News flash 7/13: Thanks to our contributor Ed for finding this great article on The History of Ceramic Filters. We'll digest it and see how it supplements Bob's old writeup below. Here's Bob:
Old (pre-1970s) tuners used inductors and caps for IF filters, called LC filters. Typically, the caps were fixed, and the inductors variable via slugs. LC filters needed complex adjustment in manufacturing, and required special test equipment to perform the adjustment. In general, LC filters were used in all tube-type FM tuners, which were sold until about 1967. Starting about 1968, solid state devices replaced tubes in tuners, but these early solid state models still used those same tuneable LC filters. In the early '70s, tuner manufacturers began using a new IF filter technology, the ceramic filter, with Murata being the most popular supplier. This technology was as different as CDs are from vinyl records. The ceramic type piezoelectric filter was small and inexpensive and needed no adjustment, so it was a huge technological breakthrough.
It should be noted that some high-end tuners from the '70s still used the more costly LC IF filters, which were later packaged in sealed block-like enclosures (i.e., no longer had tunable inductors). The reason was most likely that flat group delay, minimum phase filters were not yet available in the ceramic package. These minimum phase filters have only moderate selectivity, but the gaussian type response gives the lowest possible distortion of the IF signal's phase. This means less audio distortion after the signal is processed in the detector (whose output is phase sensitive). In general, a high-end tuner that has 2 or more selectable IF filter paths uses a minimum phase type IF filter in the wide path for best audio performance. The narrow IF filter path uses multiple narrow ceramic filters for best selectivity/reception. Later, minimum phase type filters became available from Murata in the ceramic package and were used, for instance, in the wide filter of the Kenwood KT-815. These are called "GDT type" in Murata's filter data sheets.
The early ceramic filters were encased in metal enclosures, probably for shielding purposes. Later, the enclosures were eliminated, and 4-pin filters were popular in mid-'70s models like Kenwood's KT-8300 and 600T. The last evolution of the filter package was the 3-pin, 2-stage ceramic filter, which began to be used in late-'70s models like Kenwood's KT-7500 and KT-815. The 3-pin filter package, still being used today, allows for easy swapping in of narrower filters (i.e., 150 or 110 kHz bandwidth) for better selectivity, at the minor expense of increased distortion.
See the DIY Mods page for information on adjusting single tuned LC filters.
Murata ceramic filters used to be available from Mouser with no minimum order. The part number for low-loss 150 kHz filters is SFELA10M7JAA0-B0 and for low-loss 230 kHz filters is SFELA10M7GAA0-B0, but both are listed as unavailable. Other Muratas are listed here.
Muratas can also be obtained, even more inexpensively, from Digi-Key, which used to carry only Toko filters. The low-loss 150s are only $0.59 apiece, $4.91 for 10 or $21.16 for 50, and even cheaper for larger quantities. That's the only low-loss Murata that Digi-Key carries, but we haven't checked whether they carry other types.
But our contributor Brian Beezley had a caveat for anyone who buys filters and installs them without measuring and matching them: "Last year I ordered 100 150s from Digi-Key. I was expecting parts in a more or less normal (bell-shaped) distribution around 10.700 MHz. I was really disappointed. Most of the parts were at 10.720 to 10.730, with others below 10.800. I found only one near 10.700, as I recall. The batch was essentially useless to me since I needed to match wide filters pretty well centered at 10.700 MHz. I'm sure the batch had been cherry-picked before I got my hands on it, probably even before Digi-Key did. Ten 230s I got at the same time were much better, all useful."
Other Murata filters are available from other distributors. One possible contact is:
Internet Sales Support
1-800-477-6668 ext 2-4352 (tel)
As of the last time we checked, they were willing to sell quantities of less than 500, but only certain part numbers were in stock in "broken lots." They were:
SFE10.7MJA10-A (150 kHz)
SFE10.7MS3-A (180 kHz)
SFE10.7MS2-A (230 kHz)
SFE10.7MA5-A (280 kHz)
The wider ones have lower insertion loss. Murata's 110 kHz filters are not generally available except in lots of 500, but it doesn't hurt to ask. Please note that Murata changed all their part numbers in 2001, but there is a cross reference on the site from the old numbers above to the new ones.
