TI E2E Community
Precision Data Converters
Precision Data Converters Forum
Effect of ADS1148 IDACs current source mismatch and temperature drift matching on error estimation
Hi, I am using ADS1148 in my design for data acquisition applications with 3-wire RTD - PT100 sensors.For better result I am using both the current sources (IDACs), available on-chip. Operating range of temperature is -40degC to +70degC. For error estimation, I am considering initial mismatch in IDACs, which is specified as typ +/-0.03% of FS in datasheet. Also there is Temperature drift matching parameter, which is specified as typ 10ppm/degC. So, could you please clarify on followings: #1. How should I consider the overall effect of IDACs' initial mismatch as well as temp. drift matching between IDACs for the error estimation? #2. If I refer to IDAC DRIFT curve provided in the datasheet (please refer to figure13 on page no.14), it seems that maximum IDAC mismatch is 0.004uA at 1.5mA IDAC setting. This comes to (0.004uA*100/1500uA) = 0.000267% of FS, which is very very less compared to the specified mismatch as typ +/-0.03% of FS. So could you please clarify that which data should be considered? #3. It will be of great help if you could specify the Total Unadjusted Error(TUE) for the ADS1148. Thanks and regards, Rajesh Kumar
The error due to the IDAC initial current mismatch and IDAC drift mismatch over temperature will not play a significant role in the over all error in the system when using the ADS1148 (16 bit). The IDAC initial mismatch error is +/-0.03% of full scale, and the IDAC mismatch drift Over Temperature is only 10ppm/C. The Drift and Noise errors of the excitation IDACs tends to cancel if the device is set up in a ratiometric configuration as shown in the figure below (at the bottom of this post). You may refer to the application SBAA180, which explains different RTD measurements using the ADS1248 (higher precision 24 bit version) Please find Link below:
On this 3-wire RTD application with hardware compensation, the RBIAS resistor must be a high precision low drift resistor, since this resistor is used to generate the voltage reference. The RCOMP resistor must also be a high precision, low drift resistor.
When calculating the Total Unadjusted Error, the standard technique is to perform the square root of the sum of squares of the maximum DC errors involved in the system. Typically DC Errors such as offset, INL and Gain are used to perform this estimate. One thing to emphasize, is that when the maximum error specifications given on the datasheet are used to perform the TUE estimate, this results in the worst case results; the typical performance of the device is much better.
For example, looking at the figure below, a quick rough estimate for worst case error could be implemented. If we let RCOMP= 105 Ohms, set the two IDAC's=1.5mA and VREF = RBIAS*3mA = ~2.5V; we obtain the following results for the differential voltage seen at the inputs of the PGA as the RTD impedance changes over temperature:
The Total Error can be calculated by performing the sum of squares:
TUE = Square Root( (INL Error)^2 + (Offset Error)^2 + (Gain Error)^2 + (IDAC Mistmatch Error)^2 +(RTD accuracy Error)^2 )
You may then perform the TUE error estimate for each point in temperature. For example, if a standard Class B RTD is used in the application, the error in temperature measurement can be estimated for each point in temperature. The accuracy of the class B RTD is specified as +/-0.5 Celsius at -40C. Using the worst case escenario maximum specs of the datasheet:
Offset Error = +/-1 LSB = +/- 1.192uV
INL Error = +/-1 LSB = +/-1.192uV
Gain Error = 0.5% = 31.095mV * 0.5% = +/-155.48uV
IDAC Mismatch Error = (1.5mA*(105Ohm-84.7Ohm)*0.03%) + ((25C-40C)*(10ppm/C)*1.5mA*(105Ohm-84.7Ohm) = ~30uV
Accuracy of RTD @-40C = +/- 0.2 Ohm *1.5mA = +/- 300uV
Performing the square root of the sum of squares, we could solve for the total uncalibrated error:
Total Unadjusted Error = ~+/-339.58uV; this would correspond to approximatly +/- 0.559 C error in temperature measurement.
In this simplified example, the errors are dominated by the Class B RTD accuracy and the 0.5% maximum Gain Error of the ADS1148 device; the IDAC mistmatch is not dominant source of error. A more realistic estimate of expected performance would be to use the typical specification of 0.125% gain error for the ADS1148 (16bit version). In addition, the ADS1248 (24 bit higher precision version) offers a tighter gain error specification (uncalibrated Gain Error of +/-0.005% typical and +/- 0.02% maximum)
Also, this quick error analysis did not account for errors due the accuracy RCOMP/ RBIAS precision resistors; and for the mismatches on the lead wires.
Thanks for the quick reply.
I can't go with Rcomp approach, as I have to measure the temperature of different types of RTD sensors. Please see the below table for the range of RTD resistances over the measuring range of temp -40degC to 200degC(I have listed the Rmin@-40degC and Rmax@200degC in the table below.) :
My accuracy requirement is +/-1degC over the measuring range of temp -40degC to 200degC.
