Actually the data looks like it has an aliasing problem due to a sample rate slower than the bandpass of the sensor/amplifier.
The bandpass of the sensor should be limited to 1/2 of the sampling rate (Nyquist sampling theory) by a low pass filter so the high frequency oscillations are filtered out. In most altimeter it is not.
You can take the data and pass it thought a digital low pass filter, or take a FFT of the data, perform a digital low pass filtering by getting rid of the high frequency data above the Nyquist frequency, and then do an inverse FFT to generate the original data minus the high frequency components.
You will have a much smoother curve to run through your Cd calculation program.
Bob
Actually the data looks like it has an aliasing problem due to a sample rate slower than the bandpass of the sensor/amplifier.
CJL
I agree that it is influenced by the resolution limit, however every accelerometer data I have seen has has several bits of noise on it. Neither the Parrot nor the RDAS sample the accelerometer above the accelerometer response frequency so the high frequency noise contribute to the noise in the measurement. If you had a 50-100 Hz LP filter on the input to the digitizer the noise level would be lower than what you typically obtain. You can obtain higher synthetic resolution by performing a n-point average of the data before you calculate the Cd and you will get a smoother curve.
Bob
The problem is that many accelerometers can only measure increments of 0.1G (or, perhaps slightly more accurately, the ADC can only distinguish changes of 0.1G in the output signal of the accelerometer). Slower rockets will only ever have a deceleration due to drag of 0.5G or so. As a result, there's really only about 5 or 6 discrete steps that can be measured. The reason that each step is a curve is because the conversion from measured deceleration to Cd is as follows:
1) Convert deceleration to drag force using the known mass of the rocket
2) Determine dynamic pressure from current velocity and altitude
3) Cd = drag force/(dynamic pressure times frontal area).
Since dynamic pressure scales with velocity squared, a line of constant deceleration will look like a 1/v^2 plot, which is exactly what is observed.
Now, the faster your rocket goes, the more accurate the Cd measurement is. At higher speed, the rocket has substantially higher drag, and this substantially higher drag-induced deceleration. Because of this, the relative accuracy of the altimeter's accelerometer goes up.
(I've done the analysis manually a number of times)
CJL
You can obtain higher synthetic resolution by performing a n-point average of the data before you calculate the Cd and you will get a smoother curve.
Bob
CJL
I agree that it is influenced by the resolution limit, however every accelerometer data I have seen has has several bits of noise on it. Neither the Parrot nor the RDAS sample the accelerometer above the accelerometer response frequency so the high frequency noise contribute to the noise in the measurement. If you had a 50-100 Hz LP filter on the input to the digitizer the noise level would be lower than what you typically obtain. You can obtain higher synthetic resolution by performing a n-point average of the data before you calculate the Cd and you will get a smoother curve.
Bob
Since our rockets accelerate and the negatively accelerate, is the Nyquist criteria important, as the data acquisition is faster then the periodic variation of the data?
Sandy.
I think in the context of this conversation, resolution would be limited by the ADC, not the accelerometer, correct? I assume the accelerometer is a device that would be the chip level equivalent of a mass on a beam with a strain gage, therefore it would be 'analog.'
Sandy.
Enter your email address to join: