This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TLV9052: Output @ + rail voltage on simple differential amplifier circuit

Part Number: TLV9052
Other Parts Discussed in Thread: LM10, TLV9002, TINA-TI

Hello, 

I have built a very simple differential amplifier with the TLV9052.  The inputs are 220K Ohm and the two feedback resistors are 470k ohm.  The inputs connect to two metal probes that are 12mm apart from each other and not connected.  They are in a magnetic field that a conductive fluid passed through.  The fluid contacts the probes and a current conducts across between the two.  The measured output voltage is in the 50 mv up to 200 mv range.  We get some low frequency baseline wandering of this signal for reasons not quite understood at this point.  But I digress.

We were fairly successful getting a good signal with the LM10 op amp from this arrangement.  Now that I have the TLV9052IDR in hand and the circuit assembled, when I power it up( 1.8VDC  - pin 8 at 1.8VDC, pin 4 at ground), the output is 1.8VDC.  To make sure the IC isn't completely damaged, I have used the same chip and built a simple inverting amplifier with the same pins and it works fine). 

I've tripled checked resistor values  and connections with an ohm meter and they are all good. 

What could be causing this issue?

  • Hi Walter,

    can you show a schematic?

    The TLV9052 is very fast and could oscillate:

    walter_tlv9052.TSC

    The ringing in the transient response and the peak in the frequency response clearly tell that the circuit is not stable.

    I would mount a small NP0 ceramic capacitor in parallel to the both 470k resistors:

    C1 provides a so called "phase lead compensation" and restores the phase margin again. C2 helps to keep the symmetry of differential amplifier.

    And I would add a 100R isolation resistor. Mount it close to the output of OPAmp.

    Keep in mind that you should not directly touch any signal pin of OPAmp with the scope probe. The unavoidable probe capacitance can destabilize the circuit. Always insert a suited isolation resistor.

    If the circuit is still unstable, take a slower OPAmp, like the TLV9002 for instance, as already suggested by Ron. But still use the phase lead compensation technique shown above.

    Kai

  • I can't seem to get an image to insert.  What's the trick?

  • Ok, I figured out how to load an image! (Why don't they say drag and drop in the first place? :)   )

    I sketched it on paper to provide this to you quickly.  

    The little square box with arrows pointing to it is my feeble attempt to represent the source of the input voltage.  This is described in the initial post.

    Unfortunately, I'll have to order 10pf caps which will take several days.  We have 100pf in stock  How did you arrive at 10pf?

    I would prefer to use the TLV9002 that Ron suggested but I could not get it in a package that we can handle here in our R&D lab.  The TLV9052 was available and seemed to be similar enough.

    The one major difference I see between our schematics is how we have power and ground connected.  I have pin 4 at ground.  I thought this would be ok since the input signals are not going below zero.

  • Hi Walter,

    Unfortunately, I'll have to order 10pf caps which will take several days.  We have 100pf in stock  How did you arrive at 10pf?

    By experience Relaxed With this sort of OPAmp a small phase lead capacitance will usually do the trick. It's sufficiently big to compensate the small input capacitance of TLV9052 causing -together with the 470k resistor- the unwanted phase lag, but at the same time not too big to result in too much capacitive loading of output.

    100pF will also work but will be slightly less efficient. The problem is that the unknown impedance of your "box" will change the situation and it cannot be said with full certainty whether the phase lead compensation will work es expected.

    The one major difference I see between our schematics is how we have power and ground connected.

    The dual supply voltage was mainly chosen to simplify the simulation. It also helps to make the output stage of OPAmp work in its linear operating range.

    I have pin 4 at ground.  I thought this would be ok since the input signals are not going below zero.

    It's true that your input signal doesn't seem to violate the common mode input voltage range of TLV9052, but with zero input signal the output of OPAmp very probably goes into saturation. This can cause stability issues. Need not, but can. It would eventually be better, if you would shift the ground terminal of lower 470k by a small potential (also called "pseudo ground") to allow the output to leave the saturation. Adding a small pseudo ground will do the same like using a dual supply voltage.

    Keep in mind that the LM10 and the TLV9052 are totally different amplifiers and what may work with the one OPAmp need not necessarily to work with the other OPAmp, and vice versa. You may have to "invent the wheel" a second time...

    Kai

  • So, I was able to switch to the TLV9002 today and have the same problem.   I tried 100 pf caps, and it did not make a difference.  

    I thought I understood the pseudo ground but did some reading on it, and basically, I just used two 220k resistors to create a voltage divider from positive supply to supply ground.  Then connected pin 4 to the supply ground and the 470K resistor from pin 2 to the center of the two 220k resistors.   It didn't make a difference.  Did I do this right?  

