Hafele and Keating experiment
Benjamine
in 1972 of Hafele and Keating experiment about relativity. (Sciences n°
177 - p.166 to 170).
I've found a lot of references to this paper, but not the article. Could
you help me to find it by a link, or send it to me ?
A lot of thanks
--
Benjamine, France
Kevin Aylward
This might be interesting http://www.mywebpages.com/asps/H&Kpaper.htm
Kevin Aylward
sa...@anasoft.co.uk
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.
Benjamine
oxUN9.366$BW1....@newsfep1-gui.server.ntli.net...
> This might be interesting http://www.mywebpages.com/asps/H&Kpaper.htm
Thanks
I've found this link too before, and gave it to my father.
He seems amazed because there is a lot of articles about this
experiment, and we don't find anywhere the original paper. He thinks
that he can't work on it with only comments of this article, and for
him, the original is necessary.
(sorry for my bad english)
--
Benjamine, France
Martin Hogbin
"Kevin Aylward" <ke...@anasoft.co.uk> wrote in message news:oxUN9.366$BW1....@newsfep1-gui.server.ntli.net...
> Benjamine wrote:
> > My father asked me to search on the web the original article published
> > in 1972 of Hafele and Keating experiment about relativity.>
> This might be interesting http://www.mywebpages.com/asps/H&Kpaper.htm
Your link to this pop-up infested, commercial site can, at best, be
described as mischievous. What are you trying to achieve?
Martin Hogbin
Kevin Aylward
Its called a balanced view. Look, I have no idea if the site is a *lie*
or not. The overall report seems reasonable without seeing any data to
contradict it. However, if the data presented, i.e the accuracy of the
clocks, is not a lie, than this particular H&K experiment is indeed
useless.
I have been an engineer long enough to know how accurate ones
instruments need to be to be trusted. I don't even need to see the
experiment details, knowledge of the instrument errors is enough,
especially if there are only a few intruments. If the clocks were say,
0.05ns/hr stable rather than what is claimed in the data, e.g.
+/-5ns/hr, then the experiment might be ok. *Otherwise*, there is no
hope in hell of that experiment being valid, and I don't care how one
fiddles the data.
Do you have any official data that says such clocks routinely achieve
*much* better than the +/-5ns/hr claimed?
Martin Hogbin
"Kevin Aylward" <ke...@anasoft.co.uk> wrote in message news:t_ZN9.1045$BW1....@newsfep1-gui.server.ntli.net...
> Martin Hogbin wrote:
> > "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
> > news:oxUN9.366$BW1....@newsfep1-gui.server.ntli.net...
> >
> >
> > Your link to this pop-up infested, commercial site can, at best, be
> > described as mischievous. What are you trying to achieve?
> >
>
> Its called a balanced view.
You mean the 'establishment' view is not to be trusted?
> Look, I have no idea if the site is a *lie* or not.
It is a commercial site with pop-up ads that wants to attract traffic.
The rest of the pages on that site consist of obvious crackpottery.
I could put up a web page claiming that H and K forgot to wind
up the clocks before the experiment started. No doubt some
gullible people would believe me.
> The overall report seems reasonable without seeing any data to
> contradict it. However, if the data presented, i.e the accuracy of the
> clocks, is not a lie, than this particular H&K experiment is indeed
> useless.
The author claims to have got hold of the 'original' H and K results.
He gives no indication of how he got hold of these or why H and K
changed their results in the published paper.
This alleged discrepancy has not been published in any reputable
journal and has not even been discussed on this newsgroup except
by one notable crackpot.
> I have been an engineer long enough to know how accurate ones
> instruments need to be to be trusted. I don't even need to see the
> experiment details, knowledge of the instrument errors is enough,
> especially if there are only a few intruments. If the clocks were say,
> 0.05ns/hr stable rather than what is claimed in the data, e.g.
> +/-5ns/hr, then the experiment might be ok. *Otherwise*, there is no
> hope in hell of that experiment being valid, and I don't care how one
> fiddles the data.
Are you saying H and K were idiots?
> Do you have any official data that says such clocks routinely achieve
> *much* better than the +/-5ns/hr claimed?
The agreement between the different clocks in the published data
suggests that clock drift did not invalidate the results.
Errors of around +/- 25 ns for the whole experiment are quoted in the
published results.
Martin Hogbin
Stephen Speicher
>
> "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message news:t_ZN9.1045$BW1....@newsfep1-gui.server.ntli.net...
>
> > The overall report seems reasonable without seeing any data to
> > contradict it. However, if the data presented, i.e the accuracy of the
> > clocks, is not a lie, than this particular H&K experiment is indeed
> > useless.
>
> The author claims to have got hold of the 'original' H and K results.
> He gives no indication of how he got hold of these or why H and K
> changed their results in the published paper.
>
First, to clarify, the "report" on the web site is an html
version of Kelly's monograph for "The Institution of Engineers of
Ireland." The data he used was from an internal US Naval
Observatory document prepared several months prior to the H & K
paper. I have the document and I have verified that Kelly does
not make up the figures he quotes. His interpretation of such is,
however, quite circumspect. I have responded before to several
of the allegations in Kelly's paper, and below I include a few
facts from one such response which should be kept in mind when
reading the paper.
*****************************************************************
1) Are you aware of the fact that Kelly's analysis was done with
data which represented just the initial 10% of data of the full
experiment? With 90% of the data not having yet been analyzed, I
would be somewhat cautious about using words such as
"conclusive."
2) What is called the "drift" rate in Kelly's paper is a
comparison between the individual clocks and the _mean_ of the
ensemble of clocks at the USNO. The reason one takes the mean of
the ensemble is to increase reliability, such that the mean
provides a better characteristic than evinced by the performance
of any single clock. When the given drift rates for the
individual clocks, as given in Kelly's paper, are treated as an
ensemble, the mean drift _for each of the eastward and westward
trips_ is less than 1 ns/hr, not the individual drift rates as
quoted. As stated in the internal USNO report, "MEAN (FLYING) can
be expected to produce considerably more reliable results."
3) There are a number of things which trigger a "red flag" for me
when I read any analysis in any paper or book. One such concern
of major import is selective quoting. When quotes are provided in
a manner to support one's thesis, and one leaves out parts of
quotes which do not so support, then I justifiably remain
extremely suspicious of whatever else is presented. Here are two
such examples which I discovered by checking the quotes Kelly
gives from Hafele himself, against the actual transcript of the
USNO internal document. Kelly quotes Hafele as saying:
"Most people (myself included) would be reluctant to
agree that the time gained by any one of these clocks
is indicative of anything"
But, Kelly chooses to leave out the words which follow in the
sentence, words which convey an entirely different message than
the partial quote Kelly provides. Here is the continuation of
the sentence:
"but the rather striking consistency between all four
clocks must be taken seriously. The averages of the
four final time differences are consistently negative
or near zero for the eastward trip and consistently
positive for the westward trip."
A second example of selective quoting. Kelly quotes Hafele as
saying:
"the difference between theory and measurement is
disturbing."
The full quote from Hafele is:
"Indeed, the difference between theory and measurement
in Figure 4 is disturbing, and if our final analysis
does not improve agreement, an improved version of this
experiment should be given serious consideration."
Also, this statement was made by Hafele based on only 10% of the
data, and in the "final analysis" indeed there was "improve[d]
agreement," just as reported in the final paper to _Science_ some
months later.
4) Keep in mind that one of the purposes of the experiment was
"to study deterioration, if any, in the performance of four good
clocks under traveling conditions"? This was, afterall, the very
first time an atomic clock was returned and compared for actual
elapsed time. They had just had some recent experience
transporting portable clocks in a car, and were trying to learn
and establish procedures for proper protection of the clocks
themselves.
*****************************************************************
--
Stephen
s...@speicher.com
Ignorance is just a placeholder for knowledge.
Printed using 100% recycled electrons.
-----------------------------------------------------------
Kevin Aylward
> "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
> news:t_ZN9.1045$BW1....@newsfep1-gui.server.ntli.net...
>> Martin Hogbin wrote:
>>> "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
>>> news:oxUN9.366$BW1....@newsfep1-gui.server.ntli.net...
>>>
>>>> This might be interesting
>>>> http://www.mywebpages.com/asps/H&Kpaper.htm
>>>
>>> Your link to this pop-up infested, commercial site can, at best, be
>>> described as mischievous. What are you trying to achieve?
>>>
>>
>> Its called a balanced view.
>
> You mean the 'establishment' view is not to be trusted?
>
No view is to trusted on face value. Its called scepticism. The idea
that scientists are more honest than anyone else is simple false. I have
worked long enough to know what the score is when budgets are involved.
I don't know of any students that did not, at some point, use some
creativity in doing their class experiments.
>> Look, I have no idea if the site is a *lie* or not.
>
> It is a commercial site with pop-up ads that wants to attract traffic.
> The rest of the pages on that site consist of obvious crackpottery.
>
That is irrelevant. Does the data stand up on its own or not.
> I could put up a web page claiming that H and K forgot to wind
> up the clocks before the experiment started. No doubt some
> gullible people would believe me.
>
So yo belive that any critisem of the 'establishment' is wrong?
>> The overall report seems reasonable without seeing any data to
>> contradict it. However, if the data presented, i.e the accuracy of
>> the clocks, is not a lie, than this particular H&K experiment is
>> indeed useless.
>
> The author claims to have got hold of the 'original' H and K results.
> He gives no indication of how he got hold of these or why H and K
> changed their results in the published paper.
>
Why should the author put foward reasons as to why the results were
changed, if in fact they were? Thats simple not relevent.
> This alleged discrepancy has not been published in any reputable
> journal and has not even been discussed on this newsgroup except
> by one notable crackpot.
>
So, if it has been discussed in this NG, you should be able to point to
the data that refute this claim.
>> I have been an engineer long enough to know how accurate ones
>> instruments need to be to be trusted. I don't even need to see the
>> experiment details, knowledge of the instrument errors is enough,
>> especially if there are only a few intruments. If the clocks were
>> say,
>> 0.05ns/hr stable rather than what is claimed in the data, e.g.
>> +/-5ns/hr, then the experiment might be ok. *Otherwise*, there is no
>> hope in hell of that experiment being valid, and I don't care how one
>> fiddles the data.
>
> Are you saying H and K were idiots?
>
I am quite sure H and K always knew exactly what they were doing.
>> Do you have any official data that says such clocks routinely achieve
>> *much* better than the +/-5ns/hr claimed?
>
> The agreement between the different clocks in the published data
> suggests that clock drift did not invalidate the results.
>
> Errors of around +/- 25 ns for the whole experiment are quoted in the
> published results.
The point is are the published results creative?
I note that you don't seem to dispute the claimed drift rates. I don't
accept this error quote here if the 4 clocks do indeed have the drifts
claimed. You can't do meanfull statistical analysis with only 4 clocks,
at least, if I used only four die as verification to go ahead and get a
$100k mask set done for a production run of one of my analogue chips I
would probable be fired.
Kevin Aylward
I don't follow exactly what you are saying here. What was the raw drift
of a single clock on its own. 4 clocks is simply not enough of a sample
to do any meanfull statistical analysis.
When I get a wafer back from the fab, my boss tells me to test at least
30-40 on several wafers. At $100k per mask set I am sure you can
apprciate why.
> 3) There are a number of things which trigger a "red flag" for me
> when I read any analysis in any paper or book. One such concern
{Sniped prose}
All that really matters is what the clocks accuracy was. Everyone
colours what they say to put them in the best light.
>
> Also, this statement was made by Hafele based on only 10% of the
> data, and in the "final analysis" indeed there was "improve[d]
> agreement," just as reported in the final paper to _Science_ some
> months later.
>
But why was the data much improved? Did the clocks get better?
Either the clocks drifted or they did not. One is flying a few clocks
around for several days open loop, i.e. no way of checking them.
Side note.
If one is measuring something and the measuring error/standard deviation
of an instrument is say 1%, one can use say 30 uncorrelated such
instruments and reduce the error to say 1/sr(30) by averaging the
readings. However, if the instruments are like 50% in error, and there
are only 4 of them, all bets are off, the statistice only applies if the
sample size is large. One could never ship a commercial product with
such poor verification.
I am not clear on the +/-1ns/hr quote that you give here is valid or
not. However, over several days e.g 10 days = +/-240ns, with the
expected value to be -40ns and +275ns, this dose not convice me at all
that the data means anything. Would you bet your life on medical
equipment supported by such data?
Jan
> First, to clarify, the "report" on the web site is an html
> version of Kelly's monograph for "The Institution of Engineers of
> Ireland."
Oh my God - this dunce is still around? Several years ago he claimed
an internal inconsistency of SR. His proof was based on the assumption
of light speed constancy in all frames. We exchanged a few e-mails but
when I pointed out to him that SR assumed light speed constancy in
inertial frames only, he never wrote back.
Jan Bielawski
Stephen Speicher
You _never_ follow _exactly_ what anyone says, which is why I
have continued to ignore you since your inglorious return to this
group. I learned my lesson with you in past frustrations; your
muddleheadedness makes it impossible to sustain a rational
conversation with you. Go find some other sucker to waste his
time "conversing" with you -- I have things of value to do.
Stephen Speicher
It just gets worse in a couple of other monographs he put out.
Things like the "Sagnac effect is in contradiction of special
relativity," and for the Michelson-Gale experiment "Relativity
has a difficulty with this test," and likewise confusions for
other classic experiments. From all this he concludes "that
space and time are absolute" and that the speed of light "may
have a speed, relative to an observer, that is less or greater
than c" and that the speed of light can be dependent on the speed
of the source.
Another case of what happens when an engineer runs amok in the
world of physics.
scerir
Benjamine
> I've found a lot of references to this paper, but not the article. Could
> you help me to find it by a link, or send it to me ?
One of the *very* best places is Keating's paper on the
MacGraw-Hill Encyclopedia of Physics.
s.
Kevin Aylward
Obviously you didn't, and don't. If so, you would not have responded to
this post. Please continue to ignore my further posts, you have never
had anything to say of relevance.
Kevin Aylward
> On 25 Dec 2002, Jan wrote:
>
>> Stephen Speicher <s...@speicher.com> wrote in message
>>
news:<Pine.LNX.4.33.021224...@localhost.localdomain>...
>>
>>> First, to clarify, the "report" on the web site is an html
>>> version of Kelly's monograph for "The Institution of Engineers of
>>> Ireland."
>>
>> Oh my God - this dunce is still around? Several years ago he claimed
>> an internal inconsistency of SR. His proof was based on the
>> assumption of light speed constancy in all frames. We exchanged a
>> few e-mails but when I pointed out to him that SR assumed light
>> speed constancy in inertial frames only, he never wrote back.
>>
>
> It just gets worse in a couple of other monographs he put out.
> Things like the "Sagnac effect is in contradiction of special
> relativity," and for the Michelson-Gale experiment "Relativity
> has a difficulty with this test," and likewise confusions for
> other classic experiments. From all this he concludes "that
> space and time are absolute" and that the speed of light "may
> have a speed, relative to an observer, that is less or greater
> than c" and that the speed of light can be dependent on the speed
> of the source.
>
> Another case of what happens when an engineer runs amok in the
> world of physics.
At least is far preferable than a physicists running amok in the world
of engineering. Measurement errors are of much more concern to engineers
than physicists, and they usually become much better qualified, e.g.
state certification, to make such judgments. If a bridge fails because
someone uses an instrument with errors of the order of the thing the is
trying to measure, people die. A faulty clock is only going to make a
physicist late for his next convention, so I doubt if a physicist cares
a toss whether he even has a watch or not. Engineers actually live in
the real world. Indeed, your belief that an instrument with a
measurement error equal to the measurement makes for a reasonable
measurement, speaks volumes. It is not surprising that your not an
engineer, you could never cut it.
