Rob's web

High-performance RTTY filters 2

Home - Techniek - Electronica - Radiotechniek - Radio amateur bladen - QST - High-performance RTTY filters 2

Improved designs for better reception

The "Mainline 1170" Filter (Fig. 6)

Fig 6a
Fig. 6. The Mainline 1170 bandpass input filter for 170-cycle shift. The bandwidth is about 275 cycles and the output is high impedance. Capacitances marked (*) are approximate values. Each section is tuned to the same frequency, 2200 cycles, for standard tones of 2125 and 2295.

Fig 6b

This is a narrow-shift bandpass input filter of 275 cycles bandwidth, intended for tones of 2125 and 2295 cycles. It can be arranged in conjunction with the original 850-shift bandpass input filter so that either may be switched in or out automatically as the 850- or 170-shift channel filters are selected. This could easily be done with a 6-pole multi-position switch such as the CRL PA-2022 in use in the author's unit.

For comparison, Fig. 7 shows the output curve for the original bandpass input filter for 850 shift that was included in the TT/L schematic (page 30, August 1965 QST.)

Fig 7
Fig. 7. Output curve for the Mainline TT/L bandpass input filter for 850 shift. (Circuit in August, 1965, QST.)

The mainline "2" and "3" series filters

The filters in these groups are 3-pole Butterworth designs with good skirt selectivity. They use three 88-mH toroids each.

Most commercial filters have similar input and output impedances - usually 600 ohms. The design of 600-ohm filters requires inductors of far different value than the 88-mH toroid, and to use the toroid it is necessary to allow the impedances to be much higher. This is no problem; in fact, it assists in using these particular filters in the TT/L. It also explains why the 850-cycle bandpass input filter and the Mainline 1170 bandpass input filter for narrow shift use the type of input network they do. It also accounts for the fact that the broad filters have a lower input impedance than the more narrow filters.

Every bandpass filter is potentially an impedance transformer. No matter how the filter is developed, its input and output impedances can be made different from one another. Therefore, in the following designs the output impedance is very high while the input impedance is only medium higH The filter then acts as a step-up transformer, and instead of having substantial voltage loss, as would most filters, will actually show a voltage rise. In introducing the filters into the TT/L no additional matching components or amplifier stages are needed.

Since these three-pole Butterworth filters are designed to terminate in a very high impedance, the 270K grid-return resistors shown on Pins 2 and 7 of V1 in the TT/L should either be removed entirely or else changed to 10 megohms. This change will have no effect even when the original filters are switched back in, as those 270K resistors were only included originally to protect the tubes in case no input section was plugged into the main unit. This is not likely to occur.

The "Mainline 2850" filters (Fig. 8)

Fig 8
Fig. 8. Circuit for the Mainline 2850 and 3170/3850 filters. Values are given in the table below. The capacitors should be 10 percent-tolerance Mylar or better; CAB and CBC should be 5 per cent tolerance if possible. Values for CA, CB and CC are approximate; see text for tuning procedures. Where an unusual capacitance value is shown smaller values may be connected in parallel; e.g., 0.05 + 0.008 = 0.058. Capacitances are in µF except where otherwise specified. R1 5% tolerance.

2850297520013 k.0024.0015.032.03.033
3170/385021258020 k.0025.0018.061.06.062
317022958022 k.0021.0013.053.051.053
385029758039 k910 pF620 pF.031.031.032
Oddball29058036 k0.001680 pF.033.032.033

These filters are nearly 200 cycles wide and are intended for use on general 850-shift copy either with or without the limiter. Because they have less "capture area" than the 8850 filter, for instance, they will have better noise-cancellation qualities at the input to the decision threshold computer stage. Their use will be particularly beneficial on limiterless copy, where they will allow some tolerance in incorrect shift or minor drift. Their output curve is shown in Fig. 9.

Fig 9
Fig. 9. Output curve of the Mainline 2850 filter system. Note that shifts of less than about 700 cycles would not be received, Each filter has approximately 200 cycles bandwidth.

The "Mainline 3850" and "3170" Filters (Fig. 8)

These filters, like the bandpass input and the 2850 filters, are of a basic 3-pole Butterworth design.(5) They are only 80 cycles wide at the -3 dB points. The response of the 3170 filters for 170 shift is shown in Fig. 10. The table below is representiveof theperformance of the type of filter. The figures given are for the 2125-cycle filter: the others are comparable:

Response in dBBandwidth, c.p.s.