  • Also, I may have given you the wrong impression with the phrase "50mv up to 200mv" range.  Those are examples of the nominal voltage of the inputs from the probe and not the output of the amplifier circuit.  Although there is a little noise on the amplifier output, it's in the 2-4mv range.  I would not call it oscillation from the amplifier, although I guess it could be.

  • Hi Walter,

    it's difficult to help without knowing the whole setup. Let me tell you what goes through my mind right now:

    1. Performing a DC measurement in a fluid can be tricky because chemical reactions at the electrodes induced by the potential differences can take place decreasing or increasing the measured voltages between the two electrodes. Or by other words, chemical cells can be formed at the electrodes influencing the measurement.

    2. Another issue is the generation of passivation layers on the electrodes making the current between the two plates more and more decrease after some seconds or minutes.

    Beacuse of these issues such measurements are often carried out by using an AC voltage.

    3. It's not clear to me whether you measure a voltage (potential difference) or a resistance. For a plain voltage measurement no measuring current would be necessary. But when you perform a resistance measurement you will need a measuring current. The LM10 as being a bipolar OPAmp generates a input bias current and an input offset current which can help to perform a resistance measurement. But the TLV9052 or TLV9002 as being CMOS OPAmps do not generate these currents in the same height. And if your setup needs these current to flow, the circuit will no longer work with the CMOS OPAmps.

    4. I don't know how the fluid and the circuit are electrically connected to each other. From your schematic the only connection seems to be the contact in the fluid. Or by other words there doesn't seem to be any electrical connection between the signal ground of circuit and the fluid elsewhere in your setup. But if such a connection still exists, an equalization current may flow through the OPAmp circuit which can have an unwanted impact on your measurement.

    5. This brings me to my next idea. Why do you use a differential amplifier? A differential amplifier is normally used when a differential signal is superimposed by an unwanted common mode signal. The differential amplifier can then suppress the common mode signal to a high degree, provided the whole circuit maintains a high symmetry and balance at both inputs. But when the unwanted common mode input signal is different at the both electrodes, the common mode signal cannot be fully suppressed by a differential amplifier. In this case an instrumentation amplifier would be a better choice.

    6. On the other hand, if no unwanted common mode signal is present at all and no other hidden electrical connection exists between the fluid and the OPAmp circuit, a standard non-inverting amplifier could to the job as well. 60Hz hum can be suppressed by shielding. Have you experimented with a non-inverting amplifier so far?

    7. If you have such an unwanted coupling and unwanted equalization currents, you could think about providing some isolation. You could directly connect the measuring circuit to the fluid via the electrodes but provide isolation to the following circuitry, if another electrical connection to the fluid exists there.

    I guess I know what you are thinking right now. Why all this hassle when everythig worked fine with the LM10? Relaxed

    Kai

  • Kai - thanks for the response.  I have tried to answer and provide more information interspersed with you response below.

    it's difficult to help without knowing the whole setup. Let me tell you what goes through my mind right now:

    ( I understand completely!  I am so grateful for the extra thoughts and questions!)

    1. Performing a DC measurement in a fluid can be tricky because chemical reactions at the electrodes induced by the potential differences can take place decreasing or increasing the measured voltages between the two electrodes. Or by other words, chemical cells can be formed at the electrodes influencing the measurement.

    We definitely understand this and know we have to study some of this to find the right materials for the electrodes.  We think we'll need to use an electromagnet for the magnetic field vs. permanent magnets to change the magnetic field direction periodically to prevent many problems with the electrodes.

    2. Another issue is the generation of passivation layers on the electrodes making the current between the two plates more and more decrease after some seconds or minutes.

    We've already seen some of this and are looking for alternative materials for the electrodes.

    Beacuse of these issues such measurements are often carried out by using an AC voltage.

    3. It's not clear to me whether you measure a voltage (potential difference) or a resistance. For a plain voltage measurement no measuring current would be necessary. But when you perform a resistance measurement you will need a measuring current. The LM10 as being a bipolar OPAmp generates a input bias current and an input offset current which can help to perform a resistance measurement. But the TLV9052 or TLV9002 as being CMOS OPAmps do not generate these currents in the same height. And if your setup needs these current to flow, the circuit will no longer work with the CMOS OPAmps.

    The concept is based on Faraday's Law.  We are measuring the potential between the two electrodes.   I have thought of but not experimented with measuring current or resistance instead.   When we got really good waveforms just measuring the potential from the two electrodes with a differential amp based on the LM10, we just believed this was the solution.

