RAB:
its not 3% but rather .3%, but quite frankly it really doesn't matter if it were either to me
the reason being
the number was calculable and predictable, to a pretty narrow margin to start with, and
as i am not really concerned with actual fuel consumption per kwatt/hr produced so long as his results followed other
reports closely, and they did.
what i was looking for was a relationship or rather a ratio of where the fuel consumption can be attributed to.
when you consider that the engine/generator consumes approx 1/8 (.126) gallon just to run at rated speed, no output
and .192 gallon at 1 kwatt output, it becomes apparent that the first kwatt is the most expensive kwatt you can produce with this setup.
and it is clearly not because of low loading on the engine, but rather one has to attribute all of the baseline fuel expense to that first kwatt an not divide it with the second and third kwatt (if it were to run at 3kw output).
what else did we learn?
popular belief is the engine is more efficient at 75-80% load than it is at 30%load,
the data show this is clearly not the case, each kwatt produced comes with an increase of 112% in fuel consumed
but appears to be burning significantly less because of the division of the baseline consumption.
now some of this incease can be attributed to other factors and not to a reduction in efficiency of the engine, things like
rising field current being one probably are responsible for half the increase, but
it is clear to me at the very least the engine does not improve in efficiency as the load increases in a significant way, but this
is probably irrelevent because the observed consumption does go down with increased loading. this gives an appearance if increased efficiency
but must be understood that the reality is somewhat different. things aren't always as they appear.
(note: this came as a bit of a shock to me as well, i would have expected an increase in efficiency as the load was raised, with a peak at around 2.5 kwatt)
getting back the the numbers, accuracy, significant digits and all that
i would assert that his numbers are not at all atypical, but rather typical of these machines, others might get results both over
and under his results, but
the ratio's will change very little, certainly within the margin of error
then there is this
when you are working with relational ratio's, even a +/- 3% in the underlieing numbers is adequate in my opinion.
also if you take ronmar's graph, smooth the curve it appears his plots are damned close, indicating to me that his accuracy is pretty good
if there were problems with his methodology, or test equipment one might expect to see a few points falling well outside the curve.
the fact that one can develop a forumula to predict his findings, would seem to me at least to verify the accuracy of his findings.
i suppose it does come down to how the results are interpreted, and what they are to be used for.
for the purposes i have in mind they are as accurated as needed.
bob g
