109 it fly wrong

[hitech]

Originally posted by HoHun
Hi Hitech,

>Whats a physics model?

Oh, that's just another word for my spreadsheet :-)

http://en.wikipedia.org/wiki/Model_(abstract)

"By abstract model (or conceptual model) we mean a theoretical construct that represents social or physical processes by a set of variables and a set of logical and quantitative relationships between them."

I could also call it an aircraft model but that would probably be misunderstood :-)

Regards,

Henning (HoHun)

Golly G Wizz I hope I can learn to model with spread sheets some day. Will you pretty please teach me how to do this stuff HoHun.


Hummor off.

Nice evaluation HoHun.

HiTech

[joeblogs]

If I follow the argument, everyone of these plots lies in the assumed standard error of the estimates. In that case there's no way to say that one is right or wrong.

If I understand what HoHun has done, he's fit a curve based on the effect of changes in atmospheric pressure on engine horsepower, possibly taking into effect changes in drag as well.

One way to check the robustness of his model is simply to generate several curves using a different point of the actual test data and examining the degree to which these curves vary (i.e. sensitivity analysis).

What is almost certainly is true is that the relationships are non-linear, which means the errors get bigger the further away the estimated relationship is plotted from the actual data.

-blogs

Originally posted by gripen
Well, this is going to strangest discussion for me so far; HoHun has ignored me but just after I have pointed out that his calculation is biased right from the beginning, he suddenly jumps in and gives another (longish) explanation for his actions.

We have no other starting point for the error analysis than the original and unaltered FAF data set as seen here again:


This set comes from the very same report as the chart, but this set does not contain that error which HoHun continously uses to support his agenda. The chart is made using this data set.


If we assume that these measurements have +/- 15 km/h error as HoHun suggests then we have following values for error boundaries:

If we ad them to HoHun's new version of the reality, then we have a following chart:

Now we can see that if we use real measured data and assume +/- 15 km/h error then the fastest speed the MT-215 could reach at 10100 m CINA was below errorneous 572 km/h value in the chart and actually HoHun's own version admits this.

Then we should analyze a bit HoHun's error boundary theory. For one reason or another he has choosen to use allredy altered set (manipulated during curve fitting) for his purposes ie he uses errornenous curve from the FAF chart, this is a big error because most of the measured points are lost and also real error boundary is lost from these altitudes. Therefore the error boundary is correct just at two altitudes 6420 and 8110 m CINA. It very obivious that HoHun does not understand the term error boundary or he has purposedly created errorneous chart. As an example we can look the error boundary at 9110 m CINA, real error range is 595-625 km/h but the chart gives about 580-610 km/h

But actually error boundaries have quite little to do with built in bias in the HoHun's original chart seen here:

Here HoHun uses the errorneous FAF curve as base for his analyze and actually managed to manipulate it during process. But lets assume for a moment that it's a "real" curve. For one reason or another he has chosen use the speed at highest altitude (572km/h at 10300m) for calculate new curve. Now it should be asked is there any particular reason to choose speed at highest measured altitude for calculation? If the calculation is correct then it should give very similar values despite what ever speed and altitude combination is choosed as a base for calculation. But if we look the new calculated speed curve, we can see that the only point where it crosses "real" curve is that highest altitude. Therefore it is easy to understand that by choosing any other speed and altitude combination would result worse results and therefore the method is biased right from the beginning.

Generally further HoHun goes in his explanations, deeper he sinks.

gripen

[HoHun]

Hi Blogs,

>When I saw the origonal plots I wondered about the standard errors of these estimates.

>I think its under appreciated how noisy these numbers can be when only 1 or 2 tests are run under any set of parameters.

Roger on the noise :-)

Since the 5 data points from full throttle height up are not repetitions of the same experiments, I couldn't really determine the standard deviation. (OK, it's infinite, but that's no help :-)

If (remembering that it actually isn't correct!) you consider them repetitions of the same experiment and for example figure just the difference to the "5.4.43" graph, the standard deviation is 13.3 km/h. If you discount the 10 km value, it's still 11.4 km/h.

