Thrust Meter/Test Rig Prototype Ver1.0

[stix]

After quite a few months of development, this is what I've come up with.

 

There are some things that I'm not fully satisfied with and therefore currently under review for Version 2. The mechanics of the device itself is simple enough - a tripod shape for stability, with adjustment using "wing nut" screws using a bullseye level - electronics are protected underneath.

 

The load cell is of the parallel type and as such, the "V" motor mount is able to hold the motor and record the data accurately regardless of a central position (within a reasonable offset, 10-20mm I would think, perhaps more) - much like your standard grocery store scales.

 

This device is capable of recording 1 gram resolution over 30kgs - with a sample rate of 200Hz. For simplicity, the next prototype will record 5 grams over 20kg with a smaller footprint and simpler electronics. This one measures approx. 10.5"/270mm between the wingnut posts.

 

I'll be doing tests to confirm the validity of the "V" mount. I'll post the results here. If there's an issue with it (which I don't think there will be) I'll gladly blow the device up using a pre-cato-defined motor, which will include lots of sparks - perhaps TT mixed with some dodgy Ti that I bought a while back - and a few micro dragons eggs thrown in for good measure - hmmm, that's starting to sound too appealing.

 

Software? Not up to standard yet, but the basic raw data provided can be recorded and evaluated with various excel programs that are available (I believe).

 

This device has been constructed using "non-captive methods". What I mean is that the whole device can be de-constructed with a ring spanner and a few hex keys (Allen keys). This means that the load cell can easily be replaced.

 

It's also very important to point out that the device has been conceived with this view: "The device should be used by those who have a good grasp of rocket motor theory, or at least have had no cato for some time"

 

[EDIT] An alternate view and purpose is that the device is expendable within limits - although that's hard to define.

 

With that in mind, all "constructive" criticisms are welcome, demanded and required.

 

Cheers.

Edited March 6, 2016 by stix

[schroedinger]

Looks neat, now we just need a test run (and how about some blueprints? ;-) )

[lloyd]

I don't know anything about the design of your load cell amp. Back in (IIRC) 2006 (or so), I designed one with a split supply, in order to get the full swing the A/D would accept, rather than settling for the 0-9VDC swing all the rest were supplying then. Almost all the A/Ds will accept a wider range. Having the full input swing at the A/D gives you twice the effective resolution.

 

I only made a handful back then, for friends, but Ed Brown (previously of Estes rockets) makes them now, to order; essentially at cost, plus a 6-pack.

 

Lloyd

[Fulmen]

Nice work. Have you considered a mechanical stop to prevent overloads?

[stix]

Looks neat, now we just need a test run (and how about some blueprints? ;-) )

 

Thanks Shroed's - Yeah, a test run would be good but at this stage I've still some calibrating to do and also it's fire season here. When I've finalised the calibration, the first test will be using an Estes bp motor, which I've always used as a "standard" ie. reliable and repeatable with known measured results.

 

I'm posting an image of the electronics module, and I''ll post the so called "blueprints" of the design at a later date if anyone is interested.

 

I don't know anything about the design of your load cell amp. Back in (IIRC) 2006 (or so), I designed one with a split supply, in order to get the full swing the A/D would accept, rather than settling for the 0-9VDC swing all the rest were supplying then. Almost all the A/Ds will accept a wider range. Having the full input swing at the A/D gives you twice the effective resolution.

 

I only made a handful back then, for friends, but Ed Brown (previously of Estes rockets) makes them now, to order; essentially at cost, plus a 6-pack.

 

Lloyd

 

Thanks Lloyd. The load cell amp wasn't designed by myself - It's a module I bought from China for around $20US and the load cells (40kg) 3 for $30US delivered . Maybe that could be looked at as "lazy", but I look at it as being "clever", only time will tell.

 

My approach to designing this device (especially when it comes to the electronics) is about putting pieces together from parts that are cheap and readily accessible, and therefore, with limited electronics experience (like myself) others may be able to put something similar together. Perhaps this is the "IKEA" of thrust meter/test rigs ... ouch! - I hope not.

