 to 2.6 kHz | Flat passband, good skirts |
| AM filter | 4 to 6 | 5 to 6 kHz | Smaller capacitors, low loss |
| Wide data filter | 4 | 3 to 4 kHz | Often used in receivers with a roofing role |
Choosing the Crystals
The crystals must be of the same type and as close as possible in frequency. A very popular choice is the 9 MHz crystal used in older SSB equipment, for which the filter is sometimes called a 9 MHz filter. Other builders use the 4 MHz crystals for microprocessors, the 10 MHz and the 20 MHz types, or the crystals for the colour burst frequency of old televisions, because all of them are cheap and plentiful. The frequency of the filter will be that of the crystals (our guide to AGC covers what follows the filter in the receiver chain), and the radio is designed with an intermediate frequency in the same area.
Buy more crystals than you need, at least twice as many, and measure every one. The sample will differ a little in frequency, and the filter needs the frequencies to be close together, usually within some tens of hertz for an SSB filter and less for a CW filter. A simple crystal tester is an oscillator with a counter, in which the crystal sets the frequency. Put each crystal in the same circuit, with the same capacitor, and write down the frequency. Then choose the group of crystals that are closest to each other. Wide spread in the group leads to a ripple in the passband or to an unsymmetrical response.
Matching the Terminations
The impedance at the input and the output of the filter must be correct, or the passband will show ripple and the insertion loss will rise. For a given design the value is calculated from the bandwidth and the crystal parameters, and it commonly lies between 100 and 1000 ohms. Most of the circuits around the filter are low impedance, for example the 50 ohm of a mixer, so a matching network is needed. A transformer, a small toroid with the proper turns ratio, does this job well. An L network with two components is an alternative, as shown in our article on L-network matching.
Pay attention to the signal level as well. A crystal filter is a passive component with no gain, and the insertion loss, typically 2 to 5 dB, must be made up by an amplifier. The placement of the filter in the receiver is also important. Put it where it can protect the later stages from strong signals, as in the roofing filter described in our guide to receive filter bandwidth, and give it a mixer with good strong-signal performance in front of it.
Building It
The layout is simple but must be done with care. Use a small piece of copper-clad board or a manhattan style construction, keep the leads short, and arrange the crystals in a line from input to output. The most important rule is to prevent signal leakage around the filter. If the input and the output are coupled by stray capacitance or by a common ground path, the signal bypasses the filter and the rejection far from the passband is poor. Place metal partitions or at least separate the input and the output by a ground strip. The ultimate rejection of the filter, often 80 to 100 dB with good construction, is limited by this kind of stray coupling and not by the crystals.
- Measure and sort the crystals, and choose the best matched ones for the filter.
- Calculate the capacitors and the terminations with a design program or use a published design.
- Solder the crystals in series with the shunt capacitors, short leads, in a straight line.
- Add the input and output transformers or networks, and the shielding.
- Test and adjust the capacitors to reach the required bandwidth and flatness.
Use capacitors of good stability, such as NP0 or C0G types, because the filter should not move with temperature. Trimmer capacitors help to fine tune the shape, but fixed parts selected by measurement keep the response more stable over time.
Measuring the Filter
To see what you have built, you need a signal source and a detector that follow the response. The best tool today is a vector network analyser of the NanoVNA type, which sweeps across the passband and shows the transmission as a curve. Set the span to about three times the expected bandwidth, calibrate with the termination in place and look at the S21 curve. You will see the width at 3 dB, the ripple in the passband and the skirts. If you have no analyser, a signal generator and a voltmeter or a spectrum analyser can be used point by point, which is slower but gives the same information. A small spectrum analyser with a tracking generator, such as the types mentioned in our overview of test equipment essentials, will also do the job.
Look for three things. The passband should be flat to within one or two decibels, the sides should be steep and symmetrical and the response far from the band should be well below the passband. If there is a ripple, the terminations are wrong or the crystals are not well matched. If the skirt is soft on one side, one or two crystals are off in frequency. If the far rejection is poor, look for leakage around the filter.