Dirk Van de moortel
"Benjamine" <benjamin...@hotmail.com> wrote in message news:au91lq$8fj$1...@news-reader10.wanadoo.fr...
Original paper not found on the web but an interesting thread
about HK was the one surrounding Eric Prebys' post:
http://groups.google.com/groups?&as_umsgid=3C921AAE...@fnal.gov
More about HKOP:
http://groups.google.com/groups?&q=hafele+keating+original+paper
Enjoy,
Dirk Vdm
Kevin Aylward
> On 25 Dec 2002, Jan wrote:
>
>> Stephen Speicher <s...@speicher.com> wrote in message
>>
news:<Pine.LNX.4.33.021224...@localhost.localdomain>...
>>
>>> First, to clarify, the "report" on the web site is an html
>>> version of Kelly's monograph for "The Institution of Engineers of
>>> Ireland."
>>
>> Oh my God - this dunce is still around? Several years ago he claimed
>> an internal inconsistency of SR. His proof was based on the
>> assumption of light speed constancy in all frames. We exchanged a
>> few e-mails but when I pointed out to him that SR assumed light
>> speed constancy in inertial frames only, he never wrote back.
>>
>
> It just gets worse in a couple of other monographs he put out.
> Things like the "Sagnac effect is in contradiction of special
> relativity," and for the Michelson-Gale experiment "Relativity
> has a difficulty with this test," and likewise confusions for
> other classic experiments. From all this he concludes "that
> space and time are absolute" and that the speed of light "may
> have a speed, relative to an observer, that is less or greater
> than c" and that the speed of light can be dependent on the speed
> of the source.
>
> Another case of what happens when an engineer runs amok in the
> world of physics.
At least is far preferable than a physicists running amok in the world
of engineering. Measurement errors are of much more concern to engineers
than physicists, and they usually become much better qualified, e.g.
state certification, to make such judgments. If a bridge fails because
someone uses an instrument with errors of the order of the thing the is
trying to measure, people die. A faulty clock is only going to make a
physicist late for his next convention, so I doubt if a physicist cares
a toss whether he even has a watch or not. Engineers actually live in
the real world. Indeed, your belief that an instrument with a
measurement error equal to the measurement makes for a reasonable
measurement, speaks volumes. It is not surprising that your not an
engineer, you could never cut it.
Kevin Aylward
Obviously you didn't, and don't. If so, you would not have responded to
this post. Please continue to ignore my further posts, you have never
had anything to say of relevance.
Kevin Aylward
Minor Crank
(Groan) Not again!
Cranks regularly attack shortcomings in the Hafele and Keating
experiment, trying to leverage them into an attack on relativity.
I myself am a contrarian in regard to the H&K experiment. However, I
wish to emphasize that the H&K experiment was NEVER originally
intended to be anything more than a clever demonstration of SR+GR
effects in the low speed realm. It tested 1970's atomic clock
technology to the limit, and the data analysis was, in my opinion,
overly optimistic.
The validity of relativity DOES NOT DEPEND on whether the H&K results
were as good as Hafele and Keating claimed!
I made this point in an earlier thread, and don't wish to waste
bandwidth repeating it.
http://groups.google.com/groups?selm=40bb2cea.0211250016.2b1495f%40posting.google.com
There is a link in that thread to an actual scanned copy of the H&K
papers.
There are literally thousands of experiments that have validated SR
and GR predictions to an extraordinary degree of accuracy. Here are
just a few examples:
- The GPS regularly validates the predictions of SR+GR 24 hours a day,
7 days a week.
- The union of SR an QM has resulted in the most accurately validated
theory of all time, QED. A benchmark prediction of QED, the anomalous
magnetic moment of the electron, has been validated to nearly 14
significant figures of accuracy!
- Lunar laser ranging data has validated the weak equivalence
principle to 1 part in 10^13 to 10^14, and has validated the strong
equivalence principle to 1 part in 10^5. General relativity is derived
as an inevitable consequence of the strong equivalence principle.
- The Hipparcos Space Astrometry Mission validated GR predictions
concerning the bending of light around the sun to within 0.1% of the
GR prediction. Indeed, so sensitive was the Hipparcos instrumentation,
that it was capable of detecting GR effects out to a 90 degree angle
away from the sun.
- Doppler-free laser spectroscopy on relativistic ion beams has
confirmed the relativistic time dilation factor predicted by SR to
several parts in 10^7
- For more on experimental tests of SR and GR, see the Usenet Physics
FAQ
http://math.ucr.edu/home/baez/physics/Relativity/SR/experiments.html
- Another fascinating article is "The Confrontation between General
Relativity and Experiment" by Clifford Wills.
http://www.livingreviews.org/Articles/Volume4/2001-4will/article_prep.html
The crankish obsession with the ancient Hafele and Keating experiment
is totally unreasonable. EVEN AS A SKEPTIC, I reached the following
conclusions in my independent reanalysis of the H&K results:
1) The experimental results suggested, to an 0.1 level of
significance, that the clocks on the eastward trip exhibited a
negative delta tau, the magnitude of this delta tau being consistent
with the value predicted by SR+GR.
2) The experimental results suggested, to an 0.1 level of
significance, that the clocks on the westward trip exhibited a
positive delta tau, the magnitude of this delta tau being consistent
with the value predicted by SR+GR.
In other words, even upon skeptical reanalysis, the Hafele and Keating
results strongly suggest the validity of the SR+GR prediction.
"Minor Crank"
Martin Hogbin
"Kevin Aylward" <ke...@anasoft.co.uk> wrote in message news:jDeO9.30$0j6....@newsfep3-gui.server.ntli.net...
> Martin Hogbin wrote:
> > "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
> > news:t_ZN9.1045$BW1....@newsfep1-gui.server.ntli.net...
> >> Martin Hogbin wrote:
> >>> "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
> >>> news:oxUN9.366$BW1....@newsfep1-gui.server.ntli.net...
> >>>
>
> that scientists are more honest than anyone else is simple false. I have
> worked long enough to know what the score is when budgets are involved.
> I don't know of any students that did not, at some point, use some
> creativity in doing their class experiments.
I am sure that the biggest problem H and K had was to get funding
to confirm something that was already accepted. Results that
disagreed with relativity would attract much more interest, and
therefore funding, than merely confirming what everybody
already knows.
> > It is a commercial site with pop-up ads that wants to attract traffic.
> > The rest of the pages on that site consist of obvious crackpottery.
> >
>
> That is irrelevant.
Not entirely. Do you know of any genuine scientific papers that
can be found on such sites.
> Does the data stand up on its own or not.
The question here is, 'what data?'
Do you mean all the data from the HK tests or do
you mean that selectively taken by Kelly?
> So yo belive that any critisem of the 'establishment' is wrong?
Criticism that relies on selecting data and misrepresenting the
original experimenters is wrong.
> >
> > He gives no indication of ... why H and K changed their
> > results in the published paper.
> >
> Why should the author put foward reasons as to why the results were
> changed, if in fact they were? Thats simple not relevent.
Because he is effectively calling K and K either incompetent or
fraudsters.
> >> I have been an engineer long enough to know how accurate ones
> >> instruments need to be to be trusted. I don't even need to see the
> >> experiment details, knowledge of the instrument errors is enough,
> >> especially if there are only a few intruments. If the clocks were
> >> say,
> >> 0.05ns/hr stable rather than what is claimed in the data, e.g.
> >> +/-5ns/hr, then the experiment might be ok. *Otherwise*, there is no
> >> hope in hell of that experiment being valid, and I don't care how one
> >> fiddles the data.
> >
> > Are you saying H and K were idiots?
> >
>
> I am quite sure H and K always knew exactly what they were doing.
So what are you saying. Are you suggesting that H and K deliberately
falsified their results?
> >> Do you have any official data that says such clocks routinely achieve
> >> *much* better than the +/-5ns/hr claimed?
> >
> > The agreement between the different clocks in the published data
> > suggests that clock drift did not invalidate the results.
> >
> > Errors of around +/- 25 ns for the whole experiment are quoted in the
> > published results.
>
> The point is are the published results creative?
What, to create a result that confirms relativity. If I were going to
be creative I would create something much more exciting.
> I note that you don't seem to dispute the claimed drift rates. I don't
> accept this error quote here if the 4 clocks do indeed have the drifts
> claimed. You can't do meanfull statistical analysis with only 4 clocks,
> at least, if I used only four die as verification to go ahead and get a
> $100k mask set done for a production run of one of my analogue chips I
> would probable be fired.
I have not read the original H and K paper but I see no need to do so
for the following reasons:
1) The experiment was carried out under the auspices of the USNO
which is one of best timekeeping laboratories in the world.
They have a vested interest in knowing how to set up, use, and
move about, accurate clocks.
2) The results were published in a reputable, peer reviewed , journal.
3) I am not aware of any serious criticism of the experiment being
published in a quality journal.
4) Several posters on this newsgroup, whose views on these matters
I trust, have defended the experiment and its conclusions.
5) Authoritative books on the experimental basis of relativity,
such as that by Zhang, quote the results without adverse comment.
6) There is no mileage, in terms of fame, fortune, or funding, in
discovering something everybody knows. If H and K had
published data that radically disagreed with relativity
they would be up there with Michelson and Morley (subject, of
course, to confirmation of their results).
Martin Hogbin
Bruce Richmond
> "Benjamine" <benjamin...@hotmail.com> wrote in message news:<au91lq$8fj$1...@news-reader10.wanadoo.fr>...
> > My father asked me to search on the web the original article published
> > in 1972 of Hafele and Keating experiment about relativity. (Sciences n°
> > 177 - p.166 to 170).
> > I've found a lot of references to this paper, but not the article. Could
> > you help me to find it by a link, or send it to me ?
> > A lot of thanks
>
> (Groan) Not again!
>
Calm down before you blow a gasket. He never even said why his father
wanted to see the original article. Not every question is an attack
on relativity.
Minor Crank
Thanks, Stephen, for posting this. I was completely unaware the points
that you raised. There's lots of important ammunition for the the next
time some moron tries to wave Kelly's paper around in an attempt to
disprove relativity - I'm keeping a link to your post under under my
"favorites."
"Minor Crank"
Kevin Aylward
So what, some peer reviewed stuff has to be bogus. You can't win all the
time.
>
> 3) I am not aware of any serious criticism of the experiment being
> published in a quality journal.
>
Not surprising is it. You have to be quite "courageous" to say anything
negative about Einstein, even a no comment.
> 4) Several posters on this newsgroup, whose views on these matters
> I trust, have defended the experiment and its conclusions.
Mr. Speicher claimed that the drift of the clocks was more like 1ns/hr
compared to Mr. Kellys claim of +/- 5ns. The lower figure still shows
its a no contest. You simple can not do valid measurements with only 4
clocks when the reading is supposed to be 275ns and the base error has
to be more than 240ns. I don't care how you dress the data up.
There is only *guesswork* in making assumptions about what the clocks
did or not do when not measured. To have any confidence at all, the
clocks would need to be 10 to 100 times better.
>
> 5) Authoritative books on the experimental basis of relativity,
> such as that by Zhang, quote the results without adverse comment.
>
Its a snow ball though. Eveone quotes every one else without checking
the data themselves, as you don't wish to.
> 6) There is no mileage, in terms of fame, fortune, or funding, in
> discovering something everybody knows.
There is a huge negative in wasting money.
>If H and K had
> published data that radically disagreed with relativity
> they would be up there with Michelson and Morley (subject, of
> course, to confirmation of their results).
The evidence suggests that the data is meaningless, not that it
contradicts any theory.
Look, I know how this thing works in the real, grown up world, not the
fantasy world. I was an analogue engineer at the SuperConductiong
Supercollider for a couple of years. There is simply no way you get to
spend lots of money without showing something for it. Its that simple.
Minor Crank
You're right. I emailed him an apology.
"Minor Crank"
Benjamine
news: 40bb2cea.02122...@posting.google.com...
> > I've found a lot of references to this paper, but not the article.
Could
> > you help me to find it by a link, or send it to me ?
> (Groan) Not again!
???? :-)
I'm not a scientist, and I really didn't understand anything about
relativity and else.... I don't know exactly what my father wants to
prove or not, and even if he wants to prove something. I just know that
he's very interessed by the question, and he's being wary about articles
about articles about articles...etc...
That's why he wants to read the original paper.
I've founded it on the thread, and transmitted it just now.
Thanks a lot.
--
Benjamine, France
Minor Crank
> "Minor Crank" <blue_whal...@attbi.com> a écrit dans le message de
> news: 40bb2cea.02122...@posting.google.com...
> > > I've found a lot of references to this paper, but not the article.
> Could
> > > you help me to find it by a link, or send it to me ?
>
> > (Groan) Not again!
>
> ???? :-)
> I'm not a scientist, and I really didn't understand anything about
> relativity and else.... I don't know exactly what my father wants to
> prove or not, and even if he wants to prove something. I just know that
> he's very interessed by the question, and he's being wary about articles
> about articles about articles...etc...
> That's why he wants to read the original paper.
> I've founded it on the thread, and transmitted it just now.
> Thanks a lot.
> --
> Benjamine, France
I'd like to repeat in public the apology that I sent to you privately by
email.
"Basically, I wanted to apologize if my posted newsgroup response sounded
like some sort of personal attack on you or your father. After confronting
so many cranks on these newsgroups, one tends to get rather touchy about
these things."
Thanks for understanding. I'm not always a very decent character in these
newsgroup discussions...
%-(
It is very admirable that your father wishes to read the original H&K
papers, rather than relying on secondary and tertiary accounts.
Regards,
"Minor Crank"
Martin Hogbin
"Kevin Aylward" <ke...@anasoft.co.uk> wrote in message news:2P%O9.897$Ux4....@newsfep1-gui.server.ntli.net...
> Martin Hogbin wrote:
> > I have not read the original H and K paper but I see no need to do so
> > for the following reasons:
> >
> > 1) The experiment was carried out under the auspices of the USNO
> > which is one of best timekeeping laboratories in the world.
> > They have a vested interest in knowing how to set up, use, and
> > move about, accurate clocks.
> >
> > 2) The results were published in a reputable, peer reviewed , journal.
>
> So what, some peer reviewed stuff has to be bogus. You can't win all the
> time.
Not at this level. You are suggesting that a simple inability to analyse
the clock drift data got past the referees.
> > 3) I am not aware of any serious criticism of the experiment being
> > published in a quality journal.
> Not surprising is it. You have to be quite "courageous" to say anything
> negative about Einstein, even a no comment.
There is no need to criticise Einstein. It would be quite normal to see
criticism of H an K's experimental methods or conclusions if this were
justified, most likely in the original publication.
> > 4) Several posters on this newsgroup, whose views on these matters
> > I trust, have defended the experiment and its conclusions.
>
> Mr. Speicher claimed that the drift of the clocks was more like 1ns/hr
> compared to Mr. Kellys claim of +/- 5ns. The lower figure still shows
> its a no contest. You simple can not do valid measurements with only 4
> clocks when the reading is supposed to be 275ns and the base error has
> to be more than 240ns. I don't care how you dress the data up.
None of the physicists on this group has claimed that the H and K
results are unreliable.
> There is only *guesswork* in making assumptions about what the clocks
> did or not do when not measured. To have any confidence at all, the
> clocks would need to be 10 to 100 times better.
I agree it would be nice if the clocks had been better, but the USNO
are far more expert at dealing with clock drifts than either you or I.
