INA125: Troubleshooting noise from INA125

Hi Sandheep,

One thing I noticed is that the 2.5V reference has quite a bit more noise on the INA125U than the INA125P--about 230mVPP vs 20mvPP (as measured on oscilloscope). 

I suspect that your INA125U's 2.5Vref is overloading from the transducer you are selecting. The output current of 2.5V from INA125U is: 2.5V/(700Ω||700Ω) = 7.143mA. The part Vref  output current is only rated up to 5mA (1000Ω type of transducer will be more suited for INA125 part, if you want to use 2.5Vref voltage). 

The reference voltage noise is too high. Let us assume the worst case of 75ppm/mA from the datasheet. With 5mA of load current in 2.5Vref, the worst case should have 375ppm or 2.5V*375ppm = 937.5uV or < 1mV. The measured figure seems to be very high for 2.5V reference voltage. At 20mVpp, the 2.5Vref load regulation is 10mV/2.5 * 100 = 0.4%, 230mVpp's load regulator is approx. 4.6%. 

Vref of 2.5Vdc should be a constant without large ripple voltage, when there is no load or light load. 

Please also check your 5Vdc voltage ripple at the rectified 5Vdc or the input of V+ or at pin1. If the 5Vref voltage is stable, connect it to your strain gauge in Wheatstone bridge, and observe the Vout of INA125U.   

I do not have your strain gauge's sensitivity (mV/V) figure. The circuit' gain is configured at 604 V/V. Enclosed is an example your INA125, since I do not your sensor configuration. Please make sure that these parameters work within the entire range of the application. . 

INA125 Bridge_Amp 09292021.TSC

Would you recommend a different amplifier for this application?

Please let me know your application and requirements. Is it to measure weight or other application? If you have to use 350Ω type of transducer, then we have to reduce the current load in 2.5Vref or find a different instrumentation amplifier. 

Best,

Raymond

Raymond--thanks for the response!  I think that I ran yesterday's tests with and without the strain gauge attached--and saw similiar noise regardless.  So...it doesn't seem to be the case that the Vref is overloaded (but I'll double check this afternoon).

The strain gauge is not being used to make a direct strain measurement, but an indirect measurement of force.  The linear strain gauge is mounted on the side of a cantilever beam (as far away from the neutral axis as possible).  I'm performing a 2 point calibration (with a calibration weight hanging off the beam, as well as unloaded) so that I can interpolate (and extrapolate) the load applied to the beam.  I could use 1000ohm strain gauges.  And I could reconfigure the circuit to use a chevron style 2-half bridge strain gauge.  So I'm generally flexible...but I'd like to see if it's possible to improve upon (or at least replicate) the performance of my sloppy but totally acceptable prototype.  Does this give you enough information to recommend an alternative amplifier?

Thanks, Raymond!

Sandheep

Hi Sandheep,

I see. Please check the ripple voltage in +5V supply voltage rail. INA125U has PSRR that is rated at 120dB. Its reference voltage should not see high ripple voltage under nominal load or use (120dB is equivalent 1uV/V in noise rejection). 

Please make sure that the PCB is clean as well, and the board does not have parasitic paths that may  couple unwanted noises.  Use low ESR capacitor after LDO and coupling capacitors. 

If it is inconclusive, please send me scope shots of the +5V supply and 2.5V reference voltages.

Best,

Raymond 

Raymond--

OK...did some troubleshooting.  I think that my 5V supply has too much noise (spec is 100mV ripple).  Here's how I came to that conclusion:

	Ch1: PCB1 Vref2.5 Ch2: Prototype Vref2.5 Ch3: PCB2 Vref2.5 Ch4: PCB2 5V supply As you can see in channels 3/4, the Vref2.5 has a strong dependence on 5V supply.
	Ch1: PCB1 Vref2.5 Ch2: PCB1 5V supply Ch3: PCB2 Vref2.5 Ch4: PCB2 5V supply Again, both of my PCBs show that Vref's noise is dependent on the 5V supply noise.
	Ch1: Prototype 24VDC supply (wall wart) Ch2: Prototype Vref2.5 Ch3: PCB2 Vref2.5 Ch4: PCB2 5V supply Using a cleaner power supply (a wall wart of all sources), Vref2.5's noise improves.
	Ch1: PCB1 5V supply Ch2: PCB1 Vref2.5 Ch3: PCB1 Strain gauge output Here's the PCB powered from my benchtop 5V source, noise is tolerable, I think.
	Ch1: PCB1 5V supply Ch2: PCB1 Vref2.5 Ch3: PCB1 Strain gauge output Here's the PCB powered from the original application's 5V source (with the 100mV spec ripple)--things are getting a bit noisier again.

