My question is very easy while I am working on it more than two months.
Is it possible to read 5 accelerometer, BMA180, simultaneously(exactly at the same time)?
The last idea that comes to my mind is using 5 arduino and 5 SD card. But in this way after sampling I can collect the data of 5 accelerometer not during sampling. In addition reading 5 SD card is not really in an efficient way . . .

Is it possible to read 5 accelerometer, BMA180, simultaneously(exactly at the same time)?

No, and there is no reason to try. The accelerometer has its own clock, and collects data points asynchronously, so you can't know exactly when the samples are collected. Also, it takes some time to read out the data.

If you have a way of accurately knowing the time, store a timestamp with each reading, and reconstruct the data along with the timing later.

But there is strong reason. If you sample a signal (seismic wave) at one point (same place) with a number of receivers and then stack the data, you can increase signal to noise ratio sqrt(n) times, in my case 2.236 times.

Your idea is the same of mine which is not efficient.

Is it possible to read 5 accelerometer, BMA180, simultaneously(exactly at the same time)?

No, this is completely impossible, for the reasons stated above.

jremington:
The accelerometer has its own clock, and collects data points asynchronously, so you can't know exactly when the samples are collected. Also, it takes some time to read out the data.

Does it means if the all the condition for two separate system (consist of arduino and accelerometer) start to reading data at the exact same time(turn on two arduino at the same time), the time which samples are collected are not the same?
If they are, I can use an arduino to give external interrupt with specified interval (using timer) to another 5 arduinos to sample accelerometers. In this way, at least I know the starting time of a sample not exactly the time of sampling.
In my case, reading each sample takes around 1100 micro seconds. Can I assume the sample is belong to 550 microsecond?

the time which samples are collected are not the same?

Correct.

It is never possible to know the exact time that something happens, as there is always some uncertainty in the clock.

You need to decide what time resolution is important. Can you pretend that the data points are all collected "at the same time", even if there is, for example, one millisecond uncertainty for any one data point?

jremington:
Can you pretend that the data points are all collected "at the same time", even if there is, for example, one millisecond uncertainty for any one data point?

2 (some time 4) milliseconds sampling time interval has been done for seismic waves. Therefore for each data pint the uncertainty should not be more than at least 50 microseconds it means in my case I can accept the data 550+50 and 550-50 milliseconds for each data point.