Also the most accurate RTD type is the PT-100, for which change in resistance per degree celcius is 0.38 Ohm, so configuring IDAC for 1mA, I need to measure the change of 380uV . To measure the smallest change in temp/voltage, TUE should always be less than half of 380uV, which is 190uV.
So, could you please confirm that whether ADS1148 will be suffice to meet +/-1degC of accuracy requirement or I have to go with ADS1248?
Could you please clarify my doubts which has been posted on 03 Feb, 2012?
I need to finalize the design very soon.
Sorry for the delay; I am out of the office on travel.
If you are not able to use the hardware compensated (RCOMP) compensation resistor; the device will see a larger diffferential voltage at the inputs of the ADS1148; and you will need to set up the device with a lower PGA gain in order to not exceed your Full-Scale voltage. Gain errors become a more dominant source of error at larger differential voltages. In this case, the ADS1248 will be more appropiate since it guarantees a maximum uncalibrated gain error of 0.005% (0.02% max) and the device offers 24 bit resolution (smaller LSB size).
Thanks for the clarification.
Now I am using ADS1248 for better resolution in my design, however need some more clarifications:
Firstly, typical IDAC Initial Mismatch in case of ADS1248 is specified as (+/-) 0.15 % of FS, which is much higher than for ADS1148 (for which it is 0.03%). So, could you please confirm this value for the ADS1248?
Secondly, specified Gain error is max 0.02%, so could you please clarify that whether this error includes the VREF error also?
Similarly, the specified Gain drift (From Fig 19 VDD=5V and 20SPS; extrapolated from PGA=1 and PGA=32) is 0.015%, does this error include the effect of VREF?
1) The ADS1248 datasheet has the correct IDAC mismatch current of typical +/- 0.15% error. The IDAC initial accuracy and mismatch histograms are shown on Figure 44 and 45 of the ADS1248 datasheet. The ADS1148 datasheet appears to have a typo and needs to be updated/corrected.
2) The Gain Error does not include the VREF error; this needs to be accounted for separately.
3) The Gain Error does not include the VREF error; a very accurate external reference of 2.5V is used in generating the gain error figures.
Thank you and Best Regards,
Thanks Luis for the clarifications.
As the specified gain error doesn't include the VREF error, I need to use a very accurate external reference of 2.5V. So could you please tell me how much drift and accuracy are required for the external VREF device? Also it will be very nice if you could suggest a very accurate external reference of 2.5V from the TI.
Luis, could internal bias voltage be used for measuring the current across a shunt resistor? I am not sure about this as datasheet states the common-mode input (VCMI) must be within the range AVSS + 0.1V + (V )(Gain)/2 to AVDD - 0.1V - (V )(Gain)/2.
Luis, I want to measure mA current using an external shunt resistor, however due to CM input range constrains I need to bias this input voltage, so could internal bias voltage be used for this purpose? I am not sure about this as datasheet states the common-mode input (VCMI) must be within the range AVSS + 0.1V + (V )(Gain)/2 to AVDD - 0.1V - (V )(Gain)/2.
You could use the REF5025 voltage reference. The High grade offers initial output accuracy to +/- 0.05%; Standard Grade +/-0.1%.
Yes, you are correct, the ADS1248 has a common-mode requirement where the ADS1248 internal Programmable Gain Amplifier requires headroom in order to amplify the input signals.
The common-mode voltage is defined as:
VCM = (AINP+AINN)/2
The differential signal VIN is given by:
VINdifferential = AINP – AINN
The Common-mode Voltage Range Limits are given by:
AVSS + 0.1 + [(PGAGAIN)(VINdifferential)/2] < VCM < AVDD - 0.1 - [(PGAGAIN)(VINdifferential)/2]
The requirement is that both the AINN and AINP should be more than 0.1V from the rail supplies AVSS and AVDD. Depending on the maximum PGA Gain that you plan to use and Maximum differential signal applied; the common-mode limits can be calculated.
In the case where you require to measure the current using a shunt resistor, one possible option is to bias the ADS1248 using bipolar AVDD/AVSS supplies (+/-2.5V) and ground the shunt resistor. A second option is to bias the ADS1248 with unipolar +5V supply and implement a voltage to current converter using an OP AMP and bias the device so the differential signal is optimally centered at the middle of the supplies.
Below is a link to a Design Note for the ADS1248 and ADS1148 families that discusses the common-mode requirements with some examples.
Thank you very much for the valuable inputs.
Yes, I will use REF5025 to generate reference voltage, it seems to be more precise and suitable for the design requirements.
However, as I am measuring RTD and mA currents on the same channel, of course, one by one, use of bipolar supplies is ruled out.
As I understand from the datasheet that typ. bias voltage = (AVDD+AVSS)/2 = 2.5V is generated from the on-chip bias generator, so I would like to know whether this internal Voltage Bias generator could be used to meet the CM requirements of ADS1248? From the design notes available from the link sent by you explains only about thermocouple measurements. So I am not sure that the internal bias could be used in case of mA measurement. Please clarify this one, Luis.
