• State Resolved
  • Locked Locked
  • Replies 8 replies
  • Subscribers 30 subscribers
  • Views 2189 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

FDC2112 throughput

Vasco Tesi46
Intellectual 360 points
Other Parts Discussed in Thread: FDC2112, FDC2212

Hello, 

we have to develop an application which involves capacity measurement at high speed. 

The desired speed is at least 1 frequency sample every ms. 

The capacity to be measured ranges (roughly) from 0.25 pF to 1 pF. 

Having to perform quick measurements, we preferred FDC2112 (12 to 16 bit) over the 28 bit chips, because we have not time to perform longer and more accurate measurements.  

A careful reading of your FDC2112 datasheet reveals some discrepancies or undefined quantities that make difficult to undestand which can be the performance in terms of speed. 

At page 17, it can be found : " 1.8ms (sensor-activation time) + 3.2ms (conversion time) + 0.75ms (channel-switch delay) = 16.75ms"  probably there is some typo because 1.8+3.2+0.75 = 5.75 and not 16.75. 

The value of Fref in this example was never mentioned in the document;  taking the 3.2 ms value and according to the document which reports that this conversion time is due to  8192 clock cycles of Fref,  it seems that Fref is 2.56 MHz. Is it true ? If wrong, can you explain the criterion? 

We did non crearly understand the choice criterion of the resonant circuit recommended, with 18 uH in parallel with 33 pF. 

why not , for example, 54 uH with 11 pF ? It should be have better sensitivity having greater (capacitance variation)/ (total capacitance).  Perhaps the choice recommended is for S/N optimizing ? 

Thanks, Vasco

  • 0
    TI__Expert 6480 points
    Hi Vasco,

    You're right, the numbers are a bit confusing. It would be best to refer to the tools spreadsheet calculator (www.ti.com/.../toolssoftware) for more accuracy.

    In regards to the 18uH inductor, we had run some tests determining that the 18uH bourns one has particularly good performance. Generally speaking, having a smaller capacitance has better sensitivity but sometimes, having a larger cap helps achieve more stable measurements.

    Thanks,
    Rachel
  • 0 in reply to Rachel Liao
    Intellectual 360 points

    Hi Rachel,

    thank you for your reply, but  you should be more precise. 

    Maybe the 18uH inductor performs best in some measurement configuration and not in others. 

    As regards the spreadsheet, i was not able to find it. If you refer to Webench tool,  it does not operate at this moment: it presents some alternatives,but when one presses "create design" button, the software stucks. 

    Other webench tools, for example Power design which i used many times in the past, operate correctly. 

    Thanks, 

    Vasco 

  • 0 in reply to Vasco Tesi46
    Intellectual 360 points

    Sorry, webench with capacitive sensing  stucks on Google Chrome but operates correctly with Firefox.

    Whatever, it gives very poor information, mostly geometrical: electrically, it reports only operating voltage, number of channels and resolution bits (the last datum wrong because FDC2212 does not have 28 bits).

  • 0 in reply to Vasco Tesi46
    TI__Expert 6480 points
    Hi Vasco,

    The reason we chose this particular inductor was because it is part of the Bourns CMH322522-180KL family which had the best noise performance and least amount of drift over time. We just happened to pick 18uH as our recommended value but a different inductance value could work too.

    The spreadsheet download link is at the bottom of the "Tools & Software" tab of the device page. You can download the "FDC211x/FDC221x Current Consumption Estimator (Rev. A)" spreadsheet which can help you calculate conversion time based on RCOUNT.

    Thanks,
    Rachel
  • 0 in reply to Rachel Liao
    Intellectual 360 points
    Hi Rachel, thank you very much.
    I was confused due to the name of the link (for this application consumption is not an issue).
    The spreadsheet, together with datasheet, clears the chip operation.
    only a doubt : with FDC2112 , if RCOUNT = 0x200 (count = 8192), the resolution at most is 13 bits (as reported in the datasheet) and not 15 bits. At best, for example if you have a Fref = 10 MHz reference, with a total count of 8192 , the time needed for making a measurement is 8192/10 MHz = 819.2 us and at most , if you measure a frequency only slightly less than Fref, that is, a 13 bit number.
    Is this right ?
    Thanks, Vasco
  • 0 in reply to Vasco Tesi46
    TI__Expert 6480 points

    Hi Vasco,

    Could you clarify where you got the 15 bits from?

    You are correct in saying that with an RCOUNT=0x0200, the resolution is at most 13 bits. With the default settings on the FDC2112, only 12 of those bits are reported. However, it is possible to shift the data output by up to 4 bits if needed. The way to do that is by playing with the gain and offset settings (Section 9.3.2.1 of the datasheet).

    Thanks,

    Rachel

  • 0 in reply to Rachel Liao
    Intellectual 360 points
    Hi Rachel, sorry for the delay.

    The 15 bits resolution datum comes from "FDC211x/FDC221x Current Consumption Estimator (Rev. A)" spreadsheet. The FDC2112 datasheet does not give complete insight into device operation, but if the input frequency must respect fsensor < fref/4 as reported in the device datasheet (Table 1 at page 14) , and if the DATA_CHx output is the counter output using a time window of N = CHx_RCOUNT length, the resolution will be 2 bits less than log2(N).

    The 13 bit resolution above mentioned in this discussion would be true if Fsensor could be very close to Fref.

    Ih the spreadsheet with Fref = 10 MHz I can set fsensor = 8 MHz and this will not result in an error.

    Am I wrong, or the spreadsheet needs some revision?

    Thanks, Vasco
  • 0 in reply to Vasco Tesi46
    TI__Expert 6480 points
    Hi Vasco,

    If fref and fsensor do not satisfy the requirement of fsensor < fref/4, the device will still work, however, there will be a lot more noise due to timing issues. The required RCOUNT value in this case will be about twice as much as it needs to be if the clocking requirement was met. It is recommended to keep to the clocking requirement.

    Thanks,
    Rachel