PGA970: 4-wire LVDT with PGA970 has a no-linear behavior in center area

Hi Hasan, 

How are you calculating the linearity? Is the graph you pictured for ADC1, the DAC, or the entire system?

Can you please share a schematic or a diagram of how the LVDT sensor is connected to the PGA970?

Thanks,

Maggie

Hi Maggie,

we have a calibration test system with μm steps, that move the sensor and take the samples from output of PGA970 and display linearity graph curve.
I tested a 10mm sensor that has 5-wire but for some purpose we must connect just S1 and S2 without SM.

In this case, I connected S1 and S2 to PGA970 and chose the 1mm just in the center area of the sensor(±0.5mm). 

I tested this sensor as 5-wire and connected with same filter like picture below. And tested again, there is no any jump or non-linearity behavior in center area:

Maximum linearity is: 0.07%

Hi Hasan, 

If are only using one channel for a 4-wire LVDT sensor, then you must connect P2 in between the secondaries (see configuration 2) I don't see that connection in you :

If you're measuring linearity across the whole system, do you have a compensation algorithm running in the PGA970?

Thanks,

Maggie

Hi Maggie,

I use the configuration1 that you sent, as I said, I have just S1 and S2 without SM from sensor. I can't connect SM to P2. 
And with a 5-wire sensor, I don't have a big problem. 
I use a 4-wire sensor like circuit schematic below and when I move the core of sensor in center area, output doesn't change and then jump to another value. For a sensor is 50mv jump and the other one was about 200mv jump. 

thanks

Hassan

Hi Hassan, 

I apologize - I misread your earlier message about the configuration.

Are you using a calibration algorithm in the device? What are your bandpass and lowpass filter settings? 

Thanks,

Maggie

Hi Maggie,

sorry for late answer, I couldn't be able to send my message. Because I got this error: "Your posting frequency has exceeded allowed rates. Please wait 1 minute(s) to post again, Texas instrument". 

it's ok. yes, first I calibrate the sensors positions then I wrote a function that convert the demodulator1 data to DAC_REG0 value (0-10v). And I have 3Vrms with 5Khz on P1-P2 and my filter setting is:

    DEMOD1_BPF_config(ADC12_BPF_CF5KHZ_B20HZ_B1,                      
                                         ADC12_BPF_CF5KHZ_B20HZ_A2,                      
                                         ADC12_BPF_CF5KHZ_B20HZ_A3); 

    DEMOD1_LPF_config((S2)ADC12_LPF_CF100HZ_OP128US_B1,               
                                         (S2)ADC12_LPF_CF100HZ_OP128US_A2);

Another test I performed was measuring the sensor’s secondary signal, which showed the same behavior as the PGA970's output. This means that the data from Demodulator1 exhibits a jump at short distances near the center area, and this jump is reflected in the output. How can I smooth the demodulator data in this region?

Thanks
Hassan

Hi Hasan, 

Normally, I would recommend a narrow bandpass filter centered at the excitation frequency, but it appears you are already doing that. 

In a previous message, you said:

Hassan Samadi said:

I tested this sensor as 5-wire and connected with same filter like picture below. And tested again, there is no any jump or non-linearity behavior in center area:

In this configuration, did you use the PGA970 to generate the excitation signal? You're only measuring S1P and S1N, correct?

Hassan Samadi said:

Another test I performed was measuring the sensor’s secondary signal, which showed the same behavior as the PGA970's output. This means that the data from Demodulator1 exhibits a jump at short distances near the center area, and this jump is reflected in the output. How can I smooth the demodulator data in this region?

From the schematic you've shared, it looks like the same exact filters as above, you're using the PGA970 to generate the excitation signal, and you're only measuring S1N and S1P. Please correct me if I'm mistaken. 

In short, it looks like you've taken two measurements of the sensor's secondary - in the first scenario, you didn't see a non-linearity, and in the second scenario, you did. So I'm trying to understand what was different between the two measurements. 

