Although the KSB2 is basically a good design, there are some areas that give more grief than they should.
a) Lack of microphone audio gain. This manifests itself as a poor peak RF signal as compared with the CW mode. In addition the VOX performance may be adversely effected. Some microphones give sufficient output to drive the Speech Compressor IC U3, however many do not. The solution is to add a gain stage between the microphone and U3. The gain should be between 3dB and 15dB dependent on the microphone output, and can be either an internal or external pre-amplifier. In my K2 I have added a 9dB pre-amplifier internally on a small PCB glued to a microphone jumper header (ex computer drive cable). The circuit is shown below.
b)VOX. Well, what can I say, it needs a major redesign ! Use sufficient microphone audio, set SSBC=4, SSBA=3, and goto my Anti-Vox pages.
c) Frequency Response. The low frequency cut off of the microphone audio stages is way too low. Computer modelling showed the frequency response went down to 20Hz. This LF energy carries no useful information, and can overload the Speech Compressor IC causing undesirable artefacts on the transmitted signal. Modelling a desired LF cut-off of 200-300Hz gave the following capacitor values, note how much smaller they are than the original values. The SSB signal is now very much 'brighter'.
Frequency response curves, before and after modification.
d) BFO Optimisation. It is well worth spending some time optimising the position of the BFO frequency relative to the SSB filter. The have been posts on the Reflector, and a number of application notes, detailing how using a noise generator and Spectrum Analysis software to get the frequency response correct. If the majority of received SSB signals sound poor, then it is very likely that your SSB transmission will sound just as bad.