 

Nah! a lot of work went into the physical design and I'm confident it will last many years including the occasional cato. The hardware components are easy enough to make if you have some basic tools and time. As I eluded to in my initial post, I've made this with "expendable" parts which can be replaced.

 

The "V" shaped motor holder is made out of 1.5mm Al. The theory being that if there is an unfortunate cato, this will "give way" and therefore be safer than a thick piece hurtling though the air. It would also put less stress on the whole machine. I have no scientific evidence to back this up - it's just a hunch. It can easily be removed/attached via wingnuts, therefore it's easy to enough to replace.

 

I'll post pic electronics pic to show what it consists of.

 

Nice work. Have you considered a mechanical stop to prevent overloads?

 

Thank Fulmen. Yes, I did consider a mechanical stop. In fact I had a long discussion with a good friend of mine who's also into this hobby. I was wanting the stop, but after being persuaded, the conclusion was reached that with any cato or overload, the load cell should be considered unreliable. They are cheap enough, and a re-calibration would have to be done regardless - so why not change the load cell at the same time?

 

The images: (first one attached again).

 

The electronics:

• At left is the load cell amplifier.
• Top right is an Arduino Uno development board ($20US)
• Bottom right is an ADC module that increases the range and resolution of the standard Aduino. https://www.circuitar.com/nanoshields/modules/adc/

I have actually tested the load cell and electronics a while back (before xmas) and all looked good. It's taken me a while to make the hardware side. Next I have to test and calibrate over at least 20 kilos and that's going to be tricky.

 

Finally, another aspect is that I wanted the device to look good. Where "form follows function" ie. obviously it has to work to specs, but there's no reason why it can't look good. "Old School" looks perhaps, with a slightly modern twist

 

Cheers.

Edited March 8, 2016 by stix

[stix]

It's taken me some time to create basic software just to get to this point. This test is to prove the validity of the "V" mount using this type of load cell.

 

The device uses a "parallel beam" load cell, and as such, the weight on top can be shifted off center without any appreciable difference in the measured result. To take a phrase from the mythbusters - this one is "confirmed".

 

This phenomenon can be viewed in the video @ approx. 1:50. Look, look, look at the 10kg weights (filmed front on). They were moved off-centre by at least 25mm (1inch) or more - at random!!. The video also reveals some of the construction details of the apparatus. Sorry about the 6min video but it was easier to leave it "warts and all" than to bother editing.

 

https://www.youtube.com/watch?v=t3FQD5VUo5U&feature=youtu.be

Edited March 19, 2016 by stix

[Oinikis]

Now that is nice! I really like the way it looks, nicely machined and all. I also like the parallel beam layout, and that whole thing where load being off-center doesn't mess up the measurments.

 

One more thing i'm flattered by is the resolution. You have 1 gram resolution up to 20 kilograms. That's fascinating. I'm using the Arduino analog in directly, which is 10 bits, so i'm limited to 1024 values, while you have over 20 000. Personally, I think it's a bit of an overkill, plus it being 200 Hz.

 

My test rig opperates on two modes, 40 gram resolution with 32 kg maximum, and 80 gram resolution with 64 kg maximum, both at 100Hz. Personally, it was accurate enough for me, because rocket I used on the first mode had max thrust up to 20 kg, and on the second mode I had rocket press 50 kg of force.

 

But, more is more, and this is very benefitial, because you can use the same rig to test bottle rockets, and some big rockets, on the same stand, without any adjustments, which is really neat.

 

One more thing, I'd add some shields, in case of a CATO, because your finish is so damn nice, i'd be shame to have it ruined. It should protect it with a small rocket as shown, but with a bigger one, you should want some protection.

 

I said it, and I'll say it again, the finish is soo nice and satisfying. Everything I make, for example, looks like from a post apocalyptic movie, but hey, function comes first, right?

 

Also I enjoyed the weird music.

