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.

ADS1281: How can I design the (single to differential) input circuit to achieve best precision?

Part Number: ADS1281
Other Parts Discussed in Thread: ADS1282, OPA211, OPA210, THP210, OPA227, OPA1612

Hi experts,

I'm designing a high-precision measurement system that contains a low-noise and high-precision analog OPAMP front-end and an ADC.

My front-end circuit is all in single-end design and utilizes coaxial cable to transfer signals. But when I'm going to select an ADC chip, I found that almost all high-precision ADCs require or suggest using differential input type to achieve the best INL and SNR parameters specified in the datasheets. (As I understand, the differential structure can minimize the external interference and suppress the even-order harmonics and charge-injection of the switched-capacitors inside the chip and so on...)

Due to my lack of experience, I'm really confused about how to design the input circuit in precision measurement applications. (The ADC meant to be selected is ADS1281)
There are three methods as I immaturely thought:

  1. Connect the single-end OPAMP output (with mid-supply) directly to the differential input pair of ADS1281. (which mentioned in the datasheet, but is not recommended)
  2. Use a fully differential high-precision OPAMP to perform the transform from single-end front-end output to differential ADC input. (which may insert additional noise, and suffer from mismatches of the resistors in the differential OPAMP stage?)
  3. Select another high-resolution ADC that is optimized for single-end input? I don't even know if it exists.

I really need you experts' rich experiences to make the decision. If there are tutorials or technical articles that may contain the answer, please let me know!

Thank you very much!

  • The goal of the ADC's parameters:

    Data rate >= 1k SPS

    Resolution >= 24 Bits

    SNR >= 120 dB

    INL <= 1 ppm

    Budget : 30 $, in stock.

  • Hello Robert,

    Apologizes for the delay. Your specifications pretty much guarantee using the ADS128x family, which rules out a different ADC optimized for single ended input (or any other ADC for that matter).

    In addition, you're right, a puesdo differential (one input at mid-supply and another input changes) will not give you the best performance. I would be surprised if you were able to reach the upper end of the SNR values that the ADS128x can provide in a puesdo-differential. So, true-differential is the way to go if you can. I will say, there is nothing wrong with prototyping something and trying it! It might be good enough (though it might be unlikely).

    I also want to point out the difference between resolution and Effective Number of Bits (ENOB). For 1kSPS data-rate, you're looking at an ENOB of 22 bits according to the SNR values. Adding in an input stage will make it worse, no matter what. The ADS1282, which incorporates a front in PGA, starts to hit an ENOB of 19-21 bits at higher gains and that PGA has a spectral noise density of 5 nV/sqrt(Hz). I'm pointing this out because your front-end, whatever the design is, should aim to be lower than the 5nV/sqrt(Hz) to be competitive. You're spot on with the types of error that will be incorporated with a single ended to differential stage so the design will be difficult. 

    I'm reporting from the ADC team, so I'm not going to be able to give you the advice you need to design something with this precision. Let me loop in the precision op-amp team and see if they can help out.

    I will say, the datasheet does call out the OPA211, which can help you find similar devices or use the OPA211. In addition, we have a circuit-cookbook with high level circuits, such as a single-ended to differential translation. While, the specific op-amps for that circuit are optimized for high-speed SARs, you will get some idea how to design the front end for your application by abstracting the topologies: https://www.ti.com/lit/slyy138  

    Unfortunately, that's the best I can recommend with my skill level. I'll loop in the op amp group and see what they say.

    Best,

    -Cole

    Edit: one more thing! Tell us what your sensor or signal you're trying to measure and it will really help the op amp team out a lot. So be sure to update with that information.

  • HI Robert,

    The ADS1281 is often used on geophone sensor applications as shown on the example of Figure 56 of the ADS1281 datasheet.  In this application, the signal of the sensor is fully-differential, and the geophone is located in close proximity to the ADC.  All these factors help the ADC module eliminate noise while achieving a high level of resolution on the geophone application.   In addition, on many of these sensor applications, the same voltage reference is used to excite the sensor and as a voltage reference for conversions in a ratiometric circuit.  This ratiometric configuration results in noise cancellation and helps to greatly reduce the noise contribution of the voltage reference.

    You are correct that the higher performance, higher resolution precision ADCs incorporate fully-differential inputs, and there will be always an addition of noise when performing single-ended to differential conversion. The resistors in the front-end will also have some noise and drift contribution as well. 

    As Cole has mentioned, the OPA2211 is probably among the lowest noise amplifiers at 1.1nV/sqrt(Hz) at 1-kHz.  Other low input referred noise devices that could be of interest include the OPA210 (2.2nV/sqrt(Hz)), OPA227 (3nV/sqrt(Hz)). The fully-differential amplifier, THP210 is relatively low noise at 3.7nV/sqtr(Hz).  

    Achieving ~120dB SNR performance will be likely difficult, but we could get close at the +118dB level with the OPA2211. This is without including any noise contributed by the voltage reference or any extrinsic noise injected into the device through the cables.

    You have mentioned the data rate is 1-kSPS.  What is the input signal frequency range? What is the signal amplitude and common-mode voltage?    In addition to noise/resolution, are there concerns about DC offset / drift performance? Are there any input impedance requirements?  What kind of sensor or source circuit is driving the acquisition system?

    Unfortunately, the OPA2211 and many of these products are currently are out of stock, the OPA1612 which is an audio version of the OPA2211 with higher offset/drift is available. The SOIC package version of the THP210 which is slightly higher noise is available. 

