Regarding the null test. Aren't we pretty sure that a cylindrical resonator will not thrust? It will heat and do all the other things, but if something is escaping, that something should escape symmetrically, no thrust. If nothing is escaping then still no thrust due to symmetry.
Of course getting the heating rates right may be a little tricky when it comes to comparing to a real EM Drive configuration.
I do think the idea posted much earlier by (I forget who posted it), the idea of two identical frustums, one at each end of a balance beam, would go a long way toward eliminating thermal lift from the data. The idea was to mount both frustums upward, run to steady state then turn one off. Or mount one up, one down and run synchronously, doubling the thrust effect while significantly reducing the lift effect. This one would allow data collection from initial power on of the cold system and still reduce the effect of thermal lift.
Only problem is that it takes two frustums, but rfmwguy can tell us how much difficulty and cost would be involved in building the second identical frustum. And he could likely scope out the required modifications to the test rig as well.
And the difficulty with the word "identical" meaning identical fabrication, feeding, power coupled etc. Not easy. Do-able, but not easy.
H
@glennfish, attached is the original ft 2d spreadsheet with data channels 2-4 (columns b-d) unhidden. Its in the original file as well, just hidden.
Note on these hidden data channels: The data inputs are unloaded, meaning nothing is attached to them, unlike channel 1 (column a on the spreadsheet), the LDS voltage input. This might be useful in looking at system noise in mag ON/OFF conditions.
While I cannot quantify the randomness, it will be the most variable (very sensitive) since there is no load (resistor) on the inputs. If mag ON imparts system noise, these 3 channels and over 8,100 data points will be the most sensitive to noise and will show it.
Another note. The balanced impedance of channel 1 is 470 ohms. Data on channels 2-4 (columns b-c) are open (infinite impedance by comparison) and are not directly equivalent to channel 1. However general deviation comparisons to mag ON/OFF should give you a relative indication of any system noise attributable to EMI, if present.
Thank you. Now ponder this with me (before I downloaded your data):
1. The data trend overall is a negative slope.
2. You suggested in the last 24 hours that -- the turn-on time to resonance could be seconds
3. So what I did is for each group (on or off), asked, "was the first half of the slope > the 2nd half of the slope?" i.e. was there a slope change from negative to more negative about midpoint in the interval?
Preliminary results
For off groups, that was true 14 out of 47 times
For on groups, that was true 29 out of 47 times
For results, Fisher's exact test for those wanting to learn stats...
I have to check my work a few times, and when I do, I'll upload the analysis.
I've confused myself. Would this result be a good thing or a bad thing?
Regarding the null test. Aren't we pretty sure that a cylindrical resonator will not thrust? It will heat and do all the other things, but if something is escaping, that something should escape symmetrically, no thrust. If nothing is escaping then still no thrust due to symmetry.
Of course getting the heating rates right may be a little tricky when it comes to comparing to a real EM Drive configuration.
I do think the idea posted much earlier by (I forget who posted it), the idea of two identical frustums, one at each end of a balance beam, would go a long way toward eliminating thermal lift from the data. The idea was to mount both frustums upward, run to steady state then turn one off. Or mount one up, one down and run synchronously, doubling the thrust effect while significantly reducing the lift effect. This one would allow data collection from initial power on of the cold system and still reduce the effect of thermal lift.
Only problem is that it takes two frustums, but rfmwguy can tell us how much difficulty and cost would be involved in building the second identical frustum. And he could likely scope out the required modifications to the test rig as well.
It would be quite the challenge to build 2 identical units, mechanically and electrically. Even if you obtained a matched pair of mags, small mechanic differences in the frustum could negate a match. Then there's the thermal lift variables at two different places...opposite ends of frustum.
One thing I should have done is remove the frustum from balance beam and hang dead weight in its place. Fire it up and check for system noise afterwards. Problem with this is where to put voltage wires...on the beam or off. Either way, not an easy way to quantify system noise.
Regarding thermal lift, only way I know how to negate this is in a vacuum. I think glennfish's plan of overlaying on/off data is the best answer I've seen so far for us airbreathers.
Anyway – to the request.
...
• Data which should/can be measured for either signal or noise/error determinations
As one of the data geeks here, one thing that's become apparant is that the data should be as complete and sharable as possible. The limited analysis I've done has depended on multiple other people massaging the data before I touch it. I'd like to propose some sort of data interchange standard, but I don't know squat about what we should be measuring. I don't know how fast you should sample, but as a rule of thumb, sample as fast as you can for as long as you can.
In general, include for each sample
1. time stamp (maximum possible accuracy... microsecond resolution would be cool)
2. state of the device (on/off)
3. other states, i.e. forward, backward, as many states as you can collect
...
4. measurement 1
5. measurement 2 as many simultaneous measurements as you can grab
... etc
CSV format is ideal, it works with almost everything
xls format is ok, most of us can work with that
Outstanding idea. Likewise I would also suggest to naming data files or other data artifacts in a unique and unambiguous way and record somewhere a detailed description of your test configuration with the same name or description.