Edited March 21, 2016 by Oinikis

[stix]

Cheers Oinikis, Yes, function does indeed come first. In this pursuit there would be little point otherwise. I simply wanted to make a device that looked good as well as perform at a high standard. Whether I have achieved that goal in an "aesthetic" sense is subjective and not important. I believe I have achieved both. The next model is going to look even better!!

 

Regarding your 50kg! that's very impressive! This thrust meter can measure around 32kg max. Most of the motors I've ever made would have an average peak of about 6 to 7kg (14lb) with 15gk being the highest. I like to call this the "Experimenter Series" where one can make smaller motors and evaluate the results accurately or pick up the pieces and easily rebuild if something goes wrong.

 

I'm still not convinced about the sample rate not being very important. I think that higher is better to a point. For instance, if you did a test and measured accurately that there was more pressure than you thought, then that may persuade you to take a different direction.

 

Whilst the device is part dismantled for calibration, I'm going to do some static "impact and response" tests and record the results. This has already been done before by a very knowledgeable friend of mine and I'm sure by many others before. Nevertheless, I have to do it myself.

 

The next post I make in this thread will be "impact and response" tests, hopefully minus the "weird" music, unless of course you like that sort of thing.

 

FYI. The weird music is: Hawkwind - Quark, Strangeness And Charm - Circa 1977.

[Peret]

A high sample rate is not very important. The motor, holder and load cell all have mass and inertia, and cannot respond to very rapid changes in thrust. The inertia tends to average it out and a sample rate of 100/sec is more than adequate.

[stix]

A high sample rate is not very important. The motor, holder and load cell all have mass and inertia, and cannot respond to very rapid changes in thrust. The inertia tends to average it out and a sample rate of 100/sec is more than adequate.

 

Cheers Peret. I think a higher sample rate could be worthwhile. I'm intending to finish the tests in the next few hours and post the results. Regarding the mass and inertia making a difference, yes, I absolutely agree. To what extent, I'll find out soon enough and post my findings.

[stix]

The tests are done. It's late (11.30pm) so I will do the evaluation of the data tomorrow or next day and provide meaningful graphs. I'll also have to do some "confirmation" tests to ensure the data is consistent. My preliminary observations of the data suggest that higher sample rates are indeed beneficial, but only if you want to record accurately.

[Peret]

"only if you want to record accurately", lol. Try this. Record at whatever high frequency you like, integrate and calculate the impulse. Then integrate again, this time skipping samples to bring it down to a sample frequency of 100Hz, and calculate the impulse again. See if there's any difference.

[stix]

"only if you want to record accurately", lol. Try this. Record at whatever high frequency you like, integrate and calculate the impulse. Then integrate again, this time skipping samples to bring it down to a sample frequency of 100Hz, and calculate the impulse again. See if there's any difference.

 

Yes Peret, I absolutely agree with that statement. I'm not saying, nor did I ever say that a higher sample rate would yield any noticeable difference with the calculated impulse of the motor.

 

What I am suggesting however, is that a higher sample rate will enable you to precisely scrutinise the recorded data. For example, sudden spikes due to voids in the fuel or from residue build up at the nozzle or even unusual vibrations etc. These events would be short and sharp and likely missed at 100Hz. (refer to the attached graph).

 

The device I've built is for the purposes of "experimenting" not just testing. Various fuel compositions, fuel and nozzle geometry can be tested, evaluated and reviewed with confidence. Detecting any anomalies early could help make an informed decision about which direction to take next. The device will also be coupled with intuitive Software - yet to be finished because I've become sidetracked with this sh*t of my own making

 

I don't see how having more data to review is an issue (better than not enough). It's not like it has to be hand drawn on graph paper. You can decide to ignore the data and still get the same impulse, or you can analise further - one has a choice. This device enables me to record at 250Hz and it doesn't cost any extra to do so, nor take up much extra disk space - it's not high definition video.

