TIPD193: TIPD193

Hi Peter,

Can you please share your schematic? How are you biasing the thermistor?

Have you verified the integrity of the connection to your sensor? If you have a spotty connection then this could cause the output to drift to the rail if the bias current path to ground is interrupted.

Hi Zak,

I use the REF200U for biasing.

I do have terminal blocks for the connection, that works good.

Temp-Measuring-Board-SCH_20180504.pdf

Peter,

Just to be clear, the values shown in your schematic are completely different from what you posted. What are the correct values? As shown, the INA326 is running in a gain of 1000. This means that even very small voltages can cause the device to saturate at one of the rails. Another issue I have seen running in gains this high is if you have used filter capacitors with a poor dielectric temperature coefficient, such as X5R, then very small fluctuations in the temperature of the capacitors can introduce additional offset error.

Parasisitic leakage paths resulting from flux contamination also become a major concern in high gain high impedance configurations. Have you properly cleaned the board to remove any excess flux residue?

Hi Zak,

Thank you for your update.

We do use the PCB for different configurations of PT100 and Thermistors at different temperature ranges (normally around 0...50 or -20...100 degree Celsius).

The schematic contains the values for a PT100 setup with a high gain / the thermistors setup has a gain of -1.79, so rather small.

I do have X7R capacitors on the PCB. The PCB is fully produced from a PCB manufacturer and we solder only the 4 parts to adopt the sensor/temperature range as listed in the first post (use of a SMD tweecer solderer). Do you think the flux contamination can also be a problem on that small area?

You mention saturation on the rails. Does it not recover from saturation if the sensor signal is stable?

Is it maybe needed to delay the current reference startup? (I do not measure the problem as long as no sensor is connected)

Hi Zak,
I did some tests / measurements on the problem.
When I startup and it works, I just have to measure at R1 and the output saturates on 5V.
Then I can "fix" it with detaching the GND of the Sensor for several seconds.

Voltages:
Pin7 - Vin: 4.98V
Pin3 - V+: 1V
Pin2 - V-: 0.028V
Pin1 - R1_2: 0.17V
Pin8 - R1: 1.12V
Pin5 - R2: 1.55V (good) / 5.48V (fail)
Pin6 - Vout: 1.6V (good) / 4.98V (fail)

Thank you and Best Regards,
Peter

Hi Peter,

Given that these components are soldered near the pins of the IC, it is still a possibility that flux could create an issue. I would still recommend a thorough cleaning of the boards to be safe and rule this out. You are particularly susceptible to induced leakage currents given the high resistances you are using around the device.

If I have understood correctly, it sounds like you are saying that the board functions properly until you measure at R1, correct? Are you measuring the voltage across R1 or with respect to ground? I wouldn't expect the device to latch-up as you are describing but depending on your meter it is possible you are altering the circuit when you apply your meter or introducing a stability issue on one of the internal amplifiers. This should not persist when the meter is removed though.

You also shouldn't be able to measure a voltage at R2 greater than your supply voltage, so this leads me to believe something else is going on in your system with regards to your ground connections.

Does this issue ever arise in boards that are configured for a PT100 sensor or is it only for the low gain thermistor configuration? Thermistors are most often 2-wire devices. Given that the connection to ground for the negative input is through the terminal block, how are you connecting your thermistor?

Hi Zak,

We do have an older device with 12 channels on one PCB and I was told, that the error also occured there - they just made some power cycles.

I took two devices from an other project, each device has 3 boards (6 channels) in use to test for the PT100/Thermistor - the devices are also quite new, so we do not have much experience.

I did find another board which hat that suspicious error.

Up to now all boards with the error do have a Thermistor connected. The Thermistors are connected as a PT100 in 3 wire connection (two wires are wired to one pin of the Thermistor and one wire to the other pin). We do have cable length within 1...5m.

The error occurs at power up - I would not think about it much if it is only when I measure on it :-(

--> then that channel is stuck at 5V!

As I said sometimes a power cycle helps - sometimes not. What always helped me was to detach the sensor GND connection.

On the other side, all boards which do not show the error, do not go into error when I measure at the R1.

I also tried to clean the PCB, first it looked a bit better. But then I moved the setup to a different working space and now it is back that it has the error sometimes. I do not have some kind of a special PCB cleaning solvent. I ordered one and will test with them tomorrow.

Thank you very much for your suggestions.

Peter

Hi Peter,

I think we need to keep digging to uncover the source of this problem. Did you get a chance to test after thoroughly cleaning the boards? Have you tried doing an ABA swap on the failing boards? i.e. Since only some of these boards are failing, if you swap the component on the failing board with one from a known good board and place the potentially failing unit on the known good board this will help narrow down whether it's a device issue or something with the board. You may try doing the same with your thermistors.

From the data you posted above, it looks like the output is the only node that ever shows wrong values, correct? This is interesting as it means the current source is at least operating as intended and would imply there is indeed a good connection to ground, this makes me suspect even more that some of the boards have either intermittent connections to the gain resistors, or again that there are flux contamination issues.

Does disconnecting and reconnecting the ground of the sensor solve the problem as long as everything remains on? I would try leaving a unit running for a couple hours after doing this to see if the output has drifted at all. If this is ultimately related to excessive leakage this can sometimes take a while to manifest.

Hi Zak,

As I have written above, the R2 has also a wrong value - above the power supply level. Thats the point  where I'm curious about. I'm not sure if something happens during the startup.

The value is stuck and it seems to be stable. I let it run the whole day avter recovering from the error and the value was stable (only some shift due to temperature increase - the value at the output fall about 0.2V).

I'm not sure if it is good to exchange parts - because you are also still think about the leakage currents. May I remove the R1/R2/C2 (which we soldered by hand) - the rest of the board was manufactured as serie.

Yes when I disconnect ane reconnect the GND of the sensor as long as the board is still powered brings the output back to normal value.

Best Regards,

Peter

Hi Peter,

I'm not sure what exactly is happening with this configuration. Could you try using an oscilloscope to measure on both sides of the input filter resistors during startup and once you disconnect the ground of the sensor as you have mentioned? If you could gather this for a board that fails and a board that doesn't this may help narrow down what the issue is.

Peter

Were you able to make the measurements as described by Zak in the last post, or were you able to resolve the issue?
If you need additional help just post another reply.

Dennis

Hi Dennis

Sorry for not replying so Long.

I was moved into another Project and did not find more time to look into that.

I was able to exchange the PCB with a different one and used another sensor (5k resistance instead of 10k). So the gain resistors are smaller. This PCB / sensor setup works now and that Project is not helt up.

I still would like to find out why the other Setup of PCB / 10k Sensor did not work properly - but it will take some time.

Best Regards,

Peter