    4. I don't know how the fluid and the circuit are electrically connected to each other. From your schematic the only connection seems to be the contact in the fluid. Or by other words there doesn't seem to be any electrical connection between the signal ground of circuit and the fluid elsewhere in your setup. But if such a connection still exists, an equalization current may flow through the OPAmp circuit which can have an unwanted impact on your measurement.

    The fluid is electrically isolated completely from circuits including pumps, valves and motors.   Since we have DC pumps and solenoid valves we are definitely always on the lookout for noise generating by these onto the power supplies.  But we don't have this occurring.

    5. This brings me to my next idea. Why do you use a differential amplifier? A differential amplifier is normally used when a differential signal is superimposed by an unwanted common mode signal. The differential amplifier can then suppress the common mode signal to a high degree, provided the whole circuit maintains a high symmetry and balance at both inputs. But when the unwanted common mode input signal is different at the both electrodes, the common mode signal cannot be fully suppressed by a differential amplifier. In this case an instrumentation amplifier would be a better choice.

    Good question!  Frankly, I used to live and breath this stuff but that was nearly 30 years ago!  I have been getting back in to this in our startup and I am very rusty in many aspects.  But it does come back!   I thought that the leads being in the same magnetic field but not connected electrically might pick up common mode noise from the magnetic field eddy currents (initially we used a DIY electromagnet and the windings were not exactly perfect!). 

    Can you post an instrumentation amp that you'd suggest that I try please?

    6. On the other hand, if no unwanted common mode signal is present at all and no other hidden electrical connection exists between the fluid and the OPAmp circuit, a standard non-inverting amplifier could to the job as well. 60Hz hum can be suppressed by shielding. Have you experimented with a non-inverting amplifier so far?

    I have not experimented with a non-inverting amplifier for this problem.  Wouldn't this require grounding one of the electrodes?  At the really small currents being generated, I did not think this would be feasible.  Initially the input resistors on the differential amplifier were 10K then 47k then 100k until I did the math and determined to use 220K (the closest we had in stock) and it started to work with the LM10.

    7. If you have such an unwanted coupling and unwanted equalization currents, you could think about providing some isolation. You could directly connect the measuring circuit to the fluid via the electrodes but provide isolation to the following circuitry, if another electrical connection to the fluid exists there.

    There really isn't another electrical connection to the fluid.

    I guess I know what you are thinking right now. Why all this hassle when everythig worked fine with the LM10? 

    I've wondered about this, but I really liked Ron's suggestion.  The LM10 has reference circuitry that I don't need.  And I have a second stage that is just a non-inverting amp using the LM10.  It took up a lot of space to have that reference stage in the LM10 that I wasn't using.   The TLV9002 looked like a very good alternative. I then ran in to trouble finding any TLV9002.  That's when Caroline got involved and suggested the TLV9052 and I was able to find those.  We are a small startup and have limited ability to handle SMD packages so it took a little doing to find the right parts in stock that would work with the DIP-ADAPTER- evaluation part but I finally go some in! 

    We seem so very close really!  

  • Hi Walter,

    if there's no electrical connection between the fluid (and everything being in electrical contact to the fluid) and your circuit, then you can connect signal ground of your circuit to any point you like. It's similar to using an isolation amplifier.

    (Datasheets on isolation amplifiers -just to demonstrate what I mean- you will find here:

    https://www.ti.com/isolation/isolated-amplifiers/products.html

    By this I don't want to recommend to use such an isolation amplifier, at least not at this moment Relaxed)

    A differential amplifier using a pseudo ground of 100mV could look like this:

    walter_tlv9052_1.TSC

    An instrumentation amplifier with balanced common mode impedances at both inputs could look like this:

    walter_tlv9052_2.TSC

    There are two pseudo grounds, one at middsupply to shift the common mode input voltage into the linear operating range of OPAmp and a second to prevent the output of U2 from going into negative saturation.

    Of course, the pseudo grounds can be modified and even only one pseudo ground may be used. And by using a dual supply voltage the pseudo grounds can even be omitted entirely.

    Another trick to be able to use a higher supply voltage than 1.8V is to use a voltage divider at the input of ADC or at the input of an additional buffer driving the ADC which is also supplied by the ADC's 1.8V.

    Even a negative suppy voltage could be used with this voltage divider method, eventually in combination with a diode clamp. I have discussed this technique in this thread:

    https://e2e.ti.com/support/amplifiers-group/amplifiers/f/amplifiers-forum/1097514/tlv9002-can-we-connect-ct-to-tlv9002

    There the purpose was to prevent latch-up or lock-op in the case the input voltage of an OPAmp exceeds its supply voltage. Maybe you find this discussion useful.