 

If you're able to record at a higher sample rate, then why wouldn't you? It's no extra effort and I don't see any negatives at all - only positives, and would be happy to hear contrary views.

 

-----

 

Anyway, about the Graph.

 

 

Basically the testing apparatus I set up dropped a steel ball from a consistent height onto a small metal plate which was secured to the load cell. I can assure you that the drop height was accurate within +/- 1mm. Good enough I reckon. The metal plate plus attachments weighed approx. 400gm which would be of similar weight to an average rocket motor for testing on this device.

 

A series of 20 tests were done. 10 at 100Hz and 10 at 250Hz. The 100Hz tests produced a total impact duration of approx. 4 samples (4 one hundredths or 40 milliseconds). The tests at 250Hz produced approx. 10 samples over the same impact period (40ms), which is what you would expect. The data points from each of the ten tests were averaged and a graph produced.

 

What does it mean? Well, probably not much but you can clearly see that there's a difference of around 800gm between the two. Why? Because sampling at a higher rate gives you a better chance of recording the point of impact (2.5 times in this case). There is no reason however that recording at 100Hz should not eventually record the same peak force of 1200gm that was achieved at 250Hz. It comes down to chance, probability and averages.

 

As expected the inertia of the 400 gram (pseudo motor weight) caused "over and under-throw" which accounts for the bounce (negative force). It's interesting that the "bounce" is only revealed in the 250Hz tests. It's interesting that the bounce is only "revealed" in the 250Hz tests.

 

It's also worth pointing out that the main impact duration is around 10ms and therefore would have an infinitesimal bearing on the overall total impulse of the motor or specific impulse of the fuel. Nevertheless, I like the fact that my device is capable of recording these events especially if you like to record accurately. lol indeed

 

Cheers.

Edited April 16, 2016 by stix

[Peret]

Your graph does indeed show a difference between the high and low sample rates. But now integrate those curves to determine the impulse, which is what determines the acceleration of the rocket. I think you will find that the area under the curves is so close that it makes no difference.

[stix]

Your graph does indeed show a difference between the high and low sample rates. But now integrate those curves to determine the impulse, which is what determines the acceleration of the rocket. I think you will find that the area under the curves is so close that it makes no difference.

 

Ok Peret, you've lost me. I am aware that your signature shows an interest in Physics and maths, so I'll concede that you have more experience than myself. When you say integrate, do you mean taking an average? Can you please expand on what you said in a simpler way.

[Mumbles]

What he means is to measure the total impulse from both sets of runs, and see if they are the same or similar. Impulse, as I'm sure you know, is force applied over a unit of time, and is essentially the total force applied for our purposes. Your rig directly measures force at distinct time points. To get impulse from that data, you need to determine the total force over a unit of time. This is most easily done by finding the area under the curve you generated, more commonly known as integrating the data.

[Peret]

Thanks Mumbles. Stix, I didn't mean to confuse you. A larger force will accelerate a mass more than a smaller force, but only for so long as that force is applied. So integration means taking each incremental measurement of force and multiplying it by the time increment over which it occurs. Do this for every measurement interval and add them all together for the whole measurement period. This gives you the area under the curve, which is called the impulse. The impulse is the most important single parameter of a rocket motor and happens to be equal to the momentum of a unit mass subjected to it, from which - knowing the mass of the motor - you can derive its final speed, height, lifting capacity and everything else as well..

 

https://en.wikipedia.org/wiki/Impulse_%28physics%29

 

When you present the output of your rig measurement it needs to be expressed as impulse, in Newton-seconds (kg.m/s), otherwise the rocket guys won't be able to use it.

[stix]

Thanks Mumbles & Peret.

 

The confusion (on my part) arises from my ignorance of scientific language and not from my testing or calculated results. Perhaps I've gone about things in an "arse-about" way, but that doesn't mean the results are incorrect.

 

I've done the impulse calculations as suggested - I believe it's correct.

