SparkFun Forums 

Where electronics enthusiasts find answers.

Have questions about a SparkFun product or board? This is the place to be.
By rsmurphy2
Hello out there,

I'm working on a research project to develop a rocket engine and am looking for a cheap method to collect data on vibrations generated from unstable combustion and resonance.

I'm expecting a range of frequencies from a few hundred Hz up to at least 10kHz or higher. The motor will be mounted on a rigid test stand so the motion induced by these vibrations should not be that large. Just doing a simple F=m*a calculation based on F(thrust)=2224.1N and m~5kg I get an average a of about 45g. If the combustion is very unstable the thrust will chug and produce much larger thrusts so I think its reasonable to assume maximum Acc of around 100-150g. (These calculations are assuming my rocket is not rigidly attached to the test stand, if it were the Acc would be much less due to the larger mass value in F=m*a.

I'm considering using the 250g ADXL 193 to capture this data but I'm not sure how to interpret the sensor's capabilities. The datasheet says it has a resonance frequency of 24KHz and a clock frequency of 400kHz. I'm assuming this sensor should be able to capture all the vibrations I'm looking for up to its resonant frequency? I'm really only concerned about vibrations under 15kHz as higher frequencies will not effect the stability of the rocket or cause structural design problems. ... ts_id=9332

The analog data must be fed into my LabView software for analysis so I'm looking at ADC solutions and am starting to fear that I will need to build my own due to the high price of good quality converters.
I was originally looking at a 12-bit ADC converter that ran at 115.2kBaud serial to USB.

Some other chips from Ti or AD capture 16-bit at over 1Mips for about $30 so I may end up designing the ADC around one of these chips.

Any thoughts would be greatly appreciated!
By rsmurphy2
As a side note, the ADC I was hoping to use is from National Conrol Devices. They have a 16ch 12-bit for $134. I like how their hardware is scalable and they offer lots of other stuff for I/O but I'm worried their ADC chip is not going to capture the higher frequency data.
By rsmurphy2
Is it correct to assume that the 12-bit A/D on a 115.2kBaud serial interface can only send 9600 12 bit numbers to the computer per second (assuming no stop bits)? If this is the case, one of my inputs would eat up this entire bandwidth and I'm going to have to get started on a new A/D solution...
By lyndon
I can't answer those questions, but in a past life designing data acquisition interfaces I found that Analog's applications engineers were very helpful. A trip to their site should yield an email address/phone number for someone to help you.

Odds are you'll be shunted to a Field Applications Engineer in your local area. FAE's are a great source of free samples, applications notes and general help.

By Topher
Looking at the picture of the design, it seems it uses a 16 channel multiplexer to go to a single ADC. The multiplexer alone I'm sure will bottleneck your throughput, especially at 10kHz. What kind of sampling speed do you require? If your looking for a out-of-box solution your probably gonna be spending some money.
By rsmurphy2
Topher, I got a response from their Tech support and your observation is correct. I had hoped to get away with a 10-16ch 12-16bit ADC for under $200 but it seems I will have to start from scratch to hit the desired sample rate. Some of the sensors would be sufficient to sample at <10Hz but the vibration and load cell data would be great to see at >10kHz sample rate.

I'll probably buy an out of box solution for the low sample rate inputs and build the higher bandwidth one from scratch.

Thanks lyndon, I'll be in touch with AD very soon!
By Topher
If your looking for oscilloscope quality sampling then its definitely going to cost you some money. But have you considered going with something more simple? If you know your output from the accelerometers will be sinusoidal and all you need to do is measure amplitude and frequency, there are ways around this that don't require very high end ADCs.
By UhClem
The ADXL193 is a poor choice of sensor in this application for many reasons. Number one is that its frequency response is limited by its internal 400 Hz filter. Number two is that it is designed to be mounted on a printed circuit board and not to hardware in a test stand.

A much better choice would be a piezoelectric sensor like those offered by Endevco.

For example, the 7250A has more than enough range and frequency response for your application. It is more expensive than the ADXL193 but it has the distinct advantage that it will work.

Don't forget the presampling low pass filter. You will need at least a second order filter with a corner at your highest frequency of interest. The order will depend on the sample rate with higher sampling rates resulting in lower orders. With a sixth order Bessel filter at 15 KHz you might get by with 50K SPS.
By rsmurphy2
I was unsure what the significance of the filter was, it just doesn't make sense to filter out everything over 400Hz when it seems the sensor could collect a much greater range. A quick Wiki read on bessel filters was not all that enlightening.

The Endevco sensors look great, minus the price tag. I've only come across a couple prices after about 20 minutes of online searching. About 1000 for the 500g model. I'll keep looking for info on this.

Piezoelectric seems to be the way to go for higher sample rates- I'll keep looking around for other manufacturers as well.
By rsmurphy2

I do not know what frequencies will be generated and they will likely fluctuate with chamber conditions. I can roughly estimate the modes of vibration using some equations from one of my texts but I want to compare test data to these calculations.
By motopic
As they told me when they shut down my rocket plant and moved it to Utah, "Rockets are known domain now. Thiokol Mgmt c1991".

So data acq for rockets is well known now. Reliability and oversampling are your keys, you only get one test
(or maybe 3: 1 nominal, 1 hi temp, 1 lo temp).
Also if it does go berserk and blows up, you want as much data as possible...