In my former life we tended to use something like "PROJ-TEST-TESTNUMBER-DATE/TIME" Where PROJ was a unique project name, TEST was a unique name for the overall test configuration being used, TESTNUMBER was a sequential number of tests in the TEST configuration or better yet a unique number in the whole PROJ area, and DATE/TIME were just that - a unique date and time stamp of the start or stop of the test - whatever your datalogger will support. It seems very pedantic and like a lot of extra work but 3 months later ( or 3 days sometimes) you will be able to unambiguously know which test generated each data set.
Somewhere in my archives I have a web based testing software that does all that and more. It was developed to datamine test and inject for huge cellular test labs. The company division it was developed for no longer exists and the company has been diced and sliced many times. There were 2 versions of it one in C++ and one in Java. I don't think I can release it but could pull out the key information. It was data store independent with all data available in CSV ascii. All tests recorded configuration, state, statuses, tester(s), labs, devices and gigabytes of data, GPS derived timestamps on all data and the like. You could configure, control, run and access data from any lab on the web via your browser. Informics, Oracle, MySql, Microsoft SQL and Postgres were supported database formats. It might be interesting to see something like that for distributed research like this.
Another option and one more would be familiar with is the testing software framework used by the huge national and international collider labs.
@glennfish, attached is the original ft 2d spreadsheet with data channels 2-4 (columns b-d) unhidden. Its in the original file as well, just hidden.
Note on these hidden data channels: The data inputs are unloaded, meaning nothing is attached to them, unlike channel 1 (column a on the spreadsheet), the LDS voltage input. This might be useful in looking at system noise in mag ON/OFF conditions.
While I cannot quantify the randomness, it will be the most variable (very sensitive) since there is no load (resistor) on the inputs. If mag ON imparts system noise, these 3 channels and over 8,100 data points will be the most sensitive to noise and will show it.
Another note. The balanced impedance of channel 1 is 470 ohms. Data on channels 2-4 (columns b-c) are open (infinite impedance by comparison) and are not directly equivalent to channel 1. However general deviation comparisons to mag ON/OFF should give you a relative indication of any system noise attributable to EMI, if present.
Thank you. Now ponder this with me (before I downloaded your data):
1. The data trend overall is a negative slope.
2. You suggested in the last 24 hours that -- the turn-on time to resonance could be seconds
3. So what I did is for each group (on or off), asked, "was the first half of the slope > the 2nd half of the slope?" i.e. was there a slope change from negative to more negative about midpoint in the interval?
Preliminary results
For off groups, that was true 14 out of 47 times
For on groups, that was true 29 out of 47 times
For results, Fisher's exact test for those wanting to learn stats...
I have to check my work a few times, and when I do, I'll upload the analysis.
I've confused myself. Would this result be a good thing or a bad thing?
Its a good thing, as it confirms the hypothesis that ON does not provide instantaneous results, as in a square-wave voltage change. Mag heating, frequency and resonance shift all combine (my belief) for variances shown in the points you made above. On the other hand, mag OFF is rather calm and predictable, with mother nature taking over and gently thermal lifting the DUT.
Not sure if the 50% point is the best point to sub-divide, but the ON state variances versus the OFF state variances is a decent way to look at the data subset.
@glennfish, attached is the original ft 2d spreadsheet with data channels 2-4 (columns b-d) unhidden. Its in the original file as well, just hidden.
Note on these hidden data channels: The data inputs are unloaded, meaning nothing is attached to them, unlike channel 1 (column a on the spreadsheet), the LDS voltage input. This might be useful in looking at system noise in mag ON/OFF conditions.
While I cannot quantify the randomness, it will be the most variable (very sensitive) since there is no load (resistor) on the inputs. If mag ON imparts system noise, these 3 channels and over 8,100 data points will be the most sensitive to noise and will show it.
Another note. The balanced impedance of channel 1 is 470 ohms. Data on channels 2-4 (columns b-c) are open (infinite impedance by comparison) and are not directly equivalent to channel 1. However general deviation comparisons to mag ON/OFF should give you a relative indication of any system noise attributable to EMI, if present.
Thank you. Now ponder this with me (before I downloaded your data):
1. The data trend overall is a negative slope.
2. You suggested in the last 24 hours that -- the turn-on time to resonance could be seconds
3. So what I did is for each group (on or off), asked, "was the first half of the slope > the 2nd half of the slope?" i.e. was there a slope change from negative to more negative about midpoint in the interval?
Preliminary results
For off groups, that was true 14 out of 47 times
For on groups, that was true 29 out of 47 times
For results, Fisher's exact test for those wanting to learn stats...
I have to check my work a few times, and when I do, I'll upload the analysis.
I've confused myself. Would this result be a good thing or a bad thing?
Its a good thing, as it confirms the hypothesis that ON does not provide instantaneous results, as in a square-wave voltage change. Mag heating, frequency and resonance shift all combine (my belief) for variances shown in the points you made above. On the other hand, mag OFF is rather calm and predictable, with mother nature taking over and gently thermal lifting the DUT.
Not sure if the 50% point is the best point to sub-divide, but the ON state variances versus the OFF state variances is a decent way to look at the data subset.
