Analog front end for ADS1278

Kyle,

Here's a screenshot from the ADS1278EVM-PDK plug-in with ADCPro, while I'm applying a 5V peak to peak sinusoid with no offset on channel 1:

Just looking at the EVM, I can see why you may be confused here. Without looking at the schematic the EVM certainly makes it appear as though the ADS1278 is taking a -2.5 V input. What's actually happening though is the buffering op amps on the board add a 2.5V offset to the signal (the common mode voltage from the ADS1278) and the ADS1278 resolves the voltage with respect to the common mode.

So when Tony says the input range is only within 0.1V of the supply voltages (-100mV to 5.1V), he is still correct even though the output codes you get from the ADS1278 will make a 0 V input (measured directly at the input pins) look like -2.5 V because 0V is -2.5 V from the common mode.

We see the signal clipped, as the datasheet describes...just a little bit lower voltage than what the datasheet says

Kyle,

Finding an op amp that prescribes to all of your requirements is going to be difficult. A low power, wide supply, small package, differential op amp would kind of be a "magic bullet" op amp if you catch my meaning. The OPA1632 does look very nasty on our board, but I should also mention to you that much of the problems that those amplifiers create on our board are directly related to the supply rails and power consumption that we are used to facing on the MMB0 platform.

As you may have noticed from looking at the schematic the supply rails for the OPA1632s are boosted up from 5V from the MMB0. When you boost you gain voltage, but lose current. The op amps, however, still need to draw current to drive the ADS1278 so the board gets toasty. You probably would not have as much heat present in your system, assuming you were able to provide sufficient supply rails without boosting.

We are looking into replacing the OPA1632 on our EVM with the THS4521 on single +5V supplies. That will eliminate the need for boost and 8 of the THS4521s will still be lower power than a single OPA1632. The sacrifice, however, is that we cannot apply a bipolar signal directly to the inputs of these amps like we can with the 1632. I bring all of this up because these are the same design decisions you're going to have to make with regard to your system. If you need to support a wide range of unknown sensor outputs, then you're going to have to sacrifice some power for your analog front end.

THS414X amplifiers may work well for you, they have tremendously wide supplies to support as many of your "mystery sensors" you may encounter and the current draw isn't terrible given the super wide range. THS4140, in particular, has a simple power down pin to drop your current consumption down to 1.2 mA if you're not continuously collecting data.

If you can bear the current consumption then the OPA1632 may work fine for you. If you can get some more information on the output you may see from sensors you may be able to work with the THS4521.

Have you considered these? If so, what did you not like about them. I can try to get in contact with some of our stronger op amp experts and see if they can help us find a good solution for you.

Okay, so you've got a single-ended signal coming out of your function generator which has two terminals; positive (signal) terminal and negative (ground) terminal. Precisely how do you have those connected to the EVM?

My worry is that you have your positive terminal connected to J9.2 and your negative terminal connected to J9.1 (or the reverse) as shown below:

Yes, this silkscreen is for the ADS1178. It's a quick and handy image from the EVM User's Guide, for this purpose they're the same.

If your set-up is like the picture above, then you need to make sure that your negative terminal at J9.1 is also connected to ground from the board. Alternatively you could connect J9.1 directly to ground on the board or from your power supplies and connect the ground from your function generator to your power supply grounds. If you don't the negative terminal of your function generator will be floating with respect to the ground of your power supplies and the EVM and some strange things like what you've described can happen.

Kyle,

Let me try to go item by item and address your comments in this post.

"It is strange that a floating negative terminal could cause an unipolar ADC to become bipolar. Seems like at best it would alter the voltage read a little but hard to believe that the sine wave would still come out with no distortion and the only "negative" outcome be that the ADC start reading voltages that are out of it's normal range."

The ADS1278 is not transforming from a unipolar converter to a bipolar converter, it's just the illusion of such a thing happening. The negative terminal of the device is not truly just floating. Sure, when you've applied your negative generator output to the negative terminal it is floating with respect to ground but the outcome can make something like this appear truly mysterious. Lets try to take a look through this stuff under some different circumstances.

