Post Go back to editing

LTC2400, DC573A, DC590B SPI noise issue?

I'm designing/building a small quantity of voltmeters for a customer based on the LTC2400. As a base for comparison tests I'm using your Eval boards (DC573A and DC590B) and software. If I take Vref (5V) from the DC573A and feed it back into Vin and to my board, your software and mine both say 5V. Mine has a stability figure of around 80uV. If I then put a resistor divider on Vref (900K/100K) so that Vin to both boards is 0.5V. My board is showing noise in the order of 600uV or more. As soon as I pause your software (i.e. stop the SPI bus), my noise level drops to around 30uV. Is there a known issue with the 573/590 combo that can cause SPI generated noise?

Note that I have modded the 573 as per the datasheet for lower noise operation (cut p2 track and power LTC2400 from the 5V reference ic) but I was getting similar noise anomalies before.

Thanks Ian

Parents
  • Hi Ian,

    The large source resistance from your voltage divider is a possible cause of the noise you are seeing. There is a glitch during each sample period that must settle before the start of conversion. With a large source impedance this glitch will be larger. Try reducing the source impedance and see if the noise goes down.

  • Thanks for the suggestion. I've run some tests now.

    Feeding the LT1236 5V reference directly into my circuit gives around 30uV of instability, with or without the SPI bus running. If I add my 1Gohm 10:1 divider, with the SPI bus running, the instability jumps to a massive 70mV! If I scale my divider by 10 to around 100K the noise drops to around 7mV. If I scale again to 10K it drops to 850uV. At 1K it's around 100uV and doesn't drop much below that if I decrease the resistance further, approaching the 10mA limit of the LT1236.

    The DC573A is only, as it says, an evaluation board and not a precision piece of hardware. Obviously, SPI data noise is being coupled onto Vref within it. I appreciate that what I'm trying to do and measure is beyond the bounds of this. I'll just have to make do for now.

    Thanks again. Ian

Reply
  • Thanks for the suggestion. I've run some tests now.

    Feeding the LT1236 5V reference directly into my circuit gives around 30uV of instability, with or without the SPI bus running. If I add my 1Gohm 10:1 divider, with the SPI bus running, the instability jumps to a massive 70mV! If I scale my divider by 10 to around 100K the noise drops to around 7mV. If I scale again to 10K it drops to 850uV. At 1K it's around 100uV and doesn't drop much below that if I decrease the resistance further, approaching the 10mA limit of the LT1236.

    The DC573A is only, as it says, an evaluation board and not a precision piece of hardware. Obviously, SPI data noise is being coupled onto Vref within it. I appreciate that what I'm trying to do and measure is beyond the bounds of this. I'll just have to make do for now.

    Thanks again. Ian

Children
  • Hi Ian,

    As I said before, this is not SPI noise. This is a glitch on the analog input from the input sampling switch closing. To get the lowest noise performance, buffer the resistor divider with a low noise op amp. Look at Figure 9 on page 17 of the LTC2508-32 data sheet as a possibility.

  • Hello ghoover,

    Sorry that I misunderstood you. I erroneously assumed that since I only saw excessive noise when the SPI bus was running, that was where the noise was coming from. But, of course, when the SPI bus is running, the ADC is converting. As you rightly say, that's the noise source.

    Now I know the problem I can get round it to get my testing done but will add a buffer for the next time.

    Thanks. Ian