Small (mV) negative analog input when there is no signal

[Omnistic]

Using USB-6002, I was reading a signal from a photodiode in AI1 when I realized that in complete darkness (no signal). My USB-6002 would read a signal with a mean value of about -1.6mV. I originally asked the Electrical Engineering StackExchange to know why I had this small offset. But after some troubleshooting I decided to short the + and - terminals of AI1 with a short (2cm-long) wire, expecting to read 0V. However, I read again a value around -1 to -2 mV. I am doing a differential measurement in NI MAX for this test. Is this normal, and where is this coming from?

[crossrulz]

Be sure to read the specification.  The main thing that comes out to me is the Typical Absolute Accuracy at full scale is 6mV.  But generally, 2mV of noise is not uncommon at all.

[Omnistic]

I am aware that the Typical Absolute Accuracy is 6mV. However, I was probably wrongly assuming that NI would bias their DAQs such that they are at zero when not connected, which is the role of the Input Bias Current (?). In other words, my assumption is that when there's no signal I would read zero but with a precision of 6mV. This was a wrong assumption from me, was it? It is a form of noise of course, but it is quite constant temporally. I'm attaching some screenshots of my recordings for reference.

Side question, 6mV of accuracy means +/-6mV (12mV total range) or +/-3mV (6mV total range)?

[santo_13]

Accuracy is +/- 6mV.

Any reading you get from DAQ is within +/-6mV of its real world value. In your case if you read -6mV when shorting the inputs, it is still valid as the read value is within 6mV of the real world value i.e, 0V.