On Sunday, August 4, 2013 8:00:28 AM UTC-7, Dan Marotta wrote:
I agree with you on the rate of change aspect and, in that sense, I can see

where you could develop an algorithm that would command changes of turn

radius to maximize time in the strongest lift.



But I fly purely for fun, these days, except for when I'm towing, and I'd

happily wager a beer (or two!) with you that I can out climb you with

variometers and computers powered off. Heck, I'll even cover up the

mechanical vario. No fancy seals required as proof of compliance and I

don't even care if I lose the bet - I'll still get a beer!



"jfitch" wrote in message



On Friday, August 2, 2013 6:55:31 AM UTC-7, Dan Marotta wrote:

An algorithm works for a single case and each thermal is different.







"jfitch" wrote in message











a lot of snipping Finding the lift is art. Centering it is science, and



science can be reduced to an algorithm. Computers, correctly programmed,

are



very good at algorithms - without head down time or otherwise adding to



pilot workload.







But I will try the reworked Naviter version!



I disagree - using only the information from your variometer, the algorithm

is pretty consistent or almost any thermal you will encounter.



The explanations here seem incomplete. You should tighten or loosen your

turn (angle of bank) based on the *rate of change of lift*, not the

strength. If you loosen your turn at the peak of lift, you will be 90

degrees out of phase and will not center quickly, if at all. What you want

to do is loosen your turn at the peak rate of increase of lift.

Alternatively, loosen your turn at 90 degrees prior to the peak lift. When

you experience peak lift, you are already at 90 degrees to the desired

correction direction (mod vario lag). Human perception is not all that great

and estimating rate of change, and recording that rate of change around a

complete circle, and relating that to your angular position in that circle

accurately, all advanced by the time constant (lag) of the variometer. This

is however very easy for the computer.

Dan, you're on! Bring your glider out west (and bring lots of beer - you are going to need it).

Waremark,
From observation I believe all thermal assistants do the same thing: identify circling by rate of heading change, record variometer readings around the circle, integrate the readings over angular displacement to calculate a correction vector. There is no need to account for non-round circles or correction movements (other than to the extent that it affects angular displacement), as this information does not affect what you should do.

After all, this is exactly that a good pilot does. If you are high in a blue thermal that is all you can do: your only input is rate of climb and its derivatives. You integrate lift as a function of angular displacement, and reposition the circle the direction of the resultant vector. The computer simply does it in fine grained detail, without fatigue or distraction, and without the perception errors to which human senses are so susceptible. This is particularly true in rough/uneven/strong thermals. A pilot must determine and remember, "was that huge hole I just went through exactly equaled by the huge bump I just got? where they 80 or 110 degrees apart? how does that change the correction vector exactly?" Most humans aren't very good at this, and it requires a substantial amount of attention unless the thermals are well behaved. The less well behaved the thermal, the worse are your perception errors and the more the thermal assistant helps.

What separates the good, bad, and ugly thermal assistants is not the mechanics of recording vario readings or math to calculate the vector, all very straightforward I think. The devil is in the details:

-What algorithm is used to call up and dismiss the thermal assistant? Is it there when you want it and gone when you don't, without having to access menus and poke buttons?

-Is the data presented in a way that can be interpreted at a glance, and is associated data also available (such as average rate of climb, climb in thermal, better/worse than McReady expectation, better/worse than days average)?

-Is non associated but nevertheless important data also available (such as nearby Flarm targets) and easily interpreted with minimum attention?

-Does the algorithm account for data latency (such as varo lag) and reaction time in a useful way?

In my opinion Winpilot is better in each of the details than the others even in its earliest instance (but I haven't tried the latest SYM offering).