op amp max supply voltage test?

[not1xor1]

Hi

I wonder if there is an easy and safe (for the IC) way to check the maximum supply voltage of  operational amplifiers

I could not find anything with google, so I suspect it is not as easy as just measuring BJT Vcb0 or similar parameters

I've no real need fort that, but am just curious   

thanks

[Rerouter]

I would have thought it would be measuring current through a current limited source.

At close to the breakdown voltages the current will begin dramatically increasing, if your source is current limited to just above what the spec allows, it will fold-back non destructively and you should have a clear idea of what point is too far.

By current limited, i mean with both a resistance, and the supply, without a resistance your supply may not switch modes gracefully enough to avoid latchup damages.

This is primarily if its a BJT process, CMOS can be damaged easier, but if the current is limited, it will limit the risk.

[IanMacdonald]

It may depend on the output pin voltage, since the Vcbo of one output  transistor or the other may be exceeded when it's at min or max. Therefore you might need to do more than just applying voltage under quiescent conditions.

There is also the possibility that although the opamp may work at an overrated voltage, the output s/c protection may not. (Incidentally, most opamps are only proof against an output s/c to ground, a s/c to either supply rail will still destroy it)

[wraper]

Maximum supply voltage is present in the datasheet. Anything higher than that is playing with your luck. It may not last long even if appears to work fine at certain voltage. You can reliably sort individual transistors for higher than rated voltage but not ICs.

[David Hess]

Some parts built on 30 volt bipolar processes including the 741 and 324 are or were graded for supply voltage.  The lowest grade 741 is 36 volts and the other versions are 44 volts.  The normal 324 is 32 volts but the 2902 version is 26 volts.  Tektronix used to grade their own 36 volt 741s for operation at 40 volts.

I suspect this is done by either measuring the change in quiescent current as the supply voltage is increased toward maximum or the change in bias current at high differential input voltages.

[Zero999]

Some parts built on 30 volt bipolar processes including the 741 and 324 are or were graded for supply voltage.  The lowest grade 741 is 36 volts and the other versions are 44 volts.  The normal 324 is 32 volts but the 2902 version is 26 volts.  Tektronix used to grade their own 36 volt 741s for operation at 40 volts.

I suspect this is done by either measuring the change in quiescent current as the supply voltage is increased toward maximum or the change in bias current at high differential input voltages.

I've often wondered about that.

I doubt that modern parts are graded like that. It's highly likely they're all higher grade, the only difference being the part number printed on the case.

[Cerebus]

Some parts built on 30 volt bipolar processes including the 741 and 324 are or were graded for supply voltage.  The lowest grade 741 is 36 volts and the other versions are 44 volts.  The normal 324 is 32 volts but the 2902 version is 26 volts.  Tektronix used to grade their own 36 volt 741s for operation at 40 volts.

I suspect this is done by either measuring the change in quiescent current as the supply voltage is increased toward maximum or the change in bias current at high differential input voltages.

I've often wondered about that.

I doubt that modern parts are graded like that. It's highly likely they're all higher grade, the only difference being the part number printed on the case.

I've heard it said, by someone who ought to know, that the testing time on a precision op amp production line accounts for about 50% of the cost of the part.

Tests that take a long time can push the price up. Take the LMC662 and the LMC6001; as far as I can tell they are identical ICs (i.e. the same die) apart from the testing, the LMC662 has a typical input leakage of 2fA but no guaranteed value, the LMC6001 has a 100% tested and guaranteed leakage of 25fA max. Low current leakage testing is slowwww, so the testing is expensive. A LMC662 costs £1.15, an LMC6001 costs £12.60.

[Zero999]

Some parts built on 30 volt bipolar processes including the 741 and 324 are or were graded for supply voltage.  The lowest grade 741 is 36 volts and the other versions are 44 volts.  The normal 324 is 32 volts but the 2902 version is 26 volts.  Tektronix used to grade their own 36 volt 741s for operation at 40 volts.

I suspect this is done by either measuring the change in quiescent current as the supply voltage is increased toward maximum or the change in bias current at high differential input voltages.

I've often wondered about that.

I doubt that modern parts are graded like that. It's highly likely they're all higher grade, the only difference being the part number printed on the case.

I've heard it said, by someone who ought to know, that the testing time on a precision op amp production line accounts for about 50% of the cost of the part.

Tests that take a long time can push the price up. Take the LMC662 and the LMC6001; as far as I can tell they are identical ICs (i.e. the same die) apart from the testing, the LMC662 has a typical input leakage of 2fA but no guaranteed value, the LMC6001 has a 100% tested and guaranteed leakage of 25fA max. Low current leakage testing is slowwww, so the testing is expensive. A LMC662 costs £1.15, an LMC6001 costs £12.60.

That makes sense for precision parts but for cheap ICs such as the LM741 and LM358? No way!

[David Hess]

I've heard it said, by someone who ought to know, that the testing time on a precision op amp production line accounts for about 50% of the cost of the part.

This applies to a lot of parts simply because if the testing is a majority of the cost, then it pays to invest to lower testing time.  Once testing time is a minority of the cost, then the point of diminishing returns has been reached.

Tests that take a long time can push the price up. Take the LMC662 and the LMC6001; as far as I can tell they are identical ICs (i.e. the same die) apart from the testing, the LMC662 has a typical input leakage of 2fA but no guaranteed value, the LMC6001 has a 100% tested and guaranteed leakage of 25fA max. Low current leakage testing is slowwww, so the testing is expensive. A LMC662 costs £1.15, an LMC6001 costs £12.60.

Robert Pease of National Semiconductor wrote specifically about the testing of the LMC6001 so more is known about it than would normally be the case.

I think the LMC6001 is a selected version of the LMC6081 which is an improved LMC662.  We used the LMC6081 extensively and graded them ourselves although we never needed the potential 2fA performance.  It is one of my favorite operational amplifiers although precision is relative; it is a precision part compared to typical CMOS operational amplifiers but not compared to bipolar parts.  Our alternative to the LMC6081 was the LM11 bipolar operational amplifier.

[T3sl4co1l]

The maximum voltage is where the specs are still met (assuming the specs apply over a range of supply voltages).  Otherwise, it is the point where the amp begins to take damage (avalanche breakdown), which is also the absolute maximum voltage rating.

Instead of building a tester for every parameter, you might be able to identify a few parameters which vary around the maximum supply range, and test the family for that.

Probably, which parameters fail first (e.g., input voltage/current bias/offset, GBW, supply current, etc.) varies from family to family, so don't expect to be able to grade just any amp based on the same parameter(s).

Tim

[David Hess]

That makes sense for precision parts but for cheap ICs such as the LM741 and LM358? No way!

Actually it is the opposite.  Think instead of the standard 741 being a 44 volt part with 2 or 3 different grades based on bias current, offset voltage, and temperature range while the 36 volt grade (741C) is a way to recover all of the parts which function but with looser specifications.  The same thing occurs with most of the National LM series parts; the military and industrial grade parts are rated for 44 volts but the commercial part is 36 volts.  The LM358 (and LM324) is exactly one of these parts but with an extra 4th grade; the LM158, LM258, and LM358 are all 32