Let me preface all of the figures to follow with the state of my set-up. My input signal is a +/-2.5V Sinusoid at 1KHz. I am using the EVM-PDK with the OPA1632 not powered, and switched out of the system. Additionally, I have removed R17 and R18 so the buffering amplifiers are completely removed from the system. All of my inputs will be applied to channel AIN1.

The figure below is with my positive generator output connected to AIN1+, AIN1- is floating, and the negative output of my generator is connected to EVM GND.

As another example case, here is the same test but without even sharing ground between the EVM and my generator:

Perhaps these results fall more aligned with what you were expecting to see from a true floating terminal. Now, lets revert to the set-up I showed in the silkscreen image from my previous post. I'll probe the input signals with respect to EVM GND so we know exactly what the ADS1278 is actually seeing.

Purple is the signal at AIN1+ (Positive Generator Output)
Orange is the signal at AIN1- (Negative Generator Output)
Red is the difference of these two signals.

So, what we see in this figure is the signal the converter is actually seeing. As you can tell, this signal is not truly bi-polar. There's a few things going on here. First of all, you don't have any sort of common ground between your signal source and your EVM. Because of that, the negative output of your generator is actually around 2.5V because there is no common reference voltages between these two things. Second, the ADS1278's internal behavior may be cause a little pull up on the signal applied to your negative terminal. It's hard to say since we don't have a common ground reference between these signals to begin with.

For the rest of your post...

"Have any of your colleagues seen this type of behavior before?"

Short answer, yes.

"What happens if you drive your EVM single ended in the same matter? I have 2 that perform in the exact same way using two different function generators."

The same thing happens, see the above explanations and images.

Hi Kyle,

There are four pictures here that I will comment on one by one.

- First picture, using the function generator single ended - Since you connected the ground of the function generator to the ADC, the behavior that you see is exactly what I would expect. Once your input exceed about -0.6V below ground, the ESD diode kicks in and clamps the input voltage.

- Second picture, Function generator - From my understanding, you are inputting a 2V differential signal. Each input has its own 1V amplitude 180 degrees out of phase of one another to give you your signal.

- Third picture, ADC converted result without sharing the grounds - Think of the ground not necessarily as a common ground but as a reference point for all your voltages on your board. As you have it here, the reference point (ground) for your EVM board is different than your reference point (ground) of your generator. These two grounds are essentially 'floating' from one another. Therefore, the common mode point of your input signal is floating relative to your EVM ground reference point. In testing we have seen a ground from one device 'float' as much as 30V from a ground of a second device. As Kevin showed in the previous post, the two differential input signals relative to one another will give you the sine wave you expect, but if you look at each input relative to the EVM ground, they are both going to appear a little different.

- Fourth picture - Connecting the source ground to the EVM ground - As I can see it here, the amplitude of your converted result is attenuated to about +/-1.6V. This is because when you connect the two grounds together, you are fixing the common mode input point to 0V. As you do this, either your positive or negative input signal will be exceeding the absolute input range spec (AGND-0.1V) causing the ESD diodes to kick in and clip one of the two signals. The ESD diodes kick in around -0.6V and will clip the signal. That is why your amplitude of your signal looks like +/-1.6V (1V+0.6V) as opposed to +/-2V.

I read that your concern is when you have a customer attach a sensor to the input. When they do so, that sensor will be sharing the same ground as the ADC so the fourth picture is the most accurate representation of what they can expect. You need to select a differential driving amplifier for the ADS1278, so just make sure that it has a pin used to set the common mode of the output (OPA1632 on the EVM shows a working example). The ADS1278 has a common mode out which is the midpoint of the two supplies. This will make sure the common mode of the differential input signal to the ADC is set to 2.5V. Therefore, you will not have problems with a +/-2V input differential signal which may come from your sensor.

Regards,

Tony Calabria

To continue trying to solidify this, the reason you're not seeing distortion on any of the signals you've shown here like the pictures I've shown you is because you're not doing the same thing that I described in those tests, or at least you've not described it as such in any of your text.

The distorted signals were with AIN1- actually floating. There was nothing connected to that pin at all. Not even the negative output from the generator. The only thing that was connected was the positive output from my generator to AIN1+. I did this all to try to help clarify what is happening in the case that you apply the negative terminal of a single ended function generator to AIN- and the positive terminal to AIN+. Purely for illustration purposes, it is in no way something correct to do.