So...I'll re-run some tests tomorrow with a cleaner 5V supply to see if the output signal noise improves.  Can you advise on the following:

• If things improve with a cleaner 5V supply, then I'll need to redesign the PCB.  Do you have any recommendations on low-noise regulators for this application?  My PCB has both 5V and 24V available.
• Since I'll have to redesign the PCB, would you recommend another amplifier instead of the INA125?

Thanks, Raymond!

Sandheep

Hi Sandheep,

Do you have any recommendations on low-noise regulators for this application?  My PCB has both 5V and 24V available.

If your 5Vdc and 24Vdc power sources are directly connected from wall wart adapter, you need to place some filters after the "regulated" voltages, which they are consisted large ripple voltage and high frequency noises. You may generate 5V LDO directly after 24Vdc regulator, where 5Vdc's ripple and high frequency noises maybe further attenuated by LDOs  

You may place TPS7A20 LDO after your existing 5Vdc source, it may regulate and attenuate the existing 5Vdc with lower ripple and eliminating high frequency noises. I have not used this part, but this is simply a low dropout series regulator based on the datasheet. 

From the scope shots, your 5Vdc source consists wide range of high frequency noises. You may need to place low pass filters to filter out unwanted noises in the 5Vdc (You do not need exact 5Vdc, 5Vdc ± 5 to 20% will be adequate, as long as the ripple voltage and high frequency noises are low for your application).   

Since I'll have to redesign the PCB, would you recommend another amplifier instead of the INA125?

INA125 Instrumentation amplifier (IA) will work good with 1kΩ type of strain transducer. 350Ω type of strain gauge may load down 2.5V excitation reference voltage. 

We have many high precision low voltage IAs such as INA333 INA317 etc. and huge selections of high voltage precision IAs (up to 36Vdc) for various applications,  see the link below. 

https://e2e.ti.com/support/amplifiers-group/amplifiers/f/amplifiers-forum/818894/faq-how-do-instrumentation-amplifiers-inas-fit-into-my-design/3031609?tisearch=e2e-sitesearch&keymatch=%5BFAQ%5D%20tamara#3031609

INA125 is an integrated IA, which it consists selectable excitation reference voltages.  INA333 INA317 and many other IAs do not come with built-in reference voltage. However, you can select a known voltage reference voltage, say REF5025,, REF3025  or REF6025, connect it to an op amp buffer in configuration and use the op amp buffer as a precision excitation voltage and current source for your application. If you need assistant in the circuit, I can simulate on on your behalf. 

If you are able to provide me with the strain gauge operating range in resistance (R decreases in compression, R increases in tension). I can simulate a strain gauge circuit. The above reference voltages are based on 2.5Vdc. I believe that you may want to increase the excitation voltage of the strain gauge from 2.5V to 5V, perhaps higher, which it will increase the differential responses and more sensitive to a load change in the sensing application. Please check the strain gauge operational recommendation from the supplier. 

If you have additional questions, please let me know. 

Best,

Raymond 

Raymond--

Thanks again for the response.  I ran the circuit with the cleaner supply voltage and the output was much cleaner than expected, so that problem is solved.

However, I'd like to improve the variation in the output signal. Unfortunately, it looks like I'll be stuck with the 350 ohm strain gauges, so I think that my main options are increasing the gain and increasing the voltage through the bridge.  My calcs show that the strain gauges can tolerate 15V through the bridge (which should result in a 6x increase in variation).  A 350 ohm bridge will have a 350ohm total resistance, so the current draw will be 44mA.  Do you have any reference designs for a 15V, 50mA+ voltage reference + op amp buffer?  My DAQ is expecting a 0-5V signal, so I'll need to bias the instrumentation amp's output to 2.5V (but, correct me if I'm wrong, I can use a standard voltage reference e.g. LM4030, right?). 

