so, the process of generating SSB is to first generate an AM signal and filter out one sideband and the carrier. this is way easier said than done
so we can generate an AM signal by just feeding our audio input and an RF oscillator into a mixer. the resulting double sideband signal plus carrier can be sent to the antenna and we have AM centered at the RF oscillator frequency, albeit at pretty low power if we're just directly feeding the mixer output to the antenna
problem is, to cut off the carrier and unwanted sideband, we need a razor sharp filter. all real filters have some "rolloff", they don't instantly transition from passband to stopband
we're trying to minimize this transiton consider that the usual legal requirement for SSB is to suppress the unwanted sideband to 40dB below the wanted sideband. way our filter passband starts at an audio frequency of 300Hz. this means that in the space of 300Hz, not only do we need to get rid of the carrier but we also need to reduce the other sideband by a factor of 10 000 (40dB less) -40dB in 300Hz is possible, but it takes some care first thing; we can get rid of the carrier if we are careful about our mixer design. we can use a balanced mixer to cancel out the carrier, making our filtering job much easier. balanced mixers can be built easily but need some fine adjustment to null out the carrier. this still leaves us with DSB-suppressed carrier still on board?
so I want to finish it
to cut out the opposite sideband, we can make a filter. conventional LC networks aren't really going to cut it here; hypothetically you could but it's really hard to get a high enough quality factor (Q) out of them
another part we can use that constitutes a high Q filter is a quartz crystal we can make a very sharp (high Q) filter out of quartz crystals, cascaded one after the other highqfilter.png we call these "crystal ladder filters" adrian-bandpassfilters.png more crystals -> sharper passband-stopband transition (ie higher Q) -> better unwanted sideband suppression
so, we can take our balanced mixer and input an audio frequency and a radio frequency, getting DSB-suppressed carrier. then, we can feed the output of that to a crystal ladder filter that cuts off the unwanted sideband. we've got SSB, so we're done right?
actually we're not done; the crystal ladder filter only works at one exact frequency
if the frequency of our crystal filter falls within a ham band, we can build something like VK3YE's Knobless Wonder: a fixed frequency direct conversion SSB transceiver
https://www.youtube.com/watch?v=UxDD2Qe1VX4
but if we want to be frequency agile, we need to do more
to cut to the chase, what we do is to pick a crystal filter frequency outside a ham band (typically 9ish MHz, this frequency is our IF or Intermediate Frequency because it's inbetween the audio and final RF frequencies) and then mix it with a variable frequency oscillator, resulting in an SSB signal that can be changed in frequency as easily as turning the VFO
a mixer generates two sidebands, but in this case the sidebands are several MHz apart (the spacing is determined by the VFO frequency fed to the mixer) and thus the unwanted one can be easily filtered out by a sloppy, easily built filter
as an example, we have a SSB signal generated at a 9.0MHz IF. we want to transmit on the 20m band so we use our VFO to generate 5.2MHz. the VFO mixed with the IF produces two images, the sum frequency at VFO + IF = 5.2 + 9.0 = 14.2MHz and the difference frequency at VFO - IF = 9.0 - 5.2 = 3.8MHz (in the 80m band, you'll notice; with a well chosen IF frequency we can cover multiple bands with the same VFO)
the difference in frequencies between the images is in this case VFO * 2 = 10.4MHz and this is way, way easier to filter out the unwanted image from
