Actions
Forum Thread Details
  • Replies 118 replies
  • Subscribers 313 subscribers
  • Views 14028 views
  • Users 0 members are here
Related
See a helpful answer?

Be sure to click 'more' and select 'suggest as answer'!

If you're the thread creator, be sure to click 'more' then 'Verify as Answer'!

Have a question about ADCs or DACs? Ask our Expert, Nick Gray

ChristyZ

This thread has been closed to new questions.

However, we welcome you to Post Your Question about Communications in the element14 Community Wireless Communications Technology group. You'll find many fellow members and experts who have just the answer you're looking to find! 

 

Thank You, Your Friends at element14 Community


Nick Gray

 

Nicholas Gray

Nicholas has worked in the Semiconductor industry for over 30 years and has authored a number of published articles about data converters (ADCs and DACs) and signal integrity issues.

 

Parents

  • Hi Nick,

     

    I've got a National LMP2011 opamp, non-inverting gain of 2, driving the input of a TI ADS7957SDBT.ADS7957SDBT. ADC.  Although the opamp has very low offset voltage and current, I'm seeing about 40 millivolts offset on the output of the opamp.  The offset disappears when not connected to the ADS7957.  What in the ADC input could cause such a significant offset on the opamp output? I've tried using very low resistance for the gain-setting resistors on the opamp, but that makes no difference. (This is one of those unfortunate situations where I've used two different manufacturers so they will tend to just blame each other.)

     

    Thanks,

     

    Barry Volain

  • in reply to Former Member
    Hello Barry,

    Unfortunately, the ADS7957SDBT.ADS7957SDBT. data sheet does not indicate the equivalent input circuit to the ADC. However, most SAR (Successive Approximation Register) ADCs have a sampling network at the input. The basic operation of this type of circuit would be to connect an internal capacitor to the input for maybe two or three clock cycles, then disconnect the input from the capacitor and connect the capacitor to the ADC itself and the. This is repeated for each sample converted.

    Each time the capacitor is reconnected to the input, the charge on it is different from what it was when previously connected and current is required from the source to charge that capacitor. Whether the ADC will tend to discharge the capacitor toward ground or charge it to the supply or some voltage in between depends entirely upon the converter itself and its input design. The result is that, when the capacitor is reconnected and the output of the op-amp is pulled away from its equilibrium point, after a small time delay it will react by trying to pull its output back to where the op-amp is again at equilibrium. But the capacitor is being charged to the op-amp output voltage and, by the time the op-amp reacts to its output being pulled away from equilibrium, its output is already at or near equilibrium and its reaction causes the op-amp output to overshoot where it wants to be.

    This typically continues for a anywhere from 1 to a few oscillation cycles, but before it can settle, the ADC input capacitor is disconnected from the ADC input, resulting in the ADC “capturing” a voltage that may be incorrect. At any rate, the DC voltage measured at the op-amp output may be either high or low and the amount of apparent error will depend upon the design of the ADC input, the ADC sample rate and the current capabilities and speed of the amplifier. Ironically, a slow amplifier would be better.

    This means that we need to compensate for the charge current required for the ADC input capacitor. If we put an additional capacitor at the ADC input that is about ten times the input capacitance of the ADC, most of the input current pulse required to charge the ADC input capacitor would come from that capacitor. However, most op-amps can not tolerate a capacitive load and will oscillate with a capacitive load. The answer here is to use a small series resistor between the op-amp output and the capacitor added at the ADC input. I find that this works extremely well.

    The ADS7957SDBT.ADS7957SDBT. data sheet shows its input capacitance to be 15pF, so an input capacitance of about 150pF seems to be indicated. Now, we want to maintain a reasonable time constant so that the ADC input can settle quickly enough to avoid linearity and distortion problems.

    The ADS7957SDBT.ADS7957SDBT. data sheet does say that the impedance of the driving source should be no more than 50 Ohms, but this is with the 15pF input capacitor. If we add another 150pF, we can tolerate more resistance because the input pulse is lower and voltage recovery is faster than without the additional capacitance. I would simply use a series resistor of about 75 Ohms. I would not be as concerned about the DC voltage at the ADC input as I would be about whether the circuit converts accurately. Because of circuit tolerances and gain errors, you probably will not get exactly the output you expect and I would be most concerned about circuit linearity. The best way to determine if the circuit is linear is to apply 5 to 6 different voltages to the amplifier input with each step being EXACTLY the same size. Record the digital output from the ADC for each case, then be sure that difference in the digital count of each conversion is separated from the next one by the same amount, within about 4 counts. You can improve the accuracy of this measurement process by taking 8 readings at each voltage input setting then using the average of the 8 readings.

    Please let me know how this works out.


    Nick Gray
  • in reply to nickgray

    Hi Nick,

     

    Thanks for the reply.  After looking at this a little closer, it appears that the problem is the settling time of the opamp.  Although I'm NOT overdriving the input or output, it seems to take it about 11 milliseconds to settle after a step input.  This is contrary to the mfrs spec of about 1.4 MICROseconds, but I've asked them to look into it.