Also, do you have any recommendations for a 5V LDO with an input range up to 24V?

Thank you!

Sandheep

Hi Sandheep,

Do you have any reference designs for a 15V, 50mA+ voltage reference + op amp buffer?  

Per your requirements, I selected LM4030-4.096 reference voltage. The output current in driving 350Ω Wheatstone bridge will be approx. 35mA at 12.2935Vdc in reference voltage. I configured the OPA189 with Gain at 3V/V, see the schematic below. 

OPA189 Vref 10012021.TSC

Also, since you increased the excitation voltage from 2.5V to 12.2935V, the current goes though each 350Ω is increased with power dissipation in each resistor at approx. 0.11W or (350Ω* (12.2935V/700Ω)^2). If you are able to mount the Wheatstone bridge circuit together on a thermal conductive metal, then the temperature changes in these resistors will sync together. 

My DAQ is expecting a 0-5V signal, so I'll need to bias the instrumentation amp's output to 2.5V (but, correct me if I'm wrong, I can use a standard voltage reference e.g. LM4030, right?).

Let me know which instrumentation amplifier you are going to use. We can figure it out if you need 2.5Vref for the instrumentation amplifier. Ideally, you want to use low voltage IA or up to 5.5Vdc IA for your ADC. 

Vref node in IA is a low impedance and the reference voltage is required to be source and sink in characteristics. I believe that LM4030 can only source a current, thus you are unable to use it for the application. Some of our reference voltage IC is able to source or sink, then you are able to use it (e.g. REFXXXX series). 

In Figure 33 in INA333's datasheet, you will be able to find this reference circuit. You do not have to use current source as presented in the schematic. You can use LM4030-2.5 reference voltage and configure an op amp to drive IA's Vref node, where the output of OPA333 op amp buffer is configured as low impedance node, and able to sink or source current at its output.  

Do you have any recommendations for a 5V LDO with an input range up to 24V?

Below is a search link about our fixed 5Vdc LDO. I do not support these products. If you want to have a recommendation, I would submit a request via our DC/DC or LDO support team. 

To step down from 24Vdc to 5Vdc, you need to consider the heat dissipation in LDO or linear regulator. There is 19V drop across a LDO's output to input. If LDO's operating current is high (e.g. 19V*100mA = 1.9W), you need to take care of heat dissipation in the converter (the voltage conversion efficiency in LDO or linear regulator is low).  

https://www.ti.com/power-management/linear-regulators-ldo/products.html?keyMatch=5V%20PRECISION%20LDO#p238min=0;24&p238max=24;48&p451max=0.5;0.75&p182=Fixed%20Output&p1154=5;5&sort=p634min;desc

If you have additional questions, please let me know. 

Best,

Raymond

Raymond--

Sorry for the delay in response--am still waiting from guidance from my client on how they'd like to proceed.  Thanks for your assistance with troubleshooting and design--much appreciated.

Sandheep

Hi Raymond--

I'm back to working on this, and am having issues with the INA125 again.  This time around, I'm using external excitation for the bridge (15V) and have cleaned up the 5V supply.  I'm still using the 2.5V reference to bias the reference to the middle of the output.  Below is my schematic (note that the 51ohm Rg has been replaced with a 10k pot):

The issue I'm having is that the output (10/110 signal is stuck in the 4+V range.  With no load on the strain gauge (and bridge differential voltage of 0V as measured by my DMM), I get 4.418-4.427V output (relative to ground) with Rg ranging from 0 to 10k respectively.  Loading the DUT does not change the output, but I'm able to measure a bridge differential voltage up to 6mV with a DMM. 

Troubleshooting I've done:

• I've ran the potentimeter up and down each way--only getting a few mV of change in output. 
• reversed the bridge inputs to the INA125 (e.g. swapped pins 6/7)--no change in output.
• changed out INA125 chips
• confirmed expected voltages at all the pins/locations:
  • 1/2: 5.1V
  • 3: 0V
  • 4/5/14: 2.493V
  • 6: 7.5V
  • 7: 7.5V
  • 8: 4.139V
  • 9: 4.139V (Rg=0) to 4.234 (Rg=10k)
  • 10/11: 4.427V
  • 12: 0V
  • 13: 1.23V
  • 15:2.482V (clipped since V+ is only 5.1V)
  • 16: 2.462V (clipped since V+ is only 5.1V)

Any other ideas as to what's going wrong here?

