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.

THS4631: THS4631 input current sharing (1/N) paralleling

Part Number: THS4631
Other Parts Discussed in Thread: THS3491, , OPA548


I always thanks for your great supports.

Back in the last thread, I solved problem to reduce the cable and improve Rf 30 to 800ohm. (Thanks Kai Slight smile)

But unfortunately, I can't use THS3491 in my circuit. Based on my application, I have to use 30ohm as a Rf.

and I have to use Cf to cut-off high frequency noise.

So I bought some THS4631 (Voltage Feedback Amp,) and now I found out it works I expected. (I can use 30ohm and also Cf)

But there is a problem now and I'd like to ask you the advices.

The current range we'd like to get is very big. (0~2A) 

and Frequency range is also very wide (20~1MHz with no delay (no phase delay at 1MHz))

So I tried to use THS3491 (which has 420mA output current) for my application,

But now I realized that Ths3491 can't use small Rf (like 30ohm) and also can't use Cf.

To use 3491, I tried to use big resistor inner roop, (800ohm) (please refer the circuit in previous thread.) It works, but when the load goes to high or the cable goes to long or the input of negative feedback is open,

The output voltage starts to ringing. I think inner roop Resistor makes system quite unstable. I don't know why.

So I think the best way now is use 20 parallel of THS4631 (or other Amps. please recommend)

like below.

Is it possible?

I checked that when I 2-parallel of 4631, the output voltage of each Amp. has different DC offset voltage. 

But it was small (500mV↓). and the sum of them was zero. So I think It can be ignored.

But I worried when the parallel OP-amp. are more, one of them has very big offset voltage.

anyway, I think it looks weird too. 

But I don't have other idea to meet my system requirement.Sweat smile

I want to ask you that, is it possible? and Is there an any good idea to get 0~2A current without any phase delay from 0~1MHz?

I can't add any parts in negative input line.

Actually, Money is no object in this project.


  • Hello Wonwoo,

    Just to confirm I understand your post: you cannot proceed with a previously advised design solution due to a requirement to keep Rf set to 30 Ohms, correct?

    Are you looking for any viable solution to achieve high output current at low frequency?  I am a bit confused on what aspects of your design are absolutely set and cannot be changed, vs. preferences and design targets.

    If you could elaborate or clear up the details of your proposed design, I would appreciate any extra information.



  • Hello Alec Saebeler,

    Thank you for your concern.

    Yes, I can't use 3491 because it must need over 500~600ohm of Rf.

    The reason why I have to use 30~60ohm is,

    if the current flows to each opamp, (300mA(peak) each opamp.)

    the output voltage goes to 9V. (300mA * Rf(30ohm))

    If the Rf is 100ohm, the output voltage would 30V.

    I think +-15V is resonable. not +-30V.

    I checked that when the output voltage goes over 10V, the linearity cannot meet. so I think I have to use under 10V.

    If parallel is possible, The current of each Amp. would be lower and I can use more big Rf. 

    but I don't know 20 opamp. paralleling is really best solution. 

    anyway, the design target is,

    0~2Arms (AC signal no DC offset)

    and the frequency target is  20Hz ~1MHz (no any delay).

    The priority of those two target is frequency.


  • Hello Wonwoo,

    Thank you for the clarification Slight smile.  This week is busy for the team; I will bring up your post so we can all take a look at it.  Please allow some additional time due to the holiday & busy season.

    I will update you on any ideas or questions I may have.



  • Hello Alec Saebeler,

    Thank you for that.

    I share you all of my circuit design.

    please refer it.

    the purpose of this circuit is, sensing the current which flows to Cap. (or resistor or Inductor..)


    and I'd like to ask one more thing,

    I said when I use 2-parallel of 4631, Offset voltage of each amp. appeared which is smaller than 500mV,  

    So I also said that it could be ignored. 

    but today, I tested 2-parallel of 4051, the offset voltage of each amp. goes up to 5V.

    and when I applied AC source, the output voltage shape of each Amp. is like half wave rectifier.

    I think, shape like 4631 is right waveform. 

    If possible, I'd like to know the reason why there is offset voltage when paralleling and How to reduce.

    and I want to listen your comment. 

    I really thanks for listening to my woes.

    Have a good holiday!

  • Hello Wonwoo,

    Thank you for the extra details!  I appreciate the well wishes for the holiday Slight smile.



  • Hi Wonwoo,

    unfortunately, you cannot connect TIAs with their inputs in parallel. Because the -input of TIA is on virtual ground, the input current doesn't know how to evenly distribute over the individual inputs of TIAs.