> > 5) Authoritative books on the experimental basis of relativity,
> > such as that by Zhang, quote the results without adverse comment.
> Its a snow ball though. Eveone quotes every one else without checking
> the data themselves, as you don't wish to.
Have a word with yourself Kevin. Are you seriously suggesting that a
man who sets himself the task of producing an authoritative work on
the experimental basis of relativity would not critically read
the work of H an K. It is not quoted in Zhang's books as a major
verification of relativity but it is a nice example of yet another experiment
in agreement with theory.
You might note that Zhang is quite critical of some of the other
experimental evidence for relativity.
> The evidence suggests that the data is meaningless, not that it
> contradicts any theory.
The data is, within the quoted limits of experimental error, in agreement
with relativity. No one is suggesting that this experiment alone is
the definitive verification of relativity.
> Look, I know how this thing works in the real, grown up world, not the
> fantasy world. I was an analogue engineer at the SuperConductiong
> Supercollider for a couple of years. There is simply no way you get to
> spend lots of money without showing something for it. Its that simple.
As I understand it H and K were interested in how to handle and
move caesium clocks as well as the results. I am sure that this information
valuable.
I have no doubt that results are adjusted to maximise funding in
some cases but this is not one of them. You are barking up the
wrong tree.
Martin Hogbin
Stephen Speicher
I'm glad you found the comments useful, but note that there is
much more that can be said in reference to Kelly and the
experiment. However, in a sense, it is a somewhat silly exercise
to do, if one puts this experiment into the proper perspective.
Previous "flying clock" experiments were not successful because
of the exhorbitant amount of ground time compared to flight, as
well as technical problems in transporting the clocks. H & K was
a groundbreaking experiment, and that is where its historical
significance lies. This successful experiment was the _first_,
not the last of its kind, and the further experiments were so
much better instrumented as to remove any openings for criticism
of the results being in accord with relativity.
We cannot hold the H & K experiment to the same level of
standards applied to the ones which followed; H & K were on the
periphery of detection for the relativistic phenomena, yet the
results were clear enough to demonstrate that the flying clocks
are affected in reasonable accord with the predictions of general
relativity. One can reasonably argue about quantification of that
degree of "reasonableness," but the direction is clearly
indicated.
In fact, one may wonder why anyone would choose to focus on the
H & K experiment to the degree that some do, until one realizes
that those who choose to do so, have a very big axe to grind. In
one case it may be a befuddled engineer who seeks to replace
relativity with his own pet theory -- ignoring all of the
indisputable experimental facts -- or in another case a
muddleheaded engineer who follows the lemming into the sea.
Those who lack the ability to challenge the theory directly on
principle, do so by nipping away at some exposed meat on the
bone. All they accomplish is to expose themselves, not the
theory.
Kevin Aylward
But none of them, it would seem state how H and K is a truly wonderfull
confiormation of GR as is often quoated. Its hedged a bit.
e.g, http://hyperphysics.phy-astr.gsu.edu/hbase/relativ/airtim.html
**********
"...at the U.S. Naval Observatory, should have lost 40+/-23 nanoseconds
during the eastward trip and should have gained 275+/-21 nanoseconds
during the westward trip ... Relative to the atomic time scale of the
U.S. Naval Observatory, the flying clocks lost 59+/-10 nanoseconds
during the eastward trip and gained 273+/-7 nanosecond during the
westward trip, where the errors are the corresponding standard
deviations.
*******
>> There is only *guesswork* in making assumptions about what the clocks
>> did or not do when not measured. To have any confidence at all, the
>> clocks would need to be 10 to 100 times better.
>
> I agree it would be nice if the clocks had been better, but the USNO
> are far more expert at dealing with clock drifts than either you or I.
However, I am far more expert in producing products, that work, with
minimum returns, that sell for money. I have been doing this for over 20
years. I found the ASSUME "arse out of you and me" reasonable accurate.
Again, I see drift rates variously quoted at 8ns/hr, 1ns/hr etc. The
fact that any measure at all can give the 8ns/hr quote makes my
confidance fall significantly on the data.
>
>>> 5) Authoritative books on the experimental basis of relativity,
>>> such as that by Zhang, quote the results without adverse comment.
>
>> Its a snow ball though. Eveone quotes every one else without checking
>> the data themselves, as you don't wish to.
>
> Have a word with yourself Kevin. Are you seriously suggesting that a
> man who sets himself the task of producing an authoritative work on
> the experimental basis of relativity would not critically read
> the work of H an K.
Actually. Yes. In principle. Its a lot of work to check all details. If
someone told you that Einstein and Plank had signed off on an
experiment, would you still feel as inclined to check it in detail as
you would if it was Laa Laa and Po.? People are human beings. We all
take short cuts. The scrutiny that is attributed my many, is simple not
done in practice. Again, I have just been about to long to know what
actually happens, rather than what the ideals are.
> It is not quoted in Zhang's books as a major
> verification of relativity but it is a nice example of yet another
> experiment in agreement with theory.
>
> You might note that Zhang is quite critical of some of the other
> experimental evidence for relativity.
>
>> The evidence suggests that the data is meaningless, not that it
>> contradicts any theory.
>
> The data is, within the quoted limits of experimental error, in
> agreement with relativity.
The quotes seem to vary depending on the source. I don't even know what
the "accepted" accuracy quote is. It seems about a sigma of 50% to me.
Engineers have a much more pessimistic view of things, imo. We simple
cant assume what has been done here, i.e. comparing so few clocks in
flight. It costs to much if he assumption is incorrect. Many times I
have had a circuit apparently work, but for the wrong reasons.
On the issue of H&K, my view is that a good scientist would simple state
that it was inconclusive, and appeal to better experiments. One must
error on the conservative side.
> I have no doubt that results are adjusted to maximise funding in
> some cases but this is not one of them. You are barking up the
> wrong tree.
imo, *all* results are adjust in the interests of funding. Its just a
question of how much.
If you don't get funding, you starve. Its that simple.
Martin Hogbin
"Kevin Aylward" <ke...@anasoft.co.uk> wrote in message news:pqdP9.33$N32....@newsfep3-gui.server.ntli.net...
> Martin Hogbin wrote:
> > "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
> > news:2P%O9.897$Ux4....@newsfep1-gui.server.ntli.net...
> >> Martin Hogbin wrote:
> > None of the physicists on this group has claimed that the H and K
> > results are unreliable.
> But none of them, it would seem state how H and K is a truly wonderfull
> confiormation of GR as is often quoated.
Nobody is saying that.
> >> There is only *guesswork* in making assumptions about what the clocks
> >> did or not do when not measured. To have any confidence at all, the
> >> clocks would need to be 10 to 100 times better.
> >
> > I agree it would be nice if the clocks had been better, but the USNO
> > are far more expert at dealing with clock drifts than either you or I.
>
> Again, I see drift rates variously quoted at 8ns/hr, 1ns/hr etc. The
> fact that any measure at all can give the 8ns/hr quote makes my
> confidance fall significantly on the data.
> > It is not quoted in Zhang's books as a major
> > verification of relativity but it is a nice example of yet another
> > experiment in agreement with theory.
> >
> > The data is, within the quoted limits of experimental error, in
> > agreement with relativity.
>
> The quotes seem to vary depending on the source. I don't even know what
> the "accepted" accuracy quote is. It seems about a sigma of 50% to me.
>
> On the issue of H&K, my view is that a good scientist would simple state
> that it was inconclusive, and appeal to better experiments. One must
> error on the conservative side.
I am not sure what your main point is. Is it just that H&K were a little
'over enthusiastic' in their reporting or is there a suggestion that the
foundations of relativity are beginning to crumble.
If you want to discuss the H&K paper in detail then we must
start from the paper itself, not various versions of it found on
the net.
Martin Hogbin
Kevin Aylward
More than likely.
> suggestion that the foundations of relativity are beginning to
> crumble.
>
Don't be daft. I don't see any conceivable way that the clocks can not
to do as claimed by GR/SR. Its a given with the results of a standard
E/M experiment or particle accelerator for the SR part. The equations
that determine atomic clocks are essentially the same ones.
The issue here is that too many criticise experiments with an axe to
grind. I have no axe to grind. Although I am in no way an expert in
GR/SR, I do know a bit about it,
http://www.anasoft.co.uk/physics/gr/index.html. As Fynmann points out, a
good scientist finds evidence to back up his theory, a better scientist
finds evidence to refute his theory. Any "reasonable" doubt on any
evidance at all, should cause a prudent man to reject such evidence.
My interest is only on the technicalities of what constitutes a valid
experiment. Its something I know about from a, get it wrong and you've
just lost the company $100k in a new set of i.c. masks sort of thing. As
a basic rule, if my instruments are not verifiable to say, at least 5%
right of the cuff I won't have much faith in them. There too many bits
and bobs that can go wrong. If you have to do all sorts of "corrections"
or interpretations of readings, its usually a sign of things aint
looking good matey. My experience is that if instruments are
significantly off, there could be all sorts of extraneous reasons for
duff readings, e.g. dirt, bad connectors, noise interference, temp
etc... If your instruments consistently read say within 0.01%, then you
know that the probability of there being a fault to be quite remote. To
get good data requires lots of things going right, to get bad data
requires only one thing to go wrong. Again, in my experience, do
instruments fail the same way?, often they do. If you go to a car scrap
yard looking for a particular part, Murphy says that your part is
already taken. This is in part because if that part failed for your,
chances are that is a manufacturing/design problem, that makes it fail
for someone else as well. I'm simple coming from the position of having
dealt with all sorts of strange things happening with instruments.
Minor Crank
I agree, absolutely!
To Kevin:
Why are you so focused on the Hafele and Keating experiment? I have
stated over and over in this newsgroup, the H&K experiment was NEVER
originally intended to be taken as anything more than a clever
demonstration of SR+GR effects in the low speed realm. Do you somehow
think that relativity stands or falls by its results? Don't you
understand that literally thousands of other experiments have been
conducted testing the predictions of SR and GR to extraordinary levels
of accuracy? The H&K experiment is just not that important!
"Minor Crank"
Minor Crank
> Martin Hogbin wrote:
> > I am not sure what your main point is. Is it just that H&K were a
> > little 'over enthusiastic' in their reporting or is there a
>
> More than likely.
>
> > suggestion that the foundations of relativity are beginning to
> > crumble.
> >
>
> Don't be daft. I don't see any conceivable way that the clocks can not
> to do as claimed by GR/SR. Its a given with the results of a standard
> E/M experiment or particle accelerator for the SR part. The equations
> that determine atomic clocks are essentially the same ones.
>
> The issue here is that too many criticise experiments with an axe to
> grind. I have no axe to grind. Although I am in no way an expert in
> GR/SR, I do know a bit about it,
> http://www.anasoft.co.uk/physics/gr/index.html. As Fynmann points out, a
> good scientist finds evidence to back up his theory, a better scientist
> finds evidence to refute his theory. Any "reasonable" doubt on any
> evidance at all, should cause a prudent man to reject such evidence.
In that case, what are we arguing about? My analysis states that the
H&K results were not of publication quality, since they didn't meet
the (p < 0.05) standard which is considered border-line acceptable
error level.
We can surely agree on that?
Where we differ is that, from reading Kelly, you don't believe the H&K
experiment showed anything at all, whereas I, from my independent
analysis of the H&K results, find their results consistent with the
SR+GR prediction to the (p < 0.1) significance level, and would
therefore use the words "strongly suggestive" to characterize the
results.
> My interest is only on the technicalities of what constitutes a valid
> experiment. Its something I know about from a, get it wrong and you've
> just lost the company $100k in a new set of i.c. masks sort of thing. As
> a basic rule, if my instruments are not verifiable to say, at least 5%
> right of the cuff I won't have much faith in them. There too many bits
> and bobs that can go wrong. If you have to do all sorts of "corrections"
> or interpretations of readings, its usually a sign of things aint
> looking good matey. My experience is that if instruments are
> significantly off, there could be all sorts of extraneous reasons for
> duff readings, e.g. dirt, bad connectors, noise interference, temp
> etc... If your instruments consistently read say within 0.01%, then you
> know that the probability of there being a fault to be quite remote. To
> get good data requires lots of things going right, to get bad data
> requires only one thing to go wrong. Again, in my experience, do
> instruments fail the same way?, often they do. If you go to a car scrap
> yard looking for a particular part, Murphy says that your part is
> already taken. This is in part because if that part failed for your,
> chances are that is a manufacturing/design problem, that makes it fail
> for someone else as well. I'm simple coming from the position of having
> dealt with all sorts of strange things happening with instruments.
When I started the "Hafele and Keating Revisited" thread, it was to
make the point that the H&K experiment was being blown up all out of
proportion. Anti-relativity cranks, who never read the original papers
and whose only knowledge of H&K was through Kelly, would attack the
H&K results, which would be defended by pro-relativity advocates who
likewise had never read the original papers, and whose only knowledge
of H&K was through popular accounts.
I wanted to stop that sort of nonsense. H&K is just not that important
an experiment. There are literally thousands of other experiments that
have confirmed the predictions of SR an GR to extremely high levels of
precision.
1) Is that your point as well?
2) Are you willing to acknowledge that your citation of Kelly was a
mistake, for the reasons that Stephen gave?
"Minor Crank"
Kevin Aylward
Where on earth do you get that idea from? I don't even really know what
you mean by this. I have simple seen some quotes on how amazingly
accurate the H&K experiment was, and have noted that this is rather
dubious. What is one supposed to do. Are you suggesting that one is not
allowed to criticize any experiment without being accused of being
accused of being overall "focused". Certainly if one was to criticize
any experiment, what other ones would you have in mind. Surely not
particle accelerates like CERN, or at FEMI.
I have
> stated over and over in this newsgroup, the H&K experiment was NEVER
> originally intended to be taken as anything more than a clever
> demonstration of SR+GR effects in the low speed realm. Do you somehow
> think that relativity stands or falls by its results? Don't you
> understand that literally thousands of other experiments have been
> conducted testing the predictions of SR and GR to extraordinary levels
> of accuracy?
Look, you are way, way off base on this. I have *no* problems with
either GR or SR (http://www.anasoft.co.uk/physics/gr/index.html). I have
never even *suggested* otherwise. The accuracy of the equations of SR
are simple not debatable. Questioning the validity of particular
experiments says nothing about ones views on the correctness of the
theory associated with the experiments. Unfortunately I do note that
many simply assume a criticism of an experiment is a criticism of the
theory. As I have already noted. A good scientist finds evidence to
support his beliefs, a better scientist finds evidence to contradict his
beliefs. I have already explained elsewhere, that particle experiments
automatically mean that clocks must behave as SR predicts. This would be
the case even if SR was in fact incorrect. The same basic equations
apply to both situations.
Kevin Aylward
More or less.
The issue here is that I am simply too familiar with instruments as an
analogue engineer. Its difficult to put ones finger on it explicitly,
its just one of those things you do to cover your arse. In covering my
arse, I would be quite unable to tell my boss that H&K was a reasonable
verification of GR/SR. That's all I can say really.
> Where we differ is that, from reading Kelly, you don't believe the H&K
> experiment showed anything at all, whereas I, from my independent
> analysis of the H&K results, find their results consistent with the
> SR+GR prediction to the (p < 0.1) significance level, and would
> therefore use the words "strongly suggestive" to characterize the
> results.
>
As noted, I am an engineer. I am pessemistic by nature.
