TLV2186: 0-70V to 4-20mA converter circuit problem with startup

Hello Luis,

• Do you have the high-voltage present at the INA826 input voltage divider when you turn the INA826/TLV2186 power supply on and off?
• If you drive the TLV2186 input directly from a low-voltage supply and cycle the supply do the TLV2186 outputs appear as expected?
• Do realize that the INA826 output does not go all the way to zero volts, but hits the rail about 100 mV above ground. If the output is driven into the rail it takes time for it to recover each time it is slammed.
• Take a look at the INA826 and TLV2186 outputs and the supply line with a DSO as you cycle the power supply on and off. Observe what they are doing. In no case should the TLV2186 supply be allowed to exceed 26 V. 
• What is the load at the Curr Out point of the circuit?  What voltage compliance range are you expecting?

Regards, Thomas

Precision Amplifiers Applications Engineering

Hi Thomas, thank you for the quick answer.

1) Sometimes I don't, sometimes the input is shorted to ground. In the application where it goes, it is connected to another circuit whose initial voltage can be any value between 0-70V at startup.

2) Didn't try that since I wanted to figure out what might be happening with the whole circuit.

3) But that's alright since the output is between 4-20mA. 4mA with a 147 resistance in series is 735mV, well above the 100mV that it can output. Why would it go to 0 in the first place? There's always a current path. For example, when the output is stuck at 0, the usual "solution" (I wouldn't really call it that) is to open the current path, short circuit the R5 so that the 10V are placed on the input of the TLV2186 and the output goes to 24V. After that, connect the current path again, and it goes to the 4mA. But, in the field, this procedure isn't really an option. Some other times, the output gets offset by alot, like starting at 7mA of output going all the way up to the 20mA.

4) I'm using a 24V LED power supply and with a TDS 210 I didn't see any overshoot on the supply. I did assume that because I'm using a wago connector to make the connections and these problems occur when connecting and disconnecting the 24V from the circuit messing with the cables on those connectors.

5) The curr out point of the circuit is either connected to a multimeter (for debugging purposes) or a NX-AD4203 analog input unit. It has 85 of input impedance, so adding that to the 147 ohm present on the circuit, it goes to 232 ohms. For a 4 to 20mA range, we're talking 928mV to 4.64V.

Best regards,

Luís

Hi Luis,

Regarding your responses:

1) Sometimes I don't, sometimes the input is shorted to ground. In the application where it goes, it is connected to another circuit whose initial voltage can be any value between 0-70V at startup.

Having a voltage present before the circuit is powered places a voltage at the INA826 that will have to be resolved once the circuit is powered. Until powered, or reaching the minimum operating voltage, the internal circuits inside the INA826 and TLV2186 are moving from a shutdown state to a linear bias state. It takes some time for the circuits to attain their linear operating point. It is best that the input voltage be brought up once the amplifiers are fully powered and have stabilized at their operating points.

2) Didn't try that since I wanted to figure out what might be happening with the whole circuit.

I am attempting to understand if it is the INA826, or the TLV2186, that is having difficulty attaining the correct output level relative to a specific input level. Having just the TLV2186 in the circuit would help identify which stage is not resulting in the expected voltage output.

3) But that's alright since the output is between 4-20mA. 4mA with a 147 resistance in series is 735mV, well above the 100mV that it can output. Why would it go to 0 in the first place? There's always a current path. For example, when the output is stuck at 0, the usual "solution" (I wouldn't really call it that) is to open the current path, short circuit the R5 so that the 10V are placed on the input of the TLV2186 and the output goes to 24V. After that, connect the current path again, and it goes to the 4mA. But, in the field, this procedure isn't really an option. Some other times, the output gets offset by alot, like starting at 7mA of output going all the way up to the 20mA.

My concern is that the INA826 minimum output voltage is 100 mV above the negative rail of 0 V. If the input voltage is very low, close to 0 V, the output can't follow it that low. Running the output into a rail results in saturation and there is a recovery time associated with coming out of that state. The TLV2186 op amps have a similar output swing to the rail limit and depending on the current being demanded by the load it can be upwards to a half volt above the lower rail. 

4) I'm using a 24V LED power supply and with a TDS 210 I didn't see any overshoot on the supply. I did assume that because I'm using a wago connector to make the connections and these problems occur when connecting and disconnecting the 24V from the circuit messing with the cables on those connectors.

To properly evaluate the "when the circuit is connected to the 24V power rail, the output of the ampops is completly wrong" behavior the variables need to be minimized. My thought is that the +10 V and +24 V supplies and the applied input voltage to the divider should all be brought up together in unison. If say for example, the +10 V supply comes up before the +24 V, that might set up U2 for a latch condition. If an op amp latches due an input voltage being applied before the supply is applied it may not recover until the power is turned off. A latch condition can result in an incorrect output condition. Although modern op amps are designed to minimize latches, but I wouldn't rule it out as a possibility.

5) The curr out point of the circuit is either connected to a multimeter (for debugging purposes) or a NX-AD4203 analog input unit. It has 85 of input impedance, so adding that to the 147 ohm present on the circuit, it goes to 232 ohms. For a 4 to 20mA range, we're talking 928mV to 4.64V.

The main concern is that the 4 to 20 mA has a current path to ground through the load and there is sufficient linear output voltage range for U2.

I do have additional ideas regarding your circuit.:

• If you have some flexibility to experiment with the circuit, replace the TLV2186 with an OPA2196 dual op amp. The TLV2186 is a chopper op amp of a very complex design. The OPA2196 uses more conventional design precision op amp that doesn't rely on chopping to minimize the voltage offset and drift. If you need more bandwidth than consider the OPA2197. You can find the OPA2196 datasheet at: 
• Since your input is single ended there really isn't a strong reason to use an INA826 for that part of the circuit. A single op amp such as an OPA196 could replace the INA826. The OPA196/2196 is rail-to-rail I/O op amp. The op amp could be set up as a non-inverting amplifier having the same gain as the INA826 circuit. It would be interesting to see if the issue you are trying to resolve goes away with that circuit change.

Regards, Thomas

Precision Amplifiers Applications Engineering

Hey Thomas,

Again, thank you so much for taking the time to try and help me with this issue.

In this debugging stage, I just run a wire from the input to ground. So there's no voltage level applied before the supply voltage. Also, the 10V rail is taken from the 24V rail with a voltage regulator (U3 in the circuit). So, in theory, the 24V appears before the 10V voltage rail, so the ampops should be powered before the 10V is even available.

When I just have a wire running from the input to ground, the voltage output of the INA826 is very close to 0 (17mV). And, whenever the voltages are "stuck" it's between R4 and the input of the TLV2186. It's either 8V or 7V, something weird like that. But, on the other side of that resistor (between R4 and the output of the INA826 the voltage is following the input voltage). I did add a 1uF capacitor on the C1 to see if it did anything, but alas, the problem persisted. It looks like there's a leakage on the input of the TLV2186 and that affects the R4 - R5 - 10V part of the circuit.

From what I could gather. up until R4 everything is working as expected. It's just that the input of the TLV2186 actually changes the R4 and R5 circuitry and changes the voltage of that node apparentely, which is strange since that particular input doesn't have feedback...

We did order three other opamps to test:

OPA2210ID
OPA2991IDR
LM7322MAX/NOPB

Do you think any of these might work in solving the issue? We're on a tight time schedule so we had to order some opamps to try and fix this and I've identified these which seemed capable of working in these conditions.

Best regards,

Luís