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.

ADS1255: Single ended conversion and supply

Part Number: ADS1255
Other Parts Discussed in Thread: ADS1260, INA2128, INA128, TINA-TI, ADS1235, ADS131A02, ADS1246, ADS124S06

Dear community,

I am trying to use the ADS1255 for converting two wheatstone bridge signals simultaneously, so I am using the ADS1255 in single-ended mode.

Since the excitation bridge is unipolar to +3.3V, I am wondering if the following schematic is correct. 

According to the datasheet:

"For single-ended measurements use AINCOM as common input and AIN0 through AIN7 as single-ended inputs".

"The modulator measures the amplified differential input signal, VIN = (AINP – AINN), against the differential reference, VREF = (VREFP − VREFN). The differential reference is scaled internally by a factor of two so that the full-scale input range is ±2VREF (for PGA = 1). "

So my questions are:

1) if I want my reference of 0 ADC value to correspond to +1.65V for each bridge, should I tie AINCOM to 1.65V?

2) What about the VREF pins? should VREFP be tied to +1.65V?

3) How can I read each channel being input? the ADS1255 should output the values for bridge 1 and bridge 2, correct?

Best regards,

Miguel

  • For some reason, the image was not showing, so here it is

  • Hi Miguel,

    You mentioned in your original post that you wanted to read the bridges simultaneously - can you confirm? The ADS1255 is a multiplexed ADC, so you will not be able to read each bridge at the same time. You will have to select AIN0 for example, take a measurement, switch the mux to AIN0, take a measurement, and so on. I just wanted to be clear that the device is not a simultaneous sampling ADC. If you do need a simultaneous sampling ADC, let me know and I can recommend something else.

    When your bridge is balanced, the bridge centerpoint will be at 1.65V. However, you are feeding 1.65V into both inputs on the op amp, so the resulting output from the amp should be 0V. If you set AINCOM to +1.65V, then will you measure VIN = AIN0 - AINCOM = 0 - 1.65V = -1.65V when the bridge is balanced. If that is not what you want I would modify the circuit accordingly.

    In a bridge circuit, usually a ratiometric reference is used, such that the excitation voltage (3.3V) is also used as the VREF voltage. Therefore any changes in the excitation voltage are seen equally at the reference, and basically cancel out. The ADS1255 can support a max VREF voltage of 2.5V however, so this would not work in this case. I would suggest taking a look at the ADS1260, which is the next generation version of the ADS1255. This device has a very low-noise PGA built in, which can likely replace your discrete amplifiers and simplify your circuit. This is of course assuming you can take multiplexed measurements.

    Let me know

    -Bryan

  • Hi Bryan,

    Thanks for your reply,

    So, ideally, they should be simultaneous sampling. However, if the two sampling AIN0 and AIN1 frequency cycles can be fast enough to output both channels faster than 1kHz, it should be good. But if you recommend something else I will take it into consideration as well.

    The output from the amplification is 1.65V when the bridge is balanced I set the Vref of the amplifier to be 1.65V, so the bridge signal swings from 0 to 3.3V and stays at 1.65V when balanced. Thus. if the input on AIN0 = 1.65V when bridges are balanced and I set AINCOM to +1.65V, then VIN = 0, and the ADC should be 0 correct?

    I didn't understand the VREF. The VREF voltage is fixed at 1.65V it won't swing at any moment. Maybe a formula will help but I thought VREF would work only as a reference value not as ratiometric.  

    When you say ratiometric you are saying that a reading of X will be scaled twice for a PGA = 1? (considering the supply of 3.3V and VREF of 1.65V)

    The ADS1260 has 5 channels, and space is a big constrain in my case, I only need two single-ended channels. Also, I need gains higher than 128, more in the order of 1000 and above. That is why I have instrumentation amplifiers before the ADC.

    Thanks in advance!

