Driving LTC1968 Fully Differential

Hi,

I see that LTC1968 RMS to DC converter has a differential input that can be driven in single ended or differential mode and DC or AC coupled.

If I drive the input with a fully differential op-amp, do I need to place an RC filter between the op-amp outputs and LTC1968 inputs? Would a damping resistor (two in my case) be enough to mitigate the spikes generated by the switches capacitor input of the LTC1968?

I also see that the input of the LTC1968 must have a return path to ground, therefore it is not possible to place a series capacitor on both inputs. It seems that this aspect is not modeled in the LTspice model. I see no different behavior if I place two coupling capacitors on both inputs. Am I correct?

Thanks. 

Top Replies

  • 0
    •  Analog Employees 
    on Dec 17, 2019 8:49 PM

    The differential input of the LTC1968 can be driven by a differential amplifier. Use a series 50 ohm resistor at each output driving a differential capacitor so that 1/ ( 2 pi 100 C) is the maximum input signal bandwidth (the 1% error BW of the LTC1968 is 500 kHz).

  • 0
    •  Analog Employees 
    on Dec 17, 2019 8:50 PM

    The differential input of the LTC1968 can be driven by a differential amplifier. Use a series 50 ohm resistor at each output driving a differential capacitor so that 1/ ( 2 pi 100 C) is the maximum input signal bandwidth (the 1% error BW of the LTC1968 is 500 kHz)

  • 0
    •  Analog Employees 
    on Dec 17, 2019 8:52 PM

    Hi steveone,

    To answer your questions:

    1. Input Filtering Needed? There are settling time issues associated with the 0.8pF internal input capacitance of the LTC1968. You'd want to keep the input time constant (RC) to a fraction of 125ns. For 12.5ns time constant, a 15.5k resistor is recommended (RC = 15.5k * 0.8pF = 12.5ns).

    2. LTspice model: While the model could conceivably work without a DC path to ground on either input, I've been told that the DC levels are being set by diode leakage (which could lead to it working in simulation, but never in reality). This apparently represents the part correctly. The LTC1968 model is probably representative of the part but using it without a DC path in the real,  not simulated world would be a very bad idea. So, it is recommended that you follow the datasheet advise and keep the DC path to at least one of the inputs.

    Regards,

    Hooman

  • Hi 

    Thanks for answering.

    The LTC1968 has a 125 ns sampling time. The time constant of an RC filter at the front end (or just a resistor if the input capacitance of the LTC1968 is present) must be at least 10 times smaller than 125 ns, to minimize the conversion error. If 10 time constant elapse, then error is 1/(e^10).

    If I want to drive the input of the LTC1968 differentially, I see at least three solutions:

    1) Using just two resistors at VIN1 and VIN2. This is the solution adopted in the evaluation board, where 2x 499 ohm resistors are used. These resistors with the 0.8pF input capacitance form a 398 MHz input filter with R*C = 399 ps (does R_SW = 2 kOhm need to be included?). I think this solution is mainly meant to mitigate the sampling spikes not to overload the opamp driving the LTC1968.

    2) Using two resistors and a differential capacitor

    3) Using two resistor, a differential capacitance and two common mode capacitors that will be in parallel with the 0.8 pF.

    The fully differential op-amp is the LTC6363.

    May either of the three solutions be adopted? Which solution is the recommended one, if any?

    Finally, which is the effect of the 2 kohm R_SW shown in the equivalent analog input circuit of the LTC1968? I did not consider it in the assumptions I did before, but I think I should...

  • 0
    •  Analog Employees 
    on Dec 27, 2019 10:20 PM in reply to SteveOne

    My option is:

    3) Using two resistor 50 oms each, a differential capacitance and two common mode capacitors that will be in parallel with the 0.8 pf.