LM393: The design issue for LM393

The LM393 is optimized for high supply voltages, up to 36 V. It needs a headroom of 2 V to the positive supply, so at lower supply voltages, the usable input voltage range becomes very small. At 2.5 V, the usable range is only from 0 V to 0.5 V.

In your circuit, when one input is above 0.5 V, the electrical characteristics (especially propagation delay) are no longer guaranteed. When both inputs are above 0.5 V, the output is somewhat random and depends on the internal characteristics of the die. The change you are seeing is not related to "G4" or not, but to two different versions of the die.

You need to use a comparator with rail-to-rail inputs, i.e., the LM393LV.

Hi Clemens,

Thanks for your professional comment, so you mean, LM393DR also don't good for this design requirement, we need to use the rail-to-rail comparator like LM393LV and so on? thanks.

Hi Zhang,

Yes, LM393LV is rail to rail and drop in replaceable so it can be able to handle input voltages up to the rail and has no limits on the input range. The output will be correct. 

Hello Zhang,

As Clemens said, the LM393 is actually designed for higher voltages.  Running the LM393 at 2.5V is "running on fumes". The valid input range is only 0 to 500mV.

They were lucky that the logic worked out with the older die. They were operating on the edge and I can almost guarantee them that it WILL fail at lower temperatures. Please see section 2.7 and 2.8 of the appnote that explains this in detail.

We would recommend the LM393LV which could drop-in replace the LM393, or, change the resistor values to bring the thresholds below 500mV.

The only difference between the LM393DR and LM393DRG4 is that the LM393DRG4 ensures you that you get only NiPdAu lead plating.

Ordering the "generic" LM393DR, the plating can be either NiPdAu or Sn (tin), depending on where it was assembled.

Either one could have either die at this time.

As Paul said, you can move the input voltages to below 0.5 V.

In the current circuit, R5 and R10 generate a reference voltage of about 1.67 V. R4 and R7 divide an input of 4.43 V down to this voltage.

For example, if you change to R5 = 25.5 kΩ and R10 = 4.7 kΩ, then you have a reference voltage of 0.461 V, and can use R4 = 118 kΩ, R7 = 13.7 kΩ.

Thanks Clemens for your continued excellent support of the forum.

Zhang, Thanks for posting this on the forum.  It is unfortunate that this has occurred.  Maybe if there is material that you have not opened, you can return the material and purchase the LM393LV instead as this would remove the need to change the values and is really a better solution for low voltage applications.  The LM393 can be used in lower voltage applications but it forces one to keep the inputs in the proper common mode range as everyone is eluding to.  I am sorry you have found yourself in this position, but I believe my colleagues have provided the proper guidance to move forward.

Chuck