    Kai

  • Kai,

    I built the differential amplifier with the 100mv pseudo ground.  I don't have 10pf caps in stock.  I had to use 82pf ceramics for now.  Anyway, it works beautifully on the bench with a 1.8VDC supply.   (In case I have not mentioned this, our ADC has a maximum input of 1.8VDC, so TLV90xx's support this very well. )  I have not hooked the inputs to the actual probes yet because the software developer is using the system today.  I simulate the input from a second bench supply. 

    The output appears linear as the input goes from 10mv to 410mv.  It is very stable and has low noise.  Of course, the bench supply has a good low noise output.  The input at VG1 can get to 410mv before the circuit output at VF1 reaches about 1.79V - the maximum output with a 1.8V supply.   So the gain appears to be 4.39.  The voltage at pin 7 is about 880mv.  All the components on the U1 stage in your schematic are the same on the breadboard - R5: 170K, R6: 10K, C1 100nf.   The only components that are different are the 82pf caps. 

    How is the gain set in this situation?  I don't get 4.39 when any normal gain calculations for inverting or non-inverting amplifiers, even with the pseudo ground in place.  I would really like to be able to have a maximum of 750mv input at VG1.

    What type of capacitors should I use for the 10pf and 100nf caps?

    I SINCERELY appreciate your help and patience.  You've held my hand through this as I get back into analog design beyond the call of duty!.

  • Hey Walter, 

    I ran Kai's circuit that he attached above, its a little odd that with your circuit application the output voltage is at 1.78 V with an input of 410 mV. In simulation, it looks a little different. 

    Here is Kai's file again if you would like to run the simulations yourself, I feel like something else might be wrong... 0564.walter_tlv9052_1.TSC

    All the best,
    Carolina

  • Hi Walter,

    I agree with Caro. With the circuit the output signal should not clip with a 410mV input signal. Or do you have set R5=R6=100k? Then the circuit would clip at 410mV input signal:

    The gain of the circuit is 470k / 220k = 2.14. You can also estimate the gain from the DC sweep plot: (1.7V - 0.1V) / (750mV - 0V) = 2.13.

    For the 10pF cap I would take C0G (or NP0) ceramics. For the 100nF cap I would take X7R.

    Kai

  • In the original differential amp circuit posted above the pseudoground section has R5 = 170K and R6 = 100k.  I did not have a 170K resistor so I created equivalent resistance with two resistors in parallel.   The latest one posted has R5=R6 = 100K.  I'll change R5 to 100K and see what happens.   Maybe that's the problem.    

    I don't have experience with the simulation tools but I really want to learn them.  I'll download and install TINA so I can run these simulations.  Currently, I use Fusion 360 (formerly Eagle 9.x separately) because we received a grant for it.  It has SPICE functionality but I have not learned how to use it yet.

  • Hi Walter,

    my original circuit has R5=170k and R6=10k forming a voltage divider of division factor (170 + 10) / 10 = 18 to decrease the 1.8V supply voltage to a pseudoground potential of 1.8V / 18 = 100mV.

    100mV is sligthly above the output low saturation voltage of TLV9052 to allow the OPAmp to work in its linear operating range.

    The 170k resistor doesn't exist in the E-series. But this is not important. Just make R5 seventeen times bigger than R6.

    And if you don't like the 100mV pseudoground but want to make it higher, just choose different resistor values for R5 and R6.

    The simulation software I use is TINA-TI. It's free and can be downloaded here:

    https://www.ti.com/tool/TINA-TI

    Kai

  • Thanks,    I've downloaded TINA-TI and will learn how to use it.   

    I went over my circuit build on the breadboard.  I think it helps a lot when you connect pin 4 to ground instead of the row next to it!!!!  My mistake!  Now, it seems to take a maximum value for VG1 of 800mv to an output for VF1 of 1.80V.   This should work. I'm still using a bench supply to provide the 800mv until the software developer can give me the test platform long enough to test this with real probes in the fluid.  

    Many thanks to you,  and Ron!

    Is there a best way to share how much this help has benefitted me on social media or somewhere else?  I'd like to do that!

  • Hey Walter, 

    I marked "TI thinks Resolved" on the post I think was most useful to the issue. The best way to let us know we helped is to confirm by selecting "Resolved"

    Here is an FAQ on it: https://e2e.ti.com/support/site-support-group/site-support/f/site-support-forum/694649/faq-what-do-the-resolved-indicators-on-threads-mean-on-e2e-support-forums

    All the best,
    Carolina