OK spreadsheet with analysis (no VBA)
Sheet: raw.php
raw data and processing
Sheet: Sheet1
summary & results
m1>m2 m1<=m2
Group ON 29 18
Group OFF 14 33
Fisher's exact test, two tailed p = .0035
Fisher's exact test, one tailed p = .0018
@glennfish, attached is the original ft 2d spreadsheet with data channels 2-4 (columns b-d) unhidden. Its in the original file as well, just hidden.
Note on these hidden data channels: The data inputs are unloaded, meaning nothing is attached to them, unlike channel 1 (column a on the spreadsheet), the LDS voltage input. This might be useful in looking at system noise in mag ON/OFF conditions.
While I cannot quantify the randomness, it will be the most variable (very sensitive) since there is no load (resistor) on the inputs. If mag ON imparts system noise, these 3 channels and over 8,100 data points will be the most sensitive to noise and will show it.
Another note. The balanced impedance of channel 1 is 470 ohms. Data on channels 2-4 (columns b-c) are open (infinite impedance by comparison) and are not directly equivalent to channel 1. However general deviation comparisons to mag ON/OFF should give you a relative indication of any system noise attributable to EMI, if present.
Thank you. Now ponder this with me (before I downloaded your data):
1. The data trend overall is a negative slope.
2. You suggested in the last 24 hours that -- the turn-on time to resonance could be seconds
3. So what I did is for each group (on or off), asked, "was the first half of the slope > the 2nd half of the slope?" i.e. was there a slope change from negative to more negative about midpoint in the interval?
Preliminary results
For off groups, that was true 14 out of 47 times
For on groups, that was true 29 out of 47 times
For results, Fisher's exact test for those wanting to learn stats...
I have to check my work a few times, and when I do, I'll upload the analysis.
I've confused myself. Would this result be a good thing or a bad thing?
Its a good thing, as it confirms the hypothesis that ON does not provide instantaneous results, as in a square-wave voltage change. Mag heating, frequency and resonance shift all combine (my belief) for variances shown in the points you made above. On the other hand, mag OFF is rather calm and predictable, with mother nature taking over and gently thermal lifting the DUT.
Not sure if the 50% point is the best point to sub-divide, but the ON state variances versus the OFF state variances is a decent way to look at the data subset.
OK spreadsheet with analysis (no VBA)
Sheet: raw.php
raw data and processing
Sheet: Sheet1
summary & results
m1>m2 m1<=m2
Group ON 29 18
Group OFF 14 33
Fisher's exact test, two tailed p = .0035
Fisher's exact test, one tailed p = .0018
Nice glenn, for us stat-light posters out here, what does the data tell you?
stat-light... sounds like a beer
OK, there is a
very significant difference between the slope behavior for
on vs.
off, given by the values of p. Close to 1 is bad. Close to 0 is good.
What this indicates is:
If ON, the slope is likely to become more negative during the cycle.
If OFF, the slope is likely to not change or become less negative during the cycle.
examples
define
m1 as the slope during the 1st half of the cycle and
m2 as the slope during the last half of the cycle.
if m1 = -2 and m2 = -3 then m1>m2 = TRUE
if m1 = -2 and m2 = -1 then m1>m2 = FALSE
When doing tests with multiple sensors and possibly multiple platforms timing quickly becomes an engineering challenge. Enclosed is a timing paper I had commissioned, around March 2000, so that our test tools could adequately deal with measurement. It was quite useful to us.
@glennfish, attached is the original ft 2d spreadsheet with data channels 2-4 (columns b-d) unhidden. Its in the original file as well, just hidden.
Note on these hidden data channels: The data inputs are unloaded, meaning nothing is attached to them, unlike channel 1 (column a on the spreadsheet), the LDS voltage input. This might be useful in looking at system noise in mag ON/OFF conditions.
While I cannot quantify the randomness, it will be the most variable (very sensitive) since there is no load (resistor) on the inputs. If mag ON imparts system noise, these 3 channels and over 8,100 data points will be the most sensitive to noise and will show it.
Another note. The balanced impedance of channel 1 is 470 ohms. Data on channels 2-4 (columns b-c) are open (infinite impedance by comparison) and are not directly equivalent to channel 1. However general deviation comparisons to mag ON/OFF should give you a relative indication of any system noise attributable to EMI, if present.
Thank you. Now ponder this with me (before I downloaded your data):
1. The data trend overall is a negative slope.
2. You suggested in the last 24 hours that -- the turn-on time to resonance could be seconds
3. So what I did is for each group (on or off), asked, "was the first half of the slope > the 2nd half of the slope?" i.e. was there a slope change from negative to more negative about midpoint in the interval?
Preliminary results
For off groups, that was true 14 out of 47 times
For on groups, that was true 29 out of 47 times
For results, Fisher's exact test for those wanting to learn stats...
I have to check my work a few times, and when I do, I'll upload the analysis.
I've confused myself. Would this result be a good thing or a bad thing?
Its a good thing, as it confirms the hypothesis that ON does not provide instantaneous results, as in a square-wave voltage change. Mag heating, frequency and resonance shift all combine (my belief) for variances shown in the points you made above. On the other hand, mag OFF is rather calm and predictable, with mother nature taking over and gently thermal lifting the DUT.