Well... There is no doubt that the equations of SR have been verified to
great accuracy, e.g. 0.999999c etc. There is no doubt that they describe
the world very well. GR is another story. There are of course various
ways to derive clock shifts due to gravity. Indeed, Einstein did this
many years before GR even existed, so such shifts are not inherent to
GR. MTW discuss a few others. However, I am not aware of any
confirmations of GR where the GR effect is say 100 times the Newtonian
effects. 1% accuracies on 10% variations of Newtonian predictions can
hardly be described as high levels of confirmation. However, GR is on
quite a sound footing. It is based on a very simple assumption. It is
effectively *the* *direct* mathematical embodiment that all particles
fall the same in the same gravitational field.
>
> 2) Are you willing to acknowledge that your citation of Kelly was a
> mistake, for the reasons that Stephen gave?
Not really. It pointed out significant errors in the measurements. One
only has to look at the web to see how biased the view is that H&K are
magic to SR/GR. The fact that Kelly is dissident is no reason to silence
his views. He makes good points, the fact that he does not accept SR is
not really relevant. What is one supposed to do. Ignore all criticisms?
As far as all the waffle that is used to criticise SR, I can't, off
hand, think of any experiments that you could use to reasonably
criticise SR in the sense of refuting its basic equations. Do you have
any in mind?
Minor Crank
> I don't see any conceivable way that the clocks can not
> to do as claimed by GR/SR. Its a given with the results of a standard
> E/M experiment or particle accelerator for the SR part. The equations
> that determine atomic clocks are essentially the same ones.
If that is what you believe, it was at the very least mischievous,
possibly malicious, and certainly irresponsible of you to be giving
Benjamine a link to Kelly's anti-relativity trash, when he wanted a
link to the H&K papers.
You call what you did giving a "balanced view." I think you better do
some apologizing. Yes, I know that Benjamine was already aware of that
link, but that's not the point.
My behavior towards Benjamine wasn't that great, either. But the
instant Bruce made me realize that I had acted like an ass towards
him, I dashed off an email of apology.
Are you capable of the same?
"Minor Crank"
Bob Zombiewoof
> Stephen Speicher wrote:
> > On Wed, 25 Dec 2002, Martin Hogbin wrote:
> >>
> of a single clock on its own. 4 clocks is simply not enough of a sample
> to do any meanfull statistical analysis.
You aren't doing statistics on the clocks. You're doing statistics
on the time intervals read from four clocks with known statistical
errors.
> When I get a wafer back from the fab, my boss tells me to test at least
> 30-40 on several wafers. At $100k per mask set I am sure you can
> apprciate why.
Because you're doing statistics on the wafers from a given mask set.
At $100k per mask set, you don't want to do statistics on the mask sets
themselves. clock <==> mask set; time interval <==> wafer
> > 3) There are a number of things which trigger a "red flag" for me
> > when I read any analysis in any paper or book. One such concern
>
> {Sniped prose}
>
> All that really matters is what the clocks accuracy was. Everyone
> colours what they say to put them in the best light.
Precision, not accuracy. The actual accuracy is irrelevant, since
it makes little difference if the clock runs fast or slow, so long
as you normalize the data by normalizing the time intervals to a
single common reference.
> > Also, this statement was made by Hafele based on only 10% of the
> > data, and in the "final analysis" indeed there was "improve[d]
> > agreement," just as reported in the final paper to _Science_ some
> > months later.
> >
>
> But why was the data much improved? Did the clocks get better?
Apparently, that would require reading the original paper.
> Either the clocks drifted or they did not. One is flying a few clocks
> around for several days open loop, i.e. no way of checking them.
Since any drift is a systematic error, it is, in principle, possible
to quantify and correct for it. It would appear that reading the original
paper is necessary to determine exactly to what extent this was done.
> Side note.
>
> If one is measuring something and the measuring error/standard deviation
> of an instrument is say 1%, one can use say 30 uncorrelated such
> instruments and reduce the error to say 1/sr(30) by averaging the
> readings. However, if the instruments are like 50% in error, and there
> are only 4 of them, all bets are off, the statistice only applies if the
> sample size is large.
The clocks aren't the statistic. The time intervals read from clocks
with known statistical errors are. The use of multiple clocks is a
check on systematic effects, not statistical errors. The use of a large
number of clocks would give a statistical error on the systematics,
which would be useful, but not necessary unless one makes no investigation
of the sources of systematic errors in the experiment. The equivalent in
terms of your mask set, would be to dispense with analyzing the defects
in the wafers and use a lot of mask sets designed by trial and error
to obtain the desired failure rate on the wafers. Checking the wafers
to correct the systematic effect on the failure rate for a given mask
set is exactly the same as choosing a single (or several) clocks for
reliability.
> One could never ship a commercial product with such poor verification.
Microsoft does it all the time. In fact, they do worse, and advertise
products that don't yet exist, much less been tested for reliability
in performing the functions advertised. The term "vapourware" did not
spring into existence as a result of meeting the claims advertised
for existing products that had been verified against the performace
and cpability claims in the advertisements used to sell the products.
Selling but not shipping, and then shipping something less than advertised
is not an improvement over selling something which specifies the level
of verification that actually was carried out.
An physics experiment is also not a commercial product. The results
are intended for other physicists to evaluate, draw their own conclusions
and perform further experiments based upon what they've learned from
the successes/flaws in the published articles. However, like a commercial
product, one must make trade-offs between the cost of the experiment
and what questions the experiment can answer with what degree of
confidence. The major part of the work in any experiment takes place
before and after the experiment, in analyzing the systematic effects
of different designs and the data analysis. That's why experimental
articles focus on the experimental setup and the analysis of the data
more than the data or the conclusions. There is no such thing as a
definitive, cost-is-no-object experiment.
> I am not clear on the +/-1ns/hr quote that you give here is valid or
> not. However, over several days e.g 10 days = +/-240ns, with the
> expected value to be -40ns and +275ns, this dose not convice me at all
> that the data means anything. Would you bet your life on medical
> equipment supported by such data?
If there were no other options, sure. Lots of people do bet their
lives on operations which are even more unsupported by data as
successful. The first reciepients of organ transplants bet their
lives on procedures for which there existed no data from previous
human recipients, because there were no alternatives. As for the
haefele-keating experiment, you have alternatives if you find the
data to be unacceptable in supporting relativity. The alternatives
are convincing enough that I'd bet on it long before betting on
the success of even the most well-established surgical procedure.
Furthermore, I wouldn't trust the opinion of any doctor regarding
a procedure for which he/she had read only the biased opinions of
others rather than the articles written by those who have performed/
developed the procedure.
Martin Hogbin
"Kevin Aylward" <ke...@anasoft.co.uk> wrote in message news:vlgP9.329$N32....@newsfep3-gui.server.ntli.net...
> Martin Hogbin wrote:
> > "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
> > news:pqdP9.33$N32....@newsfep3-gui.server.ntli.net...
> >> Martin Hogbin wrote:
> >>> "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
> >>> news:2P%O9.897$Ux4....@newsfep1-gui.server.ntli.net...
> >>>> Martin Hogbin wrote:
> >
> >
> > I am not sure what your main point is. Is it just that H&K were a
> > little 'over enthusiastic' in their reporting or is there a
>
> More than likely.
Then I guess we will just have to agree to disagree about that.
I think the comments in my first post in this thread stand.
Martin Hogbin
Kevin Aylward
> "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
> news:<vlgP9.329$N32....@newsfep3-gui.server.ntli.net>...
>
>> I don't see any conceivable way that the clocks can not
>> to do as claimed by GR/SR. Its a given with the results of a standard
>> E/M experiment or particle accelerator for the SR part. The equations
>> that determine atomic clocks are essentially the same ones.
>
> If that is what you believe, it was at the very least mischievous,
> possibly malicious, and certainly irresponsible of you to be giving
> Benjamine a link to Kelly's anti-relativity trash, when he wanted a
> link to the H&K papers.
Absolutely nonsense. Look, it does not mater whether someone is a
"crank" or not, all that matters is is what they say on a particular
point valid or not. We can even argue that Mr.Spiecher is a crank
because he puts forward a theory of QM that is experimentally proven to
be false. Secondly, When I read the paper, I had *absolutely* *no* idea
at all that he was anti-relativity. The paper did not state any view on
whether relativity was correct or not. The final conclusion was that H&K
proved nothing, And as far as *reasonably* proved, I agree.
You cannot guarantee a completely objective argument from someone
presenting his own position. It would be nice if people were truly
honest, but they just arnt. They delude themselves. You have to accept
the way it really is, not how one would like/dream it to be. A sceptic
should *always* look at for and against views. To suggest that someone
believes in say, SR, because 1000's of others do is nonsense. The
rational for why one view is more valid than another must be ascertained
by examination of the facts, not by common consensus.
>
> You call what you did giving a "balanced view." I think you better do
> some apologizing. Yes, I know that Benjamine was already aware of that
> link, but that's not the point.
>
There is *nothing* to apologize for.
> My behavior towards Benjamine wasn't that great, either. But the
> instant Bruce made me realize that I had acted like an ass towards
> him, I dashed off an email of apology.
>
> Are you capable of the same?
What personal comments, rather then technical ones do you have in mind
that should be apologized for?
I keep getting the feeling that you have objections simply to the fact
that you perceived the paper to be anti-relativity, because it did not
*support* relativity. I saw nothing in the paper that was
anti-relativity, it simply presented data that suggested that H&K proved
nothing, one way or the other. I saw no claim that it refuted
relativity.
Kevin Aylward
>> drift of a single clock on its own. 4 clocks is simply not enough of
>> a sample to do any meanfull statistical analysis.
>
> You aren't doing statistics on the clocks. You're doing statistics
> on the time intervals read from four clocks with known statistical
> errors.
>
But the drift of the clocks are not known. Its the accuracy of the
clocks that are being questioned. Time intervals are given by the
clocks!
>
>> When I get a wafer back from the fab, my boss tells me to test at
>> least 30-40 on several wafers. At $100k per mask set I am sure you
>> can apprciate why.
>
> Because you're doing statistics on the wafers from a given mask set.
> At $100k per mask set, you don't want to do statistics on the mask
> sets themselves. clock <==> mask set; time interval <==> wafer
>
The device <==> clock. How accurate is the device, how accurate is the
clock.
>
>>> 3) There are a number of things which trigger a "red flag" for me
>>> when I read any analysis in any paper or book. One such concern
>>
>> {Sniped prose}
>>
>> All that really matters is what the clocks accuracy was. Everyone
>> colours what they say to put them in the best light.
>
> Precision, not accuracy. The actual accuracy is irrelevant, since
> it makes little difference if the clock runs fast or slow, so long
> as you normalize the data by normalizing the time intervals to a
> single common reference.
>
Ho hum. Its a given that one is discussing variations in drift rates.
Obviously, if you have a pure, predictably drift, it is a simple
calibration error. I addressed this in another post concerning HP's 3
cheap watches.
>
>> Either the clocks drifted or they did not. One is flying a few clocks
>> around for several days open loop, i.e. no way of checking them.
>
> Since any drift is a systematic error, it is, in principle,
> possible to quantify and correct for it. It would appear that reading
> the original paper is necessary to determine exactly to what extent
> this was done.
Again, all elementary no one cares about a calibration error, it is the
change in drift rates that is being referred to. I should have thought
that this was obvious.
>
>> Side note.
>>
>> If one is measuring something and the measuring error/standard
>> deviation of an instrument is say 1%, one can use say 30
>> uncorrelated such instruments and reduce the error to say 1/sr(30)
>> by averaging the readings. However, if the instruments are like 50%
>> in error, and there are only 4 of them, all bets are off, the
>> statistice only applies if the sample size is large.
>
> The clocks aren't the statistic. The time intervals read from clocks
> with known statistical errors are. The use of multiple clocks is a
> check on systematic effects, not statistical errors.
If by this, you mean that one cannot just average 4 clocks to reduce the
error. i.e. all clocks could be in error.
>The use of a
> large number of clocks would give a statistical error on the
> systematics, which would be useful, but not necessary unless one
> makes no investigation of the sources of systematic errors in the
> experiment. The equivalent in terms of your mask set, would be to
> dispense with analyzing the defects in the wafers and use a lot of
> mask sets designed by trial and error to obtain the desired failure
> rate on the wafers. Checking the wafers to correct the systematic
> effect on the failure rate for a given mask set is exactly the same
> as choosing a single (or several) clocks for reliability.
>
The analogy is all messed up here. One does not check the mask, there is
always only *one* set of masks. One checks devices from different wafers
from the same mask set. Generally one does statistics on the devices
from several wafers.
>> One could never ship a commercial product with such poor
>> verification.
>
> Microsoft does it all the time. In fact, they do worse, and
> advertise products that don't yet exist, much less been tested for
> reliability in performing the functions advertised.
Now I know that your out to lunch. In 20 years of using MS products,
never have I had one lock up, deleted file, unexpected exception, blue
screen, unable to save file...This is all lies by people wishing to
slander thy lord, Bill Gates.
>The term
> "vapourware" did not spring into existence as a result of meeting the
> claims advertised for existing products that had been verified
> against the performace and cpability claims in the advertisements
> used to sell the products. Selling but not shipping, and then
> shipping something less than advertised is not an improvement over
> selling something which specifies the level of verification that
> actually was carried out.
>
I should have phrased this a bit clearer. Software and Hardware are
worlds apart. My statement of what the statistics is actually done in
practice for i.c.s is based on real experience. It typically costs about
$1M to develop a product. You just don't ship a product, except
exceptionally, without doing and meeting standard tests. You actually
get around not doing full statistics by a bit of a cheat. This is the
+/- 6 sigma rule. Given a part has various specifications such as CMRR,
Frequency response, Slew Rate, offset voltage etc. one typically
measures say 40 devices on a few different wafers. You cant afford to
wait months to get another batch, so instead of using a 3 sigma, which
would give 97% good parts, you double it. Since the basic process does
not change from batch to batch, and the process have been characterised
historically, this is found to work in practise. What I will say is that
if the statistics don't match the intended goal, the spec sheet *will*
change. You simple cant sign of the design if statistics don't agree. Of
course there are misc. errors that can occur that result in duff parts
being shipped, but that is not really relevant to what lengths are taken
to ensure a decent product. You don't want to get sued.
> An physics experiment is also not a commercial product. The results
> are intended for other physicists to evaluate, draw their own
> conclusions and perform further experiments based upon what they've
> learned from the successes/flaws in the published articles. However,
> like a commercial product, one must make trade-offs between the cost
> of the experiment and what questions the experiment can answer with
> what degree of confidence. The major part of the work in any
> experiment takes place before and after the experiment, in analyzing
> the systematic effects of different designs and the data analysis.
> That's why experimental articles focus on the experimental setup and
> the analysis of the data more than the data or the conclusions. There
> is no such thing as a definitive, cost-is-no-object experiment.
>
Actually, as someone who has been an analogue engineer at the
SuperConducting Supercollider for a couple of years, this is debatable.
It was great. I got to buy *anything* I wanted in terms of equipment. I
can assure you, to all intents an purposes, compared to industrial
company labs, it was a true joy. I remember we bought 10 $5,000 HP
meters, no way could I get even one of those at other places I have
worked.
>
> Furthermore, I wouldn't trust the opinion of any doctor regarding
> a procedure for which he/she had read only the biased opinions of
> others rather than the articles written by those who have performed/
> developed the procedure.
But these latter article are also just as biased. Its a given fact.
Indeed, the authors usually have much more invested. They stand to lose
a lot more if what they did was a waste of tax payers money. To get a
balanced view, you want to look at both for and against, so likewise you
should not trust the opinion of the dentist with regard to how much work
is needed to be done on your teeth, as many sting operations have
proved.