  • Hi Miguel,

    Can you provide the following information:

    • Load cell sensitivity (usually in mV/V)
      • This will give the maximum bridge output voltage as VOUT(max) = sensitivity * excitation voltage
    • Amplifiers you are using
    • Target resolution (how finely you need to resolve the bridge's maximum output voltage)
    • Confirm that you would need data at least at <0.5ms/ch in order to read both load cells in 1 ms?

    I can perform some quick calculations to see what would work best here.

    -Bryan

  • Hi Bryan,

    • The sensitivity that I am trying to achieve is 1ADC per 0.5gm. With 200gm the bridge varies 0.08mV, being excited with 3.3V.
    • INA2128 with gains of 2000 +/- 1000, depending on the application
    • If I can detect 0.5gm of change, it would be enough
    • So far, I have achieved 320SPS, but 100SPS would be the bare minimum

    Let me know if you need anything else

    -Miguel

  • Thanks Miguel

    So I understand, you are saying that the maximum weight you are trying to measure is 200g, and at this weight the bridge output voltage is 80uV using 3.3V excitation - is that correct?

    And you are trying to get 0.5gm resolution, which is 200 / 0.5 = 400 counts. So that means you are looking at a noise level of 80uV / 400 = 200 nV - please confirm.

    -Bryan

  • Hi Bryan,

    I am not necessarily trying to max the weight at 200gm. I was thinking on adjusting the gain of the amplifiers to suit the maximum weight that I am trying to measure (I know that will affect the resolution but the sensitivity is around 0.4uV/gm)

    I am targeting always for the best possible resolution, which would be 0.5gm

    Not sure if this would help, but if needed we can try fixing the maximum weight as well

    - Miguel

  • Hi Miguel,

    I was trying to get an understanding of your maximum signal input as well as how small of a signal you need to resolve. This will tell you the noise floor you need to design for. From your previous posts, it sounds like your bridge had a maximum output voltage of 80uV at 200g, and you are trying to resolve 0.5gm, which results in a noise floor of 0.2uV. Note that the INA128 noise is specified at 0.2uVPP from 0.1 to 10 Hz at G = 1000, so this is already hitting your noise floor. And you are nowhere near 1kSPS sampling rates at 10 Hz BW, so I want to make sure your system provides you the performance you require.

    The PGA inside the ADS1260 actually offers lower noise density (~6nV/√Hz) compared to the INA128. If absolute best noise performance is required, it would make sense to consider this ADC. The ADS1260 is also smaller than the ADS1255, since you said that size was a constraint in your system. You are correct that the ADS1260 only offers a gain of 128, but given the information you have already provided I am not sure you need a gain of 1000 (or anything >128). If you use a higher noise amplifier with a gain of 1000, you are just amplifying the noise as well. So the first thing to consider is if the amplifier can meet the target noise performance at the required bandwidth. I would ask you to double-check my calculations based off of the system parameters you have provided.

    I would also doubt that you could measure 2x channels on the ADS1255 or ADS1260 fast enough to meet your 0.5ms requirement while also meeting the noise requirements. Even if you relax this requirement to 5 ms, you would need to set the ADS1255 data rate to 500 SPS. This results in a 3dB BW of 221 Hz, which will pass in considerably more noise whether you use the INA128 or just the ADC+integrated PGA. So you might need to consider using 2x ADCs for this application, unless you cannot relax your conversion time/ch requirements.

    Let me know if you have any questions about this information.

    -Bryan

  • Hi Bryan,

    The amplifier that I was using was an INA2128 followed by a Sallen Key active LPF with a Q factor of 0.707. The cutoff frequency is set to 30Hz. So the noise should be filtered before entering the ADC.

    Considering the mux complexity, I might consider using two ADC's. I might consider also the ADC+PGA (since it will allow saving space as well) if I get the resolution need.

    Any particular ADC on mind? 

    Thanks,

    -Miguel

  • Hi Miguel,

    Thanks for letting me know about the filter. I attempted to simulate your signal conditioning circuitry using TINA-TI and our Filter Pro tool given the conditions that you provided. The schematic is shown in the first image below.