Not sure if the 50% point is the best point to sub-divide, but the ON state variances versus the OFF state variances is a decent way to look at the data subset.
OK spreadsheet with analysis (no VBA)
Sheet: raw.php
raw data and processing
Sheet: Sheet1
summary & results
m1>m2 m1<=m2
Group ON 29 18
Group OFF 14 33
Fisher's exact test, two tailed p = .0035
Fisher's exact test, one tailed p = .0018
Nice glenn, for us stat-light posters out here, what does the data tell you?
stat-light... sounds like a beer
OK, there is a very significant difference between the slope behavior for on vs. off, given by the values of p. Close to 1 is bad. Close to 0 is good.
What this indicates is:
If ON, the slope is likely to become more negative during the cycle.
If OFF, the slope is likely to not change or become less negative during the cycle.
examples
define m1 as the slope during the 1st half of the cycle and m2 as the slope during the last half of the cycle.
if m1 = -2 and m2 = -3 then m1>m2 = TRUE
if m1 = -2 and m2 = -1 then m1>m2 = FALSE
Ok, more confirmation of ON cycle changes. Looks like all roads leading to the conclusion that something is definitely happening, i.e. not a null test...positive not null. Fair statement?
I've been OCD like on not calling this effect thrust, and guess I will continue to do so. Without a true understanding of why this happens, it might as well be an attractive or repulsive force. Not 100% comfortable with thrust, so my shorthand is emdrive effect.
Guess a last piece of the puzzle is the open channel data points for relative system noise comparisons betwee on/off states. A quick glance, saw nothing significant in the numbers indicating an EMI induced spike or noise. Let me know if you concur.
@glennfish, attached is the original ft 2d spreadsheet with data channels 2-4 (columns b-d) unhidden. Its in the original file as well, just hidden.
Note on these hidden data channels: The data inputs are unloaded, meaning nothing is attached to them, unlike channel 1 (column a on the spreadsheet), the LDS voltage input. This might be useful in looking at system noise in mag ON/OFF conditions.
While I cannot quantify the randomness, it will be the most variable (very sensitive) since there is no load (resistor) on the inputs. If mag ON imparts system noise, these 3 channels and over 8,100 data points will be the most sensitive to noise and will show it.
Another note. The balanced impedance of channel 1 is 470 ohms. Data on channels 2-4 (columns b-c) are open (infinite impedance by comparison) and are not directly equivalent to channel 1. However general deviation comparisons to mag ON/OFF should give you a relative indication of any system noise attributable to EMI, if present.
Thank you. Now ponder this with me (before I downloaded your data):
1. The data trend overall is a negative slope.
2. You suggested in the last 24 hours that -- the turn-on time to resonance could be seconds
3. So what I did is for each group (on or off), asked, "was the first half of the slope > the 2nd half of the slope?" i.e. was there a slope change from negative to more negative about midpoint in the interval?
Preliminary results
For off groups, that was true 14 out of 47 times
For on groups, that was true 29 out of 47 times
For results, Fisher's exact test for those wanting to learn stats...
I have to check my work a few times, and when I do, I'll upload the analysis.
I've confused myself. Would this result be a good thing or a bad thing?
Its a good thing, as it confirms the hypothesis that ON does not provide instantaneous results, as in a square-wave voltage change. Mag heating, frequency and resonance shift all combine (my belief) for variances shown in the points you made above. On the other hand, mag OFF is rather calm and predictable, with mother nature taking over and gently thermal lifting the DUT.
Not sure if the 50% point is the best point to sub-divide, but the ON state variances versus the OFF state variances is a decent way to look at the data subset.
OK spreadsheet with analysis (no VBA)
Sheet: raw.php
raw data and processing
Sheet: Sheet1
summary & results
m1>m2 m1<=m2
Group ON 29 18
Group OFF 14 33
Fisher's exact test, two tailed p = .0035
Fisher's exact test, one tailed p = .0018
Nice glenn, for us stat-light posters out here, what does the data tell you?
stat-light... sounds like a beer
OK, there is a very significant difference between the slope behavior for on vs. off, given by the values of p. Close to 1 is bad. Close to 0 is good.
What this indicates is:
If ON, the slope is likely to become more negative during the cycle.
If OFF, the slope is likely to not change or become less negative during the cycle.
examples
define m1 as the slope during the 1st half of the cycle and m2 as the slope during the last half of the cycle.
if m1 = -2 and m2 = -3 then m1>m2 = TRUE
if m1 = -2 and m2 = -1 then m1>m2 = FALSE
Ok, more confirmation of ON cycle changes. Looks like all roads leading to the conclusion that something is definitely happening, i.e. not a null test...positive not null. Fair statement?
I've been OCD like on not calling this effect thrust, and guess I will continue to do so. Without a true understanding of why this happens, it might as well be an attractive or repulsive force. Not 100% comfortable with thrust, so my shorthand is emdrive effect.
Guess a last piece of the puzzle is the open channel data points for relative system noise comparisons betwee on/off states. A quick glance, saw nothing significant in the numbers indicating an EMI induced spike or noise. Let me know if you concur.
