NI-9237 and Voltage Divider

[ngthai]

I'm a civil engineer, so please forgive if I misunderstand electrical engineering stuffs.
My NI-9237: 2 RJ50 ports (each port is connected to a 10-position connector) are connected to 1 Strain Gauge and 1 PR Accelerometer (both are Full Bridge). PR Accel sensitivity is 0.03 mV/V/g, and the NI-DAQ limitation is 25mV/V, thus I'm able to measure only to 25/ 0.03 = 833 g before the signal clips. However, I need to measure acceleration up to 1600 g.
As such, I have to add a Voltage Divider (350 + 350 ohm) to prior to the P+, P- excitation ports (G and E) of the Accel. No Divider is needed for the Strain Gauge (excitations are pin G
Without power, without load, I measured Resistance across G-K (future strain gauge) on the 10-position connector, it was supposed to be infinite, but my measurement is  350 ohm. Resistance across D-E (future Accel) was measured 263 ohm. So I back-calculated and found out that it looks like NI-9237 has a 700-ohm resistance there.
Thus the parallel resistance calculations are :
D-E:  (1/1050 + 1/350)^-1 = 263 ohm (match measured value) 
G-K: (1/700 + 1/700)^-1  = 350 (also match measured value)

I need to calculate how much Excitation is reduced when I plugged Accel and Strain Gauge on.
Accel input resistance = 6000 ohm.
My calculation is:   Equivalent resistance (1/350 + 1/6000)^-1 = 330.7 ohm 
  Thus, on 1V excitation from NI-9237, the P+ P- on Accel shall be 1V * 330.7 / (350 + 330.7) = 0.4858
   Thus, whatever voltage from S+ S- measured on the Accel, I need to multiply by a FACTOR of 1/ 0.4858 = 2.06 to get the actual acceleration.
I tested this "Voltage Divider" device with another Accel that does not require Voltage Divider, and I found out that the multiplying FACTOR needs to be 1.87 for the results to match with the non-divider accel.

Question:
1) Why 1.87? is it before of some other parallelism that I missed - And that parallelism affects excitation on D-E furthermore (than my calc above)?
2) If there is some other parallelism, then does it affect Strain Gauge excitation too (on G-K)?

[ngthai]

I have another setup, where the Voltage divider is 1000ohm+1000oh. My resistance measurements on GK is 917 ohm, and on D-E is 733 ohm. Thus I was able to back calculate and found out the supposed "internal" resistance from NI-9237 is 1700 ohm. Based on this 1700 ohm, my calculation on GK is 919 ohm (very close to 917), and on DE is 730 ohm (very close to 733).
Additional question is:
3) why "internal" resistance is now 1700 ohm?
4) similar to above calc, when LOAD is attached, I'm expecting FACTOR of 2.17 to be multiplied to my acceleration, but it appears a FACTOR of 1.94 will give reasonable matching results. Why this FACTOR does not match my calc (of 2.17) and if strain-gauge excitation is also being affected by some other parallelism I don't know?

[ngthai]

I attached my accelerometer to the system, but stripped the wires so that I can measure voltage with DVM between P+P- while the load is connected. 
Supplying Vex = Vgk = 1 V
Measured Vde = 0.486 V (example 1) so it matches perfectly with the theoretical calculation I provided.
I did induce different accelerations to see if Vde stay constant. It did stay constant at 0.486 V.

So my question is why on S+S-, by output is INCREASED by approximately 9%? (thus overall Modifying Factor changes from 2.06 to 1.87 in example 1).
(e.g., S+S- supposed to read 0.486mV so that when I multiply by the supposed 2.06 factor, I get the true 1mV of signal - assuming no Dividers connected,
but it actually read more at 0.535 mV, so the Modifying Factor is 1.87).

Is there a formulas that you can figure out so that I can calculate this output S+S- change (i.e., slightly increase more than it is supposed to be)?

[ZYOng]

Voltage divider introduces error and noise. You might want to add an unity gain buffer to isolate the load. See Unity gain amplifier or voltage follower in a voltage divider

Other alternative includes:

• Use other accelerometer with higher sensitivity value.
• Use NI-9219, which has 60mV/V input range and can measure up to 2000g for your accelerometer.

[ngthai]

Thank you so much. NI-9219 won't work because I need to sample up to 50 kHz. We couldn't find accelerometer with higher sensitivity value (yet). We'll look at the Op Amp option, but that appears to require external power which again, I'm a civil engineer so I'm not familiar with Op Amp yet.

[santo_13]

To save your engineering efforts, would you consider getting different accelerometer and DAQ pair?

[ngthai]

we haven't found PR accelerometers that can measure up to 6000 g, sampling rate up to 50kHz, damped (or multimode damped), with sensitivity at or better than 0.015 mV/V/g yet. All we found so far is PR 2000 g, 50 kHz, multimode damped, sensitivity 0.03 mV/V/g. Do you have suggestions (with accel model and link)? Thank you.

[santo_13]

---

@ngthai wrote:

we haven't found PR accelerometers that can measure up to 6000 g, sampling rate up to 50kHz, damped (or multimode damped), with sensitivity at or better than 0.015 mV/V/g yet. All we found so far is PR 2000 g, 50 kHz, multimode damped, sensitivity 0.03 mV/V/g. Do you have suggestions (with accel model and link)? Thank you.

---

Does it need to be Piezoresistive? why not the IEPE accelerometers?

Single-axis IEPE accelerometer - Shock measurement ± 2500 to ± 30000g

[ngthai]

For now, it has to be PR because I use NI-9237 (and +-0.025 V/V range). Thank you.

[santo_13]

---

@ngthai wrote:

For now, it has to be PR because I use NI-9237 (and +-0.025 V/V range). Thank you.

---

If you get better performance, are you open to choose a different set of NI DAQ and Accelerometer? or are you still limited to make the best out what you have?

In short, some scenarios, the requirements have the higher authority which means do whatever it takes to satisfy the requirements, in others, making the best out of what you have though you don't meet the requirements take over.

What is your requirement constraint?