An application note is badly needed for the AFE Sensor Diagnostic Flags, because vital information is missing from the datasheet.
Specifically, Figure 18 shows that sensor diagnostic flags are collected throughout the datapath. Only the active channel is connected to the datapath.
But, scan mode 3 is described as providing the "sensor diagnostic burnout current" to an inactive channel ("The burnout current is provided in a serially scanned fashion (injected in a channel after it has undergone a conversion). Thus it avoids burnout current injection from interfering with the conversion result for the channel.")
When using scan mode 3, which diagnostics are collected during application of the burnout current, and which are collected during active conversion of the channel?
What logic can actually diagnose the various sensor failures, using ratiometric connection to a strain gage (typical bridge differentially measured) with 1kohm nominal resistance? I hope to distinguish the following cases:
1. Sensor disconnected or open-circuit in sense lines. The burnout current sources will saturate and inputs will be at their respective rails. During acquisition the inputs will float apart. Sounds like RAILS_FLAG might be useful in detecting this. Can someone verify that RAILS_FLAG is tested on the burnout channel (contrary to Figure 18)?
2. Open circuit in either excitation line. Inputs will be equal to the other excitation voltage, during both burnout and acquisition. Not sure whether RAILS_FLAG is applicable, since it tests against Va, not Vref. How is this distinguished from case 1? Reconfigure the multiplexer?
3. Sense wires shorted together. The difference between inputs will be very small during both acquisition (completely normal) and application of burnout current (abnormal). Sounds like SHORT_THLD_FLAG might be useful in detecting this if SENDIAG_THLD is properly set. I suspect that an appropriate value of SENDIAG_THLD is below the expected potential developed by the burnout current, i.e. 20% * 10uA * 1kohm = 2mV. Equation 12 says that the FGA and PGA gains apply. I'm not sure how these amplifiers can be in the signal path at the same time as burnout currents, since they are needed in the signal path of the active acquisition channel. Shorts *cannot* be detected during acquisition.
4. Either sense wire shorted to an excitation wire. Not sure whether RAILS_FLAG is applicable.
5. Sensor is operating ratiometrically but overrange.
I'd appreciate any guidance you can provide, and whether this diagnostics are possible
(a) with four channels on a single LMP90099
(b) with a separate LMP90097 for each channel, GPIO control over excitation voltage generation, and excitation signals connected to the VREF2 input pair.
