EVAL-INAMP-82RZ awful noise performance

Hi!
I've used EVAL-INAMP-82RZ as the main board for testing AD8421/AD8422/AD8221. With all of these amplifiers I've seen a lot of noise. The main setup consisted of:

- 2x18650 in series providing power to the system with: -Vs = -4.2V / +Vs = +4.2V / GND = 0V (middle point of 2 series battery setup)
- REF set to GND
- single-ended scope probe directly probing the output
- IN+ and IN- directly shorted to GND plane
- gain set on the INA to >350

With above setup we are seeing output noise in the range of 6.7mV - 8mV on the scope, while running in Normal Acquisition mode, at a sweep speed of 30kSPS, AC 1:1, BW limited.

Any idea on why such a bad noise performance?

I've also connected the INA output to a 24-bit AD7175-8 ADC and the same level of noise was observed with the sample rate being set to 31250SPS.

Parents
  • 0
    •  Analog Employees 
    on Sep 26, 2018 6:38 PM

    Hi Alincp,

    If you want to measure the noise of the instrumentation amplifier, you have to make sure that the noise of that you want to measure is higher than the noise of the equipment that you are using. Oscilloscope already have several millivolts of noise and I would think that is what you are seeing on your measurements.

    May I know why do you want to measure the noise of these inamp? Do you want to measure the pk to pk noise of specific bandwidth or the wide band noise vs frequency?

    If you want to measure the pk to pk noise, you usually add  high gain stage amplifier to make sure that the DUT noise is greater than the noise of the equipment and also with high order filter to limit the bandwidth of your circuit. Oscilloscope should be enough to get the pk to pk noise.

    If you trying to measure noise vs frequency, you might need to have several gain stage amplifier to make sure that you have enough bandwidth for high gain configuration. The equipment that you can use for this is spectrum analyzer.

    Best regards,

    Emman

  • These inamps are to be used in conjunction with a 500kg strain gauge in order to provide a high precision, very low noise analog front end for the ADC mentioned above (AD7175-8). An example circuit is provided in the datasheet of AD8422 page 23 fig. 63. Of course the ADC measures voltage levels, so there's no need for the AD8276 like the one in fig. 63.
       

    At first I've thought that the 2.5V reference voltage that I've been using, ADR431, was the culprit of the huge noise I've been seeing. To make sure that was not the case, instead of suppling the AD8422 with -Vs = 0V / +Vs = 5V, I've changed the rails to -Vs =- 4.2V / +Vs = +4.2V, so the reference could be easily tied to GND, thus eliminating ADR431 as the main source of the noise I was seeing on the output. It made no difference. The output still has ~6.8mV-8mV of peak to peak voltage noise.

    I would like to sample the signal coming from the strain gauge at 31250 SPS using the AD7175-8 with as high precision as possible. That would be 17.9 peak-to-peak noise free resolution bits at that sample rate, according to Table 7, in AD7175-8 datasheet. Such resolution requires a high performance analog front end, which is not the case with the tests I've been doing. Those inamps appear to be very noisy no matter if I look at the output using the ADC or an oscilloscope. 

    Given the above sample rate, the bandwidth of interest is between 0 - 2 x sample rate, so 0 - 62500Hz. 

    Shorting the oscilloscope probe to GND using a probe short (see bellow image), yields no more than 600uVpp noise, so the scope + probe noise floor is well beyond the one being measured at the output of the inamp, thus the measuring system is not the limiting factor over here. That's also being confirmed by the noise seen on the ADC while sampling the inamp output at 31250SPS, which is also around that same value of ~6.8-8mVpp. 

    This kind of noise appears to be present on all inamps I've tested AD8421/AD8422/AD8221.
      

Reply
  • These inamps are to be used in conjunction with a 500kg strain gauge in order to provide a high precision, very low noise analog front end for the ADC mentioned above (AD7175-8). An example circuit is provided in the datasheet of AD8422 page 23 fig. 63. Of course the ADC measures voltage levels, so there's no need for the AD8276 like the one in fig. 63.
       

    At first I've thought that the 2.5V reference voltage that I've been using, ADR431, was the culprit of the huge noise I've been seeing. To make sure that was not the case, instead of suppling the AD8422 with -Vs = 0V / +Vs = 5V, I've changed the rails to -Vs =- 4.2V / +Vs = +4.2V, so the reference could be easily tied to GND, thus eliminating ADR431 as the main source of the noise I was seeing on the output. It made no difference. The output still has ~6.8mV-8mV of peak to peak voltage noise.

    I would like to sample the signal coming from the strain gauge at 31250 SPS using the AD7175-8 with as high precision as possible. That would be 17.9 peak-to-peak noise free resolution bits at that sample rate, according to Table 7, in AD7175-8 datasheet. Such resolution requires a high performance analog front end, which is not the case with the tests I've been doing. Those inamps appear to be very noisy no matter if I look at the output using the ADC or an oscilloscope. 

    Given the above sample rate, the bandwidth of interest is between 0 - 2 x sample rate, so 0 - 62500Hz. 

    Shorting the oscilloscope probe to GND using a probe short (see bellow image), yields no more than 600uVpp noise, so the scope + probe noise floor is well beyond the one being measured at the output of the inamp, thus the measuring system is not the limiting factor over here. That's also being confirmed by the noise seen on the ADC while sampling the inamp output at 31250SPS, which is also around that same value of ~6.8-8mVpp. 

    This kind of noise appears to be present on all inamps I've tested AD8421/AD8422/AD8221.
      

Children
  • 0
    •  Analog Employees 
    on Sep 28, 2018 12:50 AM in reply to alincp

    Hi Alincp,

    I agree that the instrumentation amplifier is ideal to be used for the wheatstone bridge applications due to its superior rejection of the common mode voltage. With regards to the noise that you are seeing, sorry if I missed that you mentioned that the gain of the inamp on your circuit is 350 V/V. This means that if you do the calculation for the AD8422 with a 8nV/rtHz voltage noise, gain is 350 V/V and full bandwidth of 2.2MHz (assuming you do not put a filter on the output of the inamp), you can actually get up to around 27 mVpp RTO noise. You can check the calculation on this link

    One common way to reduce the noise is to add a post filter which usually set 10x of your highest  signal frequency of interest.  You can do the same and see how it lessen the noise of your system.

    May I know what is your expected input range of the inamp? The  8mV output noise that you are seeing  with a gain of 350 V/V means that this noise when referred to the input is only around 23 uV which still a pretty good number. 

    If you are still looking for a lower noise inamp then I would recommend you to check  the AD8421 or AD8429 especially if you are planning to use dual supply of +/- 4.2V. But if ever you settled to use single supply of 5V, then the AD8422 is best for your applications.

    I hope this helps.

    Best regards,

    Emman