Rocket Thrust Instrumentation

The Rocketry Forum

Help Support The Rocketry Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
$125 for a data collection system is as inexpensive as you will get. You will spend more cobbling together something non-DataQ based than the $125 cost. I just went through this with a University. They went the sparkfun route then ditched it after they saw how I could show up with at DataQ and get a graph and calibrate in real time.

Edward

Alpha, we recently began coordinating with USC on this, and hopefully, after these initial tests, we will have access to some of their hardware and base our set-up on theirs. I have no doubt that Dataq is better, but the marginally better performance doesn't outweigh the cost or the time it will take to get it up and running. We hope to test on the 2nd of April, so we simply didn't have the time to become acquainted with a different system by then. Which university were you working with?

-Jack
 
The $60 DataQ runs at 240 Hz, and the time it would take you to get acquainted with it is about 15 minutes. No programming, just a nice chart recorder. Calibration while the recorder is on is invaluable, as well as the ability to toggle between voltage and engineering units.

The university team I'm working with spent about 8 hours trying to get the Sparkfun working. It would quit out, have errors and the data wasn't reliable. Now, imagine that you test your M or P motor and your data collection quits on you and you get nothing. How much did that cheap data system cost you now? The time making the motor, the cost of the motor and you have nothing to show for it. Will your sponsors pony up more cash because your reason is "Our data acquisition system didn't work..."?

I've been testing motors for 15 years and working with the local university for three now. DataQ is the way to go. Others here recommend it, and there is a reason for that. I'd listen.

Edward
 
Last edited:
240Hz if run as single channel 120Hz if run as dual channel recorder. Running 4 channels on their older entry level device each channels was good for 60Hz. If one buys their software they can record at higher rates. DataQ is economical and the only additional expense is an amplifier and load cell. It's worth the investment.
Whoa, I just checked their site, the $29.00 units are replaced with $59.00 units that have kHz range capability now. Kurt
 
Last edited:
Also, the DI-194 that has a serial connection and I bought in 2003-4 still works great and has software support.

Edward
 
Just checked DATAQ web. the DI-1000 that replaces the DI-145 (I have 2 of these they are great) is available pre-order March delivery. instead of 240 HZ, you get 40kHZ if using one channel, 30kHZ for two then drops to 24kHZ for three and 20kHZ for all four. I use two, force and pressure so 30kHZ! If pressure sensor fast enough you would be able to record small pressure spikes. I put one on pre-order.
 
Agree - many pressure sensors have 1khZ BW - sampling them at something above the Nyquist rate can give very valuable information and avoid aliasing.
For those not "sampling savvy," Nyquist said ideally you need two samples per cycle. Using 3-4 is a more realistic multiplier. So >=3k samples/sec for 1KHz BW sensor.
 
Also bear in mind if you are sampling around 1kHz you generally get a fuzzy caterpillar data line. I sampled chamber pressure at 25kHz once and had a black bar ~25 psi wide for data. I have to down sample and do some maths to get something useable.

Also, the post processing takes a bit longer because instead of a couple thousand data points you now have over a hundred thousand.

Edward
 
Last edited:
Agree - many pressure sensors have 1khZ BW -
sampling them at something above the Nyquist rate can give very valuable
information and avoid aliasing.
For those not "sampling savvy," Nyquist said ideally you need two samples per cycle. Using 3-4 is a more realistic multiplier. So > =3k samples/sec for 1KHz BW sensor.

This ignores the critical details of the sensor response. It is nearly certain that the sensor response roll off is first order (6dB/octave) and the bandwidth figure gives the point at which the response is 3dB down. The attenuation at the Nyquist frequency will be minimal

In addition many sensors have a resonant frequency well above the -3dB point where the response rises. Any sensor that is basically mechanical like pressure, load, and acceleration will have this feature.

Depending on signal (and noise!) you may need a lot more attenuation to get the signal + noise below 1/2lsb of the ADC at Nyquist. For SAR type ADCs a minimum of a second order filter at 1/10 of the sample rate (giving the filter a couple of octaves to work with) and preferred would be sixth order (Bessel response). Sigma-Delta ADCs push the Nyquist frequency up quite a bit but you can't ignore the anti-aliasing filter. Its requirements are greatly relaxed but you still need one.

A search of the Internet will of course turn up lots of information on the subject. Microchip provides a decent starting point.
 
