ADS1248 External voltage reference & no. of input channel

Chandrashekar,


A typical RTD ADC measurement is ratiometric and doesn't use a specific external reference. It simply uses an excitation current to pass current through the RTD and a precision reference resistor so that the output code from the ADC is proportional to the RTD resistance divided by the reference resistor.

In one of your previous posts, I gave a link to TIPD120. It shows how this is calculated (with an error analysis). Additionally, there is another application note: "Example Temperature Measurement Applications Using the ADS1247 and ADS1248" which gives a couple of other examples. You can find it at the following link: www.ti.com/.../sbaa180.pdf.

Going back to your question, Are you sure you need a reference with a better long term drift? From the graph, the typical long term drift is about 60ppm. This translates directly to gain error. Which would be the equivalent of 0.006%. Note that this is the drift and does not account for the initial accuracy of the reference.

However, if you decide to use an external reference, I'll need to ask another group if there are any references with less of a long term drift. I'm not too familiar with entire product line.

For question 2, the input mux is a 8:2 multiplexer. If four inputs are used for two differential input channels, then there are four more inputs. You could make 3 "single-ended" measurements, if one input is used as a common input that would be tied to the ADC AINN. Note that this AINN input cannot be tied to AVSS (which would be ground for a unipolar supply) because this input would not be within the common-mode input range of the PGA.

For question 3, the same reference can be used to measure the other "single-ended" measurements. However, note that for the ADS1248 has several reference options. First, there is the internal reference that can be used to make measurements. Then there are two sets of reference pins that can be used to make measurements REFP0 and REFN0 or REFP1 and REFN1 which can be selected by the MUX1 register.

For question 4, I'm not sure what you mean. The ADS1248 doesn't take a 4-20mA input directly. The accuracy depends on what you use to convert that current to a voltage that the ADC can measure. As I'd indicated in a previous post, I don't know of any application notes that explain how to make the conversion. However, if you do have a circuit or schematic that can do that, we can discuss the accuracy.


Joseph Wu

Chandrashekar,


The accuracy can be calculated through accumulating all of the errors in the system.

First you use a 100 ohm resistor with 0.1% tolerance to convert it to voltage. This is an immediate gain error of 0.1% in the measurement. If you use the internal voltage reference, it has an error of +/-10mV. Unless you measure this reference directly and include this error into the measurement, this is about another 0.5% error. On top of that, there is the actual gain error of the ADC. This is listed in the electrical characteristics table as 0.02% max. The total gain error for this is just above 0.62% and the major portion of it is the accuracy of the internal reference. Certainly, you could use an external reference and the gain error would be significantly less. For example, the REF5020 has an initial accuracy error of 0.05% for the high grade version.

Another contributor to the accuracy is the offset. In the ADS1248, after a self calibration, the offset is 15uV max. This isn't a large amount compared to the full-scale reading, but it is something to consider in higher gains and smaller input measurement values.

Note that if are using a 100 ohm resistor to measure 4-20 mA, you may get errors if the current drops below 1mA. The input range goes only to AVSS-0.1V, and if you have the resistor connected to GND, with AVSS=GND, then a current lower than 1mA drops the input to below 100mV. It probably won't be a problem, but I thought I would mention it to you.

The amount of time it takes to shift from a ratiometric measurement with a reference measurement to a standard measurement using the internal reference should be instantaneous. The change should trigger a new conversion which should be a good conversion. If you have the internal reference turned off, and have to turn it on again, then this may take extra time to settle. There is a setting time for the internal reference, depending on the VREFOUT cap. This is listed in a table in the datasheet.

The amount of time it takes to cycle through conversions depends on the data rate of the conversion. If you look at the Data Conversion Time table in the datasheet (Table 16, page 35). It lists the conversion time required for a change in the setting. It may take a few extra microseconds to write the change from mux channel to mux channel, but the table gives you the time it takes to get the first conversion after the change. Note that there may be extra time required because of external settling. For example, if you change the excitation current from one channel to the next, you may have to wait for the input to settle because of capacitance on the measurement node. If you use large filter capacitors, this may be important to consider.

If you're interested, you might want to get an ADS1248EVM-PDK to test out how these measurements work. You won't be able to cycle through multiple channels, but you will be able to make individual measurements just to test them out.


Joseph Wu

Chandrashekar,


Could you please share a schematic of what you intend to build? It would be a little bit easier to follow. However, I'll answer your questions in order.

1. Yes you can share one reference resistor with two RTDs. TIDU191 shows something similar, where the reference resistor is shared between an RTD and a thermistor. Note that you can only make one measurement at a time.

2. Yes you can tie the VREFOUT to AIN7. There's nothing that prevents you from doing that.

3. I'm not sure what happened, but the link you posted is broken. I managed to find it, so I'll re-post it here:

e2e.ti.com/.../393973

It looks like what you want to do is take VREFOUT as the reference point and make the measurement from the 4-20mA resistors to VREFOUT. You could do that, but the accuracy may be an issue. The VREFOUT has a max error of ±10mV, this would give an initial gain error of ±0.5%. However, this max error also gives an additional offset error of ±10mV. With a small measurement, this a fairly large error.

It might be better to route the 4-20mA measurements to three resistors, but have the negative end connected together to meet at a low 25 Ohm resistor, but tap the AIN7 measurement so that all three currents are shunted to this point. The minimum for this would be 3x25x4mA, so that the AIN7 minimum would be 0.3V. The maximum current would be 3x25x20mA so that the maximum AIN7 voltage would be 1.5V. In both cases, the measurement would be in the input range. Using 100 Ohm resistors to measure each of the 4-20mA, this would be a max of a 2V input.

In this case, you do have to worry if the measurement current is less than 4mA or larger than 20mA. In either case, it may push the measurement outside the common-mode input range.

4. If you use the 100 Ohm resistors to measure the 4-20mA, the max input would be 2V. For this measurement you would use a PGA of 1. Note that you would use the internal reference to measure against, so you'd still have the 0.5% gain error (you would avoid the 10mV offset though).


Joseph Wu