Probably the first 15 db. of any filter establishes the primary effectiveness of the system, in which case the filter is still only 157 cycles wide at the -15 db. point. The 3-30-dB shape factor is about 3 to 1, which is quite good for such a simple filter.

The "Oddball" Filter (Fig. 8)

Typically, the RTTY operator who thinks he is on 850 shift is more often than not transmitting a shift of 750-8(X). This situation is improving as more operators get facilities, such as the Mainline TT/0 Semi-Counter,3 for accurately setting and checking shifts. However, many enthusiasts advocate the use of a heterodyne filter system so that two good filters can be made to tune nearly any shift likely to be received. The author has found that the addition of one "odd-frequency" (2905 cycles) filter to the 80-cycle bandwidth filters for 170 and 850 will enable the operator to copy about 90 per cent of all shifts likely to be encountered.

Fig 10
Fig. 10. Output curve of the Mainline 3170 filter system for 170 shift. Bandwidths are 80 cycles each.

As more operators add the Semi-Counter3 to their station equipment the need for both the 2905 filter and heterodyne filter system will diminish.

Installing the "2" and "3" Series Filters

Fig. 11 shows the easiest way to install the 2850 and 3850 (or 3170) filters in the TT/L. The design was chosen so that these filters could replace the original TV-coil system, thus offering the operator the choice of using or not using the limiter to improve the copy. This method gives similar output voltages without resorting to additional amplification or optional input units, such as the a.m. input section.

Fig 11
Fig. 11. Installing the 2850, 3850 or 3170 filters in the TT/I. F.S.K. Demodulator. S,, is the limiter bypass switch, allowing the limiter to be used for f.m. copy, or to be switched out for limiterless a.m. copy. The series resistors are not used for the 2850 filters, and are 150K for the 3850 or 3170 filters. A three-pole multiposition switch at points marked "X" allows quick switching from these filters to others described in the article.

Balancing the various filters

Assuming the operator will plan to use the basic broad-band 5850 or 8850 filters, he should install that system first and then carefully set the balance control for equal mark and space voltages at the input to the DTC stage. An ordinary d.c. voltmeter may be used as this is a low-impedance point and does not require a vacuum-tube voltmeter. You will get a negative voltage for mark and a positive voltage for space. It might be handy to put a small feedthrough connector from this point (cathode of V2B of the TT/L) to the top of the chassis as a perma.........

Now switch in the newfilters and adjust the control on each until once again the voltages are equal but opposite from mark to space. In the case of the 2850 or 3170/3850 types, the two controls may be adjusted to equal the voltages obtained from the more simple broadband filters. This will allow the operator to switch between the various filter systems, know each has been independently balanced correctly, and at the same time require no changes of the indicator-sensitivity controls.

Making the Filters

The filters can be built in almost any form that suits the individual operator. The author assembled them on small Vector boards and then mounted the boards in plug-in containers made from small Miniboxes, each fitted with an octal plug. Several others have built the filters into regular Vector plug-in containers. A few have built them on printed-circuit boards and mounted them beneath the chassis on angle brackets.

Tuning the filters

Few amateurs have a test bench that would have a digital audio counter such as is available to the author, but other schemes can be used. First let us list quickly the various items that will be needed:

  1. A source of audio sine waves - an audio oscillator, a tape recorder, or the receiver itself may be used.
  2. A means for determining the frequency to be used for reference. There are several possibilities:
    1. Tuning forks,
    2. Accurate audio oscillator,
    3. Musical instrument such as a piano,
    4. Tape recorder with prerecorded audio tape,
    5. Capacitor decade substitution box(3), or
    6. Digital audio counter.
  3. A means for measuring maximum output of the filter while it is being tuned, such as:
    1. Oscilloscope,
    2. Regular v.t.v.m. with a.c. scale,
    3. A.c. v.t.v.m.

The biggest problem, of course, is determining just what 2125, 2295 (for 170 shift) and 2975 "really" are. As this question was discussed in detail in the Semi-Counter article(3) it will not be gone into here, as described there, the standard is the 88 mH toroid, which has been determined to be actually 88 mH within very close tolerances. The setup for tuning a toroid to a particular frequency is shown in Fig. 3 on page 37, May QST (an a.c. v.t.v.m. may be substituted for the scope), and a step-by-step tuning procedure is given.