Ceramic filters made by Toko are available on a per-piece basis from Digi-Key. Toko filters have historically had a reputation, in DXing circles, for being not as high-quality as Murata filters, but it is not clear whether this is justified.
280 kHz filters were used in many tuners, for instance five 280 kHz filters are the stock setup in the Kenwood KT-7500. Replacement with new hand-selected ones can often yield improvements in filter and tuner performance. This is due to the fact that modern production filters appear to have less insertion loss and more consistent quality those those used in mass production 20-25 years ago. We have seen KT-7500s, for example, in which it does not appear that the supplied filters were matched at all. Using stock width filters might appeal to collectors who desire the best performance possible from a tuner, while keeping the unit's filter setup at factory specs.
The actual replacement of filters is quite easy. Mark, then remove the old ones, so you can undo the mod entirely if needed. Then install single row machine pin gold-plated sockets to hold the new filters. If you don't have a signal generator, scope, and counter to sort them, you can sort them in-situ by plugging them into the sockets and watching the signal strength meter, and/or listening. Most tuners have a bank of 4-5 filters, and they all need to match in passband frequency or you will get less than optimum performance. You are also concerned with insertion loss, the less the better. In a perfect situation, the tuner would be realigned afterwards, especially the discriminator section, for lowest distortion. But many people do the mods without the realignment and report excellent results. See Bruce Carter's page or Better FM for more detailed step-by-step instructions on doing a filter mod. Bruce has also created a new page that may become the definitive place to go for filter info. Mike Hawk's site also has a good page on filters, including links to old and new Murata catalogs and a filter identification key.
What you put in YOUR tuner depends on your goals. The best filters for your needs depends on your reception area. If you are looking for maximum sensitivity, choose low insertion loss units. For best sound, install Flat Group Delay filters, known as GDT types in Murata literature. For best selectivity, use matched 150s with perhaps a matching 110 or two.
In general, later tuners had wide and narrow filter selections, with the wide using one filter and the narrow 3-4 or more. In many cases, the tuner designer put the wide in series with the narrow, for maximum selectivity. Filters are available in 110, 150, 180, 220, 230, and 280 kHz bandwidths. In general, 220-280 was used for wide IF settings, and 180 was used for narrow. Many "modders" leave the wide setting alone, and install 150's or even one or two 110's in the narrow filter positions for excellent selectivity. See the Murata site for data on all the different types of filters available.
The only way we know to get those Murata filters that are not sold in lots of less than 500 is to "sample" them, a process well known to design engineers.
I manufacture a small PCB about 0.3" wide by 1.1" long that I use to upgrade the IF stages of tuners. It holds two ceramic filters and an IF amplifier. You use it by removing a stock ceramic filter and installing the board in its place. A pigtail is connected to +11 to +15 VDC on the PCB. It draws about 15 mA. When you use two of these in tuners like the Pioneer TX-6800, you have an upgraded unit with a total of 4 filters versus two stock. Also the IF gain is brought up so that you have a more solid limiting curve for better stereo signal-to-noise ratio with lower signal levels. To get a very selective unit, you use more narrow filters in the second IF PCB (i.e., 180 or 150 kHz ceramic filters). A modified TX-6800 has better stereo S/N at low levels than the Carver TX-11a.
The Filter Adder PCB will not change the absolute sensitivity of a tuner or receiver. That is a function of its RF stage, noise figure, gain, and intercept points. What the PCB will do in units that have little or no net IF gain is bring up the noise limiting curve of weak stations. The whole idea of the IF strip is to band-limit the signal to +/- 100 kHz or so and present to a demod stage (ratio detector, quad detector, pulse counter). For the demod to work correctly it should see a bandpass limited, fully limited (amplitude) signal.
A Low-Gain IF Chain: A fairly simple tuner like the Pioneer TX-8100 or Kenwood KT-7300 has little net IF gain. Both of these tuners (both having 4-gang MOSFET front ends) have much better than average RF front ends and mixers. In stock form, they do not stand out as great performers. Both tuners have one IF gain stage and 3 ceramic filters. What the IF Filter Adder PCB does is replace a single ceramic filter with a board that holds two ceramic filters and has an IF amplifier. This increases the number of ceramic filters by either one or two depending on the style of PCB you use, which allows you to modify a KT-7300 to have a total of 5 ceramic filters in the IF, and add about 16 dB (8 dB average per PCB) more IF gain.
