• State TI Thinks Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 1 answer
  • Subscribers 30 subscribers
  • Views 414 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.

LDC1101EVM: Open-air expected LHR output count standard deviation range

Chris Wu2
Prodigy 180 points
Part Number: LDC1101EVM
Other Parts Discussed in Thread: LDC1101

Hi,

I've been testing the noise performance of the LDC1101 EVM, and was wondering if someone could please sanity-check my numbers (they seem high): 

Configuration settings:

LDC1101 EVM placed in open-air, no metal targets used for baseline noise assessment.
Ti LDC1101 EVM GUI used in LHR mode to set registers/collect results. 

LHR RCOUNT: swept to yield different sampling rates

Clock Divider: 1 (setting clock divider >1 yields significantly better noise SD results,
but application notes say to keep clock divider = 1 when fclk > fsensor_max, which is
always the case for the EVM, with fclk = 12MHz and fsensor_max = 10MHz).

INTB Disable: Donot Report Data Ready
INTB Function: Disabled

RP_Min: 3kOhms - this yielded the best noise performance of the available RPmin settings

Optimize LHR Measurement: Enabled (this disables amplitude adjustment needed for Rp measurement, whic we don't want for LHR measurement)

Sensor FMIN: 2.7MHz - did not affect open-air measurements. This is the sensor frequency at/below which the LDC1101 checks if LC oscillation
has died

All settings were according to recommended LDC1101 settings listed in app note:
https://www.ti.com/lit/ug/snou137/snou137.pdf?ts=1593551064777&ref_url=https%253A%252F%252Fwww.google.com%252F

Results: 

SD LHR Counts:

11sps (max RCOUNT): 14.365
1ksps: 725.267
5ksps: 6623.179
10ksps: 8134.653

To calculate the sampling rate wrt RCOUNT, I used: 

tconv = (55.0+RCOUNT*16.0)/fclk_in

where fclk_in = 12 MHz

Questions:

1) do my measurements make sense for the LDC1101 EVM? I ask because these numbers are much higher than the standard deviation codes reported in Table 2 for the LDC161x (https://www.ti.com/lit/an/snoa931a/snoa931a.pdf?ts=1591912424026), and I'm wondering what factors could attribute to the differences in performance. 

2) For my application, I need <=80 counts SD sampling at 10ksps - should this be achievable with LDC1101 or LDC161x products? 

3) In the app note reporting LDC161x performance in Table 2 (https://www.ti.com/lit/an/snoa931a/snoa931a.pdf?ts=1591912424026), was this EVM powered through USB, battery, etc? Could the power source introduce noise into the SD measurement? 

4) Are there noise performance advantages of using an external crystal to drive the LDC reference clock, vs. using a microcontroller? 

Thanks,

Chris

  • 0
    TI__Mastermind 18545 points

    Hello Chris,

    For starters, I am assuming you are using SD as an abbreviation for standard deviation. Now on to your questions:

    1. That app note is specific to linear sensing and may not be the best comparison here. Additionally, the sensor coil on the EVM is not the same as the one used in that app note and the target used may also be different. These will all add to different results in your testing. I ran the same settings as you on the EVM and was able to duplicate your results.

    1. Hitting 80 counts on the standard deviation while set for 10ksps is a bit of a challenge with this device but may be doable depending on your application. You need to take into consideration the sensor coil size as well as the CLKIN frequency. Take a look at figure 4 of the Optimizing L Measurement Resolution for the LDC161x and LDC1101 app note. To test this, you may need to supply your own external clock source.
    2. For that app note, the EVM was powered using USB. Some power sources could introduce noise so a clean power source is recommended.
    3. Using an external crystal could improve the performance as the accuracy of the measurement is dependent on the CLKIN.

    Hope this helps,

  • 0 in reply to Justin Beigel
    Prodigy 180 points

    Hi Justin,

    Thank you for the response and verifying my results on your EVM.

    I'm still wondering how my noise count standard deviation numbers are so much higher than reported in the app note (~8000 vs 277 at 10kHz), and if there is a way to quantify the expected noise given coil Q factor and clock frequency. I'm testing without a target, which I would expect to give the best possible noise performance. 

    I'm wondering if you could help resolve a couple followup questions:

    1) what is the quantitative relationship (if known) between the coil parameters (L, C, Q), reference clock frequency, and noise performance? I understand from the app notes and datasheet that a higher Q and reference clock frequency improves noise performance, but I can't find equations to inform my design targets to meet my spec.  

    2) What are the main factors affecting noise performance? Going from ~8000 counts SD to 80 counts SD seems like a lot. 

    Thanks, I appreciate the help

    Chris

  • 0 in reply to Chris Wu2
    TI__Mastermind 18545 points

    Hello Chris,

    The standard deviation numbers in the table use a reference clock of 40MHz, which allows the better standard deviation for that test. Noise performance depends on many different factors and a calculation for it could be developed, but I am not aware of any that exist right now. The best way to improve the noise performance is getting a good clock signal. The next thing you could change is the sensor frequency. By lowering the sensor frequency, you can improve the resolution of the measurements. If these still aren't enough, the LDC161x has a lot more potential to minimize the Fsensor/Fref ratio but may limit you below your desired 10ksps. Depending on your application, the LDC131x might be a consideration to look at as well.

    Hope this helps,