While I don't know all the in's and out's of RC electronics, I do know how to setup and calibrate this type of canard system. Once you do a good bench calibration to determine where up is, you do not need to. or want to, change the calibration in the field because you do not have a good level reference of up in the field in the field.
Rationale. You want your rocket to go vertical regardless of the rod/rail angle, and weather cocking of the rocket after rod/rail guidance ends. Unless you level your rod/rail precisely, and make sure the mounted rocket is also mounted vertically, you can not readjust the level in the field and assume the rocket will lift-off vertically. The evidence for this is Allyssa's videos where the launch rod/rail is angled. Once the rocket leaves the launcher guidance it turns vertical, and it could not have done this if the calibrations was altered on the pad.
The Guardian itself cannot be bench-calibrated to know where "UP" is, in the sense that once you adjust the system on a bench, then turn it OFF, that it will remember any of that information.
Every time it is turned on, it is learning fresh for the first time. It knows nothing of the last time it was on.
So, every time it is turned on, the airplane needs to be level or the rocket pointed vertical, so it can learn again (for the first time) where "up" is.
But that is not hard to do on the field. Use a level to set the launcher angle vertical.
And even eyeballing it is OK if a level is not available/forgotten, because if you are 2 degrees off the worst that should happen (if the canards are otherwise centered), is that the rocket may fly 4 degrees off of vertical. Now why did I say 4 degrees? Because if it is calibrated 2 degrees off vertical (88 degree launcher angle), it will HOLD vertical anyway at first (88 degree launch angle + 2 degrees), but if it rolls 180 degrees then that 2 degree degree error causes a net 4 degree error (88 degree launch angle - 2 degrees = 86). But for the kind of rockets we fly, 4 degrees off vertical would still be fantastic. And if it kept rolling, then the net flight path would be about 88 degrees.
Now, what you meant by bench calibration probably is to have the rocket vertical on the bench, Guardian initialized to learn vertical, then adjust the servos or adjust clevises to make the control surfaces perfectly straight. And yes, that is something best done on the bench. But the system seems to have a little bit of variation to it, not exactly dead-on from power up to power up. It's nothing dangerous, just that if a person wants it to try be absolutely straight, then it may be a good idea to be able to adjust it at the pad too.
A test always worth doing, as I learned from my sunguidance, is to center it, then rotate it 180 degrees, and see if it is still centered. If not..... adjust for the middle range. That is, if a canard was zero when centered, then when rolled 180 degrees it was 1/16" off, then adjust it so it is half that, 1/32" off which then should end up as 1/32" off the other way when you roll it 180. Now that might sound scary, but as long as there are minor errors then it'll fly fine, just not dead 90.0 degrees but a little bit off from that. And of course Sunguidance had that kind of problem because of the nature of the light sensors (Photoresistors), they did not perform exactly identically and an adjustment for proper centering that was good for one day would not be good for another day due to differences in sky conditions (a hazy or milky sky causing a difference in lighting effect with the sensors than a crisp clear blue sky, for example. And Sunguidance also works in overcast skies but the sensors usually had a lot of imbalance between them that required recalibration). Of course the Guardian is not light based, but still the principle for calibration then roll it 180 and check again is a good one.
Actually, another bench test procedure to do would be to turn it on and off a few times, vertical every time, and see what variations there may be in the calibrations, if the canards are zero every time or if they are off a bit. If they are off a bit, then tilt the rocket some till the canard for that axis is zero, and note the angle of the tilt. Probably no more than 5 degrees wil lbe neede,d and probably a lot less than that. Then keep the tilt angle the same but roll it 180 and see the effect. And work towards a happy average that works OK regardless. Even if you may have to accept that the rocket won't fly 90.0 degrees but may fly 86 degrees, or whatever.
I know that may be worrisome, but it really is not.
Ad the best way to understand how it works... is to carefully build a small test model ,and fly it. By carefully I do not mean sweat over perfection, just don't build a sloppy crappy model, and DO understand how the system works (as opposed to someone just throwing a Guardian and a servo-canard system into a big rocket and flying it.... but nobody in this thread would be that careless). My old Sunguidance rockets were BT-60, about 3 feet long or so. using two staged D12 power. Sometimes more power. A Guardian would not fit cross-wise as needed, but it will fit into a BT-70 so I could retrofit my old BT-60 sunguidance model with a short BT-70 payload section to hold the Guardian. Or just simply build an all-BT-70 rocket. Alyssa's was Estes BT-80. In any case a lot of practical experience can be gained that way, and some first-time problem solving that is certainly best learned on a small "easy to build" size than something much bigger for supersonic flight. That is, the first time you build a prototype, you are running across various problems you need to solve, and might have to re-do, or end up making do with a compromise because you are in too deep to start over. That if you built it a second time it would be a lot better. That is what I mean.
A prototype guidance section for a small model can be built in a few hours, one night, once you figure out the basic layout structure, and gather the necessary parts. And you do not even need a special rocket for testing, just an existing one that will be stable enough that you can put enough power into for test flying.
- George Gassaway