Physics isn't my strong suit. Simply put, when power is on, things are different than when power is off. The data supports that.
You have shared your data, and others here have shared their data and analysis. It's fair game for anyone with an internet connection to review, critique, replicate, or ignore.
re the noise numbers in those other channels, I haven't looked extensively but a quick look shows they seem rather random. I did a couple of quick correlation analyses and the rs were in the .00x range. There are only a few discrete values and they look like the lowest possible DAC's deviation from a ground state. They are definitely not correlated with your actual data. I'll look again later, but they seem innocuous.
Regarding the null test. Aren't we pretty sure that a cylindrical resonator will not thrust? It will heat and do all the other things, but if something is escaping, that something should escape symmetrically, no thrust. If nothing is escaping then still no thrust due to symmetry.
Of course getting the heating rates right may be a little tricky when it comes to comparing to a real EM Drive configuration.
I do think the idea posted much earlier by (I forget who posted it), the idea of two identical frustums, one at each end of a balance beam, would go a long way toward eliminating thermal lift from the data. The idea was to mount both frustums upward, run to steady state then turn one off. Or mount one up, one down and run synchronously, doubling the thrust effect while significantly reducing the lift effect. This one would allow data collection from initial power on of the cold system and still reduce the effect of thermal lift.
Only problem is that it takes two frustums, but rfmwguy can tell us how much difficulty and cost would be involved in building the second identical frustum. And he could likely scope out the required modifications to the test rig as well.
And the difficulty with the word "identical" meaning identical fabrication, feeding, power coupled etc. Not easy. Do-able, but not easy.
H
As has been mentioned, identical is good, but not necessary. They just need to be close enough that the second frustum mimics the thermal lift better than a hot plate and also better than nothing. The objective is to counterbalance the thermal effects in order to reduce the unbalanced thermal lift while adding to the the EM drive effect. This should pull the EM drive effect up out of the, now reduced, thermal noise.
@glennfish, attached is the original ft 2d spreadsheet with data channels 2-4 (columns b-d) unhidden. Its in the original file as well, just hidden.
Note on these hidden data channels: The data inputs are unloaded, meaning nothing is attached to them, unlike channel 1 (column a on the spreadsheet), the LDS voltage input. This might be useful in looking at system noise in mag ON/OFF conditions.
While I cannot quantify the randomness, it will be the most variable (very sensitive) since there is no load (resistor) on the inputs. If mag ON imparts system noise, these 3 channels and over 8,100 data points will be the most sensitive to noise and will show it.
Another note. The balanced impedance of channel 1 is 470 ohms. Data on channels 2-4 (columns b-c) are open (infinite impedance by comparison) and are not directly equivalent to channel 1. However general deviation comparisons to mag ON/OFF should give you a relative indication of any system noise attributable to EMI, if present.
Thank you. Now ponder this with me (before I downloaded your data):
1. The data trend overall is a negative slope.
2. You suggested in the last 24 hours that -- the turn-on time to resonance could be seconds
3. So what I did is for each group (on or off), asked, "was the first half of the slope > the 2nd half of the slope?" i.e. was there a slope change from negative to more negative about midpoint in the interval?
Preliminary results
For off groups, that was true 14 out of 47 times
For on groups, that was true 29 out of 47 times
For results, Fisher's exact test for those wanting to learn stats...
I have to check my work a few times, and when I do, I'll upload the analysis.
I've confused myself. Would this result be a good thing or a bad thing?
Its a good thing, as it confirms the hypothesis that ON does not provide instantaneous results, as in a square-wave voltage change. Mag heating, frequency and resonance shift all combine (my belief) for variances shown in the points you made above. On the other hand, mag OFF is rather calm and predictable, with mother nature taking over and gently thermal lifting the DUT.
Not sure if the 50% point is the best point to sub-divide, but the ON state variances versus the OFF state variances is a decent way to look at the data subset.
OK spreadsheet with analysis (no VBA)
Sheet: raw.php
raw data and processing
Sheet: Sheet1
summary & results
m1>m2 m1<=m2
Group ON 29 18
Group OFF 14 33
Fisher's exact test, two tailed p = .0035
Fisher's exact test, one tailed p = .0018
Nice glenn, for us stat-light posters out here, what does the data tell you?
stat-light... sounds like a beer
OK, there is a very significant difference between the slope behavior for on vs. off, given by the values of p. Close to 1 is bad. Close to 0 is good.
What this indicates is:
If ON, the slope is likely to become more negative during the cycle.
If OFF, the slope is likely to not change or become less negative during the cycle.
examples
define m1 as the slope during the 1st half of the cycle and m2 as the slope during the last half of the cycle.
if m1 = -2 and m2 = -3 then m1>m2 = TRUE
if m1 = -2 and m2 = -1 then m1>m2 = FALSE
Ok, more confirmation of ON cycle changes. Looks like all roads leading to the conclusion that something is definitely happening, i.e. not a null test...positive not null. Fair statement?
