16 bit ADC

[bnfdvn]

I have two signals (1-5 VDC) that I need to connect to a 16-bit analog to digital converter. The signals are fairly stable (temperature and pressure) so frequency response is not an issue. The problem that I am having is that the devices that I am finding for a 16-bit ADC have differential inputs (+/- 2.048V) and I have a single ended signal. How do I convert the single-ended signal to a differential. I was thinking of running the signal through a voltage divider and then taking the reduced signal to 1) an inverting OPA and 2) a non-inverting OPA to get the differential signal. Will this work?? Anything else I need to watch out for??

Thanks!

 

[MikeMl]

Who says you must convert your sensor signals to differential?

Post the data sheet for the ADC.

Post the circuit for the signal conditioning you have done so far.

 

[bnfdvn]

This is one link to a Micrichip Technologies ADC with I2C. Maybe I am reading it wrong but I don't see where the chip can be operated with a single ended input.

https://www.electro-tech-online.com/custompdfs/2012/09/22072b.pdf

 

[Ian Rogers]

In the datasheet you posted

6.1.4 DIFFERENTIAL AND SINGLE-ENDED CONFIGURATION
Figure 6-4 shows typical connection examples for
differential and single-ended inputs. Differential input
signals are connected to the V
IN
+ and V
IN
- input pins.
For the single-ended input, the input signal is applied to
one of the input pins (typ ically connected to the
V
IN
+ pin) while the other input pin (typically V
IN
- pin)
is grounded. All device characteristics hold for the
single-ended configuration, but this configuration loses
one bit resolution because the input can only stand in
positive half scale. Refer to Section 1.0 “Electrical
Characteristics”.

 

[MikeMl]

You can tie one of the diff inputs to a fixed reference; then the voltage at the other input is measured with respect to a fixed reference.

Any chance you can find an ADC that brings the internal 2.048V reference out on a pin so that it can be used to define the reference voltage for your sensors?

 

[bnfdvn]

Sorry - I guess that I didn't read far enough. Thanks for looking all the way to the end!

That makes sense since only half of the available input signal is used. Seems like I would lose the MSB on the ADC Conversion. I would still have to scale the 1-5 VDC signal that I have to a X-2.048 Volt signal, correct??

 

[MikeMl]

Here is the real problem: The ADC reading is proportional to its own internal reference (2.048V band-gap). What is the 1-5V sensor output referenced to (derived from)?
A 5V power supply that drifts with respect to the ADC's reference.?
If so, you will be lucky to get 10bit accuracy.

The only way that you can approach 16bit accuracy is if the sensors themselves are driven off the same voltage reference (or an equally precise one) as is inside the ADC. The simple fix is get an ADC where the reference voltage is external, or brought out so it can be used by the signal conditioning circuits upstream of the ADC.

Getting 16bits accurately is an order of magnitude more difficult and complex than just hooking a sensor to a 10bit PIC.

 

[bnfdvn]

I should say that I am not limited to the Microchip Technology ADCs. I looked at TI chips and they had the same limitation of 2.048 V. I did find the MAX 195 chip which does just what I want but it is $80. That seems a little silly. It doesn't have to be I2C either. I couldn't find any PICs with 16 bit ADCs so I believe I have to do this outside the PIC. I2C seemed like an easy interface to use.

 

[ericgibbs]

hi,
Is there any special reason for requiring a 15Bit resolution [single ended] for temperature and pressure measurement.??

 

[bnfdvn]

Both signals are referenced to 0 VDC (DC Common). Both signals are generated off board.

 

[bnfdvn]

The resolution is a design specification. Realistic or not, it is still the design spec.

 

[MikeMl]

The resolution is a design specification. Realistic or not, it is still the design spec.

There are two things; resolution and accuracy. 16 bit resolution is not very useful if the accuracy is about 10bits, ie 1/1024.

Even if the sensors are conditioned externally, you still need a local reference to offset the 1-5V range so it fits within the allowed input range of the ADC. If you use the 5V supply for that, you will introduce a third variable into the drift problem. You need an ADC which outputs its reference voltage, or one that uses an external reference so you can use that reference both to offset the inputs as well as for the ADC to convert against.

 

[Mr RB]

As they are slow changing signals, sample at 10bits or 12bits using any normal hardware then average many readings and output the data as 16bit.

 

[ericgibbs]

The resolution is a design specification. Realistic or not, it is still the design spec.

hi,
As Mike says you could have a problem in using a 'fixed' internal Vref for the A2D if the other voltage supplies drift.

I have used the AD7705 dual 16bit A2D with an external Vref input thats 'derived' from the sensor supplies, so the circuit behaves in a radiometric way

I connect to a PIC using SPI.

On other projects I use a dual 12bit A2D MCP3202 and over sample in order to increase the 'apparent' resolution.

Also have used the ADS1100 16 bit A2D.

 

[bnfdvn]

I am a little confused about the mentions of power supplies drifting, etc. For the system where my board will be installed, everything will be powered from the same 12 or 24 VDC supply. For the 1-5 VDC signals, two wires will be brought in (Signal and ground). So the grounds will be tied together. Thiat being said, is there still a concern about power supplies drifting??

 

[ericgibbs]

I am a little confused about the mentions of power supplies drifting, etc. For the system where my board will be installed, everything will be powered from the same 12 or 24 VDC supply. For the 1-5 VDC signals, two wires will be brought in (Signal and ground). So the grounds will be tied together. Thiat being said, is there still a concern about power supplies drifting??

Hi,
The Input to the A2D can range from +/-2.048V which will give a 16 bit value output of +/-32768.

As your system is +1V thru +5V , this means only that the positive 15bits apply, ie 32768decimal.

You will need to reduce the +5Vin to +2.048v by using a resistive potential divider, 5/2.048= 2.441 ratio, this means the +1V lower input limit will be 0.4096V, 13,421decimal.

This is a range of 19347decimal for a 4V input change.

Any changes in the divider due to the environment will produce errors in the A2D output.

Its possible to use the full 16 bit range by using a OPA to reduce the input the input voltage and the offset and make it span -2.4 thru +2.4v, a dual supply would be required for the OPA.
E.

 

[MikeMl]

...
Its possible to use the full 16 bit range by using a OPA to reduce the input the input voltage and the offset and make it span -2.4 thru +2.4v, a dual supply would be required for the OPA.
E.

And you will have to have an external Vref which tracks the one inside the ADC!!!