ADS127L21: Sharing the results of Idle state vibration measurement using ads127l21 and requesting opinions.

Hello Cheol,

If I understand your question, you want to reduce the bandwidth of the input signal to further reduce noise?  The input anti-alias filter is set at a much higher frequency to attenuate alias signals near the ADC modulator frequency, which will be greater than 1MHz.  All of the low frequency attenuation will be due to the internal digital filter.

Can you tell me the configuration of the ADS127L21?  What is the input clock frequency, divider setting, filter selection and filter OSR setting that you are using?   What is the maximum input signal frequency that you need to measure?

Regards,
Keith Nicholas
Precision ADC Applications

Hi keith, thanks for your reply.

My ultimate question is as follows:

In the FFT waveform, it is around 70dB at 0Hz, but it seems to drop to 60dB at around 100Hz.

I would like to move the drop from 70dB to 60dB earlier, to around 50Hz, not 100Hz.

(After testing with the Siemens SCADAS measurement system, we observed that the signal dropped to below 60 dB starting from around 10 Hz.)

The reason is that we need to precisely observe signals in the frequency range below 200 Hz, which has become a requirement for our signal processing.

What steps would I need to take to do that?

External Oscilator : 13.1072MHz
CONFING2(0x07) = 0b00101100
CONFING3(0x08) = 0b10000000
FILTER1(0x09) = 0b00000011 (osr=256)
FILTER2(0x0A) = 0b00000001
CONFING1(0x06) = 0b00101110 (high_ref,24BIT)

Based on our conditions, the calculated sampling rate comes out to 25.6 kSPS.

Hello Cheol,

This is low frequency noise, and may be due to something else in the signal chain.  The VOCM2 buffer may be the cause.  Can you show me the schematic for the VOCM2 buffer?  Also, please show the reference circuit, including the reference voltage that you are using.

I would recommend changing the CONFIG1 register setting to 0b00101111 which will enable all input buffers and should provide the best overall performance.

Also, is the FFT plot shown with an IEPE sensor powered-up and connected to the inputs?  What is the reference level for the FFT measurements, 1mg?

Regards,
Keith

The circuit below includes a VOCM buffer circuit.

The circuit below is a signal conditioning circuit related to the constant current supply for an IEPE sensor

The time-series data for vibration measurement has been converted to units of m/s², and an FFT has been performed on it. When performing the FFT, the reference value is set to 10^(-6) to check the absolute dB level.

The reason for setting CONFING1(0x06) = 0b00101110 was that the last buffer seemed to not be used, so I assumed it wouldn’t matter whether it was set to 1 or 0. Would there be a significant difference?

Hello Cheol,

Regarding your last question on the pre-charge buffers, I would not expect this to make any difference at low frequency.  However, this will make a small difference at higher frequency.  Since you are driving the ADC inputs differentially with the THS4551, you will want to either enable both AINP and AINN buffers, or disable both buffers to reduce power, but again, only the AINP enabled and not the AINN should not make a difference at low frequency.

I checked the low frequency noise on the EVM, and it only shows a minimal increase in noise form 10Hz to 100Hz.  Your ADC+THS4551 is almost identical to the evaluation board, so I do not think the low frequency noise is related to this section.

My best guess is the noise is from the current source, and specifically the TLVH431.  However, the only way to track down the source is to disable sections of your signal chain and rerun the FFT with each section until you identify the source of noise.  You should be able to easily disable the current source; just do not connect it to the sensor and rerun your tests.

Start by removing all sensor connections, which should result in nothing connected to the IEPE_Signal_Out net and rerun the FFT.  This will give you a baseline for the ADC and common mode buffers.  If the noise is acceptable, then add the sensor without the current source connected.  Assuming the noise is still acceptable, then connect the current source to confirm this is the source of noise.

Regards,
Keith