I've been OCD like on not calling this effect thrust, and guess I will continue to do so. Without a true understanding of why this happens, it might as well be an attractive or repulsive force. Not 100% comfortable with thrust, so my shorthand is emdrive effect.
Guess a last piece of the puzzle is the open channel data points for relative system noise comparisons betwee on/off states. A quick glance, saw nothing significant in the numbers indicating an EMI induced spike or noise. Let me know if you concur.
Physics isn't my strong suit. Simply put, when power is on, things are different than when power is off. The data supports that.
You have shared your data, and others here have shared their data and analysis. It's fair game for anyone with an internet connection to review, critique, replicate, or ignore.
re the noise numbers in those other channels, I haven't looked extensively but a quick look shows they seem rather random. I did a couple of quick correlation analyses and the rs were in the .00x range. There are only a few discrete values and they look like the lowest possible DAC's deviation from a ground state. They are definitely not correlated with your actual data. I'll look again later, but they seem innocuous.
Thanks again for all your hard work Glenn, it is greatly appreciated. As you said, I could find no correlation in noise to power on, it did seem random to me. I did take the time to do a rough overlay against an image provided by another poster (think it was Joe). Its shows all noise datapoints highly scattered regardless of on or off.
Readers should note that it is highly exaggerated (blue dots) in the Y axis showing is extremely small voltage changes in one of the open channels of the DAQ. So, highly exaggerated noise versus a beam deflection and power ON (pink lines) of FT 2D.
Regarding the null test. Aren't we pretty sure that a cylindrical resonator will not thrust? It will heat and do all the other things, but if something is escaping, that something should escape symmetrically, no thrust. If nothing is escaping then still no thrust due to symmetry.
Of course getting the heating rates right may be a little tricky when it comes to comparing to a real EM Drive configuration.
I do think the idea posted much earlier by (I forget who posted it), the idea of two identical frustums, one at each end of a balance beam, would go a long way toward eliminating thermal lift from the data. The idea was to mount both frustums upward, run to steady state then turn one off. Or mount one up, one down and run synchronously, doubling the thrust effect while significantly reducing the lift effect. This one would allow data collection from initial power on of the cold system and still reduce the effect of thermal lift.
Only problem is that it takes two frustums, but rfmwguy can tell us how much difficulty and cost would be involved in building the second identical frustum. And he could likely scope out the required modifications to the test rig as well.
And the difficulty with the word "identical" meaning identical fabrication, feeding, power coupled etc. Not easy. Do-able, but not easy.
H
As has been mentioned, identical is good, but not necessary. They just need to be close enough that the second frustum mimics the thermal lift better than a hot plate and also better than nothing. The objective is to counterbalance the thermal effects in order to reduce the unbalanced thermal lift while adding to the the EM drive effect. This should pull the EM drive effect up out of the, now reduced, thermal noise.
True enough - it would definitely allow for a much better signal to noise ratio. But regardless of how well it could be compensated for and used to boost S/N I suspect that the non-identical nature of the two drives would become a new source of doubt for some folks. Definitely worth doing though as I think the better S/N would present a much clearer picture.
Even if I go with a rotating table approach, one set of tests I want to run is frustum v frustum - sort of a mano a mano UFC challenge for propellentless thrust. (sorry - been a long day wading through Medicare paperwork - you haven't lived until you have done that).
H
And the difficulty with the word "identical" meaning identical fabrication, feeding, power coupled etc. Not easy. Do-able, but not easy.
H
As has been mentioned, identical is good, but not necessary. They just need to be close enough that the second frustum mimics the thermal lift better than a hot plate and also better than nothing. The objective is to counterbalance the thermal effects in order to reduce the unbalanced thermal lift while adding to the the EM drive effect. This should pull the EM drive effect up out of the, now reduced, thermal noise.
True enough - it would definitely allow for a much better signal to noise ratio. But regardless of how well it could be compensated for and used to boost S/N I suspect that the non-identical nature of the two drives would become a new source of doubt for some folks. Definitely worth doing though as I think the better S/N would present a much clearer picture.
Even if I go with a rotating table approach, one set of tests I want to run is frustum v frustum - sort of a mano a mano UFC challenge for propellentless thrust. (sorry - been a long day wading through Medicare paperwork - you haven't lived until you have done that).
H
Remember, there is always the option to evaluate each frustum by itself simply by heating them both then running one or the other but not both, to detect individual EM drive characteristics, if any. And by now, I think there might well be an effect, just need to better pull it out of the noise.
for completion sake, all VBA used in my analyses is provided below.
Since this site didn't provide the ability to upload a Microsoft Exel spreadsheet with live VBA, I've provided the code for evaluation, critique, modification, or to use on birthday cards for your mother. It was rushed & hacked and I didn't cover all possible conditions which especially threw out the last data set, but the code corresponds to the spreadsheets that I uploaded to this forum and is mapped to the buttons. Sorry no comments in the code. The code depends on the data being in explicit columns. Insert or delete a column, and it dies a horrible death. The spreadsheets that I uploaded had the correct columns in the correct place.
Until the next data run, I'm going back to lurking.