    Also, when you look at two paralleled TIAs without input signal, they look like an instrumentation amplifier with infinite gain. And even the smallest input offset voltage will make the outputs saturate:


    The green curve is "VF1" and the brown curve is "VF2".

    A technique which is sometimes used is to insert small current equalizing resistors at the inputs (R2 and R4):

    Actually, you no longer parallel TIAs but inverting amplifiers. Because of this, R2 and R4 must be very small, but at the same time not too small to cause stability issues. Such a circuit is always critical and I would not recommend it. In your special case of very high input current, R2 and R4 cannot be kept small enough, I fear.

    Another approach when the input current is so high that a standard TIA's output cannot absorb it, is to connect a very small shunt directly to signal ground and to amplify the voltage drop across it:

    The advantage of this method is that you get rid of any TIA stability issues Relaxed


  • Thank you for sharing it.

    I save time.

    If so, How about this idea?

    I'm finding another Amp. which is available.

    but I don't think there is a part which meets my requirement. (3Ampare / 1MHz no delay)

    and the best option I can do is OPA548.

    (3A / 1MHz (Gain-bandwidth Frequency))

    The current rating meets the requirement. but Frequency response would not.

    maybe there are some magnitude degradation and phase delay .

    but if the frequency response and phase delay is constant by frequency, 

    I can compensate it through software.

    But When this project is end, I have to massproduct this item.

    so It is super important the characteristic would be same by every opamp. (or the difference of it should be smaller than 1%)

    Could you check the magnitude and phase delay of each amp. would almost same or not?


  • Hi Wonwoo,

    I fear the OPA548 is too slow. It's a 1MHz OPAmp and if yo really want to measure up to 1MHz, the OPA548 would not have any linearizing gain reserve at 1MHz and would not be a good candidate. But, hey, give it try...

    Are you really wanting to measure at such high currents and such high frequencies? Can't you find a compromise?

    What about the method with the very small shunt? You could even carry out a correction of voltage drop with the help of µC, when you measure the signal coming from the function generator and the voltage drop across the shunt. If you have both the amplitudes and phases of these two signals, you can easily perform a correction. I would go this way.


  • Hi, Kai,

    Your words, "Give it try" is good encouragement for me. thanks.

    Yes I also considering small shunt like 100~300mohm.

    But my DUT condition is very harsh. 

    Actually, I don't know It even would possible or not. 

    because the DUT impedance at the measuring point is only 0.005ohm (100kHz, 330uF)

    I have to apply this DUT 10mV and get a 2Ampare feedback. 

    We have some source voltage control technique which is called 4-wire or Auto Level Control.

    But If I use shunt between DUT and ground, the feedback voltage for this source voltage controller, the shunt impedance should be added with DUT.

    So actually I wrote another thread regarding parallel operating of THS3491 as a voltage source.

    If It would possible, I'll try to use Rs instead of small shunt like you suggested.

    That's why I want to reduce Rs as possible as I can.

    I think that's key point of my project. If it isn't possible, I think my considerations in this thread also useless.

    anyway, thank you for your time and effort.  

  • Hi Wonwoo,

    I would drastically decrease the measuring current, down to a value that can easily be handled by a HF-OPAmp. Of course, the voltage drop across the impedance "Z" will decrease by this. But you can easily increase the signal gain behind the first stage (TIA method or shunt method). To overcome the negative influence of broadband noise, enhance the signal filtering by the help of low pass filtering or bandpass filtering (analogue or digital). The offset will not play any relevant role, when you use AC coupling.

    Another approach when the signal is buried in broadband noise is the lock-in method (synchronous demodulation).


  • Hi Kai.

    I will keep in mind your advice.

    I'd like to ask you one more thing.

    Because If the circuit gives DUT 2Ampare,

    This DUT have to get 2Ampare. I shouldn't modify it.

    because I have to check the magnitude and phase of it when this DUT get 2Ampares.

    That's the purpose of this project.

     So I actually want to verify to use them parallel some more even now.

    You said the problem is input bias voltage of another Amp. is amplified infinite gain to another Amp.

    I totally agree that.

    but If the input bias voltage is problem,

    Is it possible to tune the Rb resistor of each Amp. like below?   

    Considering this is not worth it?

    I worried that if it operate well, if it could damage to the parts longer period of time, 

    that's not what I wanted. that's why I ask you before doing something.

    I really thank you so much againHeart eyes

  • Hi Wonwoo,

    the equalizing resistors R2 and R4 would also alleviate this offset voltage issue:

    And then you have the AC coupling of signal which will remove these offset voltages.