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.

LM2903: input VS VCC

Part Number: LM2903
Other Parts Discussed in Thread: TINA-TI, LMV393, LM393, LM339, TLV1704, LMV339

I used LM2903 to compare LVPECL signal and reference signal to get a LVCOMS output signal.

I saw Vin- should be 2V lower than VCC in datasheet. But another engineer said it's ok if Vin- is 2V when VCC is 3.3V.

I simulated in Tina-TI and got the result as my colleague said. But i got the result as datasheet described when simulated by using Rohm Pspice model.

Please confirm if TI LM2903 Vin- should be 2V lower than VCC. Thank you.

  • Hello Usar,

    I see two different devices. You are asking about the LM2903, but are simulating with the LMV393. While functionally equivalent, they are electrically different.

    The valid LVPECL range is 1.6 to 2.4V.

    The LM2903 has a specified input range of 0V to 1.5V below V+ (for room temp, or 2V for over-temp). For your 3.3V supply, the maximum input voltage would be 3.3V - 1.5V = 1.8V. So the valid input range would be 0 to 1.8V - less than the common mode voltage.

    The LMV393 used in the sim has a specified input range of 0V to 0.7V below V+ (over temperature). For your 3.3V supply, the maximum input voltage would be 3.3V - 0.7V = 2.6V. The entire 1.6 to 2.4V LVPECL range is valid and both the sim and device should work correctly.

    But for the LM2903, all but 200mV of the VOL is above the valid input range. Because both inputs are outside the common mode range for most of the time, the output may be incorrect. You may be seeing "strange" results. Even if it seems to be working at DC & low speeds, the response time would be severely impacted (much slower).

    The engineer may have been referring to the "unique feature" that the LM2903/LMx39 family has where the output will be correct if at least ONE input is within the legal voltage range - but in this case, the reference voltage and the "high" signals are both outside the range - so the output will be indeterminate.

    But all is not lost...since it is a comparator and absolute voltage value is not necessary, you could divide (attenuate) the inputs down by half to get the voltages within the legal input range.

  • Hi Paul,

    Thank you very much.

    I didn't see LM2903 in Tina. And I saw the requirements of Vin and V+ are the same for both LM393 and LM2903 in datasheet.

    It seems like only Vin- have requirements, but Vin+ don't have. Can you confirm again for me? 

    Regards,

    Wei

  • Hello Wei,

    The model you used was for the LM_V_339 - the low voltage optimized version of the LM339. It is a completely different design on a different process.

    The wording of note 2 makes it sound like just the inverting input can go above, but the other input is similar, too. See section 8.2.2.1 (pg11) of the LM339 Apps section.

    In short - it still will not allow you to "legally" operate under your stated conditions.

    As I said, it may seem to work on the bench, but unit-to-unit differences and and temperature variations will eventually cause dropouts. I am assuming since it is at the end of a PECL line, the data rate will be fast. Operating outside the input voltage range will severely slow down the response time as the input stage is not fully biased and the bandwidth reduces.

    If you cannot modify your circuit, then switch to the LMV339 which as the "extra" headroom required (and slightly faster). Or the TLV1704 which has a R-R input (full 0 to 3.3V input range).
  • Hi Paul,
    Got it! Thank you very much.

    Regards,

    Wei