INA317: Output signal behavior: part 2

Hi Jagbir

When I test the 'new' boards, I am seeing the output voltage drift over time with no input connected.

Please check the following. In the new circuit, please short pin2 and pin4 together at J2 (input pins to the INA317), you should get zero voltage output at pin 6 in the new circuit.

You should not leave the input unconnected at its input. Please let us know the result. 

Best,

Raymond

Hi Raymond,

I do get a zero voltage output at pin 6 in the new circuit. when I pin2 and pin4 together at J2 (input pins to the INA317). Why  the unconnected input pins were giving a zero voltage output with my 'old' circuit.

Thanks,

Jagbir

Hi Raymond,

 I have not yet connected the V- to -230mV, as this pin is still tied to ground via R5 jumper, meaning my reference is still tied to V-(ground). i have an option to use -230mV rail, but for that I need to populate R2  and obviously remove R5

Jagbir

Hi Jagbir,

That is what your new schematic is shown, where V- is connected to -0.232V. If V- = GND, then by shorting pin2 and pin4 together at J2 is not valid test for INA317. You have to DC biased up to have proper input within the required Vcm. 

What are the other difference between old and new circuits now per your initial inquiry? Is INA317 still working or damaged? 

Best,

Raymond 

Hi Raymond,

INA317 is still working.

In my  new circuit, there are two configurations- one with V- connected to -0.232V and another configuration where V- is connected to GND. My default configuration is  V- tied to GND. What you see in schematic is a  pcb pad without a physical resistor mounted on it (labelled as DNP aka Do Not Populate).

The list of changes to old schematic include :

• Removed the 100nF output cap, C7 since it was outside the range of allowable capacitance on the output pin
• Removed input clamping diodes(D3,D4)
• Removed the 470 Ohm series resistance.
• Provided options for input  bias currents (pin3 ,pin3 on IN317) return path to ground via pcb resistor pads but the resistors(R6 and R10)  to complete the circuit are not mounted on these pads.

• As described above , in new circuit there are options to connect V- to ground as well as -0.232V . In my default configuration is  V-  is tied to GND.
• Please note that in the new circuit there might be some resistor reference designator which are used in older circuit: R2/R5.

Hi Jagbir,

Your circuit has gain of 1001, which is high. Since your input is floating, Vout is drifting over time should not indicate anything, especially the input voltage is not within the recommended Common-mode voltage range. 

From the above described circuit. if you short the input and measure output, the following image is a correct way to do it. At Vout = 22.08mV, the input offset voltage is approx. 22.08mV/1001 = 22 uV.

If your V- = -0.232V is implemented, then you can short the input as I replied in the early email. The input common mode voltage has to be within the range below in order to get linear output response. 

If the output is drifting vs. time with unconnected input, it is not linear output response that is prescribed in the datasheet. I am unable to explain it. 

If you have other questions, please let me know. 

Best,

Raymond

Raymond,

Couple of questions:

• When I connect a 1000 Ohm strain gauge to the new circuit at J2, the output voltage drift vanishes. However, I am seeing a stable/constant 0.2V offset at the output voltage, even in absence of an input strain. What is is the reason for it.
• Why Vcm is not a problem with my old circuit and suddenly it's an issue with new circuit.

Thanks,

Jagbir

Hi Jagbir,

If you have configured the strain gauge in the following manner, you should get near zero output as shown below. 

You told me that the INA317 is working. That is why I am unable to explain your issue (see the prior reply). 

If your test circuit is configured same as the above, and Vout = 0.2V, then something is wrong with your INA317 or circuit. However, you need to verify that Vin+ and Vin- inputs have the same dc voltage of 3.3V/2 = 1.65Vdc, 

Best,

Raymond

I believe the issue is due to bias current path, as without a bias current path, the inputs float to a potential that exceeds the common-mode range of the INA317 device, and the input amplifiers saturate.