     

    Barry

  • in reply to Former Member

    Hi, Barry -

     

    Yes, the LMP2011 has an output impedance that is too large for this application. The output impedance is not specified in the data sheet, but if you look at the output current for a given output voltage, you will see that the output impedance is about 600 Ohms when sourcing current and about 50 Ohms when sinking current. This is a common problem with low power operational amplifiers.

     

    The long settling time is caused by this high output impedance and the capacitance of the ADC input, as well as the current spiking that occurs at each sampling of the ADC. Op-Amp settling time is not specified for such a load. Unless you tell the manufacturer exactly how you are using the op-amp, they will probably not know why the settling time is so long. Also, if you are not in contact with a person knowledgeable about ADC loading of op-amps, you may not get a good answer.

     

    What you need is an op-amp with a low output impedance, which, unfortunately, may mean you need to use one of a higher power than the LMP2011. ("LMP" from National Semiconductor means low power.) While I have not looked at all of the amplifier possibilities, I do believe that the LMH6645, with an output impedance of less than one Ohm at frequencies below 1 MHz, will be a better amplifier. Its power consumption is also about the same as that of the LMP2011. The "LMH" means high speed, but don't let that stop you. The primary problem with high speed amplifiers in low frequency applications is that there might be too much noise in the system because of the added bandwidth. On the other hand, lower bandwidth products might not have the phase linearity you need.

     

    - Nicholas "Nick" Gray

  • in reply to nickgray

    Thanks Nick,

     

    Although this part does have a high output impedance as you point out, the input capacitance of the ADC is only 15 pF (typ) which gives a time constant of only 9 nS for 600 ohms; I'm not sure how this would translate to the 11 mS settling I'm seeing.  Looking at this another way, say I need a 4 volt step on the output.  According to  the data sheet, the LMP2011 can source about 9mA at 4V.  Thus,  i=C*dv/dt=>

         dt=15pF*4/.009=6.7nS.  That's a million times less than 11mS.

     

    And it doesn't look like a 'normal' settling time issue where you would see an overshoot and some damped ringing-the output stays 'stuck' at some constant level for 11mS, and then rises to the 'correct' value.  It looks more like what you'd see when an opamp recovers from saturation (the output is well below saturation-about 300mV with a 5V supply).

     

    Ironically, National says in their data sheet: "The LMP201x is a great choce for an amplifer stage immediately before the input fo an ADC."

     

    I will definitely look into alternate opamps.  Also, I think I can live with setting my sample rate lower so that I can accomodate the long settling time.  At this point I'm more concerned about WHY this didn't work as expected; in my experience, when you don't understand a problem but think you've found a fix, you haven't really.

     

    Barry

Reply
  • in reply to nickgray

    Thanks Nick,

     

    Although this part does have a high output impedance as you point out, the input capacitance of the ADC is only 15 pF (typ) which gives a time constant of only 9 nS for 600 ohms; I'm not sure how this would translate to the 11 mS settling I'm seeing.  Looking at this another way, say I need a 4 volt step on the output.  According to  the data sheet, the LMP2011 can source about 9mA at 4V.  Thus,  i=C*dv/dt=>

         dt=15pF*4/.009=6.7nS.  That's a million times less than 11mS.

     

    And it doesn't look like a 'normal' settling time issue where you would see an overshoot and some damped ringing-the output stays 'stuck' at some constant level for 11mS, and then rises to the 'correct' value.  It looks more like what you'd see when an opamp recovers from saturation (the output is well below saturation-about 300mV with a 5V supply).

     

    Ironically, National says in their data sheet: "The LMP201x is a great choce for an amplifer stage immediately before the input fo an ADC."

     

    I will definitely look into alternate opamps.  Also, I think I can live with setting my sample rate lower so that I can accomodate the long settling time.  At this point I'm more concerned about WHY this didn't work as expected; in my experience, when you don't understand a problem but think you've found a fix, you haven't really.

     

    Barry

Children
  • in reply to Former Member

    Barry -

     

    I phoned a friend of my in the amplifiers group of National Semiconductor and he told me that the LMP2011 has a lot of capacitance associated with it (it is a chopper-stabilized amplifier) and the overload recovery time is typically 50 ms. See data sheet pages 3 and 4 for overload recovery time specifications. The key to getting what you need is to insure that the part never goes into overload. That is, keep the device operating within its linear range. This implies ensuring that the input does not momentarily make a jump so large that the amplifier has difficulty following it, and being sure that the output never saturates. I suggest you may need an amplifier that is NOT chopper stabilized, but such an amplifier would have a larger input offset voltage.

     

    - Nick Gray

    Barry -

     

    I phoned a friend of my in the amplifiers group of National Semiconductor and he told me that the LMP2011 has a lot of capacitance associated with it (it is a chopper-stabilized amplifier) and the overload recovery time is typically 50 ms. See data sheet pages 3 and 4 for overload recovery time specifications. The key to getting what you need is to insure that the part never goes into overload. That is, keep the device operating within its linear range. This implies ensuring that the input does not momentarily make a jump so large that the amplifier has difficulty following it, and being sure that the output never saturates. I suggest you may need an amplifier that is NOT chopper stabilized, but such an amplifier would have a larger input offset voltage.

     

    - Nick Gray