I am evaluating the energy measurement chip ADE5169 for a project where the current waveform should be analyzed in terms of shape. Therefore the fourier as well as the wavelet transformation are the basic signal processing steps that are applied to the AD converted current signal. However the converted voltage of the shunt (1mOhm) is very noisy as depicted in the following picture.
(remark: the wiring of the shunt to the ADC is the same as on the ADE5169 Eval Board but with a different shunt value)
The acquired voltage signal represents the current from a 85W ohmic load powered from the common net (230Vrms). Therefore, the peak of the sinusoideal signal should be around 0.523mVp. Thousand times smaller than the recommended maximum input voltage. In order to find the cause for that nois, I took a look into the datasheet that was not really clearly understandable for me. Following questions were popped up while studying the datasheet:
(remark: unfortunatly I was not able to find the answer neither in the analog documentation nor on the internet)
- Why is the measurement bandwith of Irms limited to 3.9kHz, although the anti aliasing LPF has a corner freq. of 10kHz?
- What is the resolution of the AD converter? (22 or 23 Bit?)
-> according datasheet p.49 (AC Transfer Function) max. value = 2²² = 4'194'304 at ± 0.794Vp and 1.2Vdcref.
-> according datasheet p.49 (Current Channel ADC) 24Bit data words in 2nd complement -> 2²³!!!
- Is the quantisation characteristic linear?
-> according datasheet p.49 (AC Transfer Function) @1.2Vref
0.5Vp ≃ 2'147'483
0.794Vp ≃ 4'194'304
-> 0.794Vp / 0.5Vp = 1.588 but 4'194'304 / 2'147'483 = 1.953
- What is the meaning of no missing codes that is declared in the datasheet on p.8 (section "Internal ADCs") with 8Bits? Indeed, I know what a missing code is, but what means the number of 8Bits for no missing codes?
I would really appreciate if anyone who is familiar with analog ADEs or ADCs could help me on this issue(s) or could at least give any hint.