What I find disconcerting is the view that scientists are in any way
*special* with regard to basic human behaviour. I see no evidence that
the average scientist is not going to be just as honest/cheat as the
average plumber, dentist, engineer, trucker, footballer, coal miner,
politician.... In fact, regarding the coal miner, arguably they are more
reliable. They certainly have a code of conduct in a mine based on their
own lives being at stake.
Minor Crank
> you mean by this. I have simple seen some quotes on how amazingly
> accurate the H&K experiment was, and have noted that this is rather
> dubious.
The difference between you and me is, I understand the correlated
rate-change analysis technique, and even wrote a computer program to verify
my understanding of the technique and its capabilities.
Correlated rate-change analysis is indeed capable of compensating for errors
introduced by sporadic quasi-permanent rate changes, which is by far the
most serious source of error affecting atomic clocks.
And now for an important message:
-----------------
I'm happy to announce that Stephen Speicher was able to knock a fundamental
hole in the arguments that I presented in the "Hafele and Keating Revisited"
thread that I had started some time ago in sci.physics
http://groups.google.com/groups?selm=40bb2cea.0211250016.2b1495f%40posting.g
oogle.com
My independent reanalysis of the H&K results was based on what I could
extract from Figure 1 of their second paper. Unfortunately, crucial data
necessary for an outside reader to verify their arguments was not published
by H&K.
I INCORRECTLY EQUATED LACK OF PUBLICATION TO BE EQUIVALENT TO LACK OF DATA.
Buried in the main text of the second H&K paper was a single sentence that
somehow I missed reading correctly: "We recorded the differences in the
times indicated by each member of the flying ensemble at regular intervals
before, during and after each trip, that is, throughout the entire data
period."
Figure 1 of the second H&K paper presents ONLY the before and after trip
data, and I misunderstood the figure to represent the entirely of their
collected data.
With data collected throughout the flight, the correlated rate-change method
could indeed be applied in a consistent manner to extract the results that
they did.
I therefore retract the conclusions that I presented in the "Hafele and
Keating Revisited" thread.
As I've written elsewhere, as well as in private emails to a few of you, I
would be happy for this matter to have been definitively resolved EITHER
WAY, and it certainly has gone the preferable direction!
"Minor Crank"
-----------------------------
Stephen Speicher <s...@speicher.com> wrote in message
news:<Pine.LNX.4.33.02122...@localhost.localdomain>...
> On 28 Dec 2002, Minor Crank wrote:
>
> >
> > | |
> > | |
> > - _ | *|
> > - _ | * |
> > - _| * *| _
> > |1 * | _ -
> > | 2 | _ =
> > | * |_ -
> > | 3|
> > | |
> > Clock behavior|Eastward | Clock behavior
> > before air |trip (no | after air
> > transport |measurements)| transport
> >
> > In the above illustration, a rate change occurred during the Eastward
> > trip, but when? If the rate change occurred at (1), then delta tau
> > amounted to -90 nsec, significantly in excess of the SR prediction. If
> > the rate change occurred at (2), then delta tau amounted to -60 nsec,
> > approximately equal to the SR prediction of -40 nsec. If the rate
> > change occurred at (3), than there would have been no delta tau at
> > all.
> >
>
> I am not sure why you label the trip section with "no
> measurements." As I mentioned in a previous post, six
> intercomparisons between the four clocks (three are redundant as
> a check) were recorded on an hourly basis throughout the flight.
> This data is what permits them to reconstruct the rate history
> for each clock during the trip, hour by hour.
I assumed "no data" during the flight, because none of the
intercomparison data was published in the papers. In Fig. 1, there are
big blank spots during the flight. Fig. 1 shows ONLY before-and-after
comparisons of the clocks against the USNO clocks!
1) It is unfortunately impossible for anyone to independently verify
the H&K results solely on the basis of the PUBLISHED data.
2) Reexamining the second H&K paper, I see a statement in the main
text I missed that indeed, data were recorded DURING the flight as
well as before and after.
3) Assuming the validity of the continuous in-flight analysis (and I
have no reason to doubt the validity of the analysis, since correlated
rate-change analysis was a well-established technique, and I
comprehend and agree with the algorithms employed), I will therefore
withdraw the set of arguments that I presented in "Hafele and Keating
Revisited" and will post a couple of retractions to that thread later
today at appropriate points.
4) You're one of the few people around here who ever bother to really
READ original literature. I'm really grateful for your contributions!
HOORAY!
"Minor Crank"
Kevin Aylward
>> Where on earth do you get that idea from? I don't even really know
>> what you mean by this. I have simple seen some quotes on how
>> amazingly accurate the H&K experiment was, and have noted that this
>> is rather dubious.
>
> The difference between you and me is, I understand the correlated
> rate-change analysis technique,
Ahmmm...
> and even wrote a computer program to
> verify my understanding of the technique and its capabilities.
Oh?. You had to that?
>
> Correlated rate-change analysis is indeed capable of compensating for
> errors introduced by sporadic quasi-permanent rate changes, which is
> by far the most serious source of error affecting atomic clocks.
What this does though is add yet another assumption, along with loads of
others, which one hopes to be valid. Well, may be it is, may be it aint.
Fundamentally this is an information problem. You cannot extract more
information than is available. If you decide to add some extra
information, that is not actually *provable*, to be there, well so be
it. The numbers themselves don't tell the whole story because they don't
contain information associated with the assumptions.
For example, what's the probabilities that over time all clocks stepped
changed together, or that there was some extra non step change drift,
and green goblins varied the drifts in-between measurements...etc...
Why are you so fixated on defending H&K now. Why do you want to confirm
that H&K *is* a decent test of GR/SR, when there are so many other much
better choices? If there is any reasonable doubt at all, a prudent
scientist, would simply say, "lets look at other tests, its no skin of
my nose if this one might be considered dubious by some, especially
those silly engineer types who no naff all..."
Minor Crank
> Minor Crank wrote:
> > and even wrote a computer program to
> > verify my understanding of the technique and its capabilities.
>
> Oh?. You had to that?
No, I didn't "have" to. I was learning Java and I wanted a real-life problem
to work on so I could become more familiar with Swing, client-server
programming, and the graphics and math packages. I had a long-standing
interest in the H&K experiment, so...
It is a "toy" program, not capable of handling the effects of shot noise,
but perfectly capable of sorting out which clock did what.
Since I was really doing this as an exercise to learn Java, the math
analysis is sort of primitive, but the Swing user interface is really,
really nice...
> > Correlated rate-change analysis is indeed capable of compensating for
> > errors introduced by sporadic quasi-permanent rate changes, which is
> > by far the most serious source of error affecting atomic clocks.
>
> What this does though is add yet another assumption, along with loads of
> others, which one hopes to be valid.
The major assumption that I made in my "toy" program was that the time
between sporadic quasi-permanent rate changes is relatively long, so that it
is unlikely that more than two clocks out of four would exhibit rate changes
within a pair of adjacent time samples.
In real life, in order to distinguish sporadic quasi-permanent rate changes
from shot noise, I imagine the requirements would have to be more stringent.
It is certainly POSSIBLE for multiple clocks to have conspired together to
fake SR+GR effects, but the probability of this occurring can be calculated
and included in the error bars.
"Minor Crank"
Stephen Speicher
>
> "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
> news:lmEP9.135$Hy5....@newsfep3-gui.server.ntli.net...
> > Minor Crank wrote:
> > > Correlated rate-change analysis is indeed capable of compensating for
> > > errors introduced by sporadic quasi-permanent rate changes, which is
> > > by far the most serious source of error affecting atomic clocks.
> >
> > What this does though is add yet another assumption, along with loads of
> > others, which one hopes to be valid.
"assumption?" "one hopes to be valid?"
What a moron. Just because that is how Aylward leads his life --
making "assumption[s]" where facts are required, and "hopes to be
valid" where rational judgment belongs -- does not mean
(thankfully!) that real scientists behave in such a muddleheaded
manner.
The USNO conducted many, many tests on a wide range of atomic
clocks, over an extended period of time. The quasipermanent rate
changes were shown to be random over time and averaged to zero
for long periods. They performed a great deal of statistical
analysis and the correlated rate-change method proved to be an
excellent way to take advantage of short-term effects.
>
> It is certainly POSSIBLE for multiple clocks to have conspired together to
> fake SR+GR effects, but the probability of this occurring can be calculated
> and included in the error bars.
>
Indeed. And, unlike Aylward, these people (USNO and physicists)
knew what they were doing. They did not function with a vacuum
between their ears.
Max Keon
Try giving your calculator some exercise.
If two clock bearer's set off from a reference point on the equator,
maintaining a 5 m/sec walking speed, following along a path which
has been excavated so that each clock bearer remains exactly at sea
level throughout the entire journey, and exactly on the line of the
equator, when they finally meet back at the reference point, their
clocks will read 4.14E-7 seconds different. The journey has taken
8004778 seconds to complete.
After synchronizing the clocks once more, they again set off on
the same journey. However, the west bound clock bearer falls asleep
along the way.
200 meters from home, the east bound clock bearer can see that his
counterpart has yet to arrive, so he waits. The west bound clock
bearer awakens, then hurries along to make up for lost time, but
the journey will obviously still take longer to complete.
When they eventually arrive in unison at their destination, the
clock difference is still 4.14E-7 seconds. Whatever happens along
the way, so long as they both arrive at their destination in unison,
and so long as the paths are exactly along the equator, the clock
difference is always 4.14E-7 seconds. Do the calculations if you
don't believe it.
Then remove the earth completely from the scene, leaving only a
circle of scaffolding where the path used to be, while maintaining
the tangential velocity of 465 m/sec. Do you think the result would
still be 4.14E-7 difference? Of course it wouldn't. The clocks would
now read (almost) exactly the same after completion of the journey.
And, according to some scientific studies, the lemming theory is
also false.
That's two out of two.
--
Max Keon
Stephen Speicher
Try giving your brain cells some exercise, for a change.
[Snip the usual ignorant drool from the Spaced-out Cadet.]
Bob Zombiewoof
> Bob Zombiewoof wrote:
> > "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
> >>
> >> I don't follow exactly what you are saying here. What was the raw
> >> drift of a single clock on its own. 4 clocks is simply not enough of
> >> a sample to do any meanfull statistical analysis.
> >
> > You aren't doing statistics on the clocks. You're doing statistics
> > on the time intervals read from four clocks with known statistical
> > errors.
> >
>
> But the drift of the clocks are not known. Its the accuracy of the
> clocks that are being questioned. Time intervals are given by the
> clocks!
>
> >
> >> When I get a wafer back from the fab, my boss tells me to test at
> >> least 30-40 on several wafers. At $100k per mask set I am sure you
> >> can apprciate why.
> >
> > Because you're doing statistics on the wafers from a given mask set.
> > At $100k per mask set, you don't want to do statistics on the mask
> > sets themselves. clock <==> mask set; time interval <==> wafer
> >
>
> The device <==> clock. How accurate is the device, how accurate is the
> clock.
Wrong comparison as I already pointed out.
> > Precision, not accuracy. The actual accuracy is irrelevant, since
> > it makes little difference if the clock runs fast or slow, so long
> > as you normalize the data by normalizing the time intervals to a
> > single common reference.
> >
>
> Ho hum. Its a given that one is discussing variations in drift rates.
> Obviously, if you have a pure, predictably drift, it is a simple
> calibration.
Exactly. And it appears that such a calibration is can be done
simply by taking the difference in drift rates at two different
points and using the slope of the line in ns/hr/hr.
> drift rates that is being referred to. I should have thought
> that this was obvious.
Nothing in that url is "obvious", although it does appear fairly clear
that such a correction is possible.
[...]
> > The clocks aren't the statistic. The time intervals read from clocks
> > with known statistical errors are. The use of multiple clocks is a
> > check on systematic effects, not statistical errors.
>
> If by this, you mean that one cannot just average 4 clocks to reduce the
> error. i.e. all clocks could be in error.
Huh? Look, each clock has a statistical uncertainty. Since the
measurements from each clock are independent, the uncertainties
add in quadrature. It's simple statistics. The statistic is
not 4 measurements. If clock A produces values of X +/- x and
clock B produces Y +/-y, then the statistics are the weighted
average of X and Y.
[...]
> The analogy is all messed up here. One does not check the mask, there is
> always only *one* set of masks. One checks devices from different wafers
> from the same mask set. Generally one does statistics on the devices
> from several wafers.
The analogy is not messed up. However, the fact that you really
do think it ridiculous to create multiple mask sets, to the extent
that you missed the point, I rest my case, since that was the point.
[...]
> > Microsoft does it all the time. In fact, they do worse, and
> > advertise products that don't yet exist, much less been tested for
> > reliability in performing the functions advertised.
>
> Now I know that your out to lunch. In 20 years of using MS products,
> never have I had one lock up, deleted file, unexpected exception, blue
> screen, unable to save file...This is all lies by people wishing to
> slander thy lord, Bill Gates.
I'll take that as agreement with what I said.
> >The term
> > "vapourware" did not spring into existence as a result of meeting the
> > claims advertised for existing products that had been verified
> > against the performace and cpability claims in the advertisements
> > used to sell the products. Selling but not shipping, and then
> > shipping something less than advertised is not an improvement over
> > selling something which specifies the level of verification that
> > actually was carried out.
> >
>
> I should have phrased this a bit clearer. Software and Hardware are
> worlds apart. My statement of what the statistics is actually done in
> practice for i.c.s is based on real experience. It typically costs about
> $1M to develop a product. You just don't ship a product, except
> exceptionally, without doing and meeting standard tests.
From my own (although somewhat limited experience) with missle guidance
system radar, I also know that the military does not trust the reliability
of "commercial products". Every single component probably cost 10 times
as much to test as it did to design and manufacture. Even the solder
used was different (80% gold, 20% tin and came in 10z rolls). A simple
mixer/multiplier [a cavity with tuning slugs and some sensistors for
temperature stability) required a week to test with
[...]
>
> Actually, as someone who has been an analogue engineer at the
> SuperConducting Supercollider for a couple of years, this is debatable.
> It was great. I got to buy *anything* I wanted in terms of equipment. I
> can assure you, to all intents an purposes, compared to industrial
> company labs, it was a true joy. I remember we bought 10 $5,000 HP
> meters, no way could I get even one of those at other places I have
> worked.
The supercollider was a political project which had a lot more
to do with politics than science. You'll notice that it never was
completed. If you perform a search for "radio frequency quadrupole"
and weapons, you might discover some additional reasons the project
was funded well initially and then allowed to crater. In general,
labs do not get funded nearly that well.
>
> >
> > Furthermore, I wouldn't trust the opinion of any doctor regarding
> > a procedure for which he/she had read only the biased opinions of
> > others rather than the articles written by those who have performed/
> > developed the procedure.
>
> But these latter article are also just as biased. Its a given fact.
> Indeed, the authors usually have much more invested.
The experimemtors do _not_ have a vested interest in the outcome of
an experiment. If anything, their interest would be in discovering
something that contradicted expectations.
> They stand to lose
> a lot more if what they did was a waste of tax payers money. To get a
> balanced view, you want to look at both for and against, so likewise you
> should not trust the opinion of the dentist with regard to how much work
> is needed to be done on your teeth, as many sting operations have
> proved.
Depending on what work I was told my teeth needed, I might not
trust the dentist. But, for something less speculative, I can tell
you from personal experience going under the knife for two knee
surgeries, a shoulder surgery and surgery on a finger, that before
each one, I went to the medical school library and read the literature
in orthopaedic journals so that I'd know exactly what my options
were and I obtained several opinions regarding those options prior
to choosing a surgeon in each case.
> What I find disconcerting is the view that scientists are in any way
> *special* with regard to basic human behaviour.