After I soldered R6 and R10 =10K  into my new circuit, I am getting 0V offset.

Hi Jagbir,

Can you upload your mods of your new circuit? What is your excitation voltage for the transducer? The simulation is specified at 3.3Vdc, which is same as your schematic. The input common mode voltage is from 100mV to 3.2V, if V-  = GND. 

Best,

Raymond

Raymond,

Wherever it says DNP , it means that component is not populated.Jumper means it is shorted with a 0 ohm resistor . So far I have been extremely disappointed with the TI 'expertise' on INA317.

Soldering R6, R10 for return path makes the output voltage zero but the output voltage has stopped responding to the input strain. 

Hi Jagbir,

All you are doing is to reduce the input signals at Vin+ and Vin- in half. You do not have input bias current path issues, unless your transducer is NOT connected to ground or has a poor ground connection. 

If you are using 3.3V as the excitation voltage for your wheatstone bride and you are mid point is 1.65Vdc, which is within the required Vcm range. Now divide the input signal with another 10kOhm, your Vin+ and Vin- input are now 0.825V, which is also within the required Vcm range. 

The latter case with Vout =0 or near 0, the first case does not. (Vout = 0.2V).  My speculation is that you likely have poor ground path at the transducer side.  Please check on the mid point of wheatstone, and make sure that the output of wheatstone bridge is measured 1.65V with respect to ground. 

In addition, please make sure that your transducer is working as it should. 

If you look at the section 8.2.2.5 of the datasheet, you do not have any configurations that is similar to Figure 34. Your transducer is providing the input common-mode current path for INA317. 

Best,

Rayond

Hi Jagbir,

Raymond is doing a good job. So please don't attack him.

Why do you expect that the INA317 is working properly and gives a correct output voltage without having connected a sensor to the input? This makes absolutely no sense. Would you drive a car without wheels?

Reading a bit between the lines of what you have posted so far, I guess that you have two operation conditions? One with a sensor connected to the input and another operation condition with no sensor connected to the input? Is this right? And with no sensor connected you expect an output voltage of 0V? Is this what you intend?

Why not just telling us what you actually want? We cannot help without knowing all the facts :-)

Kai

Hi Raymond and Kai,

Sorry for the strong language used in comments , the reason being (although not justified) that I am not getting the answers to INA317 behavior: why my output is unstable(with input disconnected) and why I have an offset voltage(after input is connected). 

I had none of these issues in my first circuit and I don' t know why its happening. Raymond explained that the unstable output is due to Vcm which is fair enough but why was this not a problem in my initial circuit. I have no idea why I am getting an offset voltage even after interfacing the sensor. The simulation results look fine. I will be posting more data after today's test.

Hi Jagbir,

in both of your circuits are several mistakes. By this it can happen that two mistakes cancel each other and the circuit seems to look working fine, but which actually does not. A new set of INA317, BAT54 and RCLAMP3321 with their manufacturing tolerances and your circuit might work totally differently and what was good is now bad and vice versa.

If you choose a gain of 1000, which is extremely high in such an application, you must only use components which have ultra low leakage currents. So, please don't use the RCLAMP3321 or, even worse, the BAT54S at the input of INA317. Their leakage currents are way too high. To provide protection only an ultra low leakage diode like the BAV199, e.g., can be used.

Provided that your bridge sensor DOES NOT output a negative offset voltage, you could use this circuit:

jagbir_ina317_2.TSC

But if your bridge sensor DOES output a negative offset voltage, then you must either provide a positive reference voltage at the REF pin of INA317 and do the offset voltage correction in software or you must do a hardware offset calibration directly at the bridge sensor which I do not recommend because of temperature and long term drift issues.

Kai

Kai,

Are you concerned with the reverse leakage current. The typical reverse Leakage current for RClamp3321P  is actually far better than BAV99.

Hi Jagbir,

I said BAV199 not BAV99 :-)

Kai