ADS1248 GAIN PROBLEM

Shaii,


Also, I need to know more about your configuration. What gain are you in? What is your input voltage and what is your input common mode voltage?

One common gain problem is that the input common mode voltage has a specific range with the ADC. It's listed in the Electrical Characteristics table, but described more in the Low-Noise PGA section on page 26 of the datasheet.


Joseph Wu

Shai,


I've never seen anything quite like this where the problem is lot dependent, so I have a few more questions.

I was going to ask what the gain error is when you have a bad measurement, but it looks like the error is mostly a random output. The graph that you provided looks a little like a digital communication error. The reason I say this is that the resulting data is a structured. It's almost like trying to read a full scale data, but the input sometimes reads full-scale, sometimes reads half-scale, sometimes reads quarter-scale etc. This looks as if the data is bit-shifting through. With the inputs at GND (which is outside the common mode input range with a single supply), you should still read something near 0. Do you have the raw data for this plot? I'd really prefer to see the digital output data instead of the converted data.

If you do get devices that perform with real output data, what is the gain error? I'd like to know if this is really isolated to this lot and if there are ones that give you some real data.

Also, I'd like you to confirm some of the digital timing. First check that the SCLK idles low and that the data is read in on the falling edge. This can cause some data that looks like it's bitshifted. I'd also like you to check the SCLK speed and the timing with relation to /CS. I could see that a timing error might be lot dependent, so I want to check this as well.

The only other things that might be in error if there is some sort of violation of common-mode input range (even momentarily) that might be seen in the next reading. If the input is over-ranging the PGA, then the input might have a long recovery. If the inputs are changed and moving from channel to channel or a reference is turned on and off or switched, there might be some settling time involved. For all of the overloads, the recovery time might be lot dependent (I find this a bit unlikely, but I still want to check).

Have you ever changed the data rate or extended the length of timing for a data read to see if it helps?

How many units of the C series do you have?

Finally, may I contact you through the email address that you supplied through the E2E signup? I'd like to get some direct communication. However, I would like you to consider all of these questions that I've written into the post and get me some specific answers. I'll need more details to even guess what the problem is.


Joseph Wu

Shai,


I'll go ahead and contact you directly through email. I don't know what the problem is, and we've never seen anything like quite like this. I have seen delta-sigma modulators become unstable, but it's usually because of some large input signal of high frequency that is out of band. In those cases though the output would stay near 0 and then give bursts of noise. This still looks different to me.

Here's what I think we'll need. I like to see a complete schematic and a bit of an explanation of what you are measuring. It should include the full signal path and whatever you are measuring. I see that you cycle through 3 of the inputs, do you use the IEXC to source current through a RTD or some other bridge type measurement?

You've taken some oscilloscope photos of the data exchange, I'd like to see them. I'd definitely like to see all of the digital lines plotted. This would include DIN, DOUT, SCLK DRDY if you can only get 4 channels. Capture /CS on another plot with the others minus /DRDY.

I saw the graph of the data when it's bad. I'd also like the data when it is good. Also, I'd like an excel file of an example of good and bad data in the output code in hex, not the converted data in Volts.

Are you using the internal oscillator or an external clock? What is your SCLK speed?

Finally, I'd like the full marking on the device. This way, I can have the history of the lot tracked. In the meantime, I'll look at how we can get a few of these units to look at and how we initiate a customer return.

Again, I'll contact you directly and we'll take this offline.


Joseph Wu