Here are the specs for the pressure transducers I have been using.
[h=2]Product description[/h] This is an industrial style transducer for use with oil, gasoline, diesel, or even air!
Specifications:
* Input: 0-1600 psi.
* Output: 0.5V - 4.5V linear voltage output. 0 psi outputs 0.5V, 1600 psi outputs 4.5V.
* Works for oil, or air pressure. Can be used in oil tank, gas tank, etc.
* Accuracy: within 1.5% of reading (full scale).
* Thread: 1/8"-27 NPT
* Wiring connector: water sealed quick disconnect.
* Mating connector is included.
* Wiring: Red: +5V. Black (yellow): ground. Blue (green): signal output.
* Working Temp: -20C - 125C (-4F - 260F)
* Over pressure: more than 200% of full scale (10+PSI)
* Response time: 50ms
* Current draw: < 4ma
* Slope = 25 and y-intercept = -12.5

So I doubt faster sampling is of much value. The fairly slow transducer probably helps by averaging the pressure some what. I have not thought about Nyquist theorem for many years, if I remember correctly it had more to do with reconstructing an analog signal or placing an analog signal on a multiplexed data link. Use to work in telephone moons ago. Before all the digital, the analog was just placed on the bus in time slots and recovered from time slots.
 
This ignores the critical details of the sensor response. It is nearly certain that the sensor response roll off is first order (6dB/octave) and the bandwidth figure gives the point at which the response is 3dB down. The attenuation at the Nyquist frequency will be minimal

In addition many sensors have a resonant frequency well above the -3dB point where the response rises. Any sensor that is basically mechanical like pressure, load, and acceleration will have this feature.

Depending on signal (and noise!) you may need a lot more attenuation to get the signal + noise below 1/2lsb of the ADC at Nyquist. For SAR type ADCs a minimum of a second order filter at 1/10 of the sample rate (giving the filter a couple of octaves to work with) and preferred would be sixth order (Bessel response). Sigma-Delta ADCs push the Nyquist frequency up quite a bit but you can't ignore the anti-aliasing filter. Its requirements are greatly relaxed but you still need one.

A search of the Internet will of course turn up lots of information on the subject. Microchip provides a decent starting point.

Thanks for this! I learned a lot from the microchip paper.
 
Those look like the E-Bay transducers that I see for <$20. The university team I'm working with this year is using them. They work well for the price range. They calibrate them using a NIST referenced gauge the University has.



Edward
 
David - yes, agree you need to know more about the signal's total frequency domain and sample/filter appropriately.
Sampling at 60 to 240Hz is really the least appropriate as you're likely to get severe aliasing with several "wraps" of the data within that limited spectrum.
And yes, signal averaging for the "quick view" along with peak detection can provide a nice combined view of what is happening....CPU time is cheap and a few second/minutes post processing is noise compared to the total test time.
 
Those look like the E-Bay transducers that I see for <$20. The university team I'm working with this year is using them. They work well for the price range. They calibrate them using a NIST referenced gauge the University has.



Edward
Yes they are 15$ free ship. have not blown any up lately! keep several in stock. I have not calibrated as yet, only go off sheet. How close are they? Linear? I just use the 4 volt they are suppose to change.
Bill
 
They were using 1000 PSI transducers. One transducer was 0.54V at ambient pressure (we are at 5,000' elevation.) 800 PSI was 3.32 V

Edward
 
I have not thought about Nyquist theorem for many years, if I remember correctly it had more to do with reconstructing an analog signal or placing an analog signal on a multiplexed data link.

It can cause other trouble too. As an example I include a plot of data recorded on an RDAS altimeter at 50SPS. The motor used on this flight was a Hypertek hybrid known to oscillate. The data looks awful and it didn't help the deployment algorithm at all.

alias.png
 
It can cause other trouble too. As an example I include a plot of data recorded on an RDAS altimeter at 50SPS. The motor used on this flight was a Hypertek hybrid known to oscillate. The data looks awful and it didn't help the deployment algorithm at all.

I read somewhere (on the web) about an algorithm that some body wrote to counter that. Not sure it works, I'll search for it.
Bill
 
Huge thank you to Steve G, sghioto, for donating an amplifier and older Dataq to our program! His generosity is extremely appreciated by everyone from our team.

And thanks to all of you for the advice; I have canceled the sparkfun order because our mentor was able to accommodate pushing back the test to April 8 th and there was no reason to risk it using faulty products. We'll test April 8th. Will be done with machining by next week. Will keep you posted.

Jack Oswald
 
[video=youtube;BJSg2elL_KQ]https://www.youtube.com/watch?v=BJSg2elL_KQ[/video]

The tests went great! As you can see, the thrust data turned out (briefly shown in the clip) very well. You guys were right; Dataq is a great system.

Jack
 
Speaking of DATAQ, does anyone know how to stop the waveform viewer from activating the keyboard scroll-lock when you open it? I've been using it lately for the 1st time in many moons on a new notebook and forgotten how to do it.

Thanks,

Troy
 
It's possible. You need to copy the waveform to the clipboard and import it to a graphics program like Windows Paint. Save file as a jpeg then you can upload to Forum.

Steve G

E9-0 #2.JPG
 
Back
Top