    The second image shows that the noise at VOUT is ~38uVRMS. If you divide by the circuit gain of 1000 you get an input-referred noise of 38nVRMS. To estimate peak-to-peak noise, you could multiply this value by 6.6 to get 251nVPP. I am not sure if this is sufficient for your system, it appears to be a little high compared to your actual requirements (~200nVPP). But if this is enough then your system should be okay. For an ADC to have negligible impact on the signal chain, you would want the ADC's input-referred noise to be 5x to 10x lower compared to the results shown in the second image, since the ADC noise will also be scaled by the gain. So if for example the ADC input-referred noise was 3.8uVRMS at gain = 1, then the ADC's noise referred to the system input would be 3.8nVRMS. For reference:

    • ADS1255 noise at G = 1 and data rate = 1000 SPS is only 2.9uVRMS
    • ADS1260 noise at G = 1, data rate = 1200 SPS, and sinc 1 filter is only 2.3uVRMS
    • ADS1235 noise at G = 1, data rate = 1200 SPS, and sinc 1 filter is only 2.3uVRMS (this device is a subset of the ADS1260 and does not include an internal reference, so it is offered at a lower price point. It does not seem like your system needs an internal VREF though)

    The third image shows how the input signal is delayed from the input to the output of the filter. Even though you are measuring a DC signal, it will be changing slowly similar to a very low frequency AC signal. I simulated this using an 80uVPP ramp signal with a frequency of 10 Hz. If you expect the signal to change much more slowly in your system, this delay will be reduced. But you don't want to sample an unsettled signal either, which will add measurement error, so I wanted to make sure you were aware of this especially if you intend to sample much faster than the 30 Hz filter cutoff.

    -Bryan

  • Hi Bryan,

    Thank you for your insight.

    • When you say "you could multiply this value by 6.6 to get 251nVPP" why the 6.6 factor?
    • " you would want the ADC's input-referred noise to be 5x to 10x lower compared to the results shown in the second image, since the ADC noise will also be scaled by the gain.  So if for example the ADC input-referred noise was 3.8uVRMS at gain = 1, then the ADC's noise referred to the system input would be 3.8nVRMS". Why is that? In my mind, the noise is amplified by the INA2128, so why multiply the ADC input-referred noise by 1000 if it is post amplificiation?
    • " It does not seem like your system needs an internal VREF though", So for the ADS1235 the how would I set the 0 counts to be +1.65V? Also the datasheet states that the supply must be 5V for unipolar configuration. My analog maximum supply is 3.3V from the microcontroller
    • The frequency of the signal that we are trying to measure is up to 25Hz, maximum, so that is why cutoff is around 30Hz. The minimum ADC sample frequency that I would like to get is at least 100SPS, to ensure that the signal is well characterized, right?

    In previous posts you told that I might want to consider other ADC's to simultaneously sample the signal or two individual ADC's. After some research I found the:

    • ADS131a02 (two channels simultaneous sampling)
    • ADS1246 (1 ch single ended)

    Both of them accept an analog unipolar supply of 3.3V, but I am not sure about the noise performance and conversion. Could you please advise?

    Thank you for your patience and time Bryan.

    Best regards,

    -Miguel

  • Hi Miguel,

    Glad this is helping. Here are the answers to your questions:

    • Since noise typically follows a Gaussian distribution, the translation from RMS to PP can be calculated using a standard crest factor of 6 or 6.6 since this is purely a statistical determination. You can read more about that in our Precision Labs ADC content, specifically module 3.1: https://training.ti.com/ti-precision-labs-adcs
    • You are not multiplying the ADC noise by 1000, you are dividing it by 1000 to refer it to the input of the INA (where you apply your signal). This is a purely mathematical effort to make sure the noise seen at the input is scaled correctly. At the output, the INA noise is scaled by 1000, the filter noise is scaled by 1, and the ADC noise is scaled by  . To refer this to the input, you would divide each term by the circuit gain (1000). That is all that I was doing.
    • As I mentioned in my original post, I would recommended a ratiometric reference for this application. An example of this is shown in the first image below, which is taken from the ADS1235's datasheet. Note that the bridge excitation voltage is routed back to the REFx0 pins on the ADC through the Sen+ and Sen- lines. If there is any slight variation in the excitation voltage due to supply issues, power spikes, etc., it will show up in the measurement and reference voltage equally and cancel out. For a system where such low noise is required, this might be important. Also, this implementation can be used on any device with external VREF inputs, it is not limited to the ADS1235 for example.
    • Regardless of your reference voltage, if the bridge midpoint is 1.65V, you apply the bridge output to AIN0 for example, and apply 1.65V to AIN1 for example, then when the bridge is balanced you will read zero code (assuming no offset/ gain errors, etc.). Your reference voltage needs to be able to accommodate the output span of your INA + filter, so whatever that may be, while also meeting the minimum VREF requirements on the ADC.
    • I am not sure the INA128 (or INA2128, which is just the dual version of the INA128) can operate on 3.3V supplies. See the second image below that is taken from our Analog Engineer's Calculator tool. Looks like it needs at least a 4.5V unipolar supply
    • If your signal has a 25 Hz BW, the filter will cause significant delays from the input to the output. In my previous post, the output signal was shifted by about 10 ms using a 10 Hz input, so you could expect even greater delays for a 25 Hz signal. And you do not want to sample an unsettled signal, so you might want to rethink this circuit a bit. But yes, if your input signal BW is 25 Hz, then Nyquist says you need to sample >2x faster for faithful reconstruction. So theoretically any sample rate >50 SPS should work
    • You can check the noise tables for any ADC you are considering based off of the criteria I gave in my last post (and expanded upon in this one). The A02 for example has an input-referred noise of 1.82uVRMS at a data rate of 1000 SPS. This is shown in Table 1 on pg. 23. The ADS1246 has a similar table you can reference in its datasheet. Again, the noise scales with data rate, so you would have to clearly define a minimum data rate value to be able to determine the ADC noise you can expect. I think in your original post you mentioned wanting to sample at 0.5ms (2 kSPS) per channel, then also 320 SPS and 100 SPS. So I am not sure what value you ultimately settled on.

    -Bryan

  • Hi Bryan,

    I took some time to analyze thins more carefully. 

    I will try to be as clear as possible with the requirements as well.

    ADC System requirements:

    • 24 bits resolution
    • Sensitivity of 0.5gm
    • Resolution of 0.5gm/ADC count
    • Need to read two bridges --> simultaneous sampling with one ADC or using two ADC's
    • input supply voltage unipolar 3.3V
    • SPS >= 10

    I agree with you that the use of a ratiometric reference seems to be a good solution. My only question is that when using ratiometric reference, do I need and should I have a pre-amplification as well? If not, what should be the gain on the PGA?

    So, to clarify my path of thoughts I made the following chart to help decide and research the most suitable ADC, that fulfills the supra mentioned requirements:

    If you could advise on this matter it would be great to ultimately decide the ADC that I will use. I am new to this matter of ADC's too, so, meanwhile, I am reading the e-books available on precision ADC's and amplifiers on the Texas website and the link you have provided.

    Thank you in advance for the time to support me and this project Bryan.

    Best regards,

    Miguel

  • Hi Miguel,

    Let me link you to our Precision Labs series. There is a lot of great material in there, but I think the section on ADC noise would help you here, and even more specifically module 4.4. This video basically steps through the process you are going through right now: selecting an ADC for a load cell application.

    When you review this video, you will notice that some of the information required to determine your noise floor and dynamic range, as well as select an ADC, is missing from the list in your last post. Specifically, the maximum bridge output voltage and weight range.

