Signal Conditioning for ADC

[newerbee]

Hi Everyone,

I am trying to design an input stage for ADC. my problem is the High input impedance sensors which can produce voltage levels from -40V to +40V. The ADC is 18bit and the required resolution is 1mV.

I am having issues when trying to "shrink" the signal to adc input levels (+-10.240V). I tried to use 20Mohms and 5Mohms voltage divider at the input of the ADC which resulted in measurements error. (without the resistor divider, the ADC can read exact voltage at its input) after I added the resistors  the inputs started to shift and ADC produced different values at each time (e.g:when the input is 1V the adc reads 1100mV or 900mV and it is not stable).

Do you have any advice?

Thanks.

[Jeroen3]

You need to know the input impedance of the ADC. You can't simply put resistors in front and hope for the best.
You need to:
- Attenuate to get a workable range.
- Filter the stuff you can't see due to Nyquist.
- Buffer to get a low impedance for the ADC.

Also, anything above 12 bits requires extra attentions since the noise is becoming a serious problem.
There is a reason not all micro controllers have >12 bit AD.

Can you tell us what the source of the signal is, and what ADC you're using?

[newerbee]

My ADC is LTC2358-18 (18 bits, not 24 sorry for the mistake) buffered adc,  datasheet says it has >1000Gohm resistance per channel. The Measured voltage can also have AC component (such as 3VDC and 10VAC).  (The AC component comes from the power grid interference, and needs to be measured as well). Further more the measurement should be taken differentially (floating ground) as the sensor and power ground connection is not permitted due to introduced errors.

I am trying to find a way to attenuate the original signal, my solution somehow didnt work

[f5r5e5d]

info about the sensor's Z and valid signal range, frequency, noise could make for better responses

[newerbee]

The source impedence is not know however it can be a few hundred K. The standards recommend a meter with at least 20Mohm input impedance to read the sensor voltages.

The AC frequency is 50Hz (line frequency interference).

The signal range can be even higher than +-40V (peak to peak) but this is the maximum range I should be able to read correctly.

There is no known significant noice source tha will create problems. without the resistor divider, the adc can read smal signal responses very accurately when they are between +-10.240Vpp range.

[Kleinstein]

There are very few, rather special ways to measure a voltage without a common ground or a some way defined relation. So even if the input is differential, one would need a defined DC level as there will be a limited common mode range. Without any ground connection it would be more like battery powered amplifier and ADC and sending the data via fiber.

For the input there are two choices:
1) A high impedance divider at the signal followed be a kind of buffer. Here input impedance is kind of a compromise between loading the source and noise. So the divider might be at 10 M, 20 M or so. For a differential input there would be two dividers, limiting the common mode suppression.

2) One or two (for differential) special high voltage buffers for the input signal(s). A divider would than be after the input buffers and can be lower impedance, though there are still limits due to self heating and thus nonlinearity. So the divider impedance might still be in the MOhms range. In case of a relatively large common mode voltage and small differential signal one might prefer a special high voltage instrumentation amplifier (e.g. build from discrete transistors / FETs) instead of the buffers.

Chances are that you don't need such a good ADC after the amplifier.

[newerbee]

... For a differential input there would be two dividers, limiting the common mode suppression.

I didn't understand the two divider aproach, is there a simple schematic for this?  Thanks...