Tri-axis accelerometer question

[victorf]

I have been looking at the DE-ACCM3D Tri-Axis Accelerometer:

http://www.dimensionengineering.com/DE-ACCM3D.htm

and would like to use it in a project. It is a 3.3v device. If I attach the outputs to three of a ZX-24s ADC inputs which measures the values relative to 5v, how do I interpret the value? This is confusing me.

Any enlightenment will be appreciated.

Vic

[dkinzer]

[liam.zbasic]

Now I am more confused about accelerometers and the microcontroller analog-to-digital converter. Some background... For the sake of this forum, let:
Vs = Accelerometer Sensitivity (mV/G units)
Vcc = Operating Voltage
Vout = Device analog output voltage
N = Acceleration (in G units)

Considering model DE-ACCM5G:
http://www.dimensionengineering.com/DE-ACCM5G.htm

Vs = 0.312V/G
Vcc = 5V
Range= +/-5G

According to their spec (DE-ACCM5G.pdf), 0G corresponds to Vcc/2, or 2.5V. Therefore, the measured acceleration is as follows:
N = (Vout - Vcc/2)/Vs

Checks:
If Vout=2.5V, then N=(2.5 - 5/2)/0.312=v0G

If Vout=3.06V, then N=(3.06 - 5/2)/0.312= 1.79G (matches spec)

Here is my confusion... the range of the model in question is +/-5G. The corresponding output voltage at +5G is only 4.06V (see calc below), short of 5V as in the accelerometer Vcc and the default microcontroller 5V ADC range.

Vout = N*Vs + Vcc/2

Vout = 5G*0.312V/G + 5V/2 = 4.06V

This means there is a further degradation in resolution beyond the 10 bit ADC quantization. Is this right? Am I missing something? If my reasoning is correct, should the ZX device reference voltage equate to the accelerometer Vcc (+5V) or more appropriately to the voltage corresponding to the accelerometer +5G range (in this case 4.06V)? Thanks.


Liam

[dkinzer]

[mikep]

Liam, I think you are right. But only you can determine if this is a problem or not.

For 5V, the range of input voltages will be between 0.94V and 4.06V i.e. between 19% and 81% of the input voltage. When the supply voltage is reduced, the range is about the same. As it says in the ADXL320 datasheet - the device is radiometric and the output senstivity varires according to the supply voltage. So for example if the supply voltage is 2.4V, the range is 0.525V to 1.875V i.e. 22% to 78%.

For the 5V supply the ADC output values will range between 192 and 832 which is about 60% of the available range. the 10-bit ADC has a resolution of about 5mV which equates to 16mG. If you make AREF 4.06V then the resolution will increase to 13mG and you are still going to lose some range at the bottom end.

You have to decide whether this resolution is sufficient or if you need something better by using a higher resolution ADC.

[liam.zbasic]

Thanks for your responses. Very much appreciated. The DE-ACCM5G was only used as an example. Other accelerometers yield better or worse resolution. That said, I'm a bit confused about your resolution in terms of G per step (dbkinzer calculated 0.004G/step). Here's why...

I think its agreed that the device analog output voltage is:

Vout = N*Vs + Vcc/2

Back to the DE-ACCM5G accel (sensitivity = 0.312V/G, Vcc=5V, Range=+/-5G), at +5G and -5G, the corresponding Vout values are: (assuming ZX reference voltage = 5V)

N=+5G, Vout = +5*.312 + 5/2 = 4.06V
N=-5G, Vout = -5*.312 + 5/2 = 0.94V

The voltage spanning from +5G to -5G (10G span) is:

delta-V = 4.06V - 0.94V = 3.12V
delta-G = +5G - (-5G) = 10G

The corresponding number of ADC steps is:

Steps = 2^10 * 3.12V/5.0V = 639 steps

Therefore, the resolution in terms of G per step is:

G/step = 10G/639steps = 0.016G/step

Am I missing something here? Again, the accelerometer model chosen is only an example, so the resolution value is not germane - its the approach that is important.

Adding linear circuitry (i.e., op-amps) was suggested as a way to improve resolution. As you may well have discovered, I'm not an electronicist, so I find a minimalist but functional approach. Adding linear circuitry probably adds more noise and more variation with temperature to the overall integrated system. I prefer setting the ZX reference voltage (Vref) to a desired value. This may require more circuitry, unless its possible to set Vref using the putDAC function and hooking this channel to AVcc? Is it possible to combine DAC channels to synthesize a voltage beyond 5V? I'd try it if I knew it wouldn't fry my ZX chip. Thanks for your comments.


Liam

[mikep]

[dkinzer]

[liam.zbasic]

I appreciate your responses very much. I now understand the scaling.

I don't have a specific application at this moment nor a specific resolution. I did want to understand how to process accelerometer output in order to apply on a myriad of projects including a simple data acquisition unit for vibration testing (up to 100Hz), tilt sensing, a balance bot, and ultimately a vibration suppression system. Some will require more resolution than others, so I was looking for a catch-all (highest system resolution).

For now, I will simply find an accel that yields the best system resolution without adding external circuitry. I'll start with a +/-3G range and at least 250mv/G sensitivity. I'll use a single voltage regulator to modify Vcc (as close as possible) to get an output of 5V at +3G.