 

[edbrown]

If I may suggest it, I would take the average of the force between the start and finish of each increment and multiply that by the duration of that increment to determine the impulse within that increment. Repeat this for each increment of time and add them to determine the total impulse under the curve.

Ed

 

In the past I was using an analog to digital board that sampled force 3500 times per second. At the time my pc was limited in speed and rather than try to use each individual sample value, I only recorded every 7th sample. This gave me an effective sample rate of 500 samples per second. For production testing, I only recorded every 35 samples for a 100 sample per second rate. Peret is correct. The values of total impulse, regardless of sample rate were essentially the same. That said, there are times when there is other information in the thrust curve that is of value with a higher sample rate.

Ed

[stix]

Appreciate your input Ed. Sounds like you had some nice equipment to work with - was that at Estes, or from your own personal pursuits?

 

The possible benefits of higher sample rates for the most part is "much ado about nothing" but obviously I agree with your statement "there are times when there is other information in the thrust curve that is of value with a higher sample rate." - and that's all I was wishing to point out. Perhaps you should be a diplomat .

 

I'm more than happy to forget about impact/response tests and move on to more interesting and constructive discussions regarding calculating the average impulse and therefore the total impulse of amateur rocket motors.

 

Ed, your suggestion of "take the average of the force between the start and finish of each increment and multiply that by the duration of that increment to determine the impulse within that increment. Repeat this for each increment of time and add them to determine the total impulse under the curve." Does that include Zero?

 

There seems to be various ways of determining the average. ie. Peret's method of "taking each incremental measurement of force and multiplying it by the time increment over which it occurs. Do this for every measurement interval and add them all together for the whole measurement period". This seems like a simpler method and that's what I used for the impact tests, which didn't take account of Zero.

 

The method I was previously using is long winded, but achieved similar results.

 

Do we actually need zero? Considering the object is at rest, then shouldn't "that" zero be included in the calculations to determine the average?

 

 

[edbrown]

Hi Stix,

 

Both at Estes and for fun myself. Hard to believe I've been retired for 10 years already.

 

The first increment does start at zero. The value to be added to zero is the force at the end of the first increment, we'll say 2 lbs for an example. The 0 plus two divided by 2 would be an average thrust of one pound for that increment. If our sample rate is 100 samples per second then each increment would be 0.01 seconds long. For this first increment, the area under the curve would be 1 pound multiplied by 0.01 seconds or 0.01 lb-seconds.

 

When I started at Estes in 1966, all we had in production was a mechanical thrust stand using adding machine paper, a spring loaded arm with a ball point to record the thrust curve, The spring loaded arm also had a damper in a container filled with antifreeze. The tape traveled at one inch/second and (if my memory is correct) the arm moved vertically one inch per pound. We had a clear plastic card with a grid of lines 0.1 inches horizontally and vertically that we laid over the thrust-time curve and manually counted the number of 0.1 inch squares underneath the curve to determine our total impulse. There is a picture of this stand in one of the articles on Vern Estes's website vernestes.com. He has copies of several articles on his website as well as a very short video of the original Mabel engine machine. Nice articles if you are interested in some of the history of Estes. Many wonderful people worked there over the years (and a few stinkers also). If you or anyone else has any questions, my email is edwindbrown@charter.net. I will try to answer as many as I can without infringing on proprietary information. I think one of the old newsletters may have had an article on determining the area under a thrust curve. If I can find it on the web and if you send me your email address I will send a copy to you.

 

Best wishes and looking forward to more info on your new thrust stand. I have one of Peret's ACME stands and several other stands. I'm kind of a nut about thrust stands and wish I was a better programmer so I could do more on that end of things.

 

Regards,

Ed

[lloyd]

Ed, if you need programming, there are folks among us to do that. You've added so much 'meat' to our pyro banquet in the couple of years I've known you, that I'd trade months of programming work to just be able to walk through your head for a while!

 

LLoyd

[lloyd]

Ed forwarded this to me for reposting: It's the original test code he wrote.

 

I will review it tomorrow, so it's "where is, as is" without any of my acerbic opinions attached! <G>

 

10 ' ***********************************

20 ' * BoxTestL.Bas *

30 ' * Lawson 141 A/D Thrust Stand *

40 ' * by Ed Brown 09 Jan 2003 *

 

Full version attached as a plain-text file. (cr/lf line terminations).

 

Lloyd

BOXTESTL.txt

[edbrown]

http://www.accur8.com/Nov64smOffice.pdf Is a link to one of the old Model Rocketry News that has an article titled "Rocket Math" on pages 4 and 5 that gives a simple explanation of finding the area (total impulse) under a thrust time curve.

Hope this is of interest.

Ed

[stix]

Thanks Lloyd and Ed for those posts - interesting reading.

 

Ed, I think you're too modest about your programming skills. I could follow (BOXTESTL.txt) to a certain point but then I got lost. Looks like you wrote a very comprehensive solution - very impressive. The last program I wrote (5yrs back) for recording rocket data was using QuickBasic - it could record, create a graph and do all the calculations but didn't have the auto trim option nor was it able to select load cell options.

 

I've programmed in Visual Basic and some other languages but I've always been a fan of the procedural nature of basic and found it quicker to get results. But I've realised that in order for others to be able to use my software, I need to be able to program for the windows, mac or linux OS. So, Basic, BasicA, QuickBasic then RealBasic and finally an extension of all that is Xojo, which took over realbasic a couple of years back.

 

Xojo has a familiarity to basic, but the bitch is that you have to pay around $100 to create an executable for others to use. There's Microsoft's free "community" editions but I couldn't be bothered learning C# or that version of Visual Basic just to create some simple software - so Xojo it seems may be the only solution.

 

I loved the "Model Rocket News" from 1964. I wished I was there. Considering that it was 5yrs before the moon landing, then what an amazing point in time that must have been - unless of course, you believe it was a hoax setup in some hollywood studio .

 

Anyway, getting back to solving the total impulse of a rocket motor, I've come to to conclusion that there is more than one method that will work, achieving the same results.

 

What I did was to take things back to basics and work through the various methods. I created a representational mock-up of a thrust/time curve of a rocket motor. The fictitious recording device was only able to record a resolution of 100grams with a sample rate of 1/10th secs (0.1). Nevertheless it serves it's purpose and can be checked.

 

The mock-up is based on the simple "slap in your face" example from Ed's posted Estes "Grid Evaluation".

 

The STIX-T1000

 

 

Working out the Total Impulse:

 

Method 1: "Counting Squares" We simply count the squares (integration of the data) - there are 210 of them. Each square is 0.1kg x 0.1 seconds which means 0.01 kg/secs (Newton-seconds can be worked from that if you like).

We end up with 2.1 kg/sec for the total impulse.

 

Method 2: Each data point (blue dot) is multiplied by the sample rate, then accumulated.

We end up with 2.1 kg/sec for the total impulse.

 

Method 3: Each data point (blue dot) is accumulated, then divided by the amount of samples, and multiplied by the total time span.

We end up with 2.1 kg/sec for the total impulse.

 

Method 4: Hard to explain, but each data point and the previous data point is averaged then multiplied by the sample rate, then accumulated. This works providing that the start ZERO and the end ZERO is taken into consideration.

We end up with 2.1 kg/sec for the total impulse.

 

Methods 1-3 are simple enough to calculate from the posted graph. Method 4 however needs further understanding. The final zero and average of it, requires that it falls at the same time as the final recorded sample - which I'm sure doesn't make sense.

 

I do realise that these tests, evaluations and questions may be more suited to "Amateur Rocketry" forums. Nevertheless, I think what I've posted can be of use. A simple set of spring scales and a video recorder and the methods (1-3) above should yield meaningful results.

 

Cheers.

 

[EDIT] Of course I realise that this has all been worked out before to a point of clear understanding - nothing wrong with a re-visit to understand some basics.

Edited April 25, 2016 by stix