Sub Button1_Click()
' find slope
Dim i As Integer
Dim col As Integer
Dim row As Integer
Dim indata As Boolean
Dim inone As Boolean
Dim inzero As Boolean
Dim outrow As Integer
Dim three As Integer
outrow = 1
indata = False
inone = False
inzero = False
col = 8
row = 1
For i = 2 To 2721
If Range("E" & i).Value <> "K" Then GoTo escape
If Range("D" & i).Value <> Range("D" & i - 1).Value Then
three = 3
If Range("D" & i - 1).Value = 0 Then three = 3
Cells(2, col).Value = "=linest(" & ColLtr(col) & three & ":" & ColLtr(col) & row - 1 & "," & ColLtr(col - 1) & three & ":" & ColLtr(col - 1) & row - 1 & ")"
If Range("D" & i - 1).Value = 1 Then
Sheets("Sheet1").Cells(outrow, 1).Value = Cells(2, col).Value
Else
Sheets("Sheet1").Cells(outrow, 2).Value = Cells(2, col).Value
outrow = outrow + 1
End If
col = col + 2
Cells(1, col).Value = "ON"
If Range("D" & i).Value = 0 Then
Cells(1, col).Value = "OFF"
End If
row = 3
End If
Cells(row, col - 1).Value = Range("A" & i).Value
Cells(row, col).Value = Range("B" & i).Value
row = row + 1
escape:
Next i
End Sub
Sub Button2_Click()
'average displacement
Dim i As Integer
Dim col As Integer
Dim row As Integer
Dim indata As Boolean
Dim inone As Boolean
Dim inzero As Boolean
Dim outrow As Integer
Dim three As Integer
Dim stuff As String
Dim started As Boolean
started = False
outrow = 1
indata = False
inone = False
inzero = False
col = 8
row = 1
For i = 2 To 2721
If Range("E" & i).Value <> "K" Then GoTo escape
If Range("D" & i).Value <> Range("D" & i - 1).Value Then
three = 3
If Range("D" & i - 1).Value = 0 Then three = 3
stuff = "=(max(" & ColLtr(col) & three & ":" & ColLtr(col) & row - 1 & ")- min(" & ColLtr(col) & three & ":" & ColLtr(col) & row - 1 & "))/" & row - 3
If started Then Cells(2, col).Value = stuff
started = True
If Range("D" & i - 1).Value = 1 Then
Sheets("Sheet1").Cells(outrow, 1).Value = Cells(2, col).Value
Else
Sheets("Sheet1").Cells(outrow, 2).Value = Cells(2, col).Value
outrow = outrow + 1
End If
col = col + 2
Cells(1, col).Value = "ON"
If Range("D" & i).Value = 0 Then
Cells(1, col).Value = "OFF"
End If
row = 3
End If
Cells(row, col - 1).Value = Range("A" & i).Value
Cells(row, col).Value = Range("B" & i).Value
row = row + 1
escape:
Next i
End Sub
Sub Button3_Click()
'coefficient of variation
Dim i As Integer
Dim col As Integer
Dim row As Integer
Dim indata As Boolean
Dim inone As Boolean
Dim inzero As Boolean
Dim outrow As Integer
Dim three As Integer
Dim stuff As String
Dim started As Boolean
started = False
outrow = 1
indata = False
inone = False
inzero = False
col = 8
row = 1
For i = 2 To 2721
If Range("E" & i).Value <> "K" Then GoTo escape
If Range("D" & i).Value <> Range("D" & i - 1).Value Then
three = 3
If Range("D" & i - 1).Value = 0 Then three = 3
stuff = "=stdev(" & ColLtr(col) & three & ":" & ColLtr(col) & row - 1 & ")/" & "average(" & ColLtr(col) & three & ":" & ColLtr(col) & row - 1 & ")"
If started Then Cells(2, col).Value = stuff
started = True
If Range("D" & i - 1).Value = 1 Then
Sheets("Sheet1").Cells(outrow, 1).Value = Cells(2, col).Value
Else
Sheets("Sheet1").Cells(outrow, 2).Value = Cells(2, col).Value
outrow = outrow + 1
End If
col = col + 2
Cells(1, col).Value = "ON"
If Range("D" & i).Value = 0 Then
Cells(1, col).Value = "OFF"
End If
row = 3
End If
Cells(row, col - 1).Value = Range("A" & i).Value
Cells(row, col).Value = Range("B" & i).Value
row = row + 1
escape:
Next i
End Sub
Sub Button4_Click()
' compare 1st half of sequence with 2nd half
Dim i As Integer
Dim col As Integer
Dim row As Integer
Dim indata As Boolean
Dim inone As Boolean
Dim inzero As Boolean
Dim outrow As Integer
Dim three As Integer
Dim half As Integer
outrow = 1
indata = False
inone = False
inzero = False
col = 8
row = 1
For i = 2 To 2721
If Range("E" & i).Value <> "K" Then GoTo escape
If Range("D" & i).Value <> Range("D" & i - 1).Value Then
three = 3
If Range("D" & i - 1).Value = 0 Then three = 3
half = (row - 4) / 2
Cells(2, col).Value = "=linest(" & ColLtr(col) & three & ":" & ColLtr(col) & half + three & "," & ColLtr(col - 1) & three & ":" & ColLtr(col - 1) & half + three & ")>linest(" & ColLtr(col) & three + half + 1 & ":" & ColLtr(col) & row - 1 & "," & ColLtr(col - 1) & three + half + 1 & ":" & ColLtr(col - 1) & row - 1 & ")"
If Range("D" & i - 1).Value = 1 Then
Sheets("Sheet1").Cells(outrow, 1).Value = Cells(2, col).Value
Else
Sheets("Sheet1").Cells(outrow, 2).Value = Cells(2, col).Value
outrow = outrow + 1
End If
col = col + 2
Cells(1, col).Value = "ON"
If Range("D" & i).Value = 0 Then
Cells(1, col).Value = "OFF"
End If
row = 3
End If
Cells(row, col - 1).Value = Range("A" & i).Value
Cells(row, col).Value = Range("B" & i).Value
row = row + 1
escape:
Next i
End Sub
Function ColLtr(ByVal iCol As Long, Optional sCol As String = "") As String
