• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 46 subscribers
  • Views 458 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

ADC124S101 results problem

Andrei Kondrashev
Prodigy 20 points
Other Parts Discussed in Thread: ADC124S101

Hello everybody,

I have a strange problem reading results from ADC124S101.  This is a 4-channel ADC and I have 4 different sensors connected to it.  When reading ANY single channel, the results completely correspond to the voltage seen on the corresponding input.  However, when I read more than one channel sequentially, the results are different (sometimes up to 50%).

I tried to read every channel 4 times and use the last reading.  This improves results, but they still differ from a single channel readings.

I am using TI F28036 DSP, SPI rate is ~4.2 MHz.

Electronics engineer says that he doesn't see any electrical issues.  Programming logic is correct and verified.

Any ideas what it could be?

 

  • 0
    TI__Intellectual 2775 points

    Hi Andrei,

    The issue here could be settling time of the internal storage cap.

    During the track time (the first 3 cycles of the SPI frame) the charge is acquired onto the internal sampling cap. The impedance of the source applied to the ADC input must be low enough to allow for complete settling of the cap voltage. The setting time is determined by the RC response, where C is the internal cap (about 33pF) and R is the sum of internal resistance in series with C (about 400-600 Ohm) and source impedance (your circuit).

    Typically we advise use of buffers between the sensors and the ADC input.

    One way to test the "settling time" theory is to extend the track time, and see whether the accuracy improves. There are a couple of ways to do that:

    1. slow down the SPI clock for the whole frame (the low limit here is the cap leakage if the clock is too slow during the HOLD/conversion period) 
    2. or slow down just the first 3 cycles of the SPI clock (extend the TRACK arbitrarily long), and then speed the clock up during the HOLD/conversion time to avoid cap leakage issues.

     

    Hope this helps.

    Sincerely,

     

    tom

  • 0 in reply to Tom Domanski24
    Prodigy 20 points

    Hi Tom,

    Thank you for the reply.

    I cannot slow down SPI clock non-uniformly, as I am using the standard SPI controller on F28035.  I tried to lower SPI rate to ~2 MHz.  I shoot four frames using SPI FIFO and use the last reading from the group of four.  The sampling rate is 20 Hz (5 Hz per channel).

    Yes, I can see the improvement.  I will ask the hardware guy to look at your other suggestions.

    As it looks to me, the problem caused by a temperature sensor (TMP36GRTZ), which is connected directly to ADC channel.  Others have amplifires.

    Again, thank you for your help.

     

  • 0 in reply to Andrei Kondrashev
    TI__Intellectual 2775 points

    Hi Andrei,

     

    Another quick and dirty thing to try is to hang a "large" cap at the output of the TMP36 - say 1nF to GND. This cap may be enough to "absorb" the ADC cap charging transients and improve your conversion accuracy.

    In the end I agree with you, TMP36 is probably not capable of driving the unbuffered ADC input by itself.

    Cheers!

     

    tom