    In our previous discussions, you had mentioned that the weight range might be 200g, but I am still unclear on the bridge sensitivity which will help determine the max output voltage. Basically, you need to know what the bridge output voltage is when the maximum weight is applied (even if you won't necessarily apply that much weight in your final design). The most common load cells we come across have a sensitivity of 1 or 2mV/V. This means that if you applied your 3.3V excitation, the maximum bridge output voltage would be 3.3mV or 6.6mV. I think you previously said your max output voltage was 80uV, which is significantly smaller than I would have expected, and suggests a very low sensitivity bridge. I am sure that such bridges exist, but I have not seen one before to be honest.

    If you can confirm these values (weight range and maximum bridge output voltage) I can help you choose an ADC, as these values will determine your noise floor. You really cannot proceed until you know the minimum noise level you need to resolve.

    Let me know if this request makes sense.

    -Bryan

  • Hi Bryan,

    I will look into the link.

    I made some experiments (file attached). The file has data regarding input weight and the differential output of the bridge

    A 3.3V excitation was used. The maximum weight tested was 2.685 kg and the sensor is prepared to hold up to 6kg (structurally the sensor can hold up to 13kg, but setting a safety coefficient of 50% gives the maximum of 6kg).

    The trend line is linear so I believe the output voltage can be estimated from the model attached.

    Best regards,

    -Miguel  

    BRIDGE_TEST_OUTPUT.xlsx

  • Hi Miguel,

    Judging from the data you sent, it would appear that the bridge sensitivity is on the order of 1mV/V, since there is about 3mV output at the max weight (13kg) given the plot trendline.

    However, you appear to only be using about 200g of this range, or ~1.5%, for your system. Is there any reason you chose a bridge with such a high max load if you are only going to use a small portion of the range? I would strongly recommend getting a different sensor, the sensor noise alone might not let you resolve such small signals (let alone the signal chain noise).

    -Bryan

  • Hi Bryan,

    The sensor is in the test phase and is being developed and adjusted. I know for sure that the sensitivity of the mechanical structure can be improved and it will be. But that was the reason why I used the INA2128 to have a controlled gain, so I could adjust it to the case of study. For now, I cannot share results, but if you could recommend some ADC's I would look into them and test some. Once I have more data than I can contact you to figure the most adjusted solution to the case.

    - Miguel

  • Hi Miguel,

    I cannot stress enough how important it is to know the noise target for your system and sensor. If that is a moving target, it will be very challenging to design your system in any meaningful way, because you might end up needed a discrete signal conditioning block comprised of an INA and a second-order filter, or you might be okay with just an ADC that has an integrated gain stage. And since you require 3.3V supplies you are not able to use the INA128 anymore, so now you need to find a lower power INA, which will definitely have higher noise. But what if you don't need an INA at all? Does it make sense to continue designing this circuit when it may end up being a sub-optimal, overly complicated solution?

    For ADCs, the ADS1235 would be an ideal option for this type of measurement, given the low noise, high gain, faster sample rates, external reference inputs, and small package. Since you want a device that can run off of 3.3V, I would instead suggest either the ADS1246 or ADS124S06. The ADS124S06 has more channels than you need, but the noise is ~30% lower compared to the ADS1246 and comes with a high performance internal reference if you decide you need one for your application. It is also in a small 5x5mm package.

    The ADS131A02 could also work, but this device would need a driver amplifier since the inputs are unbuffered. So you would likely need a wider BW amp for this application, since an INA likely would be too low BW, especially a low power op amp. The driver amp could also be configured for gain, but it would be fixed gain and would not be high input impedance, so you might also need a buffer (or you could use an INA before the driver amp).

    -Bryan

  • Hi Bryan,

    I am now more aware of the importance of having a good SNR and knowing better my system. I will make more experiments, look into data and be more accurate to get the sensitivity and noise errors as well. The sensor will be fully characterized to get these values.

    I will also look the documentation of ADC's on noise analysis as well as the datasheets for those ADC's. I believe I have everything for now and that is time to study things deeper. I will get back to you once I have more data as well.

    Thank you!

    - Miguel

  • Sounds good, Miguel, I am glad that this information was useful to you.

    Let me know if you have additional questions in the future, and we will be happy to help.

    -Bryan