If my spelling is even more erratic than usual, forgive me; I have enjoyed a champagne cocktail. (Bubbly and St-Germain, very nice, but even a little dab does for me).
A well-deserved champagne cocktail: the project I have labored on for lo, these many weeks hit the last big step this morning -- and worked. It's even working nicely.
Never mind the details; at last writing, the [device] produced adequate RF power but signal fidelity, well, the signal-to-noise really should be better than 28 deciBels and it was 19; the Error Vector Magnitude has almost got to be under 3 percent and it was 10 to 11. Despite all that, receivers were decoding it; it would have worked, ugly, awkward and inelegant but hey -- working.
But remember my old pal, the Automagical Feedback Compensator Of Stuff? I'd had to lie to it; it's supposed to get a sample of the very output and instead, I was making it look at a much earlier stage of things, where there was at least enough signal for it to work but not really enough hint of the Bad Things Happening that it could get its tiny, digital teeth into and fix.
Fast-forward to today: the tower crew did their connecting thing way up in the air,and I did mine down on the ground (and found I'd cut that very last piece of rigid transmission line 7/8" too long, necessitating hasty mods). In that piece, I had also installed a signal tap, adjusted and set by dint of much math, incantations, and the use of charts, graphs and interpolation. It should have produced a fairly blistering +15 dBm*.
Powered the thing up, about 50%, and everything worked, no horrible suprises; inched the power up to a full kiloWatt (American, none of your off-size Canadian Imperial kiloWatts here) and it still worked -- ugly and inelegant. Checked the level at the tap, +14.62 dBm, which is fancy figurin'. Checked the level, padded it, and hooked it up to the Compensator, flipped the switch and watched the analyzer.....
...And within 30 seconds (International, ain't nobody gets giddy with units of time no more, not after the Indians misplaced a whole week and had to get a U.N. bailout to get back in sync), darned if the thing wasn't pushing 26 dB signal-to-noise and 3 percent EVM -- and getting better even as I watched. (Settled at 31+ and 2.7, which for widgetry of that vintage is pretty darned good).
It worked. Weeks of improvisation, scrounging and hardware-hacks paid off: the thing actually worked and worked well.
I'd claim I knew it all along, but I gotta tell ya, last week or so, I had my doubts it would ever look that good.
There are still a few niggling loose ends, a lot of cleanup and some control tweaks to do, but the hard part (so far) is done.
You can see why I phoned ahead and asked Tam to break out the bubbly!
* Go ask a signals geek: for a monitoring sample, that's insanely hot. Alas, the thing needs to see +5 dBm at its input, and that's after a splitter; plus, up where I work (only a bit below the police calls, if your local LEOs are majorly up-to-the-minute), even cable loss starts to add up. By the time it reached where it needed to get, it only took a 3 dB attenuator to get the level right; which is also pretty close figuring.
T. R. MCELROY'S STREAMLINED TELEGRAPH KEYS
6 months ago