Of course, that wouldn't be the proper way to do a real experiment. The proper way is formulating a hypothesis on the relation of altitude and speed, determining the pivotal constants (like drag coefficients), and then using the experiment to find out the value of these constants along with their standard deviation.

A speed curve determined in accordance with scientific procedures wouldn't just be a curve fitted between a set of measured points, but actually a completely fictional curve entirely based on a calculation using the constants derived from the tests. This curve has an homogenous error as all points of it are based on the same constant (complete with its error).

It's a common misunderstanding that measured data is superior to calculated data. The FAF example shows that measured data can be subject to a considerable random error.

Fortunately, the low-altitude data is good enough to determine the drag coefficients, so it's possible to calculate the high-altitude performance from these values, ignoring the high-altitude measurements completely :-)

>U.S. engine makers always distributed power curves with a disclaimer of 5%...

No "I want my money back" that way ;-)

NACA actually pointed out that often, speed tests are based on the assumption that the engine performs up to specifications without actually checking whether that's actually correct, so that comparisons between different test aircraft of the same type with the same engine model could lead to wrong conclusions about the effects of aerodynamic changes.

(That's another reason why calculated data, based on a standard power graph, can beat measured data - which typically shows speeds achieved on an unknown power output.)

Regards,

Henning (HoHun)

[Dweeb]

HoHun

I lurk here a lot, and a search of threads in which Gripen has posted reveals two things...

Firstly, he must have the last word, even if it is wrong.

Secondly, he has difficulty understanding the reports he reads, because he lacks the analytical skills to interpret them correctly, and then attempts to win any subsequent debate by repeating his misconceptions to the point of nausea.

You offered to show him your calculations, the open actions of an honest debater who wishes to encourage transparency. Gripen refused to look at them, the defensive reaction of someone who is either afraid of exposing his inability to comprehend the physics/math involved, or is afraid that such clarity will expose his own agenda.

That is his modus operandi, you won't be the first to withdraw from debate with Gripen, due to the utter futility of reasoning to penetrate his chosen dogma, and you probably won't be the last.

Dweeb

[Dweeb]

Originally posted by hitech
Whats a physics model?

HiTech

That thing you wrote ten years ago to make confirmed kill wurk rite... Ya big tease

Dweeb

[HoHun]

Hi Dweeb!

>That thing you wrote ten years ago to make confirmed kill wurk rite... Ya big tease

LOL! I've got to admit I'm still suffering from the after-effects of oxygen starvation - after years of holding my breath for the Do 335 promised in that CGW issue :-)

>I lurk here a lot [...]

Hey, are you a FSFORUM survivor, too? CK experience and CIS terminology would fit :-)

Regards,

Henning (HoHun)

[HoHun]

Hi Hitech,

>Golly G Wizz I hope I can learn to model with spread sheets some day. Will you pretty please teach me how to do this stuff HoHun.

You probably wouldn't learn much - it's like a one-dimensional flight simulation incapable of real-time operation, and I think you're already one or two steps ahead of that with your own project ;-)

>Nice evaluation HoHun.

Thanks! :-)

Regards,

Henning (HoHun)

[HoHun]

Hi Blogs,

>One way to check the robustness of his model is simply to generate several curves using a different point of the actual test data and examining the degree to which these curves vary (i.e. sensitivity analysis).

Since I calculate the drag from the reference point, that's indeed a working option.

You can see the results from my graph:

The data points from sea level to 2.4 km would lead to the same results as the 6.4 km data point I actually used. (My calculated speed graph coincedes with the measured speed there.)

Between 2.4 km and 6.4 km, we see a systematic difference in the engine power graph. I'd have to predict a differing speed curve from these points, but that would be misleading because obviously, in this altitude band the FAF aircraft's engine was performing differently from the DB605A chart I used for my prediction. Accordingly, it's not indicative of a problem with the physics model.

Above 6.4 km, we've got 4 data points with obvious random variations, and only the use of one of them would yield the curve I predicted.

Using the others would result in considerably slower curves - and that includes speed at full throttle height.

