Opamps and 'common mode input range' ambiguities

Not all OpAmps experience phase reversal.

Datasheets vary all over the map in quality and utility, as you can see. Even
vendor to vendor for same part vary a lot.

1. All voltage values, except differential voltage, are with respect to the zero reference level (ground) of the
supply voltages where the zero reference level is the midpoint between VCC+ and VCC-

Click to expand...

Input bias current specs typically this way, to insure input stage in "healthy" bias along with
its tail current circuitry.


Regards, Dana.

Common mode range -for what- may be part of it.

You'll have one CMR for damage, something different for
phase reversal (if), yet another for "meets every electrical
table spec" (look for fine print on that, in the line item test
conditions) and so on. If you get more than one, lucky you.

As to "why?" on lousy datasheets, remember that the DS
may be started by the engineer but it's finished by MARCOM
(marketing "communications" department) and people there
couldn't tell you a diode from a dildo. Let alone which one
you should use.

Regards, Dana.

Hi,

cupoftea said:

The LM324B datasheet makes it extremely clear that your common mode range will always be from (VCC-) to (VCC+) -2V.

Click to expand...

From LM324 datasheet:
* "Common-mode input voltage range includes V–"
* " ...with features including low offset (600 μV, typical), common-mode input range to ground..."
* " VCM Common-mode voltage 0V to VCC – 2"
* "VCM_recommended: @ VS = 3 V to 36 V --> V– to (V+) – 1.5 "

My comment:
* So twice they refer to "V-". This is a clear specification.
* once they refer to 0V. But the OPAMP has no "0V" input, so how can the OPAMP know what "0V" means? One can only refer to a voltage/node hat exists.
* once they refer to GND. But the OPAMP has no "GND" input, so how can the OPAMP know what "GND" means? One can only refer to a voltage/node hat exists.

Voltages that are given without any "reference" are usually measured to GND.
Where the electronics designer puts or "GND" is the choice of the designer. One can only refer to "GND" when the IC has a dedicated "GND" pin, like logic ICs or some comparators.
It´s totally legal to use:
* GND at the center: like V- = -5V (all w.r.t GND), V+ = +5V
* GND at V- and +10V at V+
* GND at V+ and -10V at V-
--> for all these three conditions .. the OPAMP will see no difference in operation!

So for me: the datasheet talking about "0V" or "GND" is not clear at all!
******

cupoftea said:

...On page 2, the TL084C datasheet says

QUOTE...
1. All voltage values, except differential voltage, are with respect to the zero reference level (ground) of the
supply voltages where the zero reference level is the midpoint between VCC+ and VCC-
...UNQUOTE

This seems very odd and unecessary.

Click to expand...

And what does the LM324 datasheet (page 5) say:
"(2) All voltage values (except differential voltages and VCC specified for the measurement of IOS) are with respect to the network GND.
(3) Differential voltages are at IN+, with respect to IN"

Me:
How can this be better than the TL084 information? (indeed both use very similar wording) They refer to a node that the OPAMP does not know about.
The TL084 refers to th midpoint of V- and V+ ... these both nodes are known by the OPAMP. Thus this is the more "precise" way of definition.

Klaus

cupoftea said:

1. All voltage values, except differential voltage, are with respect to the zero reference level (ground) of the
supply voltages where the zero reference level is the midpoint between VCC+ and VCC-
...UNQUOTE

This seems very odd and unecessary.

Click to expand...

Why do you think that?
Do you not understand that differential voltages are measured between the two inputs
, not to ground?
The input common-mode voltage is, of course, measured to ground.

These are two very different designs. They each have different tradeoffs and the JFET front-end MUST BE defined from the mid-supply range and has strict limits on Vcm inside the rails yet wide Vdm.
Either make do with these old designs or choose better ones.

The input common-mode voltage is, of course, measured to ground.

Click to expand...

Thanks, "ground" i take it you mean "network ground"?...which of course, is often not even connected to the opamp with +/- supplies.

So shouldnt common mode voltages always be taken w.r.t. the "mid supply point", MSP..........and shouldnt that (MSP) be formally declared in the rules of "needed jargon" for opamps?

These are two very different designs. They each have different tradeoffs and the JFET front-end MUST BE defined from the mid-supply range and has strict limits on Vcm inside the rails

Click to expand...

Thanks, i hear what you are saying, but i saw a current clamp design where a TL084CDT was clamping the current through a sense resistor monitored by its inputs. The sense resistor was hanging down off the 20V rail.
The opamps supply was this same +20V rail, down to -3v3 for its NegVcc supply.

...this is way outside its common mode range, and yet in the one i saw, it worked fine.

cupoftea said:

Thanks, "ground" i take it you mean "network ground"?...which of course, is often not even connected to the opamp with +/- supplies.

