• State Resolved
  • Locked Locked
  • Replies 4 replies
  • Subscribers 64 subscribers
  • Views 2460 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

Voltage Reference for Comparator Input-Biasing (low output resistance)

Armin L.
Intellectual 480 points
Other Parts Discussed in Thread: LM4128, LM4040, REF6025, TLV3501

Hi all! ;)

I want to convert a sinusoidal signal (-5...5 V) with a frequency of 100...500 kHz into a quadrature pulse signal in the range of 0...5 V and the same frequency. For this purpose I want to use the comparator circuit in the attached picture:



To shift the input voltage into the supply range of the comparator, I use a resistive divider (R3 and R2) to attenuate the incoming sine and I use a voltage source to bias and center the signal around 2.5 V DC. I thought about using a voltage reference for that purpose because the 2.5 V source would not have to provide a big power for a load. It just has to sink and source a small current.
My question is: Would this be a suitable application for a voltage reference? Which reference (or which class of voltage references) would you recommend for this purpose? I would like the reference to have a very small serial resistance (in this case to GND) because otherwise it would influence the center potential of the voltage divider (V_in+).
What do you think?

Thanks in advance and best regards!

  • 0
    TI__Mastermind 19470 points

    Hi Armin,

    This application is suitable for a voltage reference. There is many applications where a voltage reference is used as a level shifter due to the precision that it offers over a resistor divider. One of our newer precision references is the REF3425 which can source and sink up to 10mA. Depending on your need, we also have less accurate voltage references such as the LM4128 or even the LM4040. One thing to note is that output impedance is not a typical datasheet measurent for voltage references.

    To ensure the lowest output impedance, it might be necessary to add a buffer with a known output impedance or use a REF6025. The REF6025 is a precision voltage reference with an integrated buffer with a small output impedance.

    -Marcoo Z 

  • 0
    Guru 140200 points

    Hi Armin,

    a voltage reference is no voltage regulator! A voltage reference usually has no buffered output. The output impdance can be high as hundreds of ohms and the output impedance can be highly nonlinear. As Marcoo already mentioned the output impedance of a voltage reference is seldom specified and you will hardly find an output impedance versus frequency plot in the datasheet of a voltage reference.

    Another problem of your circuit is the LT1720. This is a really really fast comparator and you must not handle it like a slow operational amplifier. Datasheet tells, that source impedances should be lower than 1k. So, R2 and R3 of your circuit are much too high for stable operation. You should also think about introducing an additional hysteresis by the help of a high ohmic resistor from output to + input.

    As the LT1720 is ultra fast and shows a maximum toggle frequency of 70MHz you need not only a 2.5V reference which is stable and free from noise up to 100MHz, but also the driver sourcing the voltage divider at input should provide a very low output impedance up to 100MHz.

    I would not use a voltage reference for the 2.5V bias. I would use a simple voltage divider. Take a 1k / 1k voltage divider, drive it from +5V and filter the 2.5V bias by the help of a low ESR 100µV/25V aluminium electrolytic capacitor paralleled by one or two 470n...1µ/0805/X7R. Keep these caps as close as possbile together. Every millimeter counts. Reduce R2 and R3 to 2k and 1k and increase C1 to 100n. If your sine amplitude is really 5V you should increase R2 to 3k or more. Otherwise you will violate the common mode input voltage range of LT1720.

    And finally, as the LT1720 is ultra fast, think about following HF-circuit design rules. Your PCB should at least have a solid ground plane.

    Kai

  • 0 in reply to kai klaas69
    Intellectual 480 points
    Thank you very much for all the mentioned points! I have several questions about them:

    1.) I do not really want to use the LT1720, I just picked it because it is one of few suitable comparators in the limited library of LTspice. I rather want to use the comparator TLV3501. But why would a high-speed comparator be such a problem for the 2.5V bias or the sinusoidal signal source? I mean the comparator's input terminals should have a very high input impedance (allowing basically no input current), no matter the toggling frequency. Or are there some parasitic effects that I am missing?

    2.) Why should R2 and R3 be reduced? I chose such high values because the sinusoidal source will not have to drive a large current.

    Best regards! ;)
  • 0 in reply to Armin L.
    Guru 140200 points
    Hi Armin,

    yes, there are parasitic effects which have to be taken in consideration. There is always a non avoidable stray capacitance from the output of a comparator to the - input and to the + input. Together with components from - input and + input to ground (the so called source impedances) feedback paths are formed, which can make the comparator oscillate if enough output signal is coupled back to the input. Datasheets mentiones, that these source impedances shall be smaller than 1k. But if you make the source impedances bigger, more from the output signal is coupled back to the input and oscillation can occur.

    The same is valid for the 2.5VDC bias generator. If the output impedance of this 2.5VDC source isn't small up to 70MHz, the total source impedance seen by the comparator at its inputs can increase above the recommended 1k. More, if furtherly the output impedance is complex, additional phase shift can occur, which can increase the instability.

    There's another reason why HF-circuits are usually low-resistance: Each resistor has a parasitic stray capacitance. Assume a stray capacitance of 1pF, then a 100k resistor begins to become a capacitor at a frequency above 1/2/pi/100k/1p = 1.6MHz. A 10k resistor becomes a capacitor above 16MHz, a 1k resistor above 160MHz, and so on. The unpleasant thing of becoming a capacitor is not so much, that the impedance of resistor decreases with increasing frequency, but that an unwanted phase shift is introduced. This can be deadly in the feedback loop of an HF-amplifier.

    Kai