Matching matters: Case study of 110 kHz narrow filters

Editor’s note: This is a guest article written by fmdxing, an WTFDA member from Australia’s east coast who has a DX blog at http://fmdxing.wordpress.com/. Additionally, the editor of this website has modified his Denon TU-1500RD radio with 110 KHz filters in the ‘narrow’ tuning mode in 2006 and saw a marked improvement in reception.

By fmdxing

DXers have modified the Intermediate Frequency (IF) ceramic filters in their radios for years. Scroll down any DXing website (this one included) and it is likely there are notes about radios being modified. Why would DXers take apart their radios and replace small, seemingly unimportant components? One damaging reason is bleedthrough or adjacent channel interference.

Bleedthrough is a radio condition where a strong, local radio signal can be heard on a neighboring frequency (i.e. a local 105.9 MHz signal booming in on 106.1 MHz). Strong bleedthrough on any frequency can make it difficult to pick up other signals — something that can ultimately make DXing impossible.

Turning a stock radio (which can only satisfactorily receive local FM stations) into a DXing ‘powerhouse’ can often be achieved by simply swapping out the wide factory IF filters (i.e. 230 or 180 kHz) for narrower filters (i.e. 150, 110 or 80 kHz). The writer of this article suggests a common solution to the problem: 110 kHz filters.

This writer prefers to use conventional component FM tuners rather than tuners with Digital Signal Processing (DSP) such as the Sony XDR-F1HD and its equivalents. It ultimately comes down to personal preference, but some of the key reasons for a preference towards conventional tuners may include:

Suitability for unattended recordings
Suitability for electronics modification projects
Suitability for external Radio Data System (RDS) decoding
Best potential FM sensitivity
No muting of weak signals
Best potential in-built RDS sensitivity
Best potential strong signal immunity
Rotary tuning dial
Manual IF bandwidth control
Preset station memory preservation
Negligible heat generation
Relatively low cost where purchased locally

For those using conventional tuners, optimizing selectivity – in particular adjacent channel selectivity (that is, 200 kHz above or below an FM broadcast) may unfortunately require considerable time and patience. One of the reasons for this is that the 10.7 MHz filter components required to achieve this selectivity exhibit asymmetrical characteristics. This term means that the two halves are not exactly the same. An example of this asymmetry is illustrated below:

One side effect is that splatter will be slightly worse on one side (illustrated above) because of the limitations of the components. Bruce Carter explains:

It is VERY important to match the filters. If they are NOT matched, your stereo receptions will suffer BADLY! The best way for an amateur without the necessary equipment is trial and error.

Conventional IF filters were first mass-produced by the Murata Manufacturing Company of Kyoto, Japan back in the 1970s. Sauerland & Blum wrote about the development of new IF filters for consumer applications in 1968. Prior to this, filters had typically been the domain of high-end military and commercial equipment.

Paralleling the increase in FM band congestion was a need to replace wide filters for narrow 10.7 MHz varieties more suited to enthusiasts’ long distance reception applications.

Although a technological marvel, it is the personal view of this writer that digital IF filter algorithms are NOT without peer. Consider the DSP algorithm included in the Sony XDR-F1HD, which is based on a Philips integrated circuit designed for car radios. The chief advantage is that filter symmetry is perfect. But disadvantages may exist. Processing artifacts, an artificial pseudo stereo sound and even distortion are often suggested amongst potential side effects of the FM broadcast audio reproduction.

Regardless of its potential limitations, these DSP systems remain immensely popular amongst enthusiasts and undoubtedly deserve considerable acclaim. For newcomers, it must be stressed that no conventional IF filter (the subject of this article) is capable of exhibiting a perfectly symmetrical response. Detailed discussion of DSP is beyond the scope of an article about the best use of conventional narrow filters.