Tuning the toroids will actually only take a few minutes each, once the equipment has been set up. When all toroids needed are tuned, they can be wired into the filter circuit along with the associated resistors and other components.

For the simple filters such as the 8850, 7850 or 8170, tune the toroids without the series resistors, and then after the toroid-capacitor combinations are finalized, install the series resistors and other components. If the toroids are tuned with the series loading resistors, the filters will be much more difficult to tune accurately because their bandwidths will be much greater.

How to tune the multi-toroid filters

Tuning the three-toroid filters, such as the Mainline 1170, 2850 or 3170/3850, is really quite simple. It involves a few more steps, but is no more difficult than tuning a simple filter.

The basic filter schematic is shown in Fig. 12. In reality, this consists of three parallel-tuned sections, as given in Fig. 13. The capacitors on top, CAB and CBC, are the "cross-coupling" capacitors which establish the primary filter characteristics. The parallel capacitors, CA, CB, and Cc, are used to tune each filter section to a specific frequency.

Fig 12
Fig. 12. Basic layout of the three-pole Butterworth filter used for the Mainline "1, "2" and "3" series.

First, set the audio source to the desired tone frequency by the tuning-fork, tape-recorder or decade-box(3) method. Then tune each section in Fig. 13 independently to that same audio tone. In Fig. 13A, do not change or remove CAB but instead add to the value of CA until the meter peaks, showing that the desired frequency has been reached. (Turns may also be removed from the coil, if more appropriate.) When this section ..........nect it in the second section as in fig. 13B. Tune this section to the identical audio frequency by varying CB or the turns on the inductor. Set aside, removing capacitor CBC, which is then used to tune the third section as shown in Fig. 13C. When this section has been tuned to the same audio frequency put the filter together as in Fig. 12, making sure no (more) turns are removed from any of the toroids in the process.

This completes the filter tuning.

Other methods of tuning the filters can be used - indeed, many different schemes are possible. It is the author's intent to attempt to convince the reader that elaborate and expensive equipment is not required, although in some instances it might make the job a bit easier. It would appear that an audio tape recorder in conjunction with a capacitor decade box and vacuum-tube voltmeter would be an ideal method available to most individuals. Just the decade box with a v.t.v.m. and a reasonably stable audio oscillator probably would work just as well.


Although intended primarily for the TT/L demodulator, some of these filters are now in use in other converters, such as the "W2JAV," the "W2PAT," the "Twin Cities" and the K6IBE (now W4DDE) TU-D. They could be adapted to any converter you may be using, to enhance its operation, although in most older units the limiter cannot be switched out as there is no threshold corrector such as the DTC circuit in the TT/L.

These filters will allow those using the TT/L to achieve the maximum performance that is inherent in it when operated with really good narrow filters in the limiterless mode. Although the TT/L with the original filters has given those using it improved copy over demodulators used previously, installing some of these filters will change it into a truly high-performance unit - one in a completely different class from anything the operator probably has used before. Frankly, it is fantastic how the limiterless unit will pull signals through unbelievable interference. It makes the difference between good copy and a hopeless muddle. Good filters and limiter-less operation can "do that" for weak signals.

Those who have never operated narrow shift with a first-rate demodulator have a most pleasant surprise in store. However, words alone will never convince you - give it a try and become a believer!

Fig 13
Fig. 13. The three-pole Butterworth filter shown in Fig. 1 2 breaks down into the equivalent of three separately-tuned circuits as shown above. Tune each of the three to the same frequency (see text), then build into the final circuit of Fig. 12.


  1. Other designs such as the Thomson linear-phase filters offer some theoretical advantages for pulse reception, such as RTTY signals, but to get comparable passband characteristics at least two extra sections (for a total of five toroids in each filter) would be needed. The author believes that the small advantage of the linear-phase types would be more than offset by the added difficulty of making them at home, plus the extra cost and size. If commercial filters are obtained, the four- or five-pole linear-phase types might well be considered at the higher cost.

Part 1 - Part 2

Irvin M. Hoff, K8DKC.