Now that the Heathkit HW-7 is working close to the original specification with all the 45 year old carbon resistors changed for 2019 metal film resistors I can now look at some modifications to bring it up to modern standards.
The first thing I want to look at is the receiver filter. The HW-7 has a simple LC low pass filter immediately after the 40673 mixer so lets see how it performs.
To do this I fed a wideband noise source into the antenna input suitably attenuated so that it didn’t overload the receiver. I then fed the audio from the headphone socket into an external sound card connected to my Windows laptop. I then opened the Spectrum Lab audio analysis program to look at the filter bandwidth and response.
The existing filter is fairly flat from 200 to 1200 Hz at the -6 dB points with the peak at 1000 Hz. It’s not a bad filter for such a simple design as the stopband from 3500 Hz is 50 dB down.
It’s probably a bit wide by modern standards and later I want to modify the transceiver so that it has a correct 750 Hz offset between transmit and receive so can the filter be modified to sharpen it up?
I modeled the existing filter in the Elsie filter program and came up with some capacitor value changes that reduced the loss between the mixer and audio amplifier by 5 dB and sharpened the filter. The peak of the filter is now at 750 Hz and the -6 dB points are at 400 and 1000 Hz. The stopband figure remains the same at 50 dB down.
The changes look very subtle on the display as the 5 dB reduction in loss tends to hide the changes but it does sound a lot better listening to CW on air.
Only three capacitors were changed, C5 is now 0.47 uF from 0.22 uF, C9 is now 0.47 uF from 0.1 uF and the coupling capacitor C52 is now 0.22 uF from 0.05 uF.
I did try experimenting by changing the inductor bypass capacitor C7 from 0.01 uF to 0.1 uF. This gave a really sharp cutoff above 1000 Hz but the stopband figure reduced from 50 dB to only 30 dB so while it would sound great working weak signals at sunspot minimum it would fail listening to a band full of strong contest signals.
With simple filters like this it’s tricky to get the right balance but you can get close with modern computer modeling and measurement tools.