Case study

Indicative of the asymmetry of conventional 10.7 MHz IF filters, the writer purchased five Murata 110 kHz filters with a date code of X from the United Kingdom in August. These differed from another five Murata 110 kHz filters purchased from the United States in June with a date code of Y. Rather than X and Y, date codes are in fact a symbol. Each filter has a symbolic printed date code visible on the lower right side of the component.

These two batches of filters with different date codes exhibited strikingly different characteristics. For this reason, ultimately these components were discarded. For the purpose of this article, these ‘outliers’ will be classified simply as bad filters.

Used in series, the first batch of 110 kHz filters exhibits a bias toward the low side. Accordingly, splatter is worse at the high side. High side means 200 kHz above a local broadcast. Because of the bias, selectivity on the other adjacent (that is, the low side) was exceptionally good. The second batch exhibits a bias toward the high side! Again, with four filters in series the selectivity is noticeably imperfect, with splatter worse at the low side.

Because neither batch was useful because of the excessive degree of asymmetry, a third batch was ordered in October. Fortunately, this batch exhibited selectivity that was roughly uniform on each side. To be clear, this batch of five 110 kHz filters exhibited little discernible difference between high side and low side adjacent channel selectivity. This experimenter’s patience was finally rewarded.

Brian Beezley explains this inherent variation in simple terms:

But the characteristics of individual filters vary. Murata specifies filter center frequency as 10.67 to 10.73 MHz. Outlier filters may make a tuner more susceptible to interference on one side of channel center than the other.

A simple way to test a filter

One crude method of testing 110 kHz IF filters may be through trial-and-error:

Use a tuner with a least four filter positions & sockets for easy swapping of filters
Use headphones — the splatter being monitored may be annoying to others
Ensure narrow & mono modes are selected
Ensure all filters (apart from the last) are known good 150 kHz filters
Put the filter required to test in the last filter position (e.g. CF04)
Tuning to a local station, check the audio on the frequency 200 kHz below
Monitor the peaks of audio (NOT signal) by watching the levels on the Volume Unit (VU) level meter for 10 seconds
Immediately tune 200 kHz above the local — ensuring the same song is playing
Again, watch audio peaks on the VU meter
It will become evident when the audio levels on one adjacent consistently peak higher than the other.

Tips relating to this particular method:

Please read the first section of the Bruce Carter website which summarizes the fundamentals of ceramic filters. It is likely to be particularly invaluable reading for newcomers.
The above steps are merely one enthusiast’s suggestion. This method is NOT meant to be prescriptive.
It is mandatory for accurate testing that the local station chosen to monitor provides the most professional audio processing and transmission facilities in the region. This writer prefers a commercial modern rock station which uses RDS. Talk stations may NOT be suitable.
If possible, position the antenna at a 90 degree null to the local station, otherwise splatter may be excessively high and testing therefore impossible. An in-line attenuator may also be used to reduce local signal levels.
As an alternative to step four, one can use any 80 kHz or 50 kHz filters instead of known good 150 kHz filters in the early stage filter positions. However, distortion may be evident in the audio.
There is an alternative method when undertaking steps six and nine. If access to a receiver or amplifier with a VU meter is NOT possible, simply record the audio to PCM format and compare the peaks of each waveform using software.
Testing should take approximately one minute per filter.
It is suggested that outliers (that is, those filters in which the test reveals an acceptable degree of asymmetry) be placed in a sealed plastic bag which is labeled appropriately.
Once all filters have been tested, it may be useful to double-check good 110 kHz filters. One suggestion is to position them in all of the filter slots using the narrow mode. Proceed to check the intelligibility of a weak station 200 kHz above a strong local, since narrow FM IF filters typically exhibit worse high side selectivity. Gradually swap positions until all filters have been tested once in the last filter position. Ensure performance does NOT significantly differ between filters.

A professional way to test a filter

In a time-consuming hobby such as long distance FM reception, outsourcing or delegating specialist tasks to professionals may be preferable. Doing this ensures maximum possible time can be spent focusing on the ‘openings’. Perhaps the most obvious example of this is when an enthusiast chooses to purchase an FM antenna from a retailer. The alternative is a very rewarding project ‘brewing one’s own’ antenna from an established design using tubing and other components sourced from local hardware stores or old superseded antennas.