Regards,

Henning (HoHun)

[gripen]

Well, I have pointed out ealier why the engine output can't be calculated using full RAM below FTH as seen here:



The engine is throttled up to FTH despite variable speed system because it is adjusted by altitude, see the curve called gebläsedruck in the graph (pressure between impeller and throttle valve). The engine output below FTH is a bit lower at high speed than at climb speed because it's throttled more below FTH due to RAM. If the full RAM is assumed then gebläsedruck would have been same as ladedruck (MAP) resulting higher output. Yes, there is a systematic error but not in the FAF graph.

gripen

[gripen]

Originally posted by HoHun

Above 6.4 km, we've got 4 data points with obvious random variations, and only the use of one of them would yield the curve I predicted.

Using the others would result in considerably slower curves - and that includes speed at full throttle height.

Well, that is what I have been saying all the time. HoHun has been using the point which gives him best results. And that point itself is 20km/h off due to error in the FAF chart.

Regarding the errors in the FAF measurements, the errors are true only for measured points, not for the fitted curve. The fitted curve in the FAF chart is errorneous due to the reason pointed out several times above and therefore it can't be used as data for the analysis.

gripen

[joeblogs]

Anyone who has taken a chemistry class can appreciate the puzzle of knowing what result you are supposed to get and yet not getting it. Sometimes it's bad procedure, but often it results from using instruments too crude for the task...

Anyone who reads my posts knows I am preoccupied with engines (a detroit boy to the end). The data I like to use most comes from the engine makers and is independent of the installation. Such data comes from the most controlled experiments one can get with a motor meant for a plane.

The cost of this precision however, is that the numbers won't match up with flight test data...

-Blogs

Originally posted by HoHun
Hi Blogs,

>When I saw the origonal plots I wondered about the standard errors of these estimates.

>I think its under appreciated how noisy these numbers can be when only 1 or 2 tests are run under any set of parameters.

Roger on the noise :-)

Since the 5 data points from full throttle height up are not repetitions of the same experiments, I couldn't really determine the standard deviation. (OK, it's infinite, but that's no help :-)

If (remembering that it actually isn't correct!) you consider them repetitions of the same experiment and for example figure just the difference to the "5.4.43" graph, the standard deviation is 13.3 km/h. If you discount the 10 km value, it's still 11.4 km/h.

Of course, that wouldn't be the proper way to do a real experiment. The proper way is formulating a hypothesis on the relation of altitude and speed, determining the pivotal constants (like drag coefficients), and then using the experiment to find out the value of these constants along with their standard deviation.

A speed curve determined in accordance with scientific procedures wouldn't just be a curve fitted between a set of measured points, but actually a completely fictional curve entirely based on a calculation using the constants derived from the tests. This curve has an homogenous error as all points of it are based on the same constant (complete with its error).

It's a common misunderstanding that measured data is superior to calculated data. The FAF example shows that measured data can be subject to a considerable random error.

Fortunately, the low-altitude data is good enough to determine the drag coefficients, so it's possible to calculate the high-altitude performance from these values, ignoring the high-altitude measurements completely :-)

>U.S. engine makers always distributed power curves with a disclaimer of 5%...

No "I want my money back" that way ;-)

NACA actually pointed out that often, speed tests are based on the assumption that the engine performs up to specifications without actually checking whether that's actually correct, so that comparisons between different test aircraft of the same type with the same engine model could lead to wrong conclusions about the effects of aerodynamic changes.

(That's another reason why calculated data, based on a standard power graph, can beat measured data - which typically shows speeds achieved on an unknown power output.)

Regards,

Henning (HoHun)

[joeblogs]

Grippen

Can you explain this chart for me? Is the top curve intake pressure (before hitting the carburetor) and the bottom the rated pressure of the engine? What are the two straight lines in the curve below?

-Blogs

Originally posted by gripen
Well, I have pointed out ealier why the engine output can't be calculated using full RAM below FTH as seen here:



The engine is throttled up to FTH despite variable speed system because it is adjusted by altitude, see the curve called gebläsedruck in the graph (pressure between impeller and throttle valve). The engine output below FTH is a bit lower at high speed than at climb speed because it's throttled more below FTH due to RAM. If the full RAM is assumed then gebläsedruck would have been same as ladedruck (MAP) resulting higher output. Yes, there is a systematic error but not in the FAF graph.

gripen

[gripen]

blogs,
Here is the entire page:



Upper part of the page is naturally speed data. Stau means RAM. Gebläsedruck is pressure between impeller and throttle  and ladedruck is MAP. Bottom part is air temperature, measured and CINA.

