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.

Precise dc 0-5V measurement from distant hot area

Other Parts Discussed in Thread: ADS1255, THS4521, ADS1278, PGA280, ADS1220, OPA333, ADS1246, THS4531, REF5025, CD4097B

Hello,

First...sorry for my English and bad grammar. I'm not native speaker.

I need to measure dc voltage signal (or get average value from slow periodic signal to 100Hz - but it's not neccesary) with biggest resolution as possible. Dynamic range is 0.2-4.8V. Let' s say 500uV resolution is neccesary and anything better like down to 100uV resolution could be awesome for me.

Measured voltage is from sensor (what has weak output buffer - max 0.5mA) and is situated in "distant" area (0.3 - 0.5 meters) where temperature reachs 120°C.

I have this design: Sensor's single-ended output voltage is converted to differental signal around middle 2.5V with differental driver (example). This differental signal is through twisted cable (long 30 - 50 cm) measured by ADC. Samples are averaged cause of better SQNR and stuff... 

Considering requirements like resolution, dynamic range, dc to slow periodic signal...I would like to choose something like 24bit sigma-delta ADS1255 - full-scale is 2*Vref (Vref=2.5V), up to 30kSPS...

There will be possibility from time to time measure "offset error" of whole block (driver, filter, twisted cable, inputs of ADC) - for software correction of measured values in mcu what reads data from ADC. Correction values will be dependent on temperature value of sensor what mcu will know.

Now my dumb questions (it's first time when I have to solve "harder" analog problem).

Is it optimal way where I'm going with my design or is there something what should I be aware? I should look for op-amp with differental output with low offset and noise...Can someone recommend me specific op-amp for my application with ADS1255?

How can I make simple low-pass filter with cutoff 100Hz for reduction of noise and anti-alias function (example here) for adc input? I mean if I consider this blog post -I can increase capacity of RC filter, but I will get too much big capacity value (R = 10ohms  -> C = 47uF -> 94Hz cutoff). This leads into big capacitive load for op-amp driver. Increasing value of resistor leads to bigger noise and dividing signal cause of input ADC impedance...Is better way to make low-pass filter in adding paraller capacitors to feedback resistors in op-amp driver circuit (Miller effect)?

I appreciate any advice you may have


Best regards

Longin

  • Longin,


    There are a couple of options that I think you have for the front end. You could use a fully differential op-amp like the THS4521. We have often used it for front-end buffering for some of our faster ADCs like the ADS1278. If you look at the ADS1278EVM, the board has these devices as buffers. Look at how it's hooked up in the schematic. The user guide is listed in the link below:

    www.ti.com/.../sbau197.pdf

    In particular look at the offset specification is ok. It's on the high side of what I'd normally use, with a typical offset of about 200uV so the max might be too high.

    Another buffer solution would be to use something like a PGA280. It's probably a bit more complicated that what you need but you can look over the datasheet and see if it is something you can use in your application.

    One alternate solution would be to change the ADC. While the ADS1255 might have a low input impedance, you might try the ADS1220. It has a much higher input impedance so you might not need the buffer at all. The input range only goes to ±VREF (not ±2VREF) so you would need to get a 5V reference to capture the entire range. If you weren't able to go a 5V reference. You could gain down the buffer to have a gain of 0.4V so that you could use a 2V reference to measure a 5V input. This might be the simplest way to go.

    Incidentally, you haven't mentioned what sensor this is. If you give more specific details about it or supply a datasheet, maybe I might have more recommendations.

    Out of curiosity, what data rate do you plan to use? You first say that the data will be limited to about 100Hz, but then you say you want to sample it at the 30kSPS rate. With such a limited bandwidth, can you do a slower data rate? This may affect the amount of noise that you see.

    When it comes to the front-end input filter, I would start by increasing the resistance. In most of the ADCs that I work with 1k isn't a problem. Here it might be worth keeping it low, because the input impedance is 150kOhms (presuming no buffer and PGA=1). However, at 1k the noise won't be much of a problem, with 25C as a temperature and a 6kHz bandwidth (for 30kSPS). This comes to a noise of .31uVrms. Even if the temperature gets to 125C, the noise is .36uVrms.