Thanks!

Sandheep

Argh, sorry, I figured it out--the Vin+ and Vin- need to be less than V+.  Works now, thanks!

Hi Raymond--

Getting closer here.  My test strain gauge setup under zero load condition has a differential output of around 7mV.  Looking through the forums (e.g. this post) and the XTR106 datasheet, I see that the best way to "zero" or balance the bridge is to use a circuit like this (from the XTR106 datasheet):

Where R1 = Vbridge * Rbridge / (4 * Vtrim) and R2<=R1.  Some questions:

1. Does the tolerance on the R2 trimmer matter? 
2. What would a reasonable temp coefficient be?
3. Can you confirm that the equation above is correct?  In particular, Vbridge is the voltage across the bridge, yes?
4. I'm using a 3 wire strain gauge (to cancel out the effects of the lead wire resistance).  Can you advise where the other lead of R7 should connect in the schematic below?  To point A or point B?  Or is this bridge balancing scheme even compatible with a 3 wire strain gauge?

Thank you as always, Raymond!

Sandheep

Raymond--

Also, FYI...the INA125 output is getting fed into a 0-10V DAQ input, so I'm using pin 15 (Vref5) to center the output at 5V.  I'll adjust the gain to ensure that the min/max signal is within the headroom of 0-15V.  Do you see any issues with this approach?

Thanks!

Sandheep

Raymond--

Sorry--one more question.  I prototyped the circuit with R1 (or R7 in the case of my schematic) connecting to point A--this seems to be working well with regards to balancing the bridge.  However, the output seems to be drifting--and not stabilizing:

Both R5 and R8 in my schematic are 10K 0.5W 25 turn cermet trimmer pots (Bourns PV36W103C01B00).  These pots have a temp coefficient of +/-100ppm/C, so I could at least replace R8 (the gain resistor) with a fixed resistor with lower temperature coefficient.  But 2 hours after letting the circuit/gauge stabilize, and it's still climbing, so I'm not sure if this is the culprit alone.

Any other ideas where the drift is coming from?

Thanks!

Sandheep

Hi Sandheep,

Let us get back to your rest of questions on Monday. 

My test strain gauge setup under zero load condition has a differential output of around 7mV. 

Your gain is 4 + 60kΩ/51Ω = 1180.47 V/V. With Vos of 50uV (RTI - refer to input), the Vos at output (RTO) will be 50uV*1180.47 =  60mV (this may not be your current gain settings)

Anyway, INA125 has Vos. Do you know the 7mV is resulted from strain gauge's imbalanced bridge or INA125's Vos?

What is your strain gauge sensing application?

If the application is used for weight or balance measurement, you can zero out the 7mV at the output. In balance, tare weight or the key on the balance is used for the specific purpose. Once Vos is removed, the remaining output signal will be linearly proportional to the input weight. 

Best,

Raymond 

Raymond--yes indeed--it's Saturday, please do take a break!  (but thanks for the prompt response)

My comment about the 7mV was unclear--if I measure the difference on the bridge with a DMM, I get about 7mV (might be -7mV).  That 7mV is the input signal into the INA125.  I'll measure the gain pot when I get to the office, but I think the gain is around 1000 (I could probably do with a little less gain), so it's been hard to zero/tare on the output with an initial output of ~7V. 

When you get to this on Monday, can you advise on how to tell whether the 7mV is due to the unbalanced bridge or the initial input offset voltage? 

FYI...the drift has increased further:

I'll troubleshoot today by replacing the gain pot with a fixed, high accuracy, low tempco (if I can find one) resistor, and then I'll monitor the voltages at pin 1,5, and 10/11 to see if I can spot the source of the drift.

Thanks!

Sandheep

Hi Sandheep, 

Does the tolerance on the R2 trimmer matter? What would a reasonable temp coefficient be? Can you confirm that the equation above is correct?  In particular, Vbridge is the voltage across the bridge, yes?

If R2 is potentiometer, the tolerances are not critical, but the resolution of wiper is important. As Kai's plot is shown, near the mid point of the potentiometer will be the best location. 

