Analog accelerometers e.g. ADXL103, have noise levels quoted as micro-g-per-root-hertz. For that one it's around 100 micro-g/rtHz. This is a spectral quantity (i.e. it is the noise in a certain bandwidth).
Digital accelerometers e.g. ADXL345 have noise levels quoted as "LSB rms". This is not a spectral quantity but just an rms quantity.
I'm guessing that for that digital accelerometer with approximate numbers of: Sensitivity = 200 LSB/g and Noise = 1 LSBrms, that it has noise level of: (1 LSBrms)/(200 LSB/g) = 5 milli-g rms.
How is the best way to compare between the two, apples-to-apples? Is it to say, I decide to electronically filter everything above 1kHz from both the ADXL103 and the ADXL345 signal, and therefore the bandwidth in both cases is 1kHz, and therefore:
ADXL103 has (100 micro-g/rtHz)*(sqrt(1kHz)) = 3 milli-g rms? Is that true?
However I notice ADXL has user-selectable bandwidth/datarate. How to factor this in?
Ultimately I just want to take a digital and an analog accelerometer and say which one is better in terms of noise.