They aren't, as the recent affair at bell labs demonstrates. The
difference is that in most cases, there is no reward for biasing one's
data one way or the other. It's only when one is paid to obtain
specific results, that the temptation to obtain them is present.
Usually, in an academic setting, the principal investigators are
tenured and have little reason to care one way or another what
anyone thinks beyond the reputation they earn through performing
experiments that are unassailable on technical merit. There is a
great incentive to _not_ fiddle with data, since the accusation
of doing so will pretty much be a career ending event. Look at
pons and fleischman and the cold fusion fiasco.
> I see no evidence that the average scientist is not going to be
> just as honest/cheat as the average plumber, dentist, engineer,
> trucker, footballer, coal miner, politician....
I never said they were less honest. The point is, that they have
no incentive to be dishonest where their data is concerned. People
cheat where there exists some incentive to do so. A "footballer"
has little incentive to play poorly, since he gets paid for playing
the best he can. If police were paid to uphold the letter of the
law rather than make arrests, they would have no incentive to harrass
people. Other than in private industry, scientists simply have little
reason to do anything that cast doubt on their integrity. No one
is going to pay haefle and keating to produce results one way or
the other. Their best shot at fame would have been to find exactly
the opposite of what they claimed.
> In fact, regarding the coal miner, arguably they are more reliable.
> They certainly have a code of conduct in a mine based on their
> own lives being at stake.
Apart from your subjectivity about "more reliable", you've just
made my point. One's behaviour is influenced by what incentives exist
in behaving a certain way. Scientists have little incentive in being
perceived as dishonest.
Kevin Aylward
> On Sun, 29 Dec 2002, Minor Crank wrote:
>>
>> "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
>> news:lmEP9.135$Hy5....@newsfep3-gui.server.ntli.net...
>>> Minor Crank wrote:
>>>> Correlated rate-change analysis is indeed capable of compensating
>>>> for errors introduced by sporadic quasi-permanent rate changes,
>>>> which is by far the most serious source of error affecting atomic
>>>> clocks.
>>>
>>> What this does though is add yet another assumption, along with
>>> loads of others, which one hopes to be valid.
>
> "assumption?" "one hopes to be valid?"
>
> What a moron.
{snip stuff now assumed to be inaccurate due to above demonstrable false
claim, and completely unrelated to science or the issues involved}
All you have Mr. Speicher is words. Nothing more. Sure, you seem to be a
somewhat competent historian, but I have yet to see any proof that you
can achieve anything practical at all. Talk is very, very cheap, and
certainly your crank belief in a theory about little bitty waves which
is experimentally proven false, tells much about you. If you were in way
shape or a form a scientist, you would keep your baseless personal
comments to yourself.
Feel free to explain how such a stupid individual can write and sell a
program consisting of 70,000+ lines of code, when software is only a
hobby of mine. Feel free to point out any "stupid" mistakes in this
software. Feel free to point out any "stupid" mistakes on my GR site.
http://www.anasoft.co.uk/physics/gr/index.html. Feel free to point out
any "stupid" mistakes on my electronics site
http://www.anasoft.co.uk/EE/index.html
The reality, is that you are nobody with nothing to show and a bigot, in
addition to having little knowledge about how the real world actually
functions. Apparently you live in a thought experiment.
Kevin Aylward
> On Sun, 29 Dec 2002, Minor Crank wrote:
>>
>> "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
>> news:lmEP9.135$Hy5....@newsfep3-gui.server.ntli.net...
>>> Minor Crank wrote:
>>>> Correlated rate-change analysis is indeed capable of compensating
>>>> for errors introduced by sporadic quasi-permanent rate changes,
>>>> which is by far the most serious source of error affecting atomic
>>>> clocks.
>>>
>>> What this does though is add yet another assumption, along with
>>> loads of others, which one hopes to be valid.
>
> "assumption?" "one hopes to be valid?"
>
> What a moron.
In a prior post Mr. Steven "I am a Bigot" Speicher claimed:
**********
"This is the last time I will address you directly, you moron"
**********
As can be plainly seen by the fact that he made a reply to this
particular post in contradiction to his claim, what he really means is
that "This is the last time I will address you directly, if I can not
refute your arguments 100%".
The post that he has indeed failed to reply to is where it was clearly
shown that the HP paper claiming that one could realistically reduce
errors of clocks by a factor of 100+, which Mr. Speicher then claimed
could be applied to atomic clocks not designed by idiots, was elementary
false.
Kevin:
> I saw no evidence
> that this technique would work for atomic clocks.
Mr. Speicher:
Well, of course you "saw no evidence," since you bury your head
in the sand and then wag your tail around as if it were a mind.
The evidence was presented as below:
Kevin:
*******
This is all very much creative writing, it should get an award as such.
The whole set-up is a cheat and can't possible work as a general
technique. If you knew anything about manufacturing electronic devices
this would be obvious. Its so obvious that is pretty amazing that you
cant see the fundamental flaw, being such an expert and all.
The watches in question were not watches with an intrinsic random and
drift error of ~1sec/day. They were much more stable watches that, for
manufacturing reasons, only been calibrated to around 1sec/day. Instead
of performing all this fancy mathematics in order to disguise the truth,
one could have simple removed the back of the watch, slapped a frequency
counter on and twiddled a screw to correctly calibrate it.
We engineers know that there is no such thing as a free lunch, or you
can't get blood out of stone. The only reason the calibration could be
done in this *particular* case is that the watches themselves had an
intrinsic better accuracy than what they were adjusted to in
manufacture. Indeed, I have one of these very cheap watches and it
achieves around 1sec/month, unadjusted. This is a *typical*
manufacturing and design *compromise* that is done all the time for
cheap products. Its not worth the time (pun) and expense to set them
correctly.
However, for expensive atomic clocks there is a snowballs chance in hell
that such a trick can be done. By its very nature, all steps to make it
as accurate as possible will have *already* been done by the vendor,
assuming of course, he is not an idiot. There is no reasonable way you
are going to re-calibrate it. If this could be done, the manufacture
would have already done so and claimed a more accurate clock that he
could sell for more money. Its that simple. The evidence that suggests
that this is indeed the case is given in Kelly's paper:
Table 1. Drift-rates of the clocks (ns per hour)
Clock No 120 361 408 447
Before the Eastward test -4.50 +2.66 -1.78 - 7.16
After the Eastward Test -8.89 +4.38 +3.22 -8.41
Before the Westward test -8.88 +6.89 +4.84 -7.17
After the Westward test -4.56 +3.97 +2.16 -9.42
The *change* in drift rates are of the order of 100%. Assuming the
validity of these figures, it makes a first order correction by assuming
a linear drift completely bogus.
Its simple not reasonable to suggest that a manufacture was as so bloody
daft as to have a decent linear, stable, drift in a finished product,
when the product costs so much, and not take advantage of that fact.
Instruments of this quality, by design, will only have residual *random*
error that can not be realistically calibrated out. The changes in drift
claimed here, are what would be typically expected in a competently
designed, high performance instrument. Comparing such an instrument with
a $10 watch is simply absurd.
***********
So, I await a rebuttal with more substance than "your a moron". Maybe
you could remove your own head from the sand and learn a bit about how
products are manufactured in the real world. Its a no wonder, that as a
rule, physicists are not looked upon for practical engineering matters
in good light, its that wind swept, tufted up hair, no socks, sort of
thing.
Kevin Aylward
> "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
> news:<wmBP9.82$Hy5....@newsfep3-gui.server.ntli.net>...
>> Bob Zombiewoof wrote:
>>> "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
>>>>
>>>> I don't follow exactly what you are saying here. What was the raw
>>>> drift of a single clock on its own. 4 clocks is simply not enough
>>>> of a sample to do any meanfull statistical analysis.
>>>
>>> You aren't doing statistics on the clocks. You're doing statistics
>>> on the time intervals read from four clocks with known statistical
>>> errors.
>>>
>>
>> But the drift of the clocks are not known. Its the accuracy of the
>> clocks that are being questioned. Time intervals are given by the
>> clocks!
>>
>>>
>>>> When I get a wafer back from the fab, my boss tells me to test at
>>>> least 30-40 on several wafers. At $100k per mask set I am sure you
>>>> can apprciate why.
>>>
>>> Because you're doing statistics on the wafers from a given mask
>>> set. At $100k per mask set, you don't want to do statistics on the
>>> mask sets themselves. clock <==> mask set; time interval <==> wafer
>>>
>>
>> The device <==> clock. How accurate is the device, how accurate is
>> the clock.
>
> Wrong comparison as I already pointed out.
I disagree. I want to know the characteristics of the device. I want to
know the characteristics of the clocks.
>
>>> Precision, not accuracy. The actual accuracy is irrelevant, since
>>> it makes little difference if the clock runs fast or slow, so long
>>> as you normalize the data by normalizing the time intervals to a
>>> single common reference.
>>>
>>
>> Ho hum. Its a given that one is discussing variations in drift rates.
>> Obviously, if you have a pure, predictably drift, it is a simple
>> calibration.
>
> Exactly. And it appears that such a calibration is can be done
> simply by taking the difference in drift rates at two different
> points and using the slope of the line in ns/hr/hr.
>
But if you don't have a reference to measure the drift, you don't what
the drift is. Secondly, as far as the atomic clocks go, some of the
drift changes are step changes, so you cant assume a linear shift.
> [...]
>>> The clocks aren't the statistic. The time intervals read from
>>> clocks with known statistical errors are. The use of multiple
>>> clocks is a check on systematic effects, not statistical errors.
>>
>> If by this, you mean that one cannot just average 4 clocks to reduce
>> the error. i.e. all clocks could be in error.
>
> Huh? Look, each clock has a statistical uncertainty. Since the
> measurements from each clock are independent, the uncertainties
> add in quadrature. It's simple statistics. The statistic is
> not 4 measurements. If clock A produces values of X +/- x and
> clock B produces Y +/-y, then the statistics are the weighted
> average of X and Y.
>
If each clock has a *random* *gaussian* uncertainty, than yes, they sum
by squares. This is a probability feature. Suppose 4 clocks are
subjected to the same relative pulse noise, they could all read wrong by
exactly the same amount, averaging won't help you.
The point is that you don't really now if everything is all nice and
neat with only 4 devices. Its not something that is usually quantified.
Its a gut, experience sort of thing. If I have only 4 devices that check
out on a wafer *perfectly*, do you really think that I am going to
believe that I got the design right?. I have been bitten by this once or
twice. You have to be very careful in what you assume. Assume nothing is
the usually the best rule.
>> I should have phrased this a bit clearer. Software and Hardware are
>> worlds apart. My statement of what the statistics is actually done in
>> practice for i.c.s is based on real experience. It typically costs
>> about $1M to develop a product. You just don't ship a product, except
>> exceptionally, without doing and meeting standard tests.
>
> From my own (although somewhat limited experience) with missle
> guidance system radar, I also know that the military does not trust
> the reliability of "commercial products". Every single component
> probably cost 10 times as much to test as it did to design and
> manufacture. Even the solder used was different (80% gold, 20% tin
> and came in 10z rolls). A simple mixer/multiplier [a cavity with
> tuning slugs and some sensistors for temperature stability) required
> a week to test with
I agree, military specification are much more severe. One designs for
the intended application/market. If a missile fails in a silo the shit
really hits the fan, but this don't mean that a lot of work don't go
into commercial products to achieve, say a net 5% failure rate.
>
> [...]
>>
>> Actually, as someone who has been an analogue engineer at the
>> SuperConducting Supercollider for a couple of years, this is
>> debatable. It was great. I got to buy *anything* I wanted in terms
>> of equipment. I can assure you, to all intents an purposes, compared
>> to industrial company labs, it was a true joy. I remember we bought
>> 10 $5,000 HP meters, no way could I get even one of those at other
>> places I have worked.
>
> The supercollider was a political project which had a lot more
> to do with politics than science. You'll notice that it never was
> completed. If you perform a search for "radio frequency quadrupole"
> and weapons, you might discover some additional reasons the project
> was funded well initially and then allowed to crater. In general,
> labs do not get funded nearly that well.
>
I do agree in general. The funding for a small sub section of the lab,
is inconsequential to the funding of the whole project, i.e. $10B.
Although, I did point out in another post that there was no such thing
as a free lunch. Well, I sort of take it back. Whilst at the collider,
there was a period where I had about, on average, 3 free lunches a week.
>
>>
>>>
>>> Furthermore, I wouldn't trust the opinion of any doctor regarding
>>> a procedure for which he/she had read only the biased opinions of
>>> others rather than the articles written by those who have performed/
>>> developed the procedure.
>>
>> But these latter article are also just as biased. Its a given fact.
>> Indeed, the authors usually have much more invested.
>
> The experimemtors do _not_ have a vested interest in the outcome of
> an experiment. If anything, their interest would be in discovering
> something that contradicted expectations.
>
I don't really agree here in general. It depends on the situation. Lets
say, a prof is getting funding to research new fuels for a military
rocket. He has to gear his proposal to induce the militry to give him
the funding. You can bet your boots he is going to do his best to give
them what they want. This is a basic feature of the real world.
>
>> What I find disconcerting is the view that scientists are in any way
>> *special* with regard to basic human behaviour.
>
> They aren't, as the recent affair at bell labs demonstrates. The
> difference is that in most cases, there is no reward for biasing one's
> data one way or the other. It's only when one is paid to obtain
> specific results, that the temptation to obtain them is present.
But that is generally the case. Professors live and die by attracting
funding from organisations with the assumption that their research will
benefit them. There is not a lot of "true" "pure" research. most
research is indeed for industrial based applications.
> Usually, in an academic setting, the principal investigators are
> tenured and have little reason to care one way or another what
> anyone thinks beyond the reputation they earn through performing
> experiments that are unassailable on technical merit.
The fact that they can't lose their jobs is not that really significant.
They still need to get funding or they wont have anything to do in their
jobs but push paper and twiddle their thumbs. Invariable, profs seek
reputations, you can a reputation without doing anything. If what you do
does not benefit the sponsor, you won't get the funding.
>There is a
> great incentive to _not_ fiddle with data, since the accusation
> of doing so will pretty much be a career ending event. Look at
> pons and fleischman and the cold fusion fiasco.
>
I agree there is also a grate incentive to not fiddle with data. This is
obvious. But its like this, if you lie once in a while, it can be very
useful if you don't get caught. If you lie all the time, you will be
rejected by everybody, hence short term gains can be offset by long time
deficits. It a selfish gene balancing act.
>> I see no evidence that the average scientist is not going to be
>> just as honest/cheat as the average plumber, dentist, engineer,
>> trucker, footballer, coal miner, politician....
>
> I never said they were less honest. The point is, that they have
> no incentive to be dishonest where their data is concerned. People
> cheat where there exists some incentive to do so. A "footballer"
> has little incentive to play poorly, since he gets paid for playing
> the best he can. If police were paid to uphold the letter of the
> law rather than make arrests, they would have no incentive to harrass
> people. Other than in private industry, scientists simply have little
> reason to do anything that cast doubt on their integrity.
But I claim that the *majority* of research is done for *industry*.
There's 100's of universities, but only a few like the Harvards and
Stanfords which get their money from trees. All the rest battle it out
in the ring. Its money that makes the world go round. The cigarette
companies are prime examples of "creative" science, but its across the
board.
Kevin Aylward
> "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
> news:<wmBP9.82$Hy5....@newsfep3-gui.server.ntli.net>...
>> Bob Zombiewoof wrote:
>>> "Kevin Aylward" <ke...@anasoft.co.uk> wrote in message
>>>>
>>>> I don't follow exactly what you are saying here. What was the raw
>>>> drift of a single clock on its own. 4 clocks is simply not enough
>>>> of a sample to do any meanfull statistical analysis.
