LM358 strange behavior !

Hello,

Excuse my english, I'm French ...

The LM 358 is a mono-voltage amplifier capable of approaching its negative supply (there is a NPN transistor in the output stage). I had to build a mono-voltage subtractor (a voltage substractor built with a LM358 powered with 0v-12v supply) and with the consent of the electronic simulator ISIS I built (wired) such a subtractor. I was surprised that it could not entirely do the job. For some subtractions, like 2v - 1.8v for example, the substractor said 0.62v instead of 0.2v. But for some other, it answered the right thing, wich was 0.3v (a voltage lower than 0.62v) sometimes. So it was ableto go under 0.62v !

Why does a 2v - 1.8v substraction gave 0.62v instead of 0.2v !


Resistors, all identical, were 2.5k one, so I do not suspect bias currents problem. I tried many other chip, it didn't change anything. I changed the 0v suply into a -12v one and the substractor worked well for every input voltage.


Does someone have an explanation?

cdlt.

Michel.

 

Le dimanche 4 novembre 2012 11:44:51 UTC+1, John Fields a écrit :

Ok. It is the basic soustractor schématic (not the instrumentation one with 3 amplifiers), with unity gain, visible here :

http://crteknologies.fr/electronique/cours/ao_soustracteur.gif

In my case, R1=R2=R3=R4=2.5 k, and Us = U2 - U1.

Michel.

 

Your feedback resistors (R1-2) are asking the LM358's
output to sink too much current. Close to ground, its
ability is quite limited. See "Output Current Sinking
Characteristics" on page 7 here:
http://www.ti.com/lit/ds/symlink/lm358-n.pdf

The reason for the limitation can be understood easily by
looking at the LM358 schematic on page 20--the output stage
is an emitter follower in parallel with a current sink. Only
12uA is guaranteed at 200mV, Vcc=+15v ("Output Current," pg. 4).

Cures: use larger resistors, or connect a load resistor
or a current sink to the output.

 

Try putting a load R on the output, like a 10k or something in that
range. You maybe in the cross over zone and that op-amp does not have
much for shoot-through current.


Jamie

 

Oh, good for you! Recommending James Arthur's suggestion as your own is
sheer genius.

 

Oops! In your vernacular, it should have been "Recommending James
Arthur's sugesstion as your own is shear genus."

 

I don't know what the **** you're talking about, I have no idea who
James arthur is and I don't give a shit. Maybe you got some infatuation
with him, I don't, who ever he is.

The fact that a 324, 358 having those issues being common knowledge
as it is, leads me to believe that you're just a sit in stooly. Yes,
one that is so full of shit that I smell it all the way over here.

Oh, btw, incase you want to expand that fake no hands on knowledge of
yours, a 358/324 etc can also oscillate if the output is brought down to
or near zero output with no load on it.

What a piece of shit you are. You belong with Phil, down under, the
table, that is.

Jamie

 

The LM358 has an input CM range that is guaranteed to go down to the
negative rail over the full temperature range. One of the things that
makes it so useful for an extremely cheap and common op-amp (it's half
an LM324). So, it's not that.

The LM358 is guaranteed to drive the output within 20mV of the
negative rail (5mV typically) over the full temperature range, with a
<=10K load (critically important, but perhaps not obvious from the
data sheet) connected to the negative rail. The subtractor has < 10K
of load but does not have that configuration of loading on the output.
That's his problem.


Best regards,
Spehro Pefhany

 

His R1-R2 combination presents 5k from -Vin to output, so he'd need to
sink 320uA for his example (Vin+ = 2.0v, Vin- = 1.8v, Vout = 200mV).

R1=R2=100k would go a long way toward fixing that.

 

Which overwhelms the 50uA nominal internal sinking capacity.

Sure, and make the other two 100K as well to kill most of the offset
due to bias current.


I suppose one could do something like this and keep the 2K5 resistors:

|\|
-|-\
| >-----+----
-|+/ |
|/| |
|
|
|
|
1M8 |
___ |/
+12 -|___|--| 2PC4081R
|>
|
|
===
GND

LOL.

 

LOL!

I nearly suggested a JFET load, Vgs=0v. Bigger resistors was easier
though.

 

a écrit :

Hey, a resistor to a negative supply would be nice too

 

Sure, if you've got one. (a negative supply, that is.)

Bob Pease did a neat trick--or was it Jim Williams?--using the e-b
junction of a powered optoisolator as a negative current generator.
The phototransistor served as a solar cell, to pull an LM358 output
all the way down. Cute.

 

a écrit :

Was joking of course. If you have a neg supply then just use it for the
opamp.

I think it was Bob Pease but I doubt this has enough current for an LM358.

 

Missed it--whoosh. I often don't use negative supplies for V- for
other reasons, like, it might be dirty, or to save power, or reduce
loading on V-, etc.

Could be. If an ordinary phototransistor isolator has a current
transfer ratio of 20%, and the phototransistor has a hfe of ~100, that
suggests i.b ~= 0.2%. So, 10mA into the IRLED could make 10 or 20uA.

 

That's not a hazard, it's a feature--crowbar. LM34, right?

Grins,
James

 

Exactly! I like Phil because his knowledge exceeds Jamie's (Maynard's)
by several orders of magnitude and he communicates it clearly. I have
learned much from him and I suggest Jamie try to do the same.

 

Well, no wonder that you never learn anything. He is a prolific poster
here and provides great information.

Maybe you got some infatuation

Infatuation? Is that your practice word of the day?

I must ask, what is a sit in a stooly? I don't think I've ever done
anything like that. It must be something with which you are familiar and
I am not.

Well, I've already replied that I would rather learn from Phil than see
your junk.