FDC1004: FDC1004

Rahul,

Are you heating the FDC1004 PC board?
If so, is it possible some of the PCB components have heat-related drift?

You mentioned an expected 0.1% drift in capacitance readings.
What calculations lead to that requirement?

Regards,
John

Hi John ,

Thanks for the message. The only components on the PCB apart from FDC1004 are a few resistors ,capacitors and a voltage regulator(5-3.3V) and none of them have any heat related drift (as mentioned in their spec sheets).

The 0.1% drift was calculated from the FDC1004 spec sheet itself.

Regards,

Rahul Shewani.

Thanks Rahul.

Can you share the details of your calculations?

Regards,
John

Hi John,

So I redid the calculations and found an error in them .According to the SNOSCY5B,the offset to be observed over 20C is 0.01 pf .My readings for the Environment electrode(Channel 1),Level Electrode(Channel 2) and Liquid Electrode(Channel 3)when the tank is empty is 0.3332,  0.9796, -0.0545 pf and when full is 1.6501,  3.7337,  1.0640pf .So the range for each of the electrodes is approx. 1.3, 2.8 and 1.1pf . Now if we were to follow the datasheet ,we should be observing 0.01 pf for every 20C change on the channels which would translate to 0.9%,0.357% and 0.76% change over Channel 1 ,Channel 2 and Channel 3.

This is not what is being observed, instead we get an offset of approx.1.3, 2.8 and 1.1pf (Channel 1,Channel 2 and Channel 3)over at 25 to 30C increase in temperature which is way higher than mentioned in the datasheet.

We have our doubts on shield drivers being affected with temperature increase and nothing about it is mentioned in the datasheet. Is there anything we can do to reduce the temperature effects ?

Rahul,

Thanks for the add'l info.

Could you share the details of your calculations?

Also, how are you inducing the temp shift in your system?
Does the temp shift apply to just the FDC1004, FDC1004 + PCB, or FDC1004 + PCB + sensors & tank?

Regards,
John

John,

The above info is the calculations we performed .The drift requirement was deduced through these calculations. The 0.01 pf/20C change is given in the datasheet and all we did was to divide this change(0.01pf) by the range that  our channel measures(1.3,2.8,1.1).

The way temperature is induced is by heating the FDC chip with a heat gun and the temperature shift only applies when heating the FDC chip and remains unchanged when heat is applied to the electrodes. Also another observation is that, when we heat the FDC upto 40C,we see a bare minimum change in capacitances of the channels but when we heat it at a range of 43 - 46C,the capacitance measured by these channels get vastly unstable and then somewhat stabalises at 50C and gives a constant offset.  

Rahul,

Thanks for the info.
I will revisit the TC specs in the data sheet and update this thread by COB tomorrow.

Regards,
John

Hi John,

I'm Rahul's colleague. Rahul is now away on holiday so I picked up the testing. We appear to have similar behaviour as reported in https://e2e.ti.com/support/sensors-group/sensors/f/sensors-forum/562860/fdc1004-temperature-related-offset-change/2148495

In our test we heated up the electronics slowly with a heat gun from ~25degC to ~50degC, then let it cool down naturally to ~28 deg C. A thermocouple was attached to the body of the FDC1004. This is the observed behaviour of temperature vs measurement:

There is a nonlinear increase in the readings that is consistent of the heating cycle as the cooling cycle (note there is barely no hysteresis in this behaviour).

Upon disconnecting two of the three channels used, the behaviour changed to this:

This appears to confirm the common mode effects from a shared driven shield. The only way forward we see is to decouple / separate the shield from the electrode but are uncertain if this is a workable reliable solution. We do have some difference in layout versus the application note but do not believe these contribute.

Is this behaviour a known 'issue' for TI, as there appears to be a number of older posts mentioning temperature related offset issues? Are other TI customers using this device in similar application in production? 

We also seem to have experienced quantization issues as reported in https://e2e.ti.com/support/sensors-group/sensors/f/sensors-forum/937271/fdc1004-fdc1004-has-large-errors-in-output/3464053 albeit only in one test (and thus far not repeated).

James,

Thanks for the update and for pointing out the previous E2E posts on this topic.

Could you expand on what is meant by "This appears to confirm the common mode effects from a shared driven shield."?

Also, could you provide a diagram of your sensor and shield arrangement?

Regards,
John

Could you expand on what is meant by "This appears to confirm the common mode effects from a shared driven shield."?

This was a reference to the linked question. Curtis button said 'The issue occurs when the sensor electrodes are capacitively coupled to a common node...'

We appear to be seeing a similar effect. We have the laid out the sensor on a FPC, so the electrodes have a small separation distance from the actively driven shield (65um). The electrodes are therefore well coupled to the shield plane, and therefore relatively strongly coupled to each over. Disconnecting two of the three electrodes from the FDC1004 eliminated the strange non-linear behaviour which would appear to suggest the effect is down to a common mode coupling between driven shield and the inactive inputs.

Note, on code review I found our code was not respecting the measurement done flag correctly. This could possibly be the cause of the quantization issue we saw. The non-linear temperature effects remain unaffected and unresolved,

Is it possible to email you the design files privately?

James,

Thanks for the add'l info.

I'm not sure about the measurement done flag. It could contribute to the problem in that data would not be stable when captured and reported.
If the quantization problem persists after the code is working correctly, let us know and we can look into it further.

Let's consider the measurement steps & sensors w/respect to the behavior over temp.
Data is gathered for the three sensors sequentially.
Shield 2 should be 180 degrees out of phase with the sensors, with the goal that the liquid under measurement is electrically neutral and at zero potential.
A follow-on requirement is the level sensor and SHLD2 are identical in area and shape, placed symmetrically with respect to the tank structure, and the liquid to be measured. 

Amplitude and time delays between the sensors and SHLD1 waveform are critical as well.

The active shields work most effectively when its waveform has amplitude and timing that is identical to the level sensor so there is no potential difference between them. If there are amplitude and timing differences, then the coupling between the sensor and SHLD1 will cause capacitive offset in the sensor readings. 

I will send you a private E2E friend request. Please accept it and we can then communicate privately & exchange files over the E2E messaging app.

In the mean-time, can you please take a look at the sensors' and shield waveforms with an oscope & high impedance probe at room temp and 50C?
Even a high impedance probe can affect measurements when the capacitances are low, so a trick that sometimes helps is to place a 1k leaded resistor between the probe tip and the test point on the board. 

Regards,
John