Thrust Test Stand question

The Rocketry Forum

Help Support The Rocketry Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

shockie

High Plains Rocketeer
Joined
Jun 2, 2009
Messages
1,185
Reaction score
625
Location
My Old Kentucky Home
Frequency response of the thrust test stand equipment shall be at least 100 Hertz, and the natural frequency of the
equipment shall be at least 5 times this number, or 500 Hertz.

Can somebody explain this to me like I'm a 5 year old? perhaps with examples.
 
https://en.m.wikipedia.org/wiki/Natural_frequency -- how fast does it want to wiggle

https://en.m.wikipedia.org/wiki/Frequency_response -- how fast can it record

Don't let it wiggle anywhere near as fast as it can record.
Does this have anything to do with the sampling rate?.

How does one determine the natural frequency of something? Is that a function of the load cell specs?

I assume that the sampling rate comes down to resolution?

Is there any online resources that describes or informs on tgrust-test stands?
 
The basic sample rate is determined primarily by the required data bandwidth. If your required data bandwidth is 100 Hz, you should be sampling at at least twice that rate, or 200+ samples per second. This is the Nyquist sampling rate to avoid aliasing. Although, if you are oversampling (e.g., 4x basic rate) to improve resolution, the data bandwidth stays the same.
The natural frequency of the test rig is essentially the dominant resonant mode of the rig when you input a disturbance, such as a motor under test. Everything vibrates when it is disturbed. You want these resonant modes to be well clear of the required data bandwidth - in your case x5.
 
Last edited:
Frequency response of the thrust test stand equipment shall be at least 100 Hertz
As with any ADC system you need to include a presample filter to limit the frequency content prior to the ADC. So the low pass filter should have a corner frequency of at least 100Hz.

The sample rate should of course be much higher than that. If your filter were perfect you could use 200SPS but there are no perfect filters. A sample rate of at least 10 times the filter corner frequency should be used. Then design the filter order so that you get the desired attenuation at the Nyquist frequency. (1/2 the sample rate)

A second order filter (easy to do with one opamp) gives you 12dB/octave of attenuation. So with a corner at 100Hz you would get 24dB of attenuation at 400Hz. If that isn't enough to remove signal or noise above Nyquist you add filter stages. (Filter response type is of course another concern. Bessel is preferred here because it preserves phase/time alignment better than Butterworth which has a flatter frequency response in the pass band.)
 
If the stationary part of the system has low mass, the thrust will jiggle it -- beef up!

If the moving part of the system ( e.g. downstream of loadcell ) has high friction and/or mass, response will be sluggish -- slim down!

I don't actually know how to find the natural frequency, so instead I'd just buy as many samples per second of both pressure and thrust as my budget allows for.
 
The reason that you want the natural frequency of the system outside the pass band of your data acquisition is because the system will have high Q at that frequency. In other words, it will tend to amplify the signal present at that frequency.

Even if calculating that frequency is difficult, testing isn't. Just fire up the data acquisition system and whack the test stand with a spanner or some such. It should be fairly obvious if you have a problem even without whipping a FFT on the data.
 
If you are using the HX711 load cell amplifier, it has selectable sample rates of 10 sps and 80 sps. You might require a different load cell amplifier if you want 100 Hz BW. The HX711 does have an external clock input, but I'm not sure whether you can get a higher rate with it.
 
As with any ADC system you need to include a presample filter to limit the frequency content prior to the ADC. So the low pass filter should have a corner frequency of at least 100Hz.

The sample rate should of course be much higher than that. If your filter were perfect you could use 200SPS but there are no perfect filters. A sample rate of at least 10 times the filter corner frequency should be used. Then design the filter order so that you get the desired attenuation at the Nyquist frequency. (1/2 the sample rate)

A second order filter (easy to do with one opamp) gives you 12dB/octave of attenuation. So with a corner at 100Hz you would get 24dB of attenuation at 400Hz. If that isn't enough to remove signal or noise above Nyquist you add filter stages. (Filter response type is of course another concern. Bessel is preferred here because it preserves phase/time alignment better than Butterworth which has a flatter frequency response in the pass band.)
UNChem: At first I will be using a HX711 load cell amplifier with an Arduino Uno at 80sps......I will only at first be using small BP motors like 1/4A thru A 13 mm Estes BP...... this should be sufficient for 2.5N-s motors and below?
 
this should be sufficient for 2.5N-s motors and below?
No. You actually need higher frequency response with smaller motors. The frequencies of things like combustion instability increase with smaller motor dimensions.

The HX711 is a fine choice for measuring static loads but the frequency response is so horrible that it is next to useless for anything else. It uses a sigma-delta converter and while the data sheet says nothing about frequency response, that architecture always has poor response. I seem to recall taking a SWAG at the number in a previous thread.
 
No. You actually need higher frequency response with smaller motors. The frequencies of things like combustion instability increase with smaller motor dimensions.

The HX711 is a fine choice for measuring static loads but the frequency response is so horrible that it is next to useless for anything else. It uses a sigma-delta converter and while the data sheet says nothing about frequency response, that architecture always has poor response. I seem to recall taking a SWAG at the number in a previous thread.
by higher frequency response you mean I need a higher sampling rate?
 
