LM393A: Frequency response

Part Number: LM393A
Other Parts Discussed in Thread: LM358, LM393, TL431, TLV3501, TLV3604, LM360, OPA855, OPA843, OPA818

Hello, I am looking for an IC to use as error amplifier in a custom voltage regulator, supposed to perform well at light loads. I am looking for an amplifier with wide band, whose dominant pole does not superpose with the power pole almost below 100Hz. I downloaded the spice model for LM393A and it seems to perform very well. Dominant pole around 10kHz and then a nice -20dB/dec slope all the way down to 0db with 80deg phase margin. 

My question now if you have a plot showing open loop gain and phase of LM393A against frequency which I can compare to simulation results.

I would also appreciate suggestions about other parts that might be suitable error amplifiers to drive a PMOS if you got some in mind.

Best regards,

Franco Maggi

  • Attached a plot of the frequency response of LM393A, configured in unity gain (voltage buffer) with a 10k pull up resistor and not capacitive load. I had expected further poles for higher frequencies before crossover. Otherwise this would be performing better than a compensated opamp!  

  • Hello Franco,

    Why are you using a comparator to make the worlds worst op-amp??

    The LM393 comparator is based off the LM358 op amp, so any gain plots should be modeled off the LM358. I doubt the creators of the (very, very simple) LM393 SPICE model were interested in open loop gain/phase accuracy. I really doubt the actual first pole is out that far...I would assume it is more like the LM358' s (see fig 7-5 in the LM358 datasheet).

    Yes, it is possible to use a comparator as an op-amp, as shown in section 21 of  AN-74,  but it really makes an inconvenient op-amp.  I could see it if you have a leftover comparator channel and have a non-critical amp need..but not from scratch..

  • Thanks Paul,

    You are actually answering another question. I am rather trying to use a comparator to implement an error amplifier, just because the internal compensation of op-amps does not fit the frequency response I need. If not for the variable voltage, I would have use a simple approach like TL431. 

    The AN-74 sec 21, is addressing something close to my problem. I got do G=1 to create a voltage buffer able to source 100mA +

    The output capacitor is 10u and is my dominant pole.

  • Hi Franco,

    I just do not believe that plot. Too good, too clean.

     I *really* doubt a 50 year old LM393 would have a zero cross at 3GHz.

    Only a few of our newer high speed op-amps would have such a plot (and the DC gain would be lower). My guess is that the 393 equivalent crossover is more like <700kHz.

    I am assuming you want the maximum available open loop gain at the mains frequencies (50/60/100/120Hz)  for minimum loop error.

    We are coming out with higher speed comparators, such as the TLV3501, TLV3604 family (soon TLV3601 single ended out), and the older ones like the LM319 or LM360/1. These, being high-speed, would theoretically have the first pole past 100Hz

    You are the first person in my many years that has really cared about the open loop gain of a comparator! Slight smile

  • That's my gut feeling too, so I decide not to trust that model. 

    Looking for an un-compensated op-amp, so I learned, one often ends up with what falls under the category comparators. So I look at "comparator's" loop gain the way I do for compensated op-amps :-) Looking at the architecture of the single amp in LM393, I recognize the elements I would expect: a bunch of bipolar transistors, arranged in a simple diff + SE amp output stage, open collector, which happens to be exactly what I need in a compact IC format. Having a reliable open loop gain plot of that would have endorsed my choice.

  • Hello Franco,

    A comparator *is* an uncompensated op-amp at heart. The difference between the LM358 op-amp and LM393 comparator, other than the output stage, is the removal of the internal compensation cap.

    If you need "flat" open loop gain pole >100Hz, then a high speed (>200MHz or more) op-amp should have that profile as their first pole moves to the right. Though the DC gain may be less (the price paid for speed).

    So what it sounds like you need for your loop amp is a high speed de-compensated op-amp, though DO expect to run into stability issues..

    Take a look at the OPA843 open loop gain plot, or the OPA818 or OPA855 for the extremes.

    I assume the actual BW of your system is much lower than this...so it may be possible to use a "composite" op-amp, that cascades two op-amps in series within the feedback loop to increase the open loop gain. These are used in high DC accuracy applications. Google can find examples.

    This is all getting outside my support wheelhouse. Any further error loop amps should be answered by the Amplifier guys (they love stability analysis :) ).

    But I can *guarantee* you we do not have any open loop gain-phase plots of the comparators...Doing that correctly would be an adventure in itself..