Assume that the RF front end assembly (RF stage and mixer) has a gain of around 30 dB. A 1.5 microvolt signal input signal would become a 45 microvolt signal at the 10.7 MHz output. With the loss of 3 ceramic filters at 6 dB each and a single transistor IF gain stage of 18 dB, the input to the quad demod chip would be about 45 microvolts. Most of the quad chips would like to see an input level of 250 microvolts to 1 millivolt for full quieting. Add 16 dB of gain (6.3 X) to the IF strip, and the 45 µV now increases to 283 µV. This will help very weak stations to have greatly improved quieting in both mono and stereo. In cases where there are much stronger signals, the IF stage and input to the quad chip become limiters, just what you want to limit AM noise riding on the signal.
This is very important in tuners like the KT-615 and KT-815, which use pulse counters as detectors. Pulse counters are excellent when you have a well-limited signal going into them. They are frequency-insensitive as compared to a ratio detector or quad demod and are quite linear (hence very low distortion, i.e., 0.008% THD). The performance drops rapidly with a noisy or poorly limited IF signal much more quickly than a narrow-band detector system.
A Higher-Gain IF Chain: The board will not change the sensitivity in a tuner with good IF gain and limiting. If you want additional filters in a wide or narrow IF chain to modify selectivity, this is an excellent application for the board. A great place to use the board is the wide IF position of tuners like the Pioneer TX-8500II and Kenwood KT-7500. Both of these tuners have a single ceramic filter in the wide mode. By adding a board, you now have two wide filters, which increases the number of stations you can pull in with the wide IF mode. A pair of low-GDT 280 kHz parts will still allow 0.05% THD and 55 dB of stereo separation at 1 kHz.
Using the KT-7500 as an example, let's look at the narrow IF filter chain. The KT-7500 has a 4 ceramic filter narrow IF path with two gain stages. If you want to narrow the IF strip for DXing or just listening to your favorite station 200 kHz away from the local blowtorch, IF modification is one of the answers. By adding filters to the narrow IF chain, you can use slightly wider filters (180 kHz bandwidth vs. 150 kHz, or 150s vs. 110s) to get much improved selectivity, while still getting a good listenable stereo signal. My formula for the serious listener who needs additional selectivity in the KT-7500 is to use two 280 kHz low-GDT filters in wide, followed by a 230-180-180-150-150 kHz chain in narrow. This will yield less than 1% stereo THD and produce at least 30 dB of stereo separation at 1 kHz.
How is it designed? The amplifier section of the PCB is a two-FET design. I chose FETs for a number of reasons over a bipolar part or MMIC. First, I can get more stable gain at 10.7 MHz using a FET vs. needing to find a bipolar device with an FT into the 700 MHz region. The FET has a great load line and is easy to match into the ceramic filter with just a 330-ohm resistor. The FET design also acts as more of a current source since the FET's are biased and running about 85% of Idss. Hence, they are gain stable over a wider range of voltages.
MMICs are, in many ways, a poor way to go. They are optimized for 50- or 75-ohm inputs and outputs. To use a 50-ohm MMIC you would need to place 280 ohms of source resistance between the input ceramic filter and the MMIC to properly load the ceramic filter. On the output side, a similar load resistor would be used to match the 330-ohm filter. That wastes a lot of gain, and calls for the uses of a higher gain MMIC which invites instability in the IF stage (re-radiation, RF ground loops, etc.).
Please note that I also make a board to fit the older style 4-pin ceramic filter's footprint. This board uses two of the 3-pin newer style filter, and works just like the 3-pin variety. Also available is a 3-filter version of the board, with one input filter, the amplifier, and two filters in series on the output. Net gain is about unity. This board is designed for narrow IF strips that have just one filter position, or in units like the Kenwood KT-815 where you want to expand the narrow IF (from 2 filters in narrow, stock, to 4 filters modified).