As you probably know that the DB 605 had variable speed supercharger with hydraulic coupling. The hydraulic coupling had two pumps of which one worked all the time making supercharger to work at fixed speed up to 2km. The second oil pump was governed by barometric valve which started to work at around 2km the to increase speed of the supercharger then it gradually increased the speed of the supercharger up to about 5km and above that altitude supercharger worked as a fixed speed system again.

A common missunderstanding is that the supercharger of the DB 605 worked at optimal speed below FTH (between 1st and 2nd FTH). But as you can see from the graph, the engine was throttled  all the way up to FTH; gebläsedruck is higher than ladedruck. Therefore it's easy to understand that with high speed RAM the engine is more throttled than at climb speed RAM and therefore the engine works more  efficiently at climb speed up to the FTH. If the supercharger had worked at optimal speed between 1st and 2nd FTH, then ladedruck and gebläsedruck would have been same at that range.

If I understand correctly, HoHun has calculated his version assuming full RAM and engine working at full throttle between 1st and 2nd FTH ie at maximal efficiency. Therefore his calculation gives higher speed between 2,4 and 6,4km than FAF data. But this is not the way the DB 605 worked; output curve is given with about climb speed RAM so assuming full RAM for high speed will result overestimated output. HoHun has apparently a systematical error in his calculations between 2,4 and 6,4km.

Regarding statistics and FAF data. There is no sense to speak about standard deviation with such a small data set:



As you can there is lot of variation so it pretty much impossible to calculate anything meaninfull. But one thing should be noted; all HoHun's analyses are based on errorneous curve with wrong speed value at 10100m CINA, he continously refuses to use real measured value for that altitude. It should also noted that if his model gives about right FTH only with that errorneous value for 10100m, then there appears to be another systmatical error in his calculation (which one could call bias).

gripen

[joeblogs]

I get the first graph - incidentally does that include a plot of a Spit V with a DB605 installed? Now that could be used for making some aerodynamic comparisons...

The second plot is ram air. Now there are a lot of ways to measure this. On a specific plane it is typically, but not always done done assuming level flight. In any case the thing to work out is the maximum contricution of RAM air to engine horsepower from the chart.

The third plot makes sense too, although it's not clear to me that the pilot would have the throttle wide open for most of the range under 6.7 km. I know some planes (FW 190?) had all these things integrated in the pilot's controls, but I don't know about the 109. And I like your description of the supercharger. Didn't know those things.

The first and third charts seem to imply a supercharger tax on the engine between 2 and 4 km. More horsepower is consumed, but I suspect the pilot has to throttle back a bit until after 4 km.
But I am just guessing here. A simple translation of a G-10 or k manual should tell us.

Is the last chart measured air temperature, a plot against a standard atmosphere, or both?

So your argument is that HoHun is projecting a speed value after setting aside a range of altitudes where the pilot is forced to throttle back the engine? In this case it really does matter what speed-altitude combination you calibrate from.

If what you care about is maximum true airspeed, you'd want a data point at or above the final critical altitude of the engine. Now there remains an obvious non-linearity - above critical altitude power falls, but so does drag as air density declines. Maximum speed is likely attained at some point just above critical altitude, depending on the airframe.

One thing I worry about is that if this flight test was run just once and the supercharger was sludging, the tax might be exaggerated.

Now I need to go back and re-read this thread...

-Blogs

[HoHun]

Hi Blogs,

>The cost of this precision however, is that the numbers won't match up with flight test data...

That's exactly what NACA meant. Engines in test aircraft are usually assumed to be performing correctly for the speed tests, but this assumption hardly ever is verified.

That's why the way of making speed predictions is to derive the drag from flight tests, then calculate a speed curve based on representative engine data.

Regards,

Henning (HoHun)