INA330 oscillation

Hi Tom,

I haven't worked on a similar issue with the INA330, but based on the block diagram of the IC there are two internal amplifiers that drive the thermistor and reference resistor.  Amplifiers generally don't behave well with capacitive loads so I could see your application experiencing issues due to the capacitance of the cable between the thermistor and the INA330.  Is there any way that you've selected nominal values of Rtherm and Rset that are less than 2 kOhms?  As noted below Figure 3 on page 7 of the datasheet, nominal values of Rset = Rtherm = 10k and should not be less than 2k or stability issues can occur.  The datasheet also mentions that the values of Rtherm/V1 or Rset/V2 should not create a situation where the excitation currents begin to dip below 125uA.  Be sure you haven't placed the INA330 in one of these configurations.

Generally we don't recommend placing the sensor far from the excitation source but if you can not avoid it, try placing a resistor between the output of internal sensor excitation source (pin 10) and the cable.  Try a value of around 100 Ohms and if that is not effective, try increasing the value to 499 Ohms.  The same value needs to be put in series with the Rset resistor to balance the system.  Depending on the accuracy requirements you may find that you need to use higher precision and low-TC type resistors.

Let me know if this solves the issue.

Regards,
Collin Wells
Precision Linear Applications

Dear Collin,

Thank you for your reply.

My set resistor is 13k, my thermistor is 20k at room temperature, dropping to 13k when the controller is running. I'm using Vexcite = 1V (V1 = V2), so I'm comfortably below the 125uA limit.

I did try putting a resistor before the cable. I found that ~1k more or less solved the problem -- it would generally be fine that then I'd poke it a bit and it'd start oscillating (more poking would stop it again). 4k7 seems to have removed the behaviour completely.

I agree that the output buffers probably don't like the capacitance. However, wanting to have the sensor a couple of meters from the controller doesn't seem like such an unreasonable thing to want (surely, this must be quite common?). The datasheet also doesn't give any specification of maximum load capacitance or any warnings about cabling. Do you know of any guidelines for this particular IC?

Tom

Hi Tom,

Thanks for the additional information, it looks like you've placed the INA330 in a good operating state with your resistor and voltage choices. 

Also, nice work on adding the additional resistance on your own.  I'm surprised that it took 4.7k to solve the issue, this probably rules it out as an effective solution.  I agree that it would be nice if the sensor input nodes could handle more capacitance, but the architecture used in the design of the INA330 is basically an instrumentation amplifier where the typical gain-set resistor, Rg, is replaced with Rtherm and Rset.  Therefore as usual with instrumentation amplifiers, additional capacitance should be avoided on the gain-set nodes to avoid stability and AC gain-error issues.  I unfortunately do not have a design guideline for this node but my educated guess would be that the capacitance should be limited to a few tens of picofarads at most to avoid issues.  Is it possible to use a different cable without heavy shielding that may reduce the effective capacitance?

Regards,
Collin Wells
Precision Linear Applications 

Dear Collin,

Thanks for the suggestions. I'll have a think about whether I mind adding some kind of load isolation resistor into the design. Otherwise, shorter cables it is...

Tom

Dear Collin,

Okay, I've had more of a think about this. I really do think it's a bad oversight of the datasheet not to mention this at all. Wanting to have a sensor a couple of meters from control circuitry isn't such an odd thing to want. Since that is likely to mean at least ~100pF, I think some room should be given to advice on capacitive loading. Without knowing much of the internal structure of the chip it's hard to have any intuition about limits on capacitance. Thus, the best thing to do seemed to take some measurements. In case anyone else runs into similar issues, I though I'd post my findings.

Conditions:

I set the chip up essentially as in figure 6 on the datasheet:

V1 = V2 = 1V

Rset = 10k

Rtherm = 10k (I just used a fixed resistor here)

Vsupply  = 5V

I added a load capacitance, Cl, in parallel with Rthem.

For Cl <= 150pF the circuit was pretty stable. It wouldn't oscillate on it's own. Prodding it with a DVM caused it to oscillate (not surprising), but it would stop when the DVM was removed.

For Cl = 180pF the circuit seemed borderline stable. It wouldn't oscillate spuriously, however if I did poke it with the DVM it would enter self-sustaining oscillation (by which, I mean that it kept oscillating even with the DVM probe removed).

For Cl = 200pF it would oscillate spontaneously.

Next, I tried adding an isolation resistor between the chip and Rtherm+Cl. For this, I took Cl = 330pF, Rl=10k. I found:

For Rl = 220R it would oscillate spontaneously.

For Rl ~470R -> 1K it would not oscillate spontaneously, but would enter self-sustaining oscillations when poked with a DVM.

For Rl >= ~2k2 It was pretty stable. It still oscillated when poked with the DVM (unsurprising since I was basically using it as a few nF of capacitance). However, oscillations stopped when the DVM was removed.

Conclusions:

For stability, try to keep Cl < 100pF. <150pF is probably okay but more than that is asking for trouble. If more capacitance is needed (e.g. the control circuitry can't be moved near the sensor easily) then it's probably best to use an isolation resistor of a few kohm. However, if that's not acceptable then ~500ohm is probably fine.

Tom

EDIT: P.S. adding an isolation resistor is probably not the best way of dealing with this issue, but it's hard to do anything cleverer without more detailed information about the inside of the IC.

Hi Tom,

Thank you greatly for your work to help the others on the community.  I agree that the resistors aren't an ideal solution but the nodes that would be required to compensate the circuit for different capacitance on the Rtherm and Rset nodes are unfortunately internal to the IC and aren't available for compensation.

Best Regards,
Collin Wells
Precision Linear Applications