>>>
>>> You aren't doing statistics on the clocks. You're doing statistics
>>> on the time intervals read from four clocks with known statistical
>>> errors.
>>>
>>
>> But the drift of the clocks are not known. Its the accuracy of the
>> clocks that are being questioned. Time intervals are given by the
>> clocks!
>>
>>>
>>>> When I get a wafer back from the fab, my boss tells me to test at
>>>> least 30-40 on several wafers. At $100k per mask set I am sure you
>>>> can apprciate why.
>>>
>>> Because you're doing statistics on the wafers from a given mask
>>> set. At $100k per mask set, you don't want to do statistics on the
>>> mask sets themselves. clock <==> mask set; time interval <==> wafer
>>>
>>
>> The device <==> clock. How accurate is the device, how accurate is
>> the clock.
>
> Wrong comparison as I already pointed out.
I disagree. I want to know the characteristics of the device. I want to
know the characteristics of the clocks.
>
>>> Precision, not accuracy. The actual accuracy is irrelevant, since
>>> it makes little difference if the clock runs fast or slow, so long
>>> as you normalize the data by normalizing the time intervals to a
>>> single common reference.
>>>
>>
>> Ho hum. Its a given that one is discussing variations in drift rates.
>> Obviously, if you have a pure, predictably drift, it is a simple
>> calibration.
>
> Exactly. And it appears that such a calibration is can be done
> simply by taking the difference in drift rates at two different
> points and using the slope of the line in ns/hr/hr.
>
But if you don't have a reference to measure the drift, you don't what
the drift is. Secondly, as far as the atomic clocks go, some of the
drift changes are step changes, so you cant assume a linear shift.
> [...]
>>> The clocks aren't the statistic. The time intervals read from
>>> clocks with known statistical errors are. The use of multiple
>>> clocks is a check on systematic effects, not statistical errors.
>>
>> If by this, you mean that one cannot just average 4 clocks to reduce
>> the error. i.e. all clocks could be in error.
>
> Huh? Look, each clock has a statistical uncertainty. Since the
> measurements from each clock are independent, the uncertainties
> add in quadrature. It's simple statistics. The statistic is
> not 4 measurements. If clock A produces values of X +/- x and
> clock B produces Y +/-y, then the statistics are the weighted
> average of X and Y.
>
If each clock has a *random* *gaussian* uncertainty, than yes, they sum
by squares. This is a probability feature. Suppose 4 clocks are
subjected to the same relative pulse noise, they could all read wrong by
exactly the same amount, averaging won't help you.
The point is that you don't really now if everything is all nice and
neat with only 4 devices. Its not something that is usually quantified.
Its a gut, experience sort of thing. If I have only 4 devices that check
out on a wafer *perfectly*, do you really think that I am going to
believe that I got the design right?. I have been bitten by this once or
twice. You have to be very careful in what you assume. Assume nothing is
the usually the best rule.
>> I should have phrased this a bit clearer. Software and Hardware are
>> worlds apart. My statement of what the statistics is actually done in
>> practice for i.c.s is based on real experience. It typically costs
>> about $1M to develop a product. You just don't ship a product, except
>> exceptionally, without doing and meeting standard tests.
>
> From my own (although somewhat limited experience) with missle
> guidance system radar, I also know that the military does not trust
> the reliability of "commercial products". Every single component
> probably cost 10 times as much to test as it did to design and
> manufacture. Even the solder used was different (80% gold, 20% tin
> and came in 10z rolls). A simple mixer/multiplier [a cavity with
> tuning slugs and some sensistors for temperature stability) required
> a week to test with
I agree, military specification are much more severe. One designs for
the intended application/market. If a missile fails in a silo the shit
really hits the fan, but this don't mean that a lot of work don't go
into commercial products to achieve, say a net 5% failure rate.
>
> [...]
>>
>> Actually, as someone who has been an analogue engineer at the
>> SuperConducting Supercollider for a couple of years, this is
>> debatable. It was great. I got to buy *anything* I wanted in terms
>> of equipment. I can assure you, to all intents an purposes, compared
>> to industrial company labs, it was a true joy. I remember we bought
>> 10 $5,000 HP meters, no way could I get even one of those at other
>> places I have worked.
>
> The supercollider was a political project which had a lot more
> to do with politics than science. You'll notice that it never was
> completed. If you perform a search for "radio frequency quadrupole"
> and weapons, you might discover some additional reasons the project
> was funded well initially and then allowed to crater. In general,
> labs do not get funded nearly that well.
>
I do agree in general. The funding for a small sub section of the lab,
is inconsequential to the funding of the whole project, i.e. $10B.
Although, I did point out in another post that there was no such thing
as a free lunch. Well, I sort of take it back. Whilst at the collider,
there was a period where I had about, on average, 3 free lunches a week.
>
>>
>>>
>>> Furthermore, I wouldn't trust the opinion of any doctor regarding
>>> a procedure for which he/she had read only the biased opinions of
>>> others rather than the articles written by those who have performed/
>>> developed the procedure.
>>
>> But these latter article are also just as biased. Its a given fact.
>> Indeed, the authors usually have much more invested.
>
> The experimemtors do _not_ have a vested interest in the outcome of
> an experiment. If anything, their interest would be in discovering
> something that contradicted expectations.
>
I don't really agree here in general. It depends on the situation. Lets
say, a prof is getting funding to research new fuels for a military
rocket. He has to gear his proposal to induce the militry to give him
the funding. You can bet your boots he is going to do his best to give
them what they want. This is a basic feature of the real world.
>
>> What I find disconcerting is the view that scientists are in any way
>> *special* with regard to basic human behaviour.
>
> They aren't, as the recent affair at bell labs demonstrates. The
> difference is that in most cases, there is no reward for biasing one's
> data one way or the other. It's only when one is paid to obtain
> specific results, that the temptation to obtain them is present.
But that is generally the case. Professors live and die by attracting
funding from organisations with the assumption that their research will
benefit them. There is not a lot of "true" "pure" research. most
research is indeed for industrial based applications.
> Usually, in an academic setting, the principal investigators are
> tenured and have little reason to care one way or another what
> anyone thinks beyond the reputation they earn through performing
> experiments that are unassailable on technical merit.
The fact that they can't lose their jobs is not that really significant.
They still need to get funding or they wont have anything to do in their
jobs but push paper and twiddle their thumbs. Invariable, profs seek
reputations, you can a reputation without doing anything. If what you do
does not benefit the sponsor, you won't get the funding.
>There is a
> great incentive to _not_ fiddle with data, since the accusation
> of doing so will pretty much be a career ending event. Look at
> pons and fleischman and the cold fusion fiasco.
>
I agree there is also a grate incentive to not fiddle with data. This is
obvious. But its like this, if you lie once in a while, it can be very
useful if you don't get caught. If you lie all the time, you will be
rejected by everybody, hence short term gains can be offset by long time
deficits. It a selfish gene balancing act.
>> I see no evidence that the average scientist is not going to be
>> just as honest/cheat as the average plumber, dentist, engineer,
>> trucker, footballer, coal miner, politician....
>
> I never said they were less honest. The point is, that they have
> no incentive to be dishonest where their data is concerned. People
> cheat where there exists some incentive to do so. A "footballer"
> has little incentive to play poorly, since he gets paid for playing
> the best he can. If police were paid to uphold the letter of the
> law rather than make arrests, they would have no incentive to harrass
> people. Other than in private industry, scientists simply have little
> reason to do anything that cast doubt on their integrity.
But I claim that the *majority* of research is done for *industry*.
There's 100's of universities, but only a few like the Harvards and
Stanfords which get their money from trees. All the rest battle it out
in the ring. Its money that makes the world go round. The cigarette
companies are prime examples of "creative" science, but its across the
board.
Kevin Aylward
Martin Hogbin
"Kevin Aylward" <ke...@anasoft.co.uk> wrote in message news:I8VP9.1420$I31....@newsfep1-gui.server.ntli.net...
> Stephen Speicher wrote:
>
> Feel free to explain how such a stupid individual can write and sell a
> program consisting of 70,000+ lines of code, when software is only a
> hobby of mine.
Writing 70k lines of code does not qualify a person to criticise a paper
on atomic clocks without reading it.
> Feel free to point out any "stupid" mistakes in this
> software. Feel free to point out any "stupid" mistakes on my GR site.
> http://www.anasoft.co.uk/physics/gr/index.html.
Dirk seems to be doing this.
> Feel free to point out
> any "stupid" mistakes on my electronics site
> http://www.anasoft.co.uk/EE/index.html
Since you ask...
In Bipolar Design Part 1 you say:
'Despite much literature that implies other wise, a transistor
is a voltage controlled device not a current controlled
device'.
At the very least this a unusual use of the term 'voltage
controlled' which needs some explanation. More likely
it shows a serious misunderstanding of how a bipolar
transistor works.
Martin Hogbin
Minor Crank
> Minor Crank wrote:
> > Correlated rate-change analysis is indeed capable of compensating for
> > errors introduced by sporadic quasi-permanent rate changes, which is
> > by far the most serious source of error affecting atomic clocks.
>
> What this does though is add yet another assumption, along with loads of
> others, which one hopes to be valid. Well, may be it is, may be it aint.
> Fundamentally this is an information problem. You cannot extract more
> information than is available. If you decide to add some extra
> information, that is not actually *provable*, to be there, well so be
> it. The numbers themselves don't tell the whole story because they don't
> contain information associated with the assumptions.
>
> For example, what's the probabilities that over time all clocks stepped
> changed together, or that there was some extra non step change drift,
> and green goblins varied the drifts in-between measurements...etc...
I generated some idealized data which my "toy" correlated rate-change
analysis program handles without a hitch.
I have an ensemble of three clocks. Clock 0 initially runs a bit fast,
while clocks 2 and 3 run a bit slow compared with the ensemble mean.
The following table presents, in columnar format, data for the three
clocks. The columns marked "t" show the clock readings, which are
compared with the ensemble means to their right.
Simple exercise:
At a certain time during the run, one of the clocks exhibited a rate
change. Which clock exhibited the rate change, when did the rate
change occur, and did the clock speed up or slow down?
t mean t mean t mean
clock 0 4 3.9650 clock 1 4 4.0052 clock 2 4 4.0294
clock 0 5 4.9564 clock 1 5 5.0065 clock 2 5 5.0367
clock 0 6 5.9478 clock 1 6 6.0079 clock 2 6 6.0441
clock 0 7 6.9392 clock 1 7 7.0093 clock 2 7 7.0514
clock 0 8 7.9306 clock 1 8 8.0106 clock 2 8 8.0588
clock 0 9 8.9220 clock 1 9 9.0120 clock 2 9 9.0661
clock 0 10 9.9134 clock 1 10 10.0132 clock 2 10 10.0732
clock 0 11 10.9081 clock 1 11 11.0129 clock 2 11 11.0789
clock 0 12 11.9028 clock 1 12 12.0126 clock 2 12 12.0845
clock 0 13 12.8976 clock 1 13 13.0123 clock 2 13 13.0902
clock 0 14 13.8923 clock 1 14 14.0120 clock 2 14 14.0959
clock 0 15 14.8870 clock 1 15 15.0117 clock 2 15 15.1015
clock 0 16 15.8817 clock 1 16 16.0114 clock 2 16 16.1072
"Minor Crank"
Minor Crank
"Minor Crank" <blue_whal...@attbi.com> wrote in message
news:40bb2cea.02123...@posting.google.com...
OK, time for the answer, which I worked out by hand to illustrate the
workings of my "toy" correlated rate-analysis program. Please note that the
above is idealized noiseless data. A "real" correlated rate-analysis must
contend with random shot noise making the analysis more difficult, requiring
use of appropriate smoothing and fitting methods.
First, a few spacing dots so those who are still working on this puzzle can
successfully avert their vision.
.
.
.
.
.
.
t Clock 0 analysis Clock 1 analysis Clock 2 analysis
------------------------------------------------------------------------
4 3.9650 4.0052 4.0294
0.9914 1.0013 1.0073
5 4.9564 0.0000 5.0065 0.0001 5.0367 0.0001
0.9914 1.0014 1.0074
6 5.9478 0.0000 6.0079 0.0000 6.0441 -0.0001
0.9914 1.0014 1.0073
7 6.9392 0.0000 7.0093 -0.0001 7.0514 0.0001
0.9914 1.0013 1.0074
8 7.9306 0.0000 8.0106 0.0001 8.0588 -0.0001
0.9914 1.0014 1.0073
9 8.9220 0.0000 9.0120 -0.0002 9.0661 -0.0002
0.9914 1.0012 1.0071
10 9.9134 0.0033 10.0132 -0.0015 10.0732 -0.0014
0.9947 0.9997 1.0057
11 10.9081 0.0000 11.0129 0.0000 11.0789 -0.0001
0.9947 0.9997 1.0056
12 11.9028 0.0001 12.0126 0.0000 12.0845 0.0001
0.9948 0.9997 1.0057
13 12.8976 -0.0001 13.0123 0.0000 13.0902 0.0000
0.9947 0.9997 1.0057
14 13.8923 0.0000 14.0120 0.0000 14.0959 -0.0001
0.9947 0.9997 1.0056
15 14.8870 0.0000 15.0117 0.0000 15.1015 0.0001
0.9947 0.9997 1.0057
16 15.8817 16.0114 16.1072
Analysis of the data showed that Clock 0 experienced a rate slowdown of 0.5%
at hour 9.9, ensemble mean time. This caused the ensemble mean rate to
decrease by 0.17%. The result is that Clock 0 showed a net slowdown of 0.33%
relative to the ensemble mean, while Clocks 1 and 2 both showed a net
speedup of 0.17% relative to the ensemble mean.
It should be apparent that correlated rate-change analysis is in essence a
very simple and straightforward concept. The devil is in the details, of
course, but the USNO had many years to work them out.
Contrary to Kevin's assertions, no attempt is being made to extract more
information from the data than is actually available. The one assumption
that I made in this three clock analysis, is that no two clocks "stepped
together." In a four clock analysis, I can handle this unlikely case.
Given that sporadic quasi-permanent rate changes are the major source of
error afflicting atomic clocks, it should be apparent how correlated
rate-change analysis applied to a small ensemble of clocks can improve the
precision of the ensemble by far more than the sqrt(n) factor that one can
achieve using a simple mean.
"Minor Crank"
Stephen Speicher
>
> Contrary to Kevin's assertions, no attempt is being made to extract more
> information from the data than is actually available. The one assumption
> that I made in this three clock analysis, is that no two clocks "stepped
> together." In a four clock analysis, I can handle this unlikely case.
>
Nicely done.
Bob Zombiewoof
> I disagree. I want to know the characteristics of the device. I want to
> know the characteristics of the clocks.
>
The statistical uncertainty is the statistical error in each clock.
The deviations from the "true" value is a systematic uncertainty.
It's not possible to perform a statistics on the clocks and obtain
a "better" value because the systematic error is not statistical.
It's only possible to perform statistics on the intervals each
clock measures. One can easily have a very precise result which
is not accurate or a very accurate result which is not precise.
Precision refers to the accumulated statistical uncertainty from
each trial. Accuracy refers to how close the result is to the
"true value" In this case, they avoid a substantial inaccuracy
by taking the difference between the two flights in opposite
directions, so that the drift rates don't depend on the drift
rates in other clocks. The uncertainties of the differences
add in quadrature.
[...]
> But if you don't have a reference to measure the drift, you don't what
> the drift is. Secondly, as far as the atomic clocks go, some of the
> drift changes are step changes, so you cant assume a linear shift.
