DDSing them Crystal Filters

I've been playing today with Crystal Filters. Firstly with one that I have made myself; this is an old design from the BITX and comprises 4 Crystals and 3 capacitors. What you would expect to find is that there is a bandwidth where RF isn't attenuated much around the design frequency and the rest of the RF spectrum shouldn't make it through the filter. There should be a slope either side of the design frequency where the attenuation increases more and more; we're looking for a kind of skirt. When generating single sideband RF signals, we want the sideband to be well within the crystal pass band and the carrier (whatever is left after the balanced mixer) and the opposite sideband need to be significantly attenuated by the filter.

Well, I used the DDS I was making recently and hooked that up to the input of my Crystal Filter then hooked the output into the scope and the spectrum analyser. Quite quickly I was able to plot a graph of the performance of the filter.

Now, the axis of this graph is in hundreds of KHz, so the bandwidth of this filter is extremely narrow, here's what it looked like on the spectrum analyser.

Now, once I'd seen this I decided to repeat the experiment with a commercial filter, this time the axis is incrementing in 2 tenths of a KHz, so the bandwidth, whilst a bit peaky is much, much wider than mine homebrew version.

I conclude that the bandwidth of my filter is way too narrow and I suspect this is due to badly (read not at all) matched crystals. I need to now hook up some method of measuring the resonance of individual crystals, but now that I have the made the DDS function and can use it as a very accurate signal generator this should not pose such a problem. So, tomorrow I will be matching crystals, remaking the filter and looking once more at the bandwidth and hoping for a rather different result.

I new that DDS would come in handy for something.

Good, egh?