Post Go back to editing

Stability with input capacitance

Category: Hardware
Product Number: ADA4084-4

Hello,

I plan to use the ADA4084 as a DAC buffer in a low noise application. The DAC is used to generate a DC biasing voltage, and the ADA4084 is used as a unity gain voltage follower.

I plan to filter the DAC noise by a using a very low frequency RC low pass filter between the DAC output and the op amp non inverting input. I will typically use 100kR with 100 nF.

My question is : can a capacitor at the non inverting input of the ADA4084 can cause unstability ? 

I know that a capacitance at the inverting input can cause unstabilty issue, but my intuition is that it not a concern for the non inverting output, as the capacitance is not in the feedback loop in that case.

Thank you,

Best regards,

Maxime Puech

  • Hi Maxime,

    it not a concern for the non inverting output, as the capacitance is not in the feedback loop in that case

    Exactly. A capacitor at the inverting input causes a phase shift in the feedback, therefore reduces the phase margin. The impadence seen by the non-inverting input can't affect the feedback loop.

    So you don't have to worry about stability.

    However, I think you do need to worry about the excessive noise introduced by the low pass filter. The thermal noise of the 100k resistor is about 40nV//√Hz. The opamp's current noise is 0.55 pA/√Hz. This flows through the 100k resistor of the low-pass filter, giving 55nV/√Hz voltage noise at the non-inverting input. These together mean about  68nV/√Hz, so more than ten times larger than the voltage noise of the amplifier. Furthermore, the current noise is specified at 1kHz, so it could be much higher at lower frequencies due to its 1/f nature. Although it is low-pass filtered, the noise from 0.1Hz to 10Hz will be much larger than the specified 0.14μV peak-to-peak value.

    You can use a 1k 10μF low-pass filter, but then connect a small series resistor between the capacitor and the non-inverting input to reduce the accidental current to less than 5mA (see absolute maximum ratings) when the opamp is turned off. In this case, the capacitor can be discharged through the input terminals of the opamp. This series resistor will add some thermal noise, but you can calculate the magnitude of it. Note that excessive input current could degrade the noise performace of an opamp.

    I do not know, how much noise can you tolerate, but consider the above to keep the requirements.

    I hope it helps.

    Zoltan

  • Hi Zoltan,

    Thanks a lot for your very detailed answer, it helps a lot. 

    About the noise concern you mentioned, I'd like to provide some context. You'll find below the noise density requirement we need to fulfil. 

    I understand that the 100k thermal noise plus the the op amp current noise will create a broadband noise of about 68 nV/sqrt(Hz), but this noise will be filtered by the 16 Hz filter (100k - 100nF). At 1kHz, this filter provides about -40 dB attenuation, which gives a filtered noise spectral density of 0.68 nV/sqrt(Hz) at 1 kHz. This is way below my requirement of 25 nV/sqrt(Hz).

    About the 1/f current noise, is there a way to estimate it? You say that it could be much larger than the specified 0.14uV peak-to-peak value, but what is the order of magnitude? As you can see above, my requirements for 1/f noise are not very strict (7 uV/sqrt(Hz) @ 1 Hz, 1 uV/sqrt(Hz) @ 10 Hz )

    I was not aware of this accidental current concern. Is it something specific to the ADA4084 or is it a general concern with op amps? My capacitor will be charged to 2.5V maximum. So if I use a 100 ohm series resistor, the maximum accitental current will be 2.5 mA, which is below the absolute maximum rating. If I use a smaller capacitor, is it still a concern? I am guessing this 5mA is the maximum DC current the input can withstand, but in the case of a small capacitor, the duration of the current peak will be very short and maybe not dangerous for the op amp?

    Best regards,

    Maxime Puech

  • Hi Maxime,

    I was not aware of this accidental current concern. Is it something specific to the ADA4084 or is it a general concern with op amps?

    It is a geneal phenomenon, all opamps have current noise. Here is a brief description about the opamp noise sources: https://www.analog.com/media/en/training-seminars/tutorials/MT-049.pdf, and another useful one: https://www.analog.com/media/en/technical-documentation/technical-articles/S64_EN-Op_Amps.pdf 

    Unfortunately, current nosie is rarely specified in as much detail as voltage noise. I think the reason is that the effect of current noise depends on the external components. So the message is: use the opamp in such conditions, when the current noise does not matter. See Figure 5 in this tutorial how to pick an opamp for different soure resistances: https://www.analog.com/media/en/training-seminars/tutorials/MT-047.pdf

    You are right, that the RC filter attenuates the noise due to the 100k resistor, that was the reason why I asked what are your requirements.

    About the 1/f current noise, is there a way to estimate it? You say that it could be much larger than the specified 0.14uV peak-to-peak value, but what is the order of magnitude? As you can see above, my requirements for 1/f noise are not very strict (7 uV/sqrt(Hz) @ 1 Hz, 1 uV/sqrt(Hz) @ 10 Hz )

    I could not find the spectral density of the current noise in the datasheet, I think the reason is what I've said above. So we do not even know what is the 1/f corner (it is typically higher than for the voltage noise). But we know the value at 1kHz. So, assuming, that the 1/f corner is less than 1kHz (it is very likely the case), then at 1Hz it probably has a value below 0.55 pA/√Hz * √(1000Hz/1Hz) = 17,4pA/√Hz. If it flows through 100k, then it means 1,74μ/√Hz. So it seems to meet your requirement.

    Zoltan

  • Hi Zotan,

    Thank you for the clarification about the current noise. I am more confident with my design now.

    One last question: I think you might have misread the latter of my previous message concerning the "accidental current". I was asking about the 5mA accidental current that might go into the op amp at switch off, not about the noise current. Here is my question again:

    I was not aware of this accidental current concern. Is it something specific to the ADA4084 or is it a general concern with op amps? My capacitor will be charged to 2.5V maximum. So if I use a 100 ohm series resistor, the maximum accitental current will be 2.5 mA, which is below the absolute maximum rating. If I use a smaller capacitor, is it still a concern? I am guessing this 5mA is the maximum DC current the input can withstand, but in the case of a small capacitor, the duration of the current peak will be very short and maybe not dangerous for the op amp?

    Maxime

  • Hi Maxime,

    Indeed, I missed this point. As a matter of fact, I can't recommend a good tutorial about this issue. I think, that for smaller capacitors it is not a problem, since the  discharge is faster, and semiconductors typically tolerate short excessive peaks. I think using a small series resistor is always a good practice, let us say, if the input capacitance exceeds 10nF. But I can only guess it.

    My capacitor will be charged to 2.5V maximum. So if I use a 100 ohm series resistor, the maximum accitental current will be 2.5 mA

    Be careful: 100 Ohm gives you 25mA. But 100 Ohm can be good, if you use Schottky diodes for input protection, see this tutorial: https://www.analog.com/media/en/training-seminars/tutorials/MT-036.pdf

    In summary, in your case I would use 10k/1uF or 1k/10uF for low pass filtering and would use an 500Ohm series resistor to protect the opamp input in any case. It will not add considerable noise, and guarantees staying within the absolute max ratings.

    Zoltan