increasing accuracy by amplifying signal?

[Forrest_Bao]

I need to measure a very weak signal in -200uV..200uV range and the accuracy requirement is 1uV.

I am thinking about USB-621x cards, of 0.088 mV accuracy in -200mV..+200mV.

I plan to amplify the weak signal to fully span into the minimum voltage range of USB-621x. I will amplify the signal by 100 or 1000 times before I send them to USB-621x card. Though the accuracy for USB-621x is 0.088mV in -200 mV..200 mV range, the accuracy for original signal can be 100 or 1000 times smaller. Thus, it's 0.88 uV or 0.088 uV. Am I right on this point?

 If i am correct, suppose I read a measurement 0.1 mV from the USB DAQ card, can I say the actual voltage is between (0.1-0.088)*100 = 1.2 uV to (0.1+0.088)*100 = 1.99 uV?

[Jason_D]

Forrest,

This may be possible but I would think the money gained from a cheaper DAQ card is going to be offset by the need for a "near perfect" amplifier.  Since you also have an email request open I have reposted here the very important info that John sent you:

To achieve the 1uV accuracy with the USB 621x you would definitely need some sort of external amplification. You should keep in mind that an external amplifier will have its own accuracy specifications that may or may not meet your requirement--make sure to double check this.

Accuracy does provide the range that the measurement will fall between. If you have an accuracy of B, given an actual voltage of A your measurement should fall between A-B and A+B.

Using an external amplifier sounds like a good idea to measure this low-voltage signal, but you will have to make sure that it meets the specifications of your application. Specifically, you should pay attention to settling time and accuracy specifications.

To compute system accuracy, you would take the RMS of the errors of the individual components (since the accuracy specifications are defined by standard of deviation). More information with some example calculations can be found here:
https://www.ni.com/en/support/documentation/supplemental/18/calculating-absolute-accuracy-or-system-...

The calculation that you preformed involves and ideal amplifier with ideal cabling.  You will have some sort of error generated by the amplfier that will probably push this back out of your range.

[AnalogKid2DigitalMan]

For such a task I would personally use an off-the-shelf nanovoltmeter with an RS232 or GPIB port.

-AK2DM

[Forrest_Bao]

What I am going to measure is scalp EEG signal, so I guess nanovoltmeter is way too accurate for me.

The frequency is not a problem for me right now, coz EEG signal is of ver low frequency, maximum 200 Hz. 

I just wonder, does NI has any signal amplifying module that I can use between the signal source and DAQ card?

[Steve.J]

NI does not have any amplifiers that are suitable for amplifying scalp EEG (or any biopotential signals).  Safety is one big concern (proper isolation and leakage current specifications) but there are performance issues as well.  The electrodes used for scalp EEG (assuming that you are using standard gold or Ag-AgCl disc electrodes with an electrolyte gel/paste) have a fairly large DC offset (in 10-100mV or more) and are unmatched (each electrode will have a different offset).  This differential voltage can block your amplifier unless it rejects DC (should be AC coupled).  Also, since the source impedance (skin/electrode) is high - on the order of 10Kohms and also not matched - you are faced with common-mode noise problems from AC line frequency interference unless you have very well balanced differential high-impedance inputs.  NI doesn't make any signal conditioners/amplifiers to deal with this measurement problem - you should look to companies like Grass Technologies, World Precision Instruments, Harvard Apparatus, and other sources for amplifiers specifically designed for amplifying biopotentials.  Most of these amplifiers will have analog outputs that are ready to be digitized by almost any NI DAQ board.

Steve