It is difficult to say what temperature coefficient should be. It will be the best if the temperature coefficient of R2 is close to temperature coefficient of strain gauge, so that changes in resistances will be tracked together vs. temperature. 

On the plot below, the null point or 0.00mV indicates the balance of the strain gauge. You may check it the Wheatstone balance over temperature as well.  

Any other ideas where the drift is coming from?

Per your specification, Gain = 1000, Vref = 5V , single supply voltage at 15Vdc:

There are three places that drifts may come from. 

1. Drift or noises from strain gauge. If the bridge is balanced and you are operating with a battery or low noise power supply, then cause of the drift may be low in this scenario  (operating in low noise environment). 

2. 5Vref voltage. Please make sure that 5V reference voltage is able to source and sink as shown in the circuit below. The 5Vref needs to be stable with low impedance node, and has to be able to source and sink such as op amp output. You could use source and sink from a reference voltage IC, but OPA333 configuration will be good as well and low in cost. You do not have to use current source as shown in Figure 33. For instance, you may use REF3312 or similar temperature dependent voltage references with OPA333 or other op amp configurations.  

3. 3rd possibility is the voltage offset in INA125 part. Since gain=1000, a small input change can cause large change at INA125's output. Say Vout = 5.146V, which is equal to Vout = Gain*(Vin+ - Vin-) + Vref. INA120U has the Vos up to ±250uV (RTI or refer-to-input) , and with Gain=1000, the Vout may see the offset voltage up to ±250uV*1000 = 0.25V (RTO, refer-to-output), where Vout will measure Vout = Vref + 0.25V= 5.25V (worst case). This fixed offset can be readily removed or zero-out from the application. There are additional temperature drift associated with INA125, but it is fairly low. 

I assumed that the your drift measurement is measured at room temperature. If this is the case, I would monitor the 5Vref voltage over time and make sure that it is stable. The Vos from INA125 is possible, but it should not drift significantly as data is shown. 

I will get your remaining questions in a separate reply. 

Best,

Raymond

Hi Sandheep, 

Can you advise where the other lead of R7 should connect in the schematic below?  To point A or point B? 

Assumed 350Ω resistors are matched; in theory, strain gauge should have nominal resistance of 350Ω. Please try this. 

1. R7 is connected Point E

2. Lead C and Lead D (point E to pin7)  length and wire gauge are the identical

3. Matching R3 (point E to R3) and R2 placement and wire length as well.   

5. Strain Gauge from point E to point A to pin7 should be identical to Lead C as indicated in line 2.  

If you do that, you may require to balance the strain gauge (I do not like to use potentiometer in commercial product because it will change over time, but it also has its advantage as well). 

Below is an image if the Wheatstone bridge is balanced:

I have a feeling that INA125's gain is too high. Your strain gauge should have a resistance specification and a sensitivity of X mV/V, which a weight transducer produces an output signal of X mV at nominal (rated) weight per Volt. With 15V excitation voltage, the input of the strain gauge should generate 15V*(X mV/V) or 15*X mV max. force per full weight scale in your application. Please make sure your Gain of 1000 is applicable for your strain gauge's specification. 

Normally, you only require to do one thing: either increase the circuit Gain or increase the excitation voltage. I am not sure that your circuit will require to do both. If you are not sure, please send me the strain gauge part number and I will simulate the linear relationship for you. 

Note: compression force (e.g.) weight will decrease in strain gauge's resistance value. Tensile force will increase the strain gauge's resistance.  

Enclosed in the INA125 simulation, please check it out.  

INA125 Strain Gauge 11152021.TSC

Please also verify that the 15V excitation voltage to the Wheatstone is stable over temperature and time. 