If they didn't have a reference, they could not have stated the
drift rates _at_all_. The assumption of a linear shift in the
drift rates is not totally unreasonable, although one could argue
for a higher order fitting function, depending upon the number
of data points. In any case, data is binned, so _all_ variations
are "steps". If you have two points, you are stuck with a line.
Even if you have 100 points, the only way to determine the "correct"
fit of the drift rate is to either (1) know, a priori, what the
functional form of the drift rate is, and/or (2) analyze the data
and make the best of it. Few, if any experiments have the luxury
of (1), or else the experimentor would correct for it prior to
performing the analysis.
>
> If each clock has a *random* *gaussian* uncertainty, than yes, they sum
> by squares. This is a probability feature.
Each clock _does_ have a random gaussian uncertainty given in
ns/Hr.
> Suppose 4 clocks are subjected to the same relative pulse noise,
> they could all read wrong by exactly the same amount, averaging
> won't help you.
Correct. That's a systematic error. It affects the overall
accuracy of the measurement, but it only affects the precision
through the number of time intervals measured by each clock.
The random noise pulse is not a statistical feature of the clock.
>
> The point is that you don't really now if everything is all nice and
> neat with only 4 devices.
By your own argument, you are no better off as far as accuracy
goes with a million clocks, since the same random noise pulse
as you descibe above will cause all 1 million clocks to be wrong
by the same amount. The uncertainty in the measurement will
be excellemt, since you'll add the same errors in quadrature,
but your accuracy won't be any better than had you used a single
clock.
> Its not something that is usually quantified.
But it is quantified.
> Its a gut, experience sort of thing. If I have only 4 devices
> that check out on a wafer *perfectly*, do you really think that
> I am going to believe that I got the design right?.
Of course not, any more than you would choose the "best"
set of masks from among 100 mask sets based upon a sample of
4 devices from each mask. 4 devices is the statistic associated
with the mask set, but it can't tell you the accuracy with which
a given mask set can produce wafers.
> I have been bitten by this once or twice. You have to be very
> careful in what you assume. Assume nothing is the usually the
> best rule.
You're missing the point. You check a lot of wafers from
a single set of masks in order to develop a _single_ set
which gives you the statstics you want and you _assume_
from the statistics, that the mask set will accurately
reproduce wafers with the same statistics. If you find some
variability (drift) in a periodic sample of 30 wafers,
every 10,000 wafers, or so, you don't redesign the mask set,
look for some systematic effect elsewhere in the process.
> I agree, military specification are much more severe. One
> designs for the intended application/market. If a missile
> fails in a silo the shit really hits the fan, but this don't
> mean that a lot of work don't go into commercial products to
> achieve, say a net 5% failure rate.
Since experiments are not done 1000 times, you have to quantify
the errors with the data you get. I designed a high pressure
(50 bar) hydrogen gas scattering chamber which consisted of a
6.00 cm diameter x 30 cm long x 1 mm Al tube, which for experi-
menatl reasons had to be concentric to about 0.01 mm, and so it
had to be ground from a rather large, heavy walled tube for a
large sum of cash. I didn't have the option of doing any statistics
to discover whether it would explode when pressurized, since I
couldn't have built another one. I had lots of opposition from
the safety engineer (who eventually lost out when taking his
argument to where the buck stopped). Sometimes you just can't
test 30-40 samples and have to live with what you have to work
with.
[...]
> I do agree in general. The funding for a small sub section of
> the lab, is inconsequential to the funding of the whole project,
> i.e. $10B.
My main point is that I suspect the initial phase of the
supercollider involved some research for reasons that went
well beyond accelerating particles in the collider. An RFQ
makes a nice injector of bunched protons, but it also makes
a rather nice particle beam for less scientific uses. Ditto
for the "gigatron".
[...]
> I don't really agree here in general. It depends on the situation.
Of course. However, we're discussing an experiment for which
no such incentive exists.
> Lets say, a prof is getting funding to research new fuels for
> a military rocket. He has to gear his proposal to induce the
> militry to give him the funding.
ALL proposals are geared toward inducing some organization
to fund it. However, what you are talking about is similar
to a facilities proposal, where the science is basically
understood well enough to at least know that there is a good
chance of building something that does what one expects.
However, the science for which the facility is ostensibly
being built is not known (or else the inducement to build
the facility would disappear). A proposal for purely scientific
research would never garauntee a result, beyond the expectation
that the funding is sufficient to address the scientific question.
> You can bet your boots he is going to do his best to give
> them what they want. This is a basic feature of the real world.
Not necessarily. (Again from personal experience, since my office
was across the hall from the person in question, and whom I knew
fairly well). During the cold fusion fiasco, companies were
throwing money (and lots of it by academic standards) at anyone
willing to run a cold fusion experiment so that they could "cash
in" on whatever patents they could weasel out of the funding
arrangement. The person in question ran such an experiment for
about 2 years. He received enough money that he ended up with
more electronics (amplifiers, cfd's, adc's tdc's logic boxes, etc)
than the lab itself owned. Now, no one else ever believed there
was a chance in hell of ever seeing any cold fusion, and I suspect
neither did he, since he never would say anything definite when I
made jokes to him about it (which was just about every time I
talked to him). Needless to say, there was never any cold fusion,
and never did he report any (although he did get mentioned in
taube's book). Now, he certainly came out ahead, since it
required essentially zero effort to let the thing just run, but
it certainly isn't how one should obtain funding. The point being,
that everyone else in the lab had similar opportunities, but didn't
think there was any wothwhile science there. Oddly enough, the
cranks will probably view such opportunism as "dissident science".
In a more serious case, there was an incident in which the faculty
(of essentially the entire college of science) went to a lot of
effort to remove, not just a tenured professor, but a distinguished
profesor, of which there were very few (< 30 out of _all_ of
the faculty, probably a couple thousand, total). Basically,
no one wants to be associated with an institution that has someone
trying to turn mecury into gold by having graduate students set
off explosions, regardless of who is gullible enough to pay for
the promise of a result.
> > They aren't, as the recent affair at bell labs demonstrates. The
> > difference is that in most cases, there is no reward for biasing one's
> > data one way or the other. It's only when one is paid to obtain
> > specific results, that the temptation to obtain them is present.
>
> But that is generally the case. Professors live and die by
> attracting funding from organisations with the assumption
> that their research will benefit them.
This has not been a major problem in physics, yet, since
most of the basic research is funded by organisations like
the doe, the nsf, the a.p. sloan foundation, etc. On the
other hand, the problem _has_ appeared in biology, where
the money for purely scientific research has been cut in
favor of corporate funding. In particular, berkeley received
a large grant from novartis pharmaceuticals in exchange for
patent rights a couple of years ago, which received a lot
of criticism.
The issue of financial reward for producing specific results
in physics was just addressed, again, in physics today,
http://www.aip.org/pt/vol-55/iss-12/p10.html
and the lucent issue the month before:
http://www.physicstoday.org/vol-55/iss-11/p15.html
> There is not a lot of "true" "pure" research. most research is
> indeed for industrial based applications.
Last I checked, no one really cared enough to offer money to
measure the lifetime of an isotope, a branching ratio, or a
validation of relativity. Most physics research is basic research.
> > Usually, in an academic setting, the principal investigators are
> > tenured and have little reason to care one way or another what
> > anyone thinks beyond the reputation they earn through performing
> > experiments that are unassailable on technical merit.
>
> The fact that they can't lose their jobs is not that really
At the moment, funding doesn't primarily originate with
corporations. When it does, you can expect ethics to go out
the window. However we are talking about research for which
no one has a vested interest in the outcome. Haefele and keating
did not have a bonus awaiting their result.
> But I claim that the *majority* of research is done for *industry*.
But, I claim that is not the case in physics. That may very
well be the case in engineering, chemistry or biology, but
industry has little use for general relativity or the standard
model to be validated.
> There's 100's of universities, but only a few like the Harvards and
> Stanfords which get their money from trees.
Still,the bulk of the funding for physics comes from doe grants
or from other non-profit organizations. Since we are discussing
physics, I'm not interested in what the engineering or other
departments do.
> All the rest battle it out
> in the ring. Its money that makes the world go round. The cigarette
> companies are prime examples of "creative" science, but its across the
> board.
All the more reason to _not_ allow private funding at public
universities unless all of the research is returned to the public
that supports the universities and all of the intellectual property
rights are owned by the public. Private universities are entitled
to do what they want, but I think the issue of property rights
would drive out better scientists, since their prestige depends
upon their work being widely disseminated.
Eric Baird
OK, in this sort of situation, where you have two camps, the reason
for the deadlock is sometimes that the real solution is a third answer
that is equally unpalatable to both groups.
My reading of the situation is this:
GROUP 1
... has historically claimed that H&K was the key experiment that
proved that SR's time dilation effect was real, and has tended to
consider that anyone not accepting this was kinda slow or ignorant, or
just plain perverse.
GROUP 2
... has historically been a bunch of people who were unconvinced by
the argument that a supposedly asymmetrical "unrelativistic" result
could correctly be used to prove a theory of relativity. Something
just didn't smell right, if "core SR" was derived by assuming
completely flat spacetime, and was considered bombproof because you
could run the same arguments in multiple ways to get the same result,
how safe was it to extend it to a case where general arguments say
that spacetime is not flat, get an asymmetrical result, and then say
that the calculations only come out "right" when we select one special
set of physical observers that are taking part in the experiment? How
legal is it to have a two-observer problem and say that the theory
works for observer A, but observer B must be forbidden from having an
opinion, because it might disagree?
These guys have tended to say that H&K actually /dis/proves special
relativity.
The problem here is that you can make a decent case for either
argument, but usually only by saying "My argument is consistent, so
the points from the other camp that don't fit can safely be discarded,
because if our argument is correct, theirs /has/ to be wrong"
====================
So here's a viewpoint that probably alienates both camps.
<personal opinion mode on>
GROUP 1 were wrong to say that this was ever a significant test of
special relativity.
It hasn't been that since Einstein's 1911 paper (deriving
gravitational time dilation from general principles, without using
SR). If you accept the idea of a light-metric, or wave/particle
duality (Newton) you can expect objects deeper in a gravitational
field to age more slowly, and then you can calculate an the same basic
effect for centrifuged or circling objects through the principle of
equivalence (or by invoking a Coriolis field).
So (as someone else has just pointed out) we could expect this result
even we'd never heard of special relativity.
There have been various types of test where the physics community have
made badly inflated claims for the significance of particular
experiments, and H&K seems like another one. Group 2 were right to
sense that they were being bullshitted, and complain.
... but GROUP 2 were also on wobbly ground when they said that the
test disproved special relativity (without introducing more involved
arguments). Just because your opponent is caught doing naughty stuff
to back up their argument, it doesn't neccessarily mean that their
favourite theory is wrong. Their theory might have been badly
misrepresented and mis-sold, but that could be more the fault of the
theory's proponents than the fault of the core logic itself (eg SR's
supposed "photographable" predictions before Terrell and Penrose).
While Group2 could also claim that the basic methodology was flawed
because it didn't take into account the existence of a more general
prediction (trying to isolate an "SR" effect without asking whether it
was distinguishable from the more general effect), and could argue
that the extension of core SR to acceleration problems had not been
proved to be logically sound, that didn't easily show that the SR core
equations themselves were invalidated. There's an additional step
between saying that the extensions to SR are illegal or untrustworthy,
being able to track back and say that the whole thing is wrong. You
might intuitively feel that H&K result should destroy the whole
structure, but actually isolating a breakpoint is not easy. I think
the mistake that people in Group 2 made was believing too much of what
they were told by the Group1-ers as to what SR should or shouldn't
say, you can disprove an extension without neccessarily disproving the
foundation structure (although it's a good start).
I notice that more mainstream guys on the group are now shifting to
the position that H&K never was claimed to be a proof of special
relativity <!>, and that shift kinda pulls the rug out from under the
Group2-ers. It acknowledges that their misgivings may have been
justified, but avoids giving them credit by saying that some of the
inflated claims that they were arguing with were never actually made.
<ouch!>
=====================
I think the next step for the Group2-ers is to advance into this new
territory and apply some logical judo.
Accepting that the H&K result itself is correct, and seems to be
correct even if we don't believe SR, and that "extended SR" does give
the right answers, is there then a problem here that the extended
theory is making predictions that are actually _too_ good? :-)
One could argue that given the strict (and somewhat unrealistic)
physical conditions imposed by SR, that perhaps if it was a true
building-block of a larger "true" theory, perhaps SR really /should/
have failed in the case of a centrifuged clock ... perhaps the fact
that it /didn't/ fail means that SR might be skimming the surface of
various correct relationships without actually being the genuine
explanation ... and it might also touch on other results that are not
so correct.
If a structure supplies good answers in a realm where it has no right
to be able to this, one sometimes suspects that it is a shortcut
method and not the "proper" calculations -- the correct answer might
have been a null answer.
When the guys behind the first laboratory "centrifuged clock" test
published their results, they scrupulously pointed out that you could
get the same answers either by using SR or by applying the
gravitational arguments ... but then the "gravitational" argument
seemed to get discarded and forgotten -- presumably because of that
apparent logical conflict with SR.
<faint sound of possible alarm bells ringing in the distance>
After some initial controversy, it seems that experts realised that
the "extended SR" and "gravitational" arguments weren't actually
dual, and realised that this suggested that at least one of the two
descriptions therefore had to be inappropriate or "wrong" -- if you
describe the experiment in terms of flat spacetime, straight
lightbeams and velocity-dependent time dilation, and get the right
result, there's no effect left to explain using acceleration and
curvature effects. On the other hand, if we use the gravitational
approach, the "flat" description has to be wrong, because if the
region really was flat, there would be no absolute clock-difference --
the entire final clock-difference effect is attributable to curvature
(/acceleration), and there's no room left for a residual
velocity-based effect like the one applied by extended SR.
Through a deep quirk of the geometry, you can get the same final;
numerical answers both ways, but at least one of these two methods has
to be getting the right answers by applying the wrong physical model.
So, if we took the more general equivalence principle arguments to be
the "reliable" ones, and also accepted that apparent lack of duality
with extended SR, what would the consequences be?
Firstly, we might lose confidence in the validity of more extreme
"extended SR" results like the "abrupt turnaround acceleration"
version of the twins problem, until someone could produce a consistent
gravitational-domain description of the problem and show that the two
methods do still agree. For that particular example, nobody yet seems
to have managed to do a "gravitational" version that agrees
consistently with the "extended SR" predictions, and there doesn't
seem to be any usable experimental data, so it /might/ still be
conceivable that the "abrupt distant acceleration" predictions of
extended SR are wrong ... SR-critics' unhappiness about this
prediction /might/ yet turn out to be justified.
If "extended SR" does turn out to be unreliable, and we decide that it
manages to work as well as it does in centrifuge-class tests because
of some happy mathematical coincidences (Lorentz relationships also
show up when we do projections of curved surfaces onto a flat
background), then if the SR core equations and core logic are still to
be regarded as fundamentally correct, we probably have to find some
way of imposing a logical firewall between the "extended" and "core"
implementations of SR, so that we can say that the stuff on _that_
side of the wall is allowed to be wrong without affecting the stuff on
_this_ side.
But some pro-SR people argue that there isn't a wall, and that the
validity of the "extended" arguments is the same as that of the "core"
arguments, which would tend to mean (if you believed it) that if the
"extended SR" stuff was fundamentally incompatible with a full
gravitational theory, SR's core arguments would be fundamentally
incompatible too ...
<personal opinion mode off>
=Erk= (Eric Baird)
(I'm not supposed to be here.)
(In fact, I'm not here.)
(If you think you saw me here, you imagined it.)