• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 1 answer
  • Subscribers 31 subscribers
  • Views 363 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.

TMCS1126-Q1: Noise mitigation on the output voltage of the hall sensor

Janus Meinert
Prodigy 30 points
Part Number: TMCS1126-Q1
Other Parts Discussed in Thread: TMCS1126

Tool/software:

Hello TI

I want to use this hall sensor to measure the AC 50Hz input current to my vienna rectifier. 

Previosly with a hall sensor from another manufacturer, we have experienced that it was necessary to add an op amp as differential amplifier to smooth out the measured signals before the ADC.

We never understood why it was necessary and the project was parked, maybe some common mode issues. 

Have you encountered some situations with your TMCS1126 where it was necessary to have a differential amplifier step between the hall sensor output and the ADC? 

Or is a simple lowpass filter enough? 

In regard to the lowpass filter, should its cutoff frequency be set to half the sampling frequency of the ADC? I guess this would also help bring down the Input Noise Density 150 μA/√Hz. 

Best regards Janus 

  • 0
    TI__Mastermind 23805 points

    Janus, 

    As you say, in general, Hall effect sensors are inherently noisy (this is true of the technology itself, so all Hall sensors typically will face this challenge, regardless of manufacturer. Take a look at analysis below on figure 6-15 from the datasheet:

    This shows that for the TMCS1126, for full bandwidth (ie, no filtering on the output), the noise alone can account for nearly 1.75App of random error on a measurement in an application. You can also see that for the device, the noise is directly linked to the BW of the device, and application of a filter at a given cutoff frequency will help attenuate the noise proportional to where the cutoff is applied. There are also other methods that can be employed to combat this, such as oversampling and averaging to average out the noise. 

    The answer to your question of difference amplifier vs. simple low pass filter will depend on the level of attenuation you are looking for here. A passive LPF is going provide a 20dB/decade attenuation beyond the cutoff to these values over BW (as well as HF noise), but active components off the output can provide even higher attenuation factors. 

  • 0 in reply to Carolus Andrews
    Prodigy 30 points

    Thank you for the quick reply!

    I have some follow up questions

    1. Is the unit for the noise A or Arms. The datasheet does not say it is Arms, but I can see that you use it in the calculation

    2. The factor 6.6 that you are multiplying the noise with, where does that come from?

    3.  You write 1.764 APP,RTI. I guess APP is Ampere peak-to-peak. What is RTI short for? Referred to input?

  • +1 in reply to Janus Meinert
    TI__Mastermind 23805 points

    Janus, here is some context on this:

    1) the unit shown in the graph is actually "Amps per square root Hertz". Noise is inextricably linked to frequency, and therefore will vary in magnitude across band, and your total noise will be the integral beneath that curve. You can see in my calculations that to calculate the integral under the curve, I am multiplying the relatively flat magnitudes by the sqrt of the frequency band of each color. Strategic implementation of a filter can be used to reduce the total noise signature, as you are effectively limiting the noise by truncating bandwidth, but of course you are doing this at the cost of latency and delay in the signal.

    2) For the calculation I describe above, once you have a value, that value is the RMS value of the noise. As noise theory will teach you, the RMS value of random noise is in fact equal to standard to deviation, so the multiplication of the RMS by 6.6 will give you the total range of noise you could potentially see statistically, which is the peak to peak noise (some folks multiply by just 6 sigma, it comes down to how many standard deviations you want to represent.)

    3) Yes, as I mention above, App stands for "Amps peak to peak." RTI stands for referred to input, ie, I am looking at the error in terms of amperage of my measurement rather than output voltage of the part. An RTI measurement is simply the error in output voltage divided by the sensitivity of the part to return the measurement to the inputs. 

  • +1 in reply to Carolus Andrews
    Prodigy 30 points

    Thank you!
    I hope you do no take it ill, but I find some of what you write a bit confusing and I hope you will elaborate a bit more. 

    It is this Amp and AmpRMS that confuses. You say that the unit on the y-axis of the graph is Amps per square root Hz. Then to get the noise you calculate the area, so "Amps per square root Hz * square root Hz. Wouldnt that just be Amps?

    But in comment 2, you say that once I get a value it is Amp RMS of the noise and not just Amps. (But amps was the unit we got when we calculated the area)

    What am I missing? 

    Br Janus

  • 0 in reply to Janus Meinert
    TI__Mastermind 23805 points

    Janus, 

    My apologies for the confusion!

    Lets start from the beginning. 

    It is this Amp and AmpRMS that confuses. You say that the unit on the y-axis of the graph is Amps per square root Hz. Then to get the noise you calculate the area, so "Amps per square root Hz * square root Hz. Wouldn't that just be Amps?

    - Yes, from the graph, the units on the y-axis are in A/sqrt(Hz):

    To get the total noise, you need to integrate under this curve, but for each area, you use the square root of the value of the x-axis. So, for example, you can see in the above calculation for the blue area I estimate my y-axis value as 170 uA/sqrt(Hz), and I multiply this by sqrt(480kHz):

    Once this calculation is performed, the value you will receive is the total RMS noise for the frequency, as this noise is based on total power, which of course is based on RMS. I think the disconnect here is that the curve lists amps, because it is an instantaneous capture across the spectrum for each value, where RMS is due to total power, ie, integration under the curve.  

    Check out this video, which walks you through how I performed this calculation and the background on RMS and peak to peak noise, and let me know if this makes it more clear.