National Instruments time signal measurement is inaccurate. Why?

We are using the National Instruments NI 9234, 4-Channel, ±5 V, 24-Bit Software-Selectable, IEPE and AC/DC Analog Input Module in a WLS-9163 carrier. We connect the module to our computer via the Universities internet LAN. A 32 bit XP operating system is installed on the laptop with a 2011a Matlab version. We measure vibration with a 100mV/g PCB acceleration meter at an industrial gearbox.

The problem we have is with the accuracy of the time domain measurement (time signal), from the Matlab program written with the help from Matlab’s data acquisition (DAQ) toolbox in the legacy interface. Please see the Matlab Code below.

To calibrate the accuracy of the first Matlab Code a Diagnostic Instruments (DI) 2200 FFT analyzer was used to compare corresponding signals with each other. A sample rate of 2048 is used for a 0.8 second duration. From the comparison it is clear that the Matlab measured signal are less populated, or have less spikes compared to the DI signal. What parameters can influence this? Sample rate and time are the same.

Another problem is experienced with time signals on other channels of the NI module running with an offset from the 0 or x-axis. On channel 0, the time signal is symmetrical around the 0-axis but the other channels give the signal with an offset. What can the reason be for this?

Program:

AI = analoginput('nidaq','ENET-150359D') 
chan = addchannel(AI,1) 
duration=0.8 

sensitivity=100 
set(AI,'SampleRate',2048) 
ActualRate = get(AI,'SampleRate') 
set(AI,'SamplesPerTrigger',duration*ActualRate) 
set(AI,'TriggerType','Manual') 
blocksize = get(AI,'SamplesPerTrigger') 
Fs = ActualRate 
start( daqgetfield(AI,'uddobject') ) 
trigger(AI) 
wait(AI,duration + 2) 
[d,t]=getData(AI); 
data=d; 
d=(d*9.81*100); 

plot(t,(d)) 
grid on 
title('New Matlab NI Peak readings') 
xlabel('Time [s]') 
ylabel(' Acceleration [m/s^2]') 

delete (AI)
clear AI