    Thank you and Regards,

    Luis 

  • Hi Luis and Cole,


    Apologize for haven't introduced my front-end design. It's a 2-stage-OPAMP-based precision small-DC-current amplifier. Because of the feedback OPAMP structure and the low-speed application, impedance matching is not a problem (I guess). Besides, the gain and amplitude of this true-bipolar single-ended front-end amplifier are configurable.

    My front-end output waveform is in sawtooth shape. Firstly, I use the ADC to sample the amplitude of each point. Then I calculate the slope of my rising edge, which represents the current to be measured.
    (In specific, I'm using the 1ksps-data-rate ADC to sample a 100Hz sawtooth wave, i.e. 10 points per period.) The DC offset is insignificant in my slope measurement.
    The specific scheme illustrates in the figure below.

    According to your suggestions, I'll insert a low-noise single-to-differential converter between the front-end and the ADC, and a simple RC anti-aliasing filter with a cut-off frequency of about 75kHz. (refer to figure 56 in the datasheet). I'll try if it can meet my requirement of noise.

    Here are my new questions: since you recommended a series of precision OPAMP, are you suggesting I utilize them to achieve the single-to-differential conversion or buffer the input signal? If the former, why you didn't mention the full-differential OPAMP?

    Thanks for your professional guide!

  • HI Robert,

    I suggested above the lowest noise amplifiers available. 

    Both Fully-Differential Amplifiers (FDAs) or op-amps circuits can be used to perform single-ended to differential conversion.  Some sensors require a high-impedance front-end and are unable to drive directly a fully-differential amplifier; therefore we need to ask questions about your front-end circuit prior suggesting an FDA or op-amp circuit.  Among the different devices, I mentioned the THP210, which is a fully-differential amplifier.  In this case, since you are using a 2-amplifier front end, there is most likely no problem driving the 1kΩ input resistance of the THP210 FDA.  Among fully-differential amplifiers, the THP210 offers high DC precision and very low 1/f (low frequency) noise.  

    Keep in mind, in properly designed multiple amplifier stage applications, the noise contribution of the first amplifier stage typically dominates.  

    What is the noise expected in microvolts RMS at the output of the 2 op-amp front-end?  Can you please share this circuit?  Depending on the noise contribution of the first 2 op amp stage; you may not need such a low noise buffer or single-ended to differential circuit; and you may be able to relax your ADC SNR requirement. 

    Thank you and Regards,

    Luis

  • Hi Luis,

    You guided me to calculate my front-end circuit's noise first and then select the proper ADC, and I'm working on it these days.

    My front-end circuit is a two-stage OPAMP-based small DC amplifier, which requires as low INL and noise parameters as possible. It designs for a high-precision instrument.
    Sorry, I can't provide my detailed circuit for now, but I build a noise model of my practical design, whose results show in the table below:

    Front-end output-referred noise voltage density (white noise) 182 nV/√Hz
    Front-end output-referred noise voltage density (1/f noise) 2.70×√(1/f) μV/√Hz
    RMS Voltage (after a 1-order RC AAF with 75kHz cut-off) 63.4 μV

    (I'm not sure that the up-limit frequency in RMS calculation should be from 1.57×RC filter or the ADC's digital filter) 


    ADS1281's 124dB SNR (for 1000SPS) seems to be wasteful. However, I don't expect the ADC overly increase my total noise floor. From your professional sight, should I relax my SNR requirement or simplify the ADC driver? Do I still need an ultralow-noise single-to-differential circuit?

    Thank you!!

  • HI Robert,

    If the noise density of your 2-op-amp front-end is at 182nV/√Hz; the noise contribution of the THP210 single-ended to differential converter will not be significant. The THP210 is a high-precision, low noise fully-differential amplifier with only 3.7nV/sqrt(Hz) at 1-kHz; and very low 1/f noise, about 10nV/sqrt(Hz) @1-Hz per figure 6-8 of the THP210 datasheet. 

    Keep in mind, the f(-3dB) bandwidth of the ADS1281 when set at a data rate of 1-kSPS is 0.412*fdata = 412-Hz.  This bandwidth may be too low to properly convert the 10ms period sawtooth signal, because the sawtooth signal has a high-frequency step component.

    Below is one possible example of the THP210 single-ended to differential converter, set up in a second-order, G=1V/V, Butterworth, low-pass filter with a f(-3dB) corner frequency of 5kHz. The total output noise of the circuit is 3.33uVRMS. The single-ended bipolar input sawtooth signal is assumed to be centered at ground, with a maximum of ±4.5V amplitude.  The fully-differential ADC is assumed to be setup with a reference voltage of 5V, where the VOCM pin of the THP210 is driven to VREF/2 = +2.5V.  The THP210 is powered with bipolar +/-5V supplies; and high quality COG/NPO ceramic capacitors are used in the signal path.

    Please also note that the R-C-R filter between the ADC and THP210 driver is a charge-kickback filter with a much higher corner frequency.  This charge-kickback filter is primarily used to help the amplifier to drive/recharge the ADC's sample-and-hold capacitor and settle when the ADC is not buffered.

    There are multiple choices for Delta-sigma's or SAR ADCs that can be used for this application. The optimal choice of ADC will be dependent on the on the expected noise target, which is of course a function of the complete analog front end.

    There are many resources to learn about ADCs below:

    Analog-to-digital converters (ADCs) | TI.com Training Series

    Hope this helps on your project.

    Thank you and Regards,

    Luis Chioye

    TINA Simulation Files:

    Stability_3_21_22_butterworth_5000Hz.zip