MSP430FR2422: 10bit ADC performance comparison with G2433 ADC10

Please read ~7 ans ~7K counts

Hello,

Tanmay Sutaria said:

We have been using the F2xxx and G2433 with 10 bit SAR ADC for over 7 years now and looking to move the products to FR2422 devices. A detailed reading of the FR2422 ADC gives the impression that it is less capable than the ADC10 on the older F2xx and G2433 devices.

It seems like a different module. A good comparison between the modules can be found in Section 8.3.1 ADC10 to ADC in the Migrating from the MSP430F2xx and MSP430G2xx families to the MSP430FR4xx and MSP430FR2xx family user's guide.

Tanmay Sutaria said:

We are able to achieve ~7counts with 1% accuracy using a TrueRMS AC voltage measurement algorithm on the G2433.However, the gain, offset and integral non-linearity parameters on the FR2422 seem to suggest that it would be difficult to implement the same on the device. Please advise, would be great if we could save some time on the evaluation process.

Those specs don't seem to be as good as the G2433, but the total unadjusted error (TUE) is much better for the FR2422. Have you considered using the MSP430i20xx device such as the MSP430i2021? These MSP430i20xx devices feature an integrated sigma-delta ADC that would allow you to achieve nearly 16 bits ENOB. The Energy Measurement Design Center supports these devices too.

Tanmay Sutaria said:

Please read ~7 ans ~7K counts

I'm assuming you mean 7k ADC samples. For a 10-bit ADC, the maximum count would be 2^10 - 1 (1023).

Hello James,

Thanks for your response and apologies for a delayed reply.

First let me elaborate on our application and the ~7k counts. We are using the G2433 for True RMS AC signal measurement. There is a 50% bias to capture both halves and we get approx+/- 500 count raw data form the ADC. After oversampling and applying a RMS calculation we are able to display achieve 7K counts which is equivalent to the full scale input signal of +/- 1.25V. e.g. an input of 700Vac RMS is divided to 2.5Vp-p as ADC input and then is calculated and displayed as 700.0Vrms. So basically a 7000 count indicator.

Since the oversampling is quite high, the ADC performance is quite critical as a 0.5 bit lower performance would not allow us to achieve the 7000 counts. While the TUE may be better on paper, it would be great of i can get an practical / actual use case feedback of the performance to be expected in terms of ENOB and noise after calibrating the midpoint/offset and gain. I have read the UG you have mentioned and it lists the migrations details but not a performance comparative between the two ADCs.

The ADCs are different and the FR2422 spec list a 8-bit ADC mode which further gives the impression that it is less capable and may reduce the oversampling headroom we have with the G2433 ADC10.

The i20xx, we are already using this for higher accuracy applications . The G2433 / FR2422 are for lower cost and basic applications.

Thanks and regards.