# Can a differential ADC replace an instrumentation amplifier?

Greeting Everyone ,

I am doing a project of data acquisition system Let me explain project in simple terms Step 1) Sensor : A strain gauge type PRESSURE sensor (Wheatstone bridge)Specs : 2mV/V output I am applying +- 5Volts (10 Volts) The output swings from 0mV to 20mV(full pressure)

Step 2) Instrumentation amplifier : Output of the sensor is given to an instrumentation amplifier which i made using 3 OP07 opamps i have also added a filter at the end of amplifier to reduce noise.Specs : the circuit has two ZERO pots (to adjust Zero output)(Connected at zero net ) and an Gain adjustment pot .Gain adjustment pot is connected at R5

Step 3) ADC : The output from this circuit is gain adjusted (typically gain =100) and given to ADC ADS1115. ADS1115 has capability to take the differential input directly and also has provision for programmable gain Upto 128 Which looks something like this :

so i just interface my controller device with the above setup to get data (Data Acquisition ) and plot graphs etc

------Question-------
Can we  Totally  bypass instrumentation amplifier and directly (maybe add a filter)connect the output of sensor to ADC ?
1) adc will do work of converting the differential signal into binary data
2) We can avoid the instrumentation amplifier circuitry
3)ADC has programmable gain which we can set to 100


Any suggestions and reviews are welcome ! Thank you in advance !!

Think of it this way. The ADC is like a ruler, the instrumentation amplifier is like a magnifying glass. The amplifier takes a small signal and "magnifies" to a size that the ADC can easily "measure".

Most ADCs like their inputs to sit around some voltage (the common mode voltage) so when you have a differential signal, one pin swings above the common mode voltage while the other swings below it. ADCs tend to have limited common mode ranges.

Instrumentation amplifiers are a special class of differential amplifier. They have very, very high input impedances (good for really weak signals). Instrumentation amplifiers are designed to amplify tiny differential signals which can have very large common mode voltages, something that is very hard with regular single ended amplifiers. The amp separates the part of the signal we want from the part we don't want and scales it to a range that the ADC can use.

E.g. Driving the sensor would yield a 5V common mode voltage. Your signal is differential so at full scale you'll see 4.990V on one pin and 5.010V at the other. this gives 20mV as 5.01 - 4.99 = 0.02 or 20mV.

Now, we want to convert that 20mV differential signal to single ended and scale that to 0-3.3V as that is the default input range of our ADC. That's where the instrumentation amp comes in, it will pick out the tiny 20mV signal and by setting it's gain to ~1600 that will let the ADC use it's full range i.e. the full 32k counts. As an added bonus, now that your signal is in the range of 0-3V or so, just about any micro can read it as just about all of them have internal ADCs anyway.

Yes, it is possible, however you will have to arrange things such that the voltage remains within the common-mode range of the chip, and certainly within the absolute maximum input voltage of about 0-5V.

It may be sufficient to operate the sensor from 0V/5V so both inputs are biased near 2.5V.

Note: Keep in mind that you can lower the voltage energizing the resistive bridge sensor and the output voltage (bias and signal) will simply be reduced proportionally. The drift and offset of the ADS1115 is not great compared to a good amplifier, but may still be good enough for your application.

• Well thank you !! That would reduce too much of the circuitry ! its a 15 bit ADC and i want minimum 20000 counts ! i think it will work then ..ill have to adjust the output somewhere around 2.14 volts ..... m i doing it right ?? – Mr.Sky Sep 24 '16 at 19:12
• How about the CMRR? and if he uses a 15bit ADC he gets only 0.15mV resolution, which might give him the liberty to obtain the accurate pressure measurement. – ammar.cma Sep 24 '16 at 19:22
• @spehro Can i power the sensor using gnd to +10 volt supply ? That would solve most requirement problems as i need full scale output to be 20mV – Mr.Sky Sep 24 '16 at 19:25
• You can. @sohamkulkarni – ammar.cma Sep 24 '16 at 19:26
• @ammar.cma Well the adc has range from 0v to 3.3 v and is 15 bit so we get 32768 counts 3.3/32768 =0.0001 – Mr.Sky Sep 24 '16 at 19:28

The ADC has adequate CMMR for low frequency but will need a Balun or CM choke for HF. The CMRR is typ. 100dB and must having incoming bandstop filter above 8Hz to result in noise below your desired resolution. You must determine the level of interference to determine the required filter and ensure on power up this filter does not exceed input safe levels.

Your input ADC error from all sources may be 0.15%. Equivalent input noise is specified on the data sheet, but to achieve this, you must design EMC filters on ambient noise determined by your unique situation.

ADDED: - Due to switched capacitor input noise is the reason they say... "Unless the input source has a low impedance, the common-mode input may affect the measurement accuracy" For this reason I suggest you include an INA front-end.

• Thank you so much for the insight ..i will start working on designing the filter right away...!!I am using a raspberry pi so i have to power adc using 3.3 v also i need a resolution of 20000 counts is it possible using 0-5volt supply to the sensor ??? ADS 1115 is 15 bit adc btw – Mr.Sky Sep 24 '16 at 20:04
• resolution is possible, accuracy is not – Tony Stewart Sunnyskyguy EE75 Sep 24 '16 at 20:30
• how can we achieve the resolution ? dont we need an amplifier ?? – Mr.Sky Sep 24 '16 at 20:34
• 1st define ambient noise on sensor cable spectrum, then determine band stop filter and determine if you can do with high RC differential filters or active high order filter using INA. – Tony Stewart Sunnyskyguy EE75 Sep 24 '16 at 21:01