Improving a Poor Filter
A systematic approach finds the cause. Begin by checking the terminations: change the resistor or the transformer ratio a little and see whether the ripple falls. Then replace the crystal that seems to cause the problem with another from the group, and measure again. If the passband is too narrow or too wide, change all the shunt capacitors in the same proportion, since they act together. Finally, improve the shielding if the far response is not good. Keep a note of each change and its effect, because the filter is an excellent way to learn how the parts influence each other.
A Worked Example: a Six-Pole SSB Filter at 9 MHz
As an illustration, consider the usual first project, a six-pole filter for single sideband at 9 MHz. You buy twenty crystals of the 9 MHz type, measure them in a test oscillator and sort them by frequency. From the sorted list you choose the six that lie closest together, perhaps within 20 to 30 hertz of each other. A design program gives shunt capacitors in the range of a few hundred picofarads for a bandwidth near 2.4 kHz, and a terminating impedance of a few hundred ohms. You build the filter on a small board, with a transformer at each end to match 50 ohms, and connect it to the analyser.
The first sweep rarely shows a perfect curve. A little ripple in the passband and a slightly tilted shape are normal. You adjust the terminations until the ripple is small, and then you check the width. If it is too narrow, you reduce the shunt capacitors a little; if it is too wide, you increase them. After a few rounds the passband is flat and the shape factor, the ratio of the width at 60 dB to the width at 6 dB, is about 2 or better, which is good for such a simple filter. You then place the filter in the receiver and listen: the voices on the band sound clean, and the neighbouring station that used to leak through is gone.
The experience teaches why professional filters are made with such care. The choice of the crystals, the spread of their frequencies and the quality of the termination are all visible in the sweep. A builder who completes this one project understands the behaviour of every IF filter in a commercial radio, and is better prepared to read the specifications of a transceiver.
Where to Use It
A ladder filter is suited to the first or second IF of a superheterodyne receiver and to the single-sideband generation stage of a transmitter, where it removes the unwanted sideband. For CW, a filter with a narrower bandwidth can be switched in. In a transceiver the same filter may serve for both receiving and transmitting, switched with diodes. The filter is also used in the simpler direct conversion designs as a part of the sideband selection. Whichever application you choose, a home-built ladder filter gives a lasting pleasure: it is a useful component that you have made, understood and measured yourself.
The Bottom Line
A crystal ladder filter uses several matched crystals in series with shunt capacitors to make a narrow, steep-sided IF filter. The capacitor values set the bandwidth and the number of crystals sets the steepness. Measure and group the crystals by frequency, terminate the filter with the correct impedance, keep the construction free of stray coupling and check the result with a network analyser. With 5 to 8 poles you get an SSB filter of about 2.4 kHz, and with 4 to 6 poles a CW filter of a few hundred hertz, at a price far below that of a ready-made unit.
Frequently Asked Questions
What is a crystal ladder filter?
It is a filter made of several quartz crystals connected in series, with capacitors from each junction to ground. It gives a narrow bandwidth with steep sides at a low cost.
How is the bandwidth of the filter set?
Mainly by the shunt capacitors. Larger values give a narrower passband, smaller values a wider one, and the number of crystals decides the steepness of the sides.
Why do the crystals have to be matched?
Because the filter needs the crystals to have almost the same frequency. If they differ too much, the passband has ripple and the response becomes uneven.
How many crystals do I need?
Four to six for CW and five to eight for SSB are common. Buy at least twice as many as you need so that you can select the best matched group.
How do I measure a crystal filter?
Use a network analyser such as a NanoVNA to sweep the response, or a signal generator with a detector. Check the flatness, the width at 3 dB and the far rejection.
Why is the rejection far from the band poor in my filter?
Usually because of signal leakage around the filter through stray capacitance or a common ground path. Shield the input from the output and keep the layout in a straight line.