Best,

Raymond

Raymond--

Thanks for the response.  A few notes:

• My VPN went down over the weekend, so I was not actually able to monitor the voltages (Vref5 and Vout), but it looks like Vref is staying solid at 4.993V.  Vout drifted from 5.007V to 5.031V, where I think it has stabilized.  I'll give it another 12 hours of monitoring and we'll see.
• The system is running off of a 24V power supply, which I'm using a precision regulator to take down to 15V for the amplifier circuit.  So no issues with battery voltage dropping.  The 15V is outside the range of my DAQ analog input, but I'll do a voltage divider and monitor the 15V indirectly.
• The 5V ref is coming directly from the INA125, which I previously mentioned is super steady (and low noise).
• I'll see if I can monitor the ambient temperature as well.
• Though I am using a strain gauge, my objective is not actually measuring strain.  I have some equipment of an odd shape that has loads applied to it.  The equipment is such that I cannot install a load cell, so I've mounted a strain gauge on the part of the equipment that will see the greatest strain.  I'm then doing a 2 point "calibration" (one with no load, one with a known load) to determine the load applied to the equipment.  This approach is working well--I'm able to resolve down to 20gf of load, but the drift is throwing results off. 
• I was unclear about how the 3 wire strain gauge is connected to the bridge.  Please see the updated drawing below.  For that reason, I cannot attach R7 to point E from your drawing.

So...wish me luck with the drift--hopefully all will be good by tomorrow.

Thanks!

Sandheep

Hi Sandheep,

The 5V ref is coming directly from the INA125, which I previously mentioned is super steady (and low noise).

Yes, the issue is that Vref5 port is only designed to source the current, but IAref or pin5 node is required to be able to source and sink current. If the application is required to have fast settling time, then you may need to buffer an op amp, such as OPA187, a zero drift op amp over time and it will be very stable, see the captured image below. 

 I'm then doing a 2 point "calibration" (one with no load, one with a known load) to determine the load applied to the equipment. 

This is a minimum requirements to guarantee the quality of the acquired data. No load point calibration can eliminate the Vos from INA125; full load will verify the known loaded calibration. The strain gauge should perform linearly between the two points. 

I was unclear about how the 3 wire strain gauge is connected to the bridge.  Please see the updated drawing below.  For that reason, I cannot attach R7 to point E from your drawing.

If you want to configure the strain gauge in the following schematic, it will be ok as well. 

You need to make sure that following resistance has the same ratios as R2/R3,=1 which is on the right

On the left side of the Wheatstone bridge R_L1 + R_G = R_L2 + R4, where R_L1 = R_L2 in wire length and gauge wire. If the Wheatstone bridge is balanced, ( R_L1 + R_G) / ( R_L2 + R4) =1. 

R_L3 leads does not play a role for Wheatstone bridge balance, but I will make sure that the leads between two green marked box have the same resistance as well, which it will eliminate the differential mode signal into the INA125's inputs. 

If the strain gauge's operating environment is noisy, you may consider to implement differential and common mode filters in front of INA125. As it is simulated, the differential and common mode LPF are configured approx. 800Hz. 

Note: If the differential mode filter is implemented, then matching lead resistance in two green marked box is not critical any longer. A few Ohm over 10kΩ generates very little errors. 

INA125 Strain Gauge 11162021.TSC

If you have additional questions, please let me know. 

Best,

Raymond

Raymond--

I think the drift is temperature-related:

Note the voltages and temperature are normalized to their respective averages.  I've got the circuit somewhat enclosed (to reduce effects from circulating air), but the AC/heater must have kicked on around data # 1750.  You can see that Vsupply and Vref are steady with minimal noise.  Vout and temperature change a good amount.  Let's look at how they are related to each other:

So...I'll do the following:

1. use gain and zero resistors with the lowest tempco I can find.
2. locate these resistors as far away from the bridge resistors (which seem to generate some heat)
3. reduce the bridge voltage from 15V to 10V to reduce the heat generated by ~45%
4. use a thermal camera to identify other warm components--and keep the gain and zero resistors away

Please let me know if you have any other recommendations.

Thanks!

Sandheep

Hi Sandheep,

Yes, your resistor may have self heating issues. With 15V/700Ω = 0.02143A or 21.43mA and heat dissipation on 350Ω is 0.161W. ((0.02143^2)*350), which is a lot for thin film or laminated type of strain gauge. 

With 10V excitation voltage, the stain gauge and resistors are dissipating approx. 0.071W.

You may short out the INA125's input and see if the INA125's circuit is stable over time and temperature. If it does, then the issues are likely at the sensor's front end. 

Please let me know the latest results and I hope that you are able to resolve the drift issues. 

