LM311 with out-of-spec input signal.

Input common mode range does not include V- rail (only 0.5V + V-), so you need an offset for the signal, or a different part like LM393 that does include -0.3V from V-.

Tim

-- Seven Transistor Labs Electrical Engineering Consultation Website:

The data sheet says it works for a single 5V supply, and has example circuits that also avoid the negative supply. By "below the specified minimum" do you mean you're violating the input common mode range? That would likely be a problem.

Hi, S.

I think it will work fine as long as one input is within the CM range and both are above GND, but that said, I'd test it. Response time might be a bit different, but probably not disastrously so.

--Spehro Pefhany

Yes, I hadn't thought about the response time. With the input pulled down, the input transistor becomes saturated, which begs the question of how long it might take to unsaturate :(

Sylvia.

The input pnps will not saturate, when input is at 0V it is same potential as their collectors - so a bit like a diode but not saturated. The real problem is the second stage npns are starved of drive - hence the cm restriction to 300mV above 0V. However if at least one input is above 300mV then I think normal comparator action will be OK.

piglet

This seems remarkably difficult to express properly in words...

Suppose that both the + and - inputs are within the common mode range, with the + input higher than the - input. So the output is high. Now let the + input gradually fall. As it falls below the - input, the output goes low. Now, let the + input continue to fall. I need the output to stay low, even when the + input falls to zero, which is outside the common mode range.

Similarly, in reverse, as the + input rises from zero, going into the common mode range, and eventually rising above the unchanging - input.

I notice in the Texas Instruments datasheet

Figure 15, TTL interface

does appear to envisage operating with one input sometimes having a level equal to the -ve power input (both being ground).

Sylvia.

You are correct. This will happen just as you want so long as the - (inverting) input is above the 300mV CM lower limit.

By inspecting the internal circuit given in the datasheet you can see that things only get risky when *both* inputs are below 300mV when the second stage npns get starved of base current.

piglet

Can you do this? (one diode to make a virtual ground with the chip negative supply one diode drop below it)

V+ ----------------+ | Gnd ---+ |\| | --| \ | | \ [diode] | / | --| / | |/| | | V- ----+-----------+

Ed

I'd have several amps going through the diode, which makes the "diode drop" rather device dependent, so I'd end up having to make a manual adjustment to the threshold.

I could replace the diode with a voltage regulator (which would probably need a heat sink) - but then I might as well use the cheaper charge pump solution to provide a negative rail to the comparator.

Sylvia.

Can you use LM339, (or the dual LM393) instead? These parts behave well even with input slightly below the -ve supply

Probably could, actually. I was thinking it would be too slow, but it's fast enough. Thanks.

Sylvia.

If one input is above the minimum, but within the specified common mode range, no untoward current diversions occur, so you're OK.

However, the "rising" input signal may not exceed the positive common mode range or there will be a possible phase flip. ...Jim Thompson