ADC interface

[abq_guy]

I am using a BIpolar ADC from Linear technology (LTC1859) to detect +/-10V. The input of the ADC is connected to LT1885 which has been configured as a sallen key filter with 1kHz cut off frequency.

The ADC is experiencing some problems in detecting 0V. Around Zero volts, the ADC starts spitting out random values to the microcontroller. This happens for +/-30mV around zero. For all other voltages, the ADC works fine. What could be causing this weirdness in the ADC. All answers are appreciated. Also around +10V and -10V, the ADC saturates at about 9.998V.

Thanks,

abqguy

 

[dknguyen]

How random are the random values? It could be some sort of crossover occuring inside the op-amp.
Crossover distortion - Wikipedia, the free encyclopedia

 

[crutschow]

Yes, you need to quantify the nature of the random values so we can analyze the problem. The term "random values" is much too general to tell what's happening.

 

[abq_guy]

Hello Dknguyen and Crutschow,

Thanks for replying. The values are all over the place, like -9, -7. -4, 3, 5, 8, 6, 8, and then as soon as i put 35mV, the ADC starts giving like 45-50mV, so it stabilises.

I also tried replacing the 1000pF capacitor with a 500 Ohm resistor then I got better results, in the sense that the randomness disappeared around zero, but still exists for zero. Also I witnessed that for all voltage values, there is a 20mV offset in the readings. The ADC also saturates at 9.9998V for 10V input and -9.998V for -10V.

What am I doing worng? The ADC impedance in bipolar mode is 31k and 5pF.

thanks,

abqguy

 

[Mikebits]

The data sheet suggest a 3000pf cap right at the input pin.

 

[abq_guy]

The data sheet suggest a 3000pf cap right at the input pin.

Hello Mike,

Yes it does, and I tried that too. It still gives me the same problem. Also in the eval circuit board, they are using a 1000pF capacitor with no problems, so I just used that.

Thanks,

abqguy

 

[crutschow]

9 counts is only 2.7mV for a 16-bit converter with a ±10V input range. You may just be seeing noise in your system.

How many counts jitter between readings do you see with a 1V input for example?

 

[abq_guy]

9 counts is only 2.7mV for a 16-bit converter with a ±10V input range. You may just be seeing noise in your system.

How many counts jitter between readings do you see with a 1V input for example?

Hi Crutschow,

Did not understand the question? Which 9 counts are you talking about ?

I am attaching some more waveforms of the ADC output.

The first three plots were obtained when 12mV was applied to input of the ADC. The readings on the computer were totally out of whack. The Last two plots were obtained with 18mV applied to the same ADC, same inputs and this time the readings are better.

All these waveforms were obtained with 1000pF capacitor on the input pins of the ADC.

Thanks

 

[dknguyen]

WHat he is asking is, if you input 1V, how much do the readings jitter around (go above and below) 1000mV? It may be doing the exact same thing at higher voltages, except they are much larger than the jitter so you do not notice it. But at 0mV, an error/jitter of +/- 9mV is easily visible.

 

[abq_guy]

9 counts is only 2.7mV for a 16-bit converter with a ±10V input range. You may just be seeing noise in your system.

How many counts jitter between readings do you see with a 1V input for example?

Ohh I understand yur question now.. Its not counts, those are Voltage values, so lets say when I apply 12mV, I get values like 6V, 5V etc. I have uploaded some plots for reference.

Thanks,

Abqguy

 

[crutschow]

I'm talking about the maximum -9 count deviation you stated was occuring with 0V input.

Your plots seem to show about the same amount of deviation at 12mV as with 18mV. Why do you say the 18mV plot is better?

Certainly a given noise will be a smaller percentage of the total with a higher input, but that doesn't change the absolute noise level deviation.

You do understand the nature of random noise do you not?

 

[abq_guy]

WHat he is asking is, if you input 1V, how much do the readings jitter around (go above and below) 1000mV? It may be doing the exact same thing at higher voltages, except they are much larger than the jitter so you do not notice it. But at 0mV, an error/jitter of +/- 9mV is easily visible.

Hello Dknguyen,

I am uploading some more plots. These are for various values that I input to the ADC, like +10, -10, +5, -5 etc.

 

[dknguyen]

Hello Dknguyen and Crutschow,

Thanks for replying. The values are all over the place, like -9, -7. -4, 3, 5, 8, 6, 8, and then as soon as i put 35mV, the ADC starts giving like 45-50mV, so it stabilises.

Wait...when you apply 0V are you getting voltages ranging from -9 to 8mV or V? Because mV might be noise like crustchow suggested. But...Volts?! I first read it as mV because right after listing them you said 35mV, but reading it more carefully you seem to be saying volts.

 

[crutschow]

If those indeed are voltages and not counts, then you may have something oscillating near zero volts. Have you looked at the input with an oscilloscope when this is hapening?

 

[abq_guy]

I'm talking about the maximum -9 count deviation you stated was occuring with 0V input.

Your plots seem to show about the same amount of deviation at 12mV as with 18mV. Why do you say the 18mV plot is better?

Certainly a given noise will be a smaller percentage of the total with a higher input, but that doesn't change the absolute noise level deviation.

You do understand the nature of random noise do you not?

Hello Carl,

On the 12mV plots, the amplitude I get is between 0.5V to 6.5V which is no where near the 12mV i applied to the input.

On the 18mV input, I am getting the result between 0.01 and 0.0225 (10mV to 22mV range), hence I stated that 18mV plots are better.

Thanks,

Abq_guy

 

[abq_guy]

Wait...when you apply 0V are you getting voltages ranging from -9 to 8mV or V? Because mV might be noise like crustchow suggested. But...Volts?! I first read it as mV because right after listing them you said 35mV, but reading it more carefully you seem to be saying volts.

Hi Dknguyen,

Those are volts. Please refer to my attachment (plots.pdf). When I apply 12mV, I get between 0.5V to 6.5V

When I apply 18mV to the same ADC, same input, I get between 10mV to 22 mV.

Around zero, the ADC just goes out of whack.

Thanks,

abq_guy

 

[crutschow]

Another consideration is how you handle the sign bit in the computer. Near zero the sign bit will be rapidly alternating which may give a false appearance of large voltage changes if you aren't handling the sign bit properly.

EDIT: According to the data sheet, the bipolar output code is two's complement format and the counts alternate between all ones and all zeros from noise with a zero input. Does the computer interpret that correctly?

 

[abq_guy]

Hello Carl,

the problem is that 12mV doesnt work and 18mV works fine. If it would have been the sign, then I guess both wouldn't have worked. Does my ADC interface circuit look fine? Should I remove the 1000 pf capacitor on the input of the ADC and replace it with some resistor ?

Thanks,

abqguy

 

[abq_guy]

Anyone has any ideas on why this is happening ?

thanks

abqguy

 

[Mikebits]

As a test can you disconnect the op amp output and put a 50 ohms resistor right at the ADC input. This way you can determine if your ADC is the problem or the op amp.