Some temporary notes on using the Q series steplock JavaScript calculators:
===========================================================================
by Don Lancaster 3 December 2001 version2.1
copyright c 2001 by Don Lancaster and Synergetics.
All commercial and media rights ~fully~ reserved.
Please view http://www.tinaja.com/glib/msinprop.pdf
mailto:don@tinaja.com http://www.tinaja.com/magsn01.asp for further support
(0) Always start with the Steplock-28Q calculator since it often will give you
the "best" results.
(1) Note that the fundamental amplitude is a normalized PEAK value that
is porportional to the applied dc supply voltage. The harmonic values are
all RELATIVE to the fundamental amplitude. Note further that ONLY the THD is
expressed as a percentage.
(2) Calculators are intended for IE use and will cause errors on certain NetScape
browsers. The PostScript calculator generating sourcecode can be recompiled
for NetScape use, but the color "critical harmonic" autopositioning feature
will be lost.
(3) The second most common cause of a JavaScript error is not waiting till the
load is complete before clicking any button or box.
(4) "Mode Jumping" may take place on certain high amplitudes and longer sequences.
Always run a reasonableness check on your pulse results. All pulses must
be greater than zero but less than ninety degrees, must start before they
end, must have "daylight" between themselves and their neighbor, must be
sequential, and must not overlap. Total amplitude must be between 0 and 1.002.
All delta triad harmonics must be less than the 10^-15 calculation noise floor.
To avoid mode jumping, "sneak up" on large amplitude values by starting with
smaller values and working up one amplitude step at a time. Use only the FINE
ADJUST rather than the COARSE ADJUST button when doing so.
(5) It takes much longer to do all eight decimal points at once, than to do
one or two at a time. The reason is that if the seventh has to clean up
something the sixth missed on its first pass, then it will take ten times
longer. The primary use of the higher decimal points is to prove that a
solution is in fact optimal.
(6) If a very high or very low amplitude is slow converging, try temporarily
asking for a slightly higher or lower amplitude to help it along.
(7) Extremely rarely, a solution may "oscillate" between two amplitude values.
The workaround is to temporarily ask for a very slightly higher or very
slightly lower amplitude. Later, you return to your desired target
amplitude.
(8) Calculation time goes up as the FOURTH (!) power of the number of pulses
in use, so longer sequences will take much longer to determine. The
Delta Friendly sequences normally calculate much faster than others.
(9) Certain small results may kick into scientific notation mode. You can
scroll right to verify any suspiciously large answers.
(10) The quantizing buttons tell you how much the magic sinewave will degrade
when you limit the number of bits in your stored table lookup amplitude.
Always fine adjust again when finished viewing a quantize button's results.
Repeated coarser quantizing may give errors.
(11) Quantizing is conservatively based upon 20 degrees being "full scale".
This may be reduced on longer sequences for lower true distortions.
(12) Prototype code has been significantly improved over code now on the website
and awaits your funding for a rollout. Speed is increased by eliminating
extra reporting, by replacing evals with direct code, by improving
the Delta amplitude adjustment and the initial guess routines, and, very
significantly, by eliminating unneeded continuous Delta calculations. The
scientific notation flipout has also been eliminated.
(13) Further significant speedup can be accomplished by linking to compiled
Java or C++ code. This awaits funding.
(14) There is no known theoretical limit to the number of pulses that are
calculatable, but at present, zeroing out more than 200 harmonics will
be slow and become unbearable beyond 400 harmonics. By applying very
fancy tricks and parallel processing, 1000 zeroed harmonics can be
approached. This also awaits funding.
==================================================
by Don Lancaster 23 December 2001 version2.2
===================================================
(1) Released a few of the "R" series interim calculators. These are significantly
faster on longer sequences (though not yet optimal). They also include an optional
timer for speed studies and eliminate the "floating point" display nuisance.
(2) Other calculators in the series are easily and quickly generated using the sourcecode
available to partners and associates.
(3) A very significant additional speedup can be made by using partial differential
techniques that will allow solutions to 2000 pulses per quadrant and beyond. This
awaits your funding.