' shg 2012
If iCol = 0 Then
ColLtr = sCol
Else
sCol = Chr(65 + (iCol - 1) Mod 26) & sCol
iCol = (iCol - 1) \ 26
ColLtr = ColLtr(iCol, sCol)
End If
End Function
A while back, somebody noted an experiment that showed thrust by bouncing a laser back and forth. Was this a photonic laser rocket (cool tech, well understood) or did you mean it showed thrust by bouncing a laser off several mirrors and measuring the thrust at the last mirror. If so, can you post a link to the paper or report?
That was probably me. I have suspected Since blundering across David Bae's work most of a year ago that it contains part of the key as to what is going on with the EM Drive.
Bae's website:
http://ykbcorp.com/tech_precFormation.htmlArticle:
http://www.centauri-dreams.org/?p=29341Test result from May 2015:
http://www.popularmechanics.com/space/a15545/photonic-laser-thrust-space-engine/Links to papers:
http://www.thelivingmoon.com/41pegasus/02files/PLT_Photonic_Laser_Thruster_01.htmlWhat gets me here is this 'photon recycling scheme' produces output power on the order of 5000 times that of the input power - which as I understand such things, is either 'free energy territory' or dang close to it. Yet, the effects are well supported by laboratory work, and supposedly COE is not violated.
Which leads me to wonder: if the EM Drive is somehow doing something similar to Bae's device, would that constitute a COE violation?
A while back, somebody noted an experiment that showed thrust by bouncing a laser back and forth. Was this a photonic laser rocket (cool tech, well understood) or did you mean it showed thrust by bouncing a laser off several mirrors and measuring the thrust at the last mirror. If so, can you post a link to the paper or report?
That was probably me. I have suspected Since blundering across David Bae's work most of a year ago that it contains part of the key as to what is going on with the EM Drive.
Bae's website:
http://ykbcorp.com/tech_precFormation.html
Article:
http://www.centauri-dreams.org/?p=29341
Test result from May 2015:
http://www.popularmechanics.com/space/a15545/photonic-laser-thrust-space-engine/
Links to papers:
http://www.thelivingmoon.com/41pegasus/02files/PLT_Photonic_Laser_Thruster_01.html
What gets me here is this 'photon recycling scheme' produces output power on the order of 5000 times that of the input power - which as I understand such things, is either 'free energy territory' or dang close to it. Yet, the effects are well supported by laboratory work, and supposedly COE is not violated.
Which leads me to wonder: if the EM Drive is somehow doing something similar to Bae's device, would that constitute a COE violation?
Thanks, had not studied this concept. 5K power increase? Going to have to read his papers closely, also beam confinement seems unlikely, but am open to read up. Trying to think of any energy that would play by these rules. Rf would need guides to contain it or it would instantly scatter. Additive? Now that's interesting.
Where is our friend Rodal? It seems he has been quiet for half a month.
Where is our friend Rodal? It seems he has been quiet for half a month.
I believe he is around just in lurking mode. Dr. Rodal follows EmDrive debate from the start, so in my opinion he also needs some time to focus on different things and some rest from this debate
. I also believe we may see many folks returing once we see data from NASA EW by the end of this year.
What gets me here is this 'photon recycling scheme' produces output power on the order of 5000 times that of the input power - which as I understand such things, is either 'free energy territory' or dang close to it. Yet, the effects are well supported by laboratory work, and supposedly COE is not violated.
Which leads me to wonder: if the EM Drive is somehow doing something similar to Bae's device, would that constitute a COE violation?
There's no COE violation because as the photons bounce between the two mirrors, the speed of the mirrors relative to each other will increase as per the momentum change imparted by the reflected photon. This in turn means that the photons will get more and more red-shifted. This ensures that total kinetic energy gained can never exceed the energy spent on producing the photons.
So, yes, it is more efficient than a simple photon drive but it only works if the mirrors are actually moving away from each other faster and faster after each bounce and this increase of speed must match the imparted momentum by each bounce. In the EMdrive case the photons are bouncing between the walls of the drive and the walls clearly don't increase their speed relative to each other in such a fashion, so it is a completely different situation.