Some DX or tuner enthusiasts offer filter matching services using expensive bench equipment (typically designed for commercial electrical engineering firms) such as an RF Network Analyzer to inspect each ceramic filter’s response curve. Such equipment can accurately measure the -3 dB Bandwidth of any IF filter and other variables such as insertion loss and the center frequency.

This may be a valuable service to consider as the degree of accuracy exceeds a simplistic trial and error testing method such as the one described above. Inquire with the appropriate regional mailing list or radio club forum. This writer has used these services in the past (for a reasonable cost built into the price of supply of each filter component) and can recommend this route.

Last resort workaround

If after buying four or five batches of five 110 kHz filters one still encounters the frustration of obtaining significantly out-of-specification filters, these outliers may still be salvaged from a potential ‘death by garbage bin’!

By using a combination of two off-center batches significant asymmetrical anomalies (peculiar to each batch) may be minimized. Simply use one batch in filter position one and the other batch in the second position (and so on in an alternating fashion). In this way, the imperfections of each batch may be potentially offset or compensated for. As simple as this trick seems, the author can confirm it certainly works. It must be stressed that this workaround is suggested as a last resort. Superior performance is still obtained by sourcing a batch of good conventional filters.

Cascading filters improves performance

An improvement in asymmetric responses can be achieved by cascading filters. The result is superior selectivity. The block diagram below shows the IF section of the writer’s Sony ST-S505 ES tuner to illustrate this concept. An extra filter (CF231) is added in narrow mode to improve the selectivity of the three filters (CF201-3) used in wide mode. In total, four filters are used in narrow mode.

Photographs of the interior of the Sony ST-S505 ES tuner are included throughout this article.

It may be beneficial to add additional filter slots into the narrow IF chain inside the tuner. Perhaps one’s tuner has only four filters in the narrowest mode? Adding more may mask anomalies in filter symmetry. Build Brian Beezley’s Do It Yourself (DIY) IF amplifier circuit. Commercially, Bill Ammons offers a Filter Adder PCB.

These experimental modifications are an inexpensive method of maximizing the capabilities of one’s existing tuner without buying a more expensive tuner. Furthermore, these projects provide ‘hands-on’ learning opportunities and the self-satisfaction that accompanies ‘home brewed’ electronics projects. Top-rated tuners such as the Yamaha T-85 or Onkyo T-9090II both incorporate five filters into the narrowest IF mode according to the Tuner Information Centre.

Purchasing good 110 kHz filters

In order to find good filters, it is suggested enthusiasts considering filter modifications purchase the required number of components from at least two different sources. For example, if four filters are needed to replace wider stock filters, buy a minimum of eight in total. Do NOT buy everything from the same supplier. In this writer’s example, it took three batches to get acceptable quality. Further, when in receipt of outliers, please don’t blame retail suppliers. Quality-control of components is exclusively the manufacturers’ responsibility.

In this writer’s experience, it is true that the search for good 110 kHz filters will result in a surplus of filters. Consider however, how inexpensive these components cost! This writer likens narrow filters to a bottle top on a bottled beer. Although bottle tops cost virtually nothing, the fundamental importance of the role (that is, sealing the beer properly in order to prevent spoilage) means the bottle top does NOT represent a trivial component!

Noticeable variation between batches of 110 kHz filters (the principal reason for writing this article) has never been observed with 80 kHz filters. However, it is dangerous to generalize because as filters become narrower in bandwidth, the degree of inherent asymmetry becomes increasingly less obvious.

A little extra effort to better match narrow filters for your long-term needs is likely worth it. If a DIY route is chosen, the entire process may take a few hours. Remember matching is a one-off process. Thereafter, one can be much more confident that the best selectivity performance (taking account the imperfections of conventional filter technology) can be realized.

Additional contributions by David.