Click to expand...

I meant the circuit signal ground, whatever you want to call it.
The op amp doesn't have to connected to it as the common-mode spec is referenced to the supply voltages, which of course, are connected to "ground".

Anything else you want to be picky about?

As the supplies are specified "ground referred" means there is
a tie between rails and signal sources / input pin voltages, albeit
"remote" (connect the dots).

Thanks, any idea why the circuit in #8 worked so well?
If there's anything in all this talk of CM range, then that cct should not have worked so well.

cupoftea said:

any idea why the circuit in #8 worked so well?

Click to expand...

What circuit?

..the current clamp, with inputs to a TL084CDT which were both well outside the common mode range.
..Also here..

Do you mean that the clamp works well in a real implementation,
or that the LTSpice schematic gives a satisfying result in LTSpice?

cupoftea said:

So shouldnt common mode voltages always be taken w.r.t. the "mid supply point", MSP..........and shouldnt that (MSP) be formally declared in the rules of "needed jargon" for opamps?

Click to expand...

The common mode input voltag range is limited by the rails.
Example: (upper limit) when
V+ = +5V .. it might be limited to +2V
V+ = +10V .. it might be limited to +8V
V+ = +15V .. it might be limited to +12V
So the upper limit does not care about V- .. and does not care about MSP.
--> Thus I vote for the terminology "V+ -3V".
But if it cares about V- then one could use a formula including V- or one could draw a chart.

V+ is the pin name of the OPAMP. One could use a (virtual on simulation) voltmeter and connect one wire to V+ and the other wire too the input.
MSP is a virtual point. You can not connect a voltmeter to a virtual point.

****

As an experienced designer .. I can not recommend to go to the very limit. Always leave some headroom. (for drifts, temperature, supply ripple ..)
So either chose a bit higher supply, use a more suitable OPAMP or change the circuit.

Klaus

Do you mean that the clamp works well in a real implementation,
or that the LTSpice schematic gives a satisfying result in LTSpice?

Click to expand...

Thanks, the one on the bench works fine.

As an experienced designer .. I can not recommend to go to the very limit. Always leave some headroom. (for drifts, temperature, supply ripple ..)
So either chose a bit higher supply, use a more suitable OPAMP or change the circuit.

Click to expand...

Thanks, i know what you mean..the thing is, using any other circuit (than the above current clamp) means more components. Using a suitable opamp instead means a cost increase of several euros. because opamps that can work with vcc+=20V and VCC- = -3V3, AND that have a common mode input range that extends to VCC+ are extremely expensive....whereas TL084CDT is 0.3 Euro's.
When you have a working TL084CDT circuit on the bench, try telling the customer that "no you cant have this for 0.3 Euros, you must spend an extra 3 Euros on a more expensive opamp".

1. I would like to remind everyone that "ground" only means 0V reference point but does not imply impedance relative to any voltage source or insulation to earth. e.g. the -Vcc or Vee in a unipolar supply (shown for LM324 schema) or somewhere in-between of a bipolar supply where gnd is not used by the OA internally.

2. A primitive analysis on random P-type JFET suggests it should work to and below Vcc


3. Your conclusion of a binary option on pricing/performance choices is only limited by your imagination.
4. If you respin the design, there are many cheap ways to sense current on the high side or low side and with any Vref thus limit the current and voltage.
e.g. using a 60 mV drop sensor. (1 ohm) On the low side even the LM324 will work.

cupoftea said:

try telling the customer that "no you cant have this for 0.3 Euros, you must spend an extra 3 Euros on a more expensive opamp".

Click to expand...

honestly .. from 0.30 € to 3.00 € is nonsense. There are plenty cheaper OPAMPs.

On the other side: you may ask your customer standing on the base of a bungee jump
wheter he would chose the rope that is specified with less of his weight but tested "good" with a stone ...
or wether he more relies on rope that is truely specified for his weight.

My recommendation:
You need to give your custormer a written information that using the exixsting circuit has reliability problems. It may work now, but it may fail, because it´s operatin beyond specification.
He needs to decide which way to go: the chaep one or the reliable one.
Otherwise you are in risk to get responsible for future failure.


Klaus

e.g.

No need to use a more expensive part, when you can get one that is guaranteed for Vcm = Vcc+.

honestly .. from 0.30 € to 3.00 € is nonsense. There are plenty cheaper OPAMPs.

Click to expand...

Thanks, ...for spec 0-20V rail, and CM input voltage to the pos rail, there is nothing in the world below $3.
Except the TL084H or TL074H , which Tony has kindly found...but thats just one part so you wouldnt dare
use it in case it goes nil stock.