ADS1258 - Questions about Reference Circuit in Datasheet

James,


Sorry I didn't get to this earlier, but I was out of the office on Friday. Here are my answers:

1. I'm not sure I understand this question. Let me explain it this way, and hopefully this answers this question. In the setup in Figure 67, there is a 10:1 voltage divider to the input. Note the 9.09k and 1k resistors on the front end. By using this voltage division, you can use +/-10V inputs to get it into a range that is usable. To adjust this to work for +/-2.5V, you would replace the resistor divider with 1.5k and 1k. This would divide the input by 2.5V instead of 10. With the gain stage, you would get a max of 2V out, just like the 10V input case with a 10:1 divider.

2. The REF3125's GND pin is connected to -2.5V, the REF3125's IN pin is connected to +2.5V. The output of the REF3125 (OUT) will be 2.5V above the GND pin so it will output a voltage near 0V. With the OPA359 as a buffer, the ADS1258 REFP will be 0V and the REFN will be -2.5V. Note that the REF3125 has quite a fair amount of 1/f noise, so they have used the 10k resistor and 100uF cap to create a low pass filter to remove low frequency noise. It will take a significant amount of time to start this reference up because of this long RC time constant.

If they don't have much experience with ADCs in general, I'd highly recommend that they get and ADS1258EVM-PDK to play with it. It has quite a bit on board that can show them a basic setup and how it works. At the very minimum, the EVM schematic is a good reference to draw from.


Joseph Wu

Ming,


I did get the latest schematic from James, and I'll post the responses here.

1. I mentioned this before, but there are multiple ground designators. I would make sure that this is cleaned up to make sure that the layout is done smoothly and there aren't any problems laying out the board.

2. There are 2 C32s and one is not connected. There are also multiples of many reference designators and there are others that are missing names or values.

3. There are two different 3.3V symbols on the schematic. 3V3 and 3.3V, make sure this is resolved before starting layout.

4. I've also mentioned this before, but C2 too large. I think that 10uF is way too large as a bypass capacitor and I'd use something like a 1uF or 0.1uF. Additionally, the supplies for any op-amp should have bypass capacitors as well.

5. OPA703 is connected to the noisy supplies, so generating the VREF- signal might have noise associated with the power supply. If the rejection is ok, this might work for the noise. However, you still have the offset and the drift associated with this op-amp as well. Since this is a reference input, you'd want to keep it low noise and accurate. That would include the two 10k resistors. If you want to have a precision inverter, then the resistors would need to be precision as well.

6. The ADS1258 is specified with a 4.096V reference. I mentioned before that I would have used the REF5040 and set VREF- to -2.5V and done a Kelvin connection to the negative reference input. If you want closer to the full range, you could use the REF5045 and count on the fact that the input range would be +/-4.5*1.06667. That gets you very close to the +/-5V range. Note that the REF3125 is probably not what you want to use. I don't remember the noise level, but it's probably a bit high and if you could use the REF3125, the ADS1158 may work in the application.

As a side note that you mentioned, you could use the schematic of Figure 67 but make a few changes. I would still use a REF5025 and change the load capacitance from 100uF to 10uF. You don't need the 10k resistor for the output so change that to 10Ohms (or replace it with a 0Ohm). What are you measuring? You might be able to use a 1k/1k voltage divider on the front end to get 2:1 division instead of 10:1 division, but I don't know what the output impedance of what it is that you are measuring. I would need to know more about your system.

7. OPA2376 may not be be the best part to use as the buffer. It's not really not a rail-to-rail input and CMRR is only specified to 1.3V from the upper supply. Inputs would only go from -2.5V to +1.2V. If you want the buffer, you would need a rail-to-rail input and a rail-to-rail output. Since a rail-to-rail input/output op-amp would probably be limited to close to the rails, the REF5045 might be ok.

8. Is ADC_START connected to anything else? Is GPIO9 used to keep START high?

9. Someone else should look over the power circuitry. I'm not too familiar with any of these devices and it might be important to keeping the supply noise down.

When this gets closer to finalizing the schematic, post back and I'll respond. If you need to, forward James the schematic again.