    The bandwidth comes from the bandwidth of the digital filter of the ADC. Check the Frequency Response section of the ADS1255 datasheet. If you use a lower bandwidth, then there's even less noise that goes into the measurement.

    If you're not as concerned about the noise, and more concerned about the gain error (since there's now a voltage divider between the input impedance of the ADC with the added series input resistance), then there's not much you can do. I wouldn't add more and more complicated filters for the input.

    Anyway, if you want reply with some more details and I'll try to respond further.


    Joseph Wu
  • Joseph,

    thank your for your reply and advices.

    I've decided to go with this solution: multiplexer - voltage follower with OPA333 - ADc driver and filter with THS4531 - twisted cable - ADc ADS1246

    I' m not concerned about gain error since I can measure it with known temperature of are and compensate in software.

    Sensors are current transducers from different manufacturers what I want to compare.

    Since not all manufactures specify max. load for voltage output of the sensors I use buffer with OPA333. Maybe I should just increase values of resistors in the driver and give OPA333 away.

    Here is circuit with driver with 1/2 gain. There have to be delay between multiplexed measurement because of settling time.

    Frequency response with modef of THS4531, twisted cable and input of ADc

    and ADc. Reference is 2.5 V from REF5025

  • examples of used sensors are here
    www.winson.com.tw/products_e_current_sensor.html
    www.allegromicro.com/.../Zero-To-Fifty-Amp-Integrated-Conductor-Sensor-ICs.aspx

    As you can see I'm trying to measure with sensor sensitivity about 200uV/mA to 1mV/mA...these values are just "drowned" in noise and it is not application what they are used.
  • Maximillian,


    Normally, I'd try to get away from using so much buffering on the front end, but if you are using the ADS1246, you'll need the THS4531 to get off the ground node for a measurement. Just be aware that the THS4531 will add some amount of offset and offset drift and it may be difficult to calibrate out. Depending on the resistors used for the THS4531, you may also see some additional gain drift.

    I only had a few other comments for the rest of the schematic. For the ADS1246 supply decoupling capacitors, I'd use just a single 0.1uF capacitor and skip the 1uF capacitor.

    The load cap for the REF5025 is going to be more than the 100nF shown on the ADS1246 REFP and REFN pin. You need at least 1uF with an ESR less than 1.5 Ohms to ensure stability. Check the REF5025 datasheet for recommendations.

    I would increase the differential capacitance in the input filter so that it is at least 10x the common-mode cap. Any mismatching in in the common-mode filter capacitors will convert common-mode signals to differential signals. A larger differential filter capacitor will help attenuate any such differential signals that result from common-mode capacitor mismatches. If you want to look at it, the ADS1248EVM has some modest input filtering which you can examine or copy.

    I'd be concerned about the leakage of the BAT54 diodes on the front end of the ADC. Any leakage combined with series filter resistance may add additional offset to the measurement. In most makes of the BAT54, the max leakage current is 2uA, so this times the series filter resistance adds to the measurement. This current is likely much worse at higher temperatures.


    Joseph Wu
  • Joseph,

    thank you for valuable advices what I applied to the design.

    Add THS4531 :

    I was thinking about using a differental multiplexer (like CD4097B) or just two 1:8 single channel multiplexer what can hold temperature up to 110-120°C. One channel will be the signal of sensor and second will be the wire from the ground node of the sensor separated from the ground plane of whole device. This pseudo-differental signal will be multiplexed into the input of driver with THS4531. Could it be solution for what you was talking about?

    Add BAT54 diodes:

    Maybe I will just remove them and let internal ESD diodes of the ADS1246 do the job...

    Just for assurance - the measuring device with ADc will be in lab area (20 - 30°C) and the device with sensors, multiplexers and driver will be in a temperature chamber (testing up to 110 - 120°C). I am just saying this, because you was talking about leakage currents at higher temperatures.

    Max