Best,

Raymond

Raymond--

The strain gauge that I'm using is a Micro Measurements (VPG) MMF404148 ( seems to also be known as the C4A-13-235SL-350-39P).  Per suggestion by VPG's application notes, 15V is within a tolerable range, especially since the strain gauge is mounted to an aluminum component with decent thermal mass.

Thanks!

Sandheep

Hi Sandheep,

Ok, I see, If the transducer is adhere to the surface of sensing object, then it should not have heating issues. As you indicated, 350Ω balancing resistors should have low temperature temperature coefficient so the Wheatstone bridge is kept balanced over time and temperature.  

Enclosed is VPG tech note about pros and cons of 350Ω vs. 1000Ω transducer, see in gauge resistance in p. 8 of the article. 

http://www.vishaypg.com/docs/11055/tn505.pdf

If you have additional questions, please let me know. 

Best,

Raymond

Raymond--

Yeah, I would have liked to have gone with 1000ohm transducers, but the size/shape of the DUT limited my options significantly.  I've been running tests to identify which component in the circuit is causing this negative temperature-based drift.  The DUT/gauge combo, zeroing resistor (R7) showed increases in Vout with temperature, so those aren't the culprits. 

The gain resistor had a negative effect on Vout with increasing temperature, but it has a ±5ppm/°C tempco, so I don't see how it could result in such variations in Vout (assume a 5 deg C increase, that would be a change in resistance from 100ohms to 100.0025 ohms, or a gain change from 604 to 603.985 (0.0025%)--that doesn't explain that ~15mV swing I'm seeing over the same temp difference.

SO...my question is: what parameters on the INA125 should I be looking at to understand temperature sensitivity? 

Thanks!

Sandheep

Raymond--

I was wondering if perhaps the INA125's offset drift is the issue here...but I'm not sure.  The spec sheet says 2uV/C max--and with my gain of 604, that would be 1.2mV/C, but my most recent measurements are consistently showing more like 3.6mV/C. 

To confirm what you had said previously, I could measure the input drift by shorting pins 6&7 and seeing what Vout looks like as a function of temperature, correct?

Also, if it is indeed the offset drift that is the problem here, then I need to find a new amplifier.  The TI/Burr-Brown selection guide shows the INA118 as a good option--Vsupply will accommodate the 15V I already have, low Vos, low offset drift, low noise.  I'll couple that with your 12V bridge voltage circuit plus a LM4030-5 and op-amp for the Vref.  Regarding zeroing, I think I'll stick with Kai's zero adjustment circuit.

Actually, it looks like the INA818 replaces the INA118--any concerns or thoughts on either?

Thanks!

Sandheep

Hi Sandheep,

Would be a change in resistance from 100ohms to 100.0025 ohms, or a gain change from 604 to 603.985 (0.0025%)--that doesn't explain that ~15mV swing I'm seeing over the same temp difference.

Where is the 100ohms is placed within the circuit. Then I can simulate and identify the issues. 

The placement of strain gauge is very important. And this determines if the differential input signal is positive or negative, where ΔV_differetial_input = Vin+ - Vin- pins. The total transfer function is Vout =  Gain*(Vin+ - Vin-) + Vref, where Gain = 604 V/V and Vref = 5.00V per your setup. 

If (Vin+ - Vin-) > 0mV, then the output will increase as (Vin+ - Vin-) differential signal increases. Otherwise, Vout will decrease if (Vin+ - Vin-) differential signal decreases, if (Vin+ - Vin-) < 0 mV.

The gain of 604 should not change significantly with temperature. As you indicated, it is changed 25ppm or 0.0025%at the output. But I do not know where the 100Ohm is placed in the circuit. Your strain gauge's resistance is 350ohm. 

To confirm what you had said previously, I could measure the input drift by shorting pins 6&7 and seeing what Vout looks like as a function of temperature, correct?

This is correct. I'd like to separate if the ~15 mV issues are from INA125 or your input stage. If I know that, I can troubleshoot more effectively. Please describe the issues and let me know where _15mV is came from. Yes, INA125 will have small offset voltage, but it is near a constant over time and temperature, where you can zero it out when you perform a no load calibration.

If you are not sure, please send me the scope shot, which it will be helpful. 

Best,

Raymond