Joseph Wu

Ming,


It took me some time to go through the circuit. There have been a lot of questions today and I wanted to make sure that I understood what each part was doing.

First let's start with the reference. With the REF3125 the inputs are set so that VREFN=-2.5V and VREFP=0V. That means the reference voltage is 2.5V.

The ADC inputs (ADCINP and ADCINN) should be able to read +/-2.5V (it will be a little more since the full scale range is 1.0667*VREF and I'll address it more in the next paragraph).

With the instrumentation amplifier set with the two OPA365s, the gain is set to 2 V/V. I would note that with single ended inputs MUXOUTN is set to 0V. So the maximum signal that could be seen from MUXOUTP would be 1.25V. Since the ADS1258 range is 1.0667*VREF the real range would be about 1.3333V as a max (or really +/-1.3333V for the max).

With the inputs before the multiplexer, there are the 9.09k/1k resistor dividers that divide the input 10:1. So the input would really be about +/-13.3V.


Joseph Wu

Ming,


Assuming that you are running in some sort of single-ended mode with a AINCOM set to ground, you could just use a 2:1 divider of 5k/5k. However, I don't know what you are measuring. If the output impedance of what you are measuring is high, then you may not be able to do this.

I would replace the REF3125 with a REF5025 and you could leave off the 10k/100uF low pass filter. I think you'll need some capacitance, but you could just use 1 Ohm and 10uF so that it settles faster.

You'll also need to do some testing and analysis too. the 10k resistors in the gain stage need to be precision resistors or there will be a gain error relating to the accuracy of the resistors. I'd also experiment with removing the OPA350 to see if it's really necessary. You might be able to do without it, but since the input impedance is so low, you still probably need it anyway.

I highly recommend getting an ADS1258EVM-PDK to play with. You'll be able to put some of this to the test with the EVM and the software.


Joseph Wu

Ming,


With AVDD to +2.5V and AVSS to -2.5V and AINCOM to ground is all ok. There's nothing wrong with the setup. You'll be able to measure the voltages from and AINx to AINCOM.

I would still use a 2.5V reference. The input voltage from AINx to AINCOM won't be larger than +2.5V to -2.5V, so there's no need to set VREFP to +2.5V, and VREFN to -2.5V. This gives a precision reference without having to use a noisy supply as the reference.

Also, you could just attach MUXOUTP/N to ADCINP/N directly. You wouldn't be using the amplification stage shown in Figure 67.

If you are making measurements from any AINx to any other AINx, then you might have voltages where one input is +2.5V and the other input would be -2.5V. In that case you might need a larger reference. However, if that were the case, then I might still use a 2.5V reference and build in a gain stage of 0.5V/V.

Again, I would just get an evaluation module and test this setup.


Joseph Wu

Thanks again for the reply. I will limit the bi-polar input to be -2.4 to +2.4V. According to the data sheet (also spec'd based on 2.5v and -2.5v bi-polar operation), the Vref = VREFP - VREFN, so it'd be 2.5-(-2.5) = 5V. The typical is 4.096v and the max is AVDD-AVSS which is 5V.

Are there specific reasons that you'd prefer 2.5V reference ?

Ming,


The reason I mentioned the 2.5V reference is because you can easily generate the 2.5V reference using the +/-2.5V supplies. You can't generate a 5V reference from +/-2.5V. You would need a reliable difference of 5.2V to generate the 5V reference with a REF5050.


Joseph Wu

Ming,


First, I need to know exactly how you want to make measurements with this ADC. Will you be measuring inputs from AINCOM to any of the other AINx inputs? Or will you be measuring from any AINx input to any other AINx input? This makes a big difference.

Case 1 - With measurements from AINCOM to AINx, this means the ADC input will always be somewhere between +2.5V and -2.5V. If this is the case, you can use the 2.5V reference. If this is the case, this would be relatively simple, using a single 2.5V reference. I would use a low noise reference like the REF5025, with the reference input at 2.5V, and the reference ground at -2.5V, so that the reference output is near ground.