Have you tried measuring the excitation signal itself? If you're using the PGA970 excitation signal, it is generated from a look-up table. I'm wondering if there is a value missing from the look-up table - that would explain why there's a non-linearity at that one point. 

Thanks,

Maggie

Hi Maggie,

Maggie Lee said:

In this configuration, did you use the PGA970 to generate the excitation signal? You're only measuring S1P and S1N, correct?

Yes, I use the PGA970 to generate the excitation signal 3Vrms-5KHz:

I have 5-wires sensor that with connecting the middle pin of sensor to SM, S1 of sensor to S1P and S2 to S2P and using both demodulators, the linearity is perfect but without SM as a 4-wires sensor without middle pin of secondary and S1 to S1P and S2 to S1N. Non-linearity behavior happens in center, I connected sensor like this:

Maggie Lee said:

From the schematic you've shared, it looks like the same exact filters as above, you're using the PGA970 to generate the excitation signal, and you're only measuring S1N and S1P. Please correct me if I'm mistaken. 

Yes exactly.

Maggie Lee said:

In short, it looks like you've taken two measurements of the sensor's secondary - in the first scenario, you didn't see a non-linearity, and in the second scenario, you did. So I'm trying to understand what was different between the two measurements.

No I mean. In first config with connecting the 5-wires sensor to PGA970 and without using phase detection technique, I got acceptable linearity. But using this sensor with just connecting S1 and S2 without SM (middle pins of sensor) to PGA970, with moving the core in short distance before and after of center signal doesn't change. 

Maggie Lee said:

Have you tried measuring the excitation signal itself? If you're using the PGA970 excitation signal, it is generated from a look-up table. I'm wondering if there is a value missing from the look-up table - that would explain why there's a non-linearity at that one point. 

Yes, I did, and I'm using the look-up table to generate this sinus signal. And my signal from connector side after U3 is like this:

Thanks
Hassan

Hi Maggie,

I tested the sensor like configuration2 and with this settings:
S1_GAIN_SINGLE_END
S2_GAIN_SINGLE_END
S1_S2_CONFIG(BIAS_VCM_CTRL_0_0 | VCM_ENABLE);

and I tested the linearity that is realy worse and more than 3% and same problem in center area.

in this test, demodulator1,2 did not change big. I mean the maximum value for demodulators is about half in compare with configuration1.
I know that the 4-wire sensors has problem in center. but I have circuit with AD698 and I tested this sensor with this IC and i works without problem and linearity test even in +-0.5mm in ther center area is less than %0.06.

And I have calulate both phase of demodulators and I get this values with moving the core like this:
start -> center : phase1 = 0 and phase2=180

center:              phase1 = 0 and phase2=0

center->end :    phase1 = 180 and phase2=0

could you please tell how can I solve this problem.

Thanks

Hassan

Hi Hassan, 

You mentioned in a previous post that you have a calibration algorithm running in your PGA970. Can you tell me more about this algorithm? How many data points did you use for calibration?

Thanks,

Maggie

Hi Maggie,

Firstly, I calibrate the positions and save them to FRAM. These positions are start-, center- and end position. Then with another function(scaler) I scale the inputs(demodulator1,2 and phase1,2) to DAC value.

Steps:

1. Check Phases:
   • Left: abs(phase1) ≤ 45 && abs(phase2 - 180) ≤ 45.
   • Right: abs(phase1 - 180) ≤ 45 && abs(phase2) ≤ 45.
   • Center: Otherwise.
2. Left Side: Map demod1 using out = -0.0085 * demod1 + 7780 (fixed-point: (-85 * demod1 / 10000) + 7780).
3. Right Side: Map demod2 using out = 0.0093 * demod2 + 7413 (fixed-point: (93 * demod2 / 10000) + 7413).
4. Center: Interpolate between left and right outputs using weight = |demod1 - demod2| * 100 / 100000.
5. Clamp: Ensure out is in [1280, 12803].

Do you have a better Idea for scaling function or calibration function?




Thanks

Hassan