David, you know I respect and value your opinion. You are obviously a lot smarter than I am, but this guy seemed to get a decent T-T curve from a C6:
Presumably you are doing this for some other reason than to replicate (badly) a thrust curve you can find on the NAR web site.
 
Presumably you are doing this for some other reason than to replicate (badly) a thrust curve you can find on the NAR web site.
just build and play with a relatively inexpensive thrust test stand to gather some motor data ... I was going to put up a gridded backstop so I can measure such things as the width, length and diamter of the exhaust plume, get a simple T-T curve, video it at around 1000 fps.....just stuff like that.....
 
An easy way to find resonance of a structure is to tap it with a hammer and listen with a microphone. An oscilloscope or other frequency measurement tool can be used to provide the frequency. If the structure is simple this works well. For complex structures I guess other methods like simulation, would offer better results.
 
DATAQ DI-1100 - https://www.dataq.com/products/di-1100/
High sample rates per channel (dependent on the number of channels enabled):
  • 40 kHz per channel, 1 enabled channel
  • 30 kHz per channel, 2 enabled channels
  • 24 kHz per channel, 3 enabled channels
  • 20 kHz per channel, 4 enabled channels


INA125BB-BP Amplifiers - https://www.picom2.com/Phillips_Instruments.html - need one for each channel.

PX319-2kgV - https://www.omega.com/pptst/px309.html - Pressure sensor - I have not gotten this yet, but Scott Szympruch from MDRA has this item.

Load Cell - https://www.robotshop.com/en/micro-load-cell-5-kg.html - also available in 0.78kg, 20kg, 50kg.

No need for all the math... AND these components are not expensive. Remember to set up your load cell to fire horizontally to that changing motor mass does not effect the curve.

Load Cell 1.jpgLoad Cell 2.jpgLoad Cell 3.jpgTest Stand 1.jpgTest Stand 2.jpgTest Stand 3.jpgTest Stand Calibrate.jpg
 
DATAQ DI-1100 - https://www.dataq.com/products/di-1100/
High sample rates per channel (dependent on the number of channels enabled):
  • 40 kHz per channel, 1 enabled channel
  • 30 kHz per channel, 2 enabled channels
  • 24 kHz per channel, 3 enabled channels
  • 20 kHz per channel, 4 enabled channels


INA125BB-BP Amplifiers - https://www.picom2.com/Phillips_Instruments.html - need one for each channel.

PX319-2kgV - https://www.omega.com/pptst/px309.html - Pressure sensor - I have not gotten this yet, but Scott Szympruch from MDRA has this item.

Load Cell - https://www.robotshop.com/en/micro-load-cell-5-kg.html - also available in 0.78kg, 20kg, 50kg.

No need for all the math... AND these components are not expensive. Remember to set up your load cell to fire horizontally to that changing motor mass does not effect the curve.

View attachment 451631View attachment 451632View attachment 451633View attachment 451634View attachment 451635View attachment 451636View attachment 451637
I don't suppose you have a wiring diagram for the various components?

whats the purpose of the pressure sensor when you have a load cell? $275 isn't cheap.
 
I don't suppose you have a wiring diagram for the various components?

whats the purpose of the pressure sensor when you have a load cell? $275 isn't cheap.
Pressure and thrust together gives you a more comprehensive look at how the motor performs and allows you to characterize the propellant for how it will perform in other sized motors.
 
@shockie , I think your arduino and the HX711 are probably just fine for what you want to accomplish. I have been using a homemade arduino DAQ with an HX711 with a load cell and an analog 5v pressure sensor and it works just fine, unless you are certifying motors for Tripoli or really need to see really high resolution data. I've been meaning to upgrade the HX711, as it only provides 80 sps, but I'm lazy and my rig works great to characterize my motors. You don't need pressure AND thrust (usually just one or the other), but having both gives confirmation that each sensor is performing accurately. Also, pressure is usually running milliseconds ahead of load and trails off after load (especially if testing vertical). Belts and suspenders.

Below is a graph from a 2 grain 75mm motor I static tested yesterday that burned for about 4 seconds. The blue is pressure and the orange is load. The direct analog pressure reading is sampling at the same 80hz and you can see that there is really very little latency with the HX711 at this resolution. I scaled the load (blue) line up by 2x in the second graph, so you could better see the tracking between pressure and load. What looks like noise in the burn (lulls, peaks, and valleys) is actual performance data, confirmed by both sensors. Again, this is likely to work just fine for what you are trying to do and/or using an Arduino and a load cell is a great place for you to start learning.

Screen Shot 2021-02-21 at 12.51.32 PM.png

Scaled up load (by 2X) for better comparison
Screen Shot 2021-02-21 at 12.51.49 PM.png
 
I don't suppose you have a wiring diagram for the various components?

whats the purpose of the pressure sensor when you have a load cell? $275 isn't cheap.
Although the chamber pressure can be estimated from thrust, actual measurements of pressure are more reliable.
 
I have a few extra pressure sensors if anyone needs one.
1000 PSI, 5V, 1/8 NPT. Datasheet: M3431-000006-01KPG
MSRP is $139. Digikey is $90+.
I'll include shipping for $60 each.

I'll follow up here soon with info on how to combine thrust and pressure data to better characterize your propellant & motor configuration.
 
Back
Top