Case 2 - With measurements from any AINx to any other AINx, you have the possibility that one input would be at -2.5V and the other would be at -2.5. With that case, the min and max inputs would be -5V and 5V. In that case you would need to have a 5V reference. After considering it some more I would use the REF5050, with the reference input at +5V, the reference ground at -2.5V, so that the reference output is about +2.5V. I'd have concerns about how noisy the +5V supply is, and what kind of rejection the REF5050 would have but it could work. I would do this instead of using U4 and the resistors to giving you the reference doubling. It's going to be simpler than using the +/-5V supplies to run the U4.

So are you using case 1 or 2?

Another comment that I had is that the OPA2376 that you use to give you the buffer (between MUXOUTP/N and ADCINP/N) won't give you the full range. If you look at the OPA2376 datasheet, the input range for the CMRR only goes to 1.3V to the top supply. In this case, the buffer MUXOUTP/N wont go above 1.2V. I think I had mentioned this before. I think you are better off using the OPA2365, where the inputs would go to at least +/-2.4V where the output runs out of range. These are shown in some of the figures in the ADS1258 datasheet.

Other than these comments, I don't see too much else. I probably would have used the other ground symbol, since this one represents an Earth ground. I don't know if you've changed your power section, but I generally don't know as much about that as other precision circuits.


Joseph Wu

Ming,


VIN is the power supply. VOUT is the output of the reference. The VIN min shows how much the power supply needs to be above the output. It's similar to the dropout voltage specification for an LDO. The VIN min specification of 2.5V+0.2V is based from the reference ground.

With the reference input at 2.5V, the output near ground, and the reference ground at -2.5V, the VIN is well above the 0.2V above VOUT needed to operate the reference. It will operate this way similar to the way the REF3125 operates in Figure 67 in the ADS1258 datasheet.


Joseph Wu

Ming,


If you are using Case 1, then the input voltage that you measure is between +2.5V and -2.5V and you only need reference input of +2.5V. You don't need to put in a +2.5V to -2.5V reference. Again, I'd use the REF5025.


Joseph Wu

Ming,


Unlike many of the other devices in the product line, the ADS1258 does not have a calibration function or command. With 16 channels, calibration may be difficult because there may be different inputs and signal paths that may affect the calibration.

If I wanted to try a calibration, I'd try to input 0V for each input and record an averaged offset. Then, since the input range goes to 1.0667xVREF, you could try to measure VREF with the ADC and record the value and call that the relative full scale. In general, this may not be necessary for each of the input channels, but if there is some sort of gain stage on the front end if the multiplexer, that is different for each channel, it may make a difference.


Joseph Wu

Ming,


I don't believe that the LDO regulator accuracy will be a factor. However, when you do finish the schematic, send it back through James and I'll look at it.

I'm not sure what you are asking about for measuring the VREFP and VREFN for calibration. The ADC will be as accurate as the reference being using. If you want you can measure the reference voltage and roll that into the calculation of the measurement. As I mentioned in the previous post, you could measure the reference voltage with the ADC. However, this only compares the accuracy between the input sampling and reference sampling.


Joseph Wu

Ming,


I believe you can do this with the Gain Reading monitor selection in the device. It is described on page 24 of the ADS1158 datasheet.

For the gain monitor, the input is connected to the reference through the internal multiplexer and then measured against the reference. Remembering that the full scale range is about 1.07% of the reference, you should be able to get an exact value for the scale factor.

Also in that section, the device gain is given in equation 6.


Joseph Wu

Ming,


First, I thought I'd recorded a reply about this post before I left work today, but I must have forgotten to press the reply button. Sorry about that.

Also, I missed that you were talking about the ADS1258 instead of the ADS1158. However, my comments will apply to both devices. I misunderstood what you were trying to accomplish with the measurement. I'll explain which each configuration will be used for.

First, the Gain Reading will use the external reference to measure the reference. This is used to make sure the sampling capacitors for the input and the sampling capacitors for the reference match each other. Based on equation 8, this shows that the input to the reference has a ratio of 1.066667.

In comparison, the Reference Reading can be used to measure the external reference by comparing it to an internal reference. However, this internal reference isn't very accurate. I'd note that in the datasheet it's listed in the Electrical Characteristics (on page 3) under the System Parameters. This reference measurement is typical 1% and max of 3%.


Joseph Wu