LM358 Offset Voltage..Is it always Positive?

[SuzyC]

I think, despite the poor offset voltage specs for a LM358, that I can use it with a single Vdd=18V  Vss=0 power supply, as  a non-inverting AV= 200 current shunt (.01-ohms) amplifier and use a MCU to be able to compensate for offset and reasonably accurately monitor positive currents 0.0 to .5-Amp.

The Motorola and ON Semi datasheet specifies the typical Voffset = 2mV. No polarity is shown, positive is assumed.

 Now this would work fine, if this Voffset is always a positive offset, over a -10C to +50C temperature range.

 Is this so?

[magic]

Unfortunately not, offset voltage of opamps is random (and thermal drift is random too).
The X mV specification means no worse than ±X mV.
Y µV/°C also is ± Y.

[GigaJoe]

all pdf define offset as positive, regardless + or - ; with dial power supply
with single supply ,you cant reach '0' at input , it not RRI, offset will be positive, with amplification  X200 the output noise would be a few millivolts p-p

[magic]

For a 0~5mV signal I think a more realistic option (including stability over 60°C range of temperature) is a "zero-drift"/"auto-zero"/"chopper-stabilized" type of opamp.
But they are considerably more expensive than LM358. The cheapest I know is probably MCP6V01.

Standard question: how much accuracy is really needed?

[TimFox]

Although many op-amps are capable of working with common-mode voltages slightly below the negative rail, the output (with very few exceptions) cannot go negative with respect to that rail.  For measuring small positive voltages, you would be better off with at least a few volts of negative power supply voltage with respect to the input common node.

[SuzyC]

Fascinating, a TI specsheet shows typ Voffset to be +/- .3 mV for the 358B, their super-duper LM358.

In the same specsheet they show the original 358 partnum having a 2 mV typical offset.

I have 20 (Motorola and with the same '88 date code) PDIP-8 LM358 op-amps.

Their offset is not random, but dependent on the low-pin or high-pin numbered op-amp section.

On all the pin1-3 numbered op-amps the offsets are positive and on the pin5-pin7 are negative.

The Motorola LM358 (in PDIP) offsets do vary quite a bit from unit to unit.

I also have about 100 small-outline LM358's and their offsets are also positive on the pin 1-3 opamp and negative on the other side of the fence.

-------------

Having a negative supply rail does more harm than good, the MCU A/D cannot handle negative inputs and latchup is a threat.

[Kleinstein]

They don't have an intentional offset in one or the other direction. It is possible that one manufacturer may have a tendency for one sign and possibly different for the 2 OPs in the chip. However one can not count on this.

For single supply use it may indeed be a slight andvanage to have a known positive offset, but I don't know of any sich OP.

A relaitively cheap AZ OP like MCP6V01 or similar is likely the best choice, especially with a 5 or 3 V supply.

[magic]

If you want a nasty "solution":
Put a resistor in series with IN+. The input bias current flowing through it will lift IN+ a little higher than the actual shunt voltage.
Or if it's a one-off build, just select a "good" chip.

But it will still do nothing about thermal drift. Typical spec for LM358: 7µV/°C, i.e. 0.42mVpp over your range.

[SuzyC]

Sage advice, buy new, but I have all these jelly-bean op-amps and I can very quickly select those among them that have a positive offset.

I think its safe to say that offset voltage doesn't change much with time, opamps are born that way.


A .42mV  offset drift over temperature range is easily handled by s MCU measuring offset periodically and correcting measured values.

---------
Adding a series resistor in series with the + input greatly increases the chance of oscillation and also noise pickup and unbalances the Kelvin connected differential inputs that have identical resistor values connected to the shunt.

[GigaJoe]

offset of any opamp can be eliminated by external adjustment circuit , but need a negative rail (let say adding ICL7660 )

again depend on precision needed , consider noise especially on x200, and Vout min  20-30mV, and max Vout = Vsupl -(2-4)V

so measure floor would be 30-50мА approx. with precision at 3-5мА due to high noise and amplification.

and pick a 358, it really no discussion about long therm stability \ long therm drift \ e.t.c.  ... that simply not the case for 358

(on a budget same OP07, it should have dual PS, and as an option a clamping diode to eliminate a negative V, any ADC tolerable to -0.5V at input. )

[SuzyC]

I don't mean to switch horses in the middle of a stream, nor beat a dead horse, but I think a very interesting topic here would be about the apparent very long life and stability of chips in general. I mean, like horses, if they aren't over stressed by heat, high voltages, etc.

My choice for a mission-critical instrument would be to find a horse that could handle the race.

There's the issues of resolution, repeatability, stability, precision, accuracy and drift, but these are all horses of different colors.

Mainly, I am horsing around in robotics, DIY power supplies, battery charging circuits and battery-use monitors.

Adding an additional chip to challenge the offset of a cheap opamp is something like beating a dead horse.
--------------
Of course noise is always a problem and it is not usually something to be measured, but something that distracts measurement. Currently, Amp measurement is not easily handled by an A/D without signal conditioning(low-pass filter). Even horses are made to wear blinders.

[magic]

Long term offset stability is only specified on precision oriented parts. Usually a matter of single digit µV over a time frame of months (for those parts).

LM358 is cheap junk, but if you pick a unit with +1mV minimum offset it seems unlikely that it will ever go negative, even over temperature with some luck.
Make the MCU report an error when it sees that offset voltage is less than 0.5mV or whatever limit you can reliably detect, problem solved.

[GigaJoe]

sure ...  horses  ...
1mv min  x200 - 0.2V out, at 1V limit , what are we talking about ?

[David Hess]

The offset voltage can go either way.

Sometimes designs can be found where the offset is externally adjusted to always be on one side of zero so that the output can always reach zero.  Below is an example from a Tektronix power supply where they did this by 10 millivolts with the error amplifier, guaranteeing that the output of the power supply can reach zero even with the worst case input offset voltage.

[16bitanalogue]

The real answer is this:

The OnSemi LM358 is a second source to the National LM358 and the datasheet simply copied the same format. No reason to confuse the customer when you are sending them the same part at a lower price.

Offset is always +/-.

[Hiemal]

If you don't mind waiting for parts from china, gainsil makes some pretty good chopper stabilized op amps you can get from LCSC that are inexpensive.

https://lcsc.com/product-detail/Precision-OpAmps_Gainsil-GS8552-SR_C157716.html

Under 5 uV offset, and comes in the same pinout as standard LM358s too.