Problem with a/d converter inputs drifting up

Hi Ron,

I think part of what you are seeing is the low impedance of the ADC drawing current in parallel with the external resistance.  This is particularly true for the ADS1278 and ADS1243 if the buffer is off (default).  The bias current will place a parallel current path on the resistor string. 

In general this is not a very good way to evaluate the devices, because of the impedance but also the placement and layout of the devices will affect performance. It may give you a general idea as to operation but not performance.  In this type of situation the ADS1278 requires an input buffer (op amp or instrumentation amp) and the ADS1243 should have the buffer enabled (ACR register bit 4).  When using a buffered input stage you will not be able to make single ended measurements as the common mode input range will not go all the way to ground.

Best regards,

Bob B

We are not trying to evaluate conversion performance, just high-temp operation.  But a somewhat stable input is still required.  The ADS1243 datasheet says that the input impedance is 5 M-ohms/PGA, and we are using PGA=1.  I don't see how 5 Mega-ohms could pull down the input.  I can enable the buffer to see if there is any change.  [By the way, the datasheet says "The buffer can be enabled using the BUFEN pin or the BUFEN bit in the ACR register."  There is no BUFEN pin.]  

For the ADS1278, from p. 36, effective input impedance = 14k * 6.75 MHz / fmod, where fmod = modulator sampling frequency = fclk / 8 [we are using low speed, CLKDIV=0], where fclk  = 625 kHz.  So Zeff = 14000 * 6750000 / 78125 = 1209600 ohms.  Again, how could this pull down the inputs, when we are supplying the voltage with an output impedance on the order of 1k?

Ron

Hi Ron,

Let's back up a bit.  There are only a few things that can actually affect the measurement.  One is a varying input and another is the reference.  The are others of course, but those are the two main considerations. I apologize that I read your post so quickly.  You said you held the MCU in reset.  How did you determine the output?  Or, are you externally measuring across the resistors?

As I don't have a schematic, I'm only guessing at how you have things connected and at what points you are measuring on your resistor string.  You originally stated that the board warms up.  Is the board in a room temperature environment or is board in a temperature chamber?

I apologize for the datasheet.  This was undoubtedly a cut and paste error when creating the datasheet for the ADS1242/43.  The ADS1240/41 has the buffer enable pin.

Now back to my original comments.  You asked if there were in any way the devices could cause an issue.  Remember that the values you quote are impedances and not resistances. The inputs to the devices have switched capacitor inputs that switch in  and out of circuit at the modulator rate. There will be some movement of charge. When the charge is available relative to the desired input voltage will determine what is actually measured.  Depending where the input voltage is relative to the reference, there actually can be charge that kicks back from capacitor as a discharge as well as charge.  You are most likely correct that the severity of the problem is not the impedance.

With the ADS1243 there is one other possibilty for an error and that is if the burn out current sources are accidently turned on.  There are most likely some other potential issues as well, but given the limited information it is difficult to tell. If it is a single issue or combinational.

Best regards,

Bob B

I am thinking that my problem may be that the ADS1243 is getting into a state where it is driving the analog inputs, because they can also be digital I/O (why someone would want this, I don't know).  Since I am not using chip select (not enough pins for everything), I suspect communication was getting out of sync and I was setting the registers to change the analog inputs to digital outputs.  It seems that after being in this state for a long time the chip may be damaged and driving these lines all the time.  This is really all speculation at this point, but seems the most likely scenario, since none of the other 3 ADC's can drive the analog inputs.

Ron

Hi Ron,

Yes, you make a good point and it is something I missed.  However, you would actually have to write to two and possibly three registers to see this happen.  This is because IOCON register defaults to analog inputs, and the DIR for the digital is set to inputs.  In your case the connection is not a direct short to the supply, so I would think it would take a very long period of time for the driver stage to be damaged. If you have connected 4 devices together and you are not using CS, how are you getting this to work without driver contention on the SPI MISO line?  How do you keep your communication in SYNC (from your previous post description)?

There are number of possible uses for digital I/Os from the ADC.  One is the case where the micro has limited GPIOs available, and another is in an isolated system where digital I/Os are limited across the isolation barrier to just communication.

Best regards,

Bob B

Those are some good points there.  To answer your questions, the only lines connected to all 4 adc devices are power & the analog inputs.  All SPI control lines are independent IOs.  We made two different boards:  1) ADS1243, ADS1243-HT, AD7714, LTC2498;  2) ADS1243-HT, ADS1278, AD7714, LTC2498.  Both types are having the problem of drifting analog inputs.  On the devices where I don't use the chip select, perhaps I was not keeping them in sync.  Ideally they would stay in sync, barring any spurious pulses on SCLK.  Now I am sending a reset (using the SCLK special sequence) to the ADS1243 and not using the other ADC's (not sending them SCLK).  The reason I think something is damaged--I have written firmware that provides no serial clocks to the ADC's, so 3 are not running (LTC2498 still runs on internal osc); and does not exercise the SPI lines to any of them, and I still get the voltage drift on one of the boards.  I am continuing to experiment to rule in/out the ADS1243, but that is where I'm leaning.

Ron

On our board that has ADS1243, ADS1243-HT, AD7714, & LTC2498--We removed both 1243's and cut the power trace to the LTC2498.  The problem remains--the analog inputs are being driven to much higher levels than they should be.  For example, what should be 1.2V based on the resistor ladder measures 2.3V.  I have seen digital switching on the lines.  The only active devices that these lines are connected to are AD7714 and LTC2498.  So TI is exonerated, but if you have any ideas, I'm all ears.  Our next step is to remove these two ICs one at a time and see what happens.  Bob, I can send you my schematic off this forum.

Ron