PO Box 582
Pine, CO 80470
Some time ago I received a Kiwa Filter Module for review from the Kiwa Electronics Company. Kiwa produces these filters as an accessory item for receivers using a 455 kc IF, and they are available from stock with 6 dB bandwidths ranging from 3.5 to 8 kc in 500 cycle increments. Their literature points out that Kiwa acn supply any custom bandwidth from 2.5 to about 8 kc. This would allow an operator to tailor IF response to ones own needs.
The filter I received is a small, black epoxy-potted rectangle measuring 1.875 x .875 x .680 inches. It is specified as having a bandwidth of 3.2 kHz at -6 dB, came with clearly identified red and black power leads and input/output miniature coax cables.
There is another model available which uses solder pins for the same leads, and is designed for mounting on an optional filter switch board. The filter switch board allows users to select from several bandwidths.
Also shipped with the review package were two self-adhesive Velcro mounting strips and some small tie wraps for securing the filter and its leads once placed inside a receiver.
For experimenters, the beauty of these active filters is that Kiwa has already done the hard work for us! Electrically, the modules consist of cascaded ceramic filters with input and output buffer amplifiers. The buffer amps result in very light loading of the IF channel, and produce lots of output drive. Kiwa selects the ceramic filters in production to provide the required bandwidth and shape factor.
The stated specifications from Kiwa regarding these filters
1. -6/-60 dB shape factor of 1.65 to 1, which is excellent performance.
2. Zero insertion loss.
3. Negligible passband ripple.
4. No impedance matching required.
5. DC power requirement is 4.5 to 36 VDC at 10 ma.
To check the filters performance, I swept them using a Tektronix 7L5 spectrum analyzer/tracking generator in a 7633 storage mainframe. Figs 1 and 2 are photos of the resulting response curves.
As seen from Fig 1, the tracking generator output level is )dBm, and zero reference level (the top line of the graticule) is between -16 and -15 dBm, center frequency is 455 kc. The vertical displacement is 10 dB per division at 3 kc resolution, and horizantal dispersion is l kc per large division. The overall passband shape is very nice, quite an improvement over the passband of single ceramic filters (see ER#57 for a review of Toko ceramic filters). The filter response from the curve is shown as 3.2 kc at -6 dB and 6.5 kHz at -60 dB. This gives a shape facytor of 1:20. This is an 18% difference in shape factor from what was quoted, but I dont feel the difference is significant. The response curve would probably improve if I didnt drive the filter at 0 dBm, but I wanted to simulate the voltage present in a typical tube IF amplifier.
There is a spurious response on the high side at 458.600 kc which is 50 dB down from the peak response and is 1.4 kc wide. This is a response in the filter, not the analyzer, as 3 dB of generator attenuation produced a 3 dB reduction in the spurious level. The spurious response is nothing to worry about, as it is 50 dB down and will be attenuated by additional selectivity in the following IF stages. Some high priced crystal filters have similar spurs.
Figure 2 is a plot of the filters peak response, only this time using 2 dB per vertical division at 300 cycle resolution and 500 cycles per horizontal division. As can be seen, the peak response is relatively smooth, increasing by 1 dB for a 1 kc increase in frequency near the design center. This meets the spec of negigible ripple in my opinion. In an actual installation, the top of the response curve could probably be made almost flat by careful adjustment of the IF transformer trimmers, at a sacrifice of some overall gain.
Using the spectrum analyzer, I measured an insertion loss of .12 dB, which in reality is zero loss. Probably the loss I measured resulted from slight impedance mismatches at the analyzer ports.
The filters were terminated in my testing as specified by Kiwa, 10K in and 100 ohms out. Changing the terminations +/- 100% made no difference with shape factor. I agree that the filters are termination insensitive.
The input voltage was varied between 4.5 and 36 VDC. The current stayed at 10.45 ma, and insertion loss did not change by any significant amount.
My conclusions are that the Kiwa Filter Modules are well made, meet specifications, and are a good value. I would recommend them to anyone wishing to upgrade the selectivity in a receiver with a 455 kc IF, and I would use them in my receivers.
I temporarily installed this filter in my 1939 Hallicrafters S-20R, which I have owned since 1965 and am rather familiar with. On SSB phone it always was fairly good if the band wasnt crowded with kilowatt splatter. The S-20R has differential temperature compensation of the L.O., and if the bottom plate is left off and allowed to warm up for 30 minutes it is as stable as any Collins beast. Tuning across the 80 meter phone band showed just how nice the filter performs, which is just as expected from the measurements. Now I wish I had Kiwas 6 kc version of the filter!
Printed by permision of Electric