OPA892: Sine wave to square wave

Part Number: OPA892
Other Parts Discussed in Thread: TLV3901, TLV3811, TLV3801

 

Hello TI Engineer,

We have a project that requires a clock waveform conversion circuit, with the following specifications:

  • Input frequency: 100 MHz

  • Signal waveform: sine wave

  • Input level: 3 dBm to 5 dBm

  • Input DC offset: < ±50 mV

  • Output impedance: 50 R

Here's my initial approach:

  1. Since this is a 100 MHz clock and we need to preserve the higher-order harmonics, a high-speed op amp or comparator is required. I'm considering a device with a bandwidth greater than 1 GHz.

  2. The output needs to be a 3.3 V LVCMOS square wave to interface with a CPLD, so I plan to use a comparator to convert the sine wave into a square wave.

  3. From my calculations, 3 to 5 dBm into a 50 R coaxial cable gives a voltage swing of roughly 1.3 V.

  4. Because there may be a DC offset up to ±50 mV, I intend to use AC coupling to remove it.

My proposed implementation:

  1. At position 1 in the diagram, place a capacitor for AC coupling to strip off any DC offset.

  2. At position 2, apply a DC bias to eliminate the negative portion of the sine wave (please refer to the two sine waveforms in the drawing).

  3. At position 3, set a comparator threshold voltage using a resistive divider. However, since the input amplitude varies between 3 dBm and 5 dBm, a fixed threshold (I used 1 V) causes the output duty cycle to deviate from 50%.

Note: The intended behavior is that when the input voltage is above 1 V, the output goes high, and when it's below 1 V, the output goes low.

Could you please review this approach and let me know if it's feasible, or if there's a better solution you would recommend? If a comparator-based design is the way to go, what aspects could be optimized? Thank you.

 

image.png
 
 
 
 

image.png

  • User,

    Try using a comparator; it is made for analog to digital output. 

    TLV3901 is fastest.

    TLV3811 is fast at a lower cost. 

    At position 3, set a comparator threshold voltage using a resistive divider. However, since the input amplitude varies between 3 dBm and 5 dBm, a fixed threshold (I used 1 V) causes the output duty cycle to deviate from 50%.

    The DC bias is 1.65V, so using that voltage should keep 50% 

  • Hello Ron,

    Thank you very much for your reply.

    Below is the circuit with the newly selected chip according to your suggestion. Please take a look.

    My question is: the input signal is only 3 to 5 dBm, which is very small. Can the TLV3801QDSGRQ1 detect it correctly?

    In addition, is the design in the figure below correct?

    1. Can the DC offset be removed via C3?

    2. By using the resistor divider R805 and R807 to bias the signal, can the sine wave be pulled above 0V? I calculated that the bias voltage should be around 1.65V. Is that correct?

    In summary, can this circuit convert an input sine signal of 3–5 dBm (with a 50 mV DC offset) at 100 MHz into a 100 MHz LVDS signal?

    Please kindly review and advise. Thank you.

  • TLV3801 is high speed and has high input current.

    So this should be a better starting point if input impedance (at 100MHz) should be 50 ohms.

    TI tina sim.

    TLV3801.TSC