I hope you are doing well.
I have few queries regarding ADS1248, listed below:
1. To meet the required accuracy of +/-1degC for RTD measurement, IDAC initial mismatch, (+/-) 0.15 % of FS, which is significant. I think this can be taken care by calibrating the ADS1248. For this either I drive both the IDACs through a highly precision resistor and measure the mismatch(as in case of 3-wire RTD measurement)? Other way of finding mismatch may be driving the IDAC one by one through this precision resistor and calculate the mismatch. But I am not sure about these methodologies. So, could you please comment on these methodologies
2. In case the TI have have a better idea of knowing the IDACs initial mismatch, could you please share them?
It is possible to calibrate the mismatch of the IDAC currents by alternating the two IDAC's through the positive and negative Inputs of the ADC and performing two measurements. By alternating the IDAC's and averaging the two measurements (performing two samples), essentially the error of the IDAC mismatch cancels out. The slides attached in the power point below explain this procedure in more detail. The absolute value of the IDAC current is not critical provided the device is configured in a ratiometric measurement.
If you are performing RTD measurements using the IDAC's; in general you will be configured in a ratiometric measurement; as described on the application note SBAA180 attached below. In a ratiometric measurement the IDAC current flows through the external resistor RBIAS across VREFP /VREFN and generates the voltage reference. In this configuration, the errors due to drift and some of the noise of the excitation current is seen both at the input signal path and reference path and tends to cancel. the RBIAS resistor must be a high precision, low drift resistor.
Can you please clarify: are you trying to measure other sensors that require an external REF5025? Is the shunt current measurement described on your post related to another sensor or was this an attempt to measure the IDAC's? If an additional sensor measurement is required where you would like to use the REF5025; the ADS1248 provides two external reference inputs for flexibility.
Thank you very much for sharing the slides regarding correction of IDAC mismatch and drift.
Now I am more confident of using ADS1248 in ratiometric configuration for RTD measurements.
Luis, as the RBIAS resistor must be a high precision, low drift resistor, can you quantify the precision and drift in this case?
Clarification: I want to measure mA current using other channel of the ADS1248 and voltage across the shunt varies from 8mV to 1920mV. So, I want to know that whether internal Vbias could be used to meet common mode voltage input constraint? As from datasheet specified Vbias = (AVDD+AVSS)/2, which comes to 2.5V (I am connecting AVDD to 5V and AVSS to AGND), so I think that this Vbias could be used. Please comment on this, Luis, I need to finalize the design very soon.
It will be very nice if you could suggest a low cost solution to provide Vbias to meet CMI voltage requirement of ADS1248.
Also, if I plan to use internal reference for this purpose, how should I calibrate the error introduced by drift and load regulation? Is it necessary to use external 2.5V VREF device?
My only concern with internal reference is drift of 15ppm/degC and accuracy on voltage output is 0.976% of FS. I am not clear about how to calibrate this error.
I am waiting for your valuable and quick response.
After measuring the voltage accross the RTD using the A to D converter; you will need to calculate the impedance of the RTD in order to solve for the temperature. In order to perform this calculation, the value of the RBIAS resistor must be known. The precision of your RTD impedance calculation is directly tied to the precision of the RBIAS resistor. I recommend that you go through the analysis in order to quantify how much error is allowed in your application.
The Bias Voltage Generation Function (VBIAS) is intended to bias ungrounded thermocouples where the ungrounded thermocouple uses negligible or very low bias current. In applications where the user needs to measure a sensor with current output or current through a shunt resistor; a current to voltage converter may be implemented using an op-amp.
The reference drift can not be easily corrected for. The reference drift error becomes important for applications where the ADS1248 device is exposed to large temperature changes. The 2.048V internal reference offer a typical of 2 ppm/C (max 10ppm/C) from +25C to +105C which is a relative low drift spec. The REF50XX series of voltage references guarantee a max 8ppm/C for the standard grade or max 3ppm/C for the high grade.
Thanks Luis for your inputs.
Now, I am using an external current to voltage converter with proper DC shifting to meet the CMI requirement of ADS1248.
Luis, do I need any buffer to drive ADCs from REF5025? Because I will be driving two ADCs as well as DC shift circuitry.
All content and materials on this site are provided "as is". TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with regard to these materials, including but not limited to all implied warranties and conditions of merchantability, fitness for a particular purpose, title and non-infringement of any third party intellectual property right. TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with respect to these materials. No license, either express or implied, by estoppel or otherwise, is granted by TI. Use of the information on this site may require a license from a third party, or a license from TI.
TI is a global semiconductor design and manufacturing company. Innovate with 100,000+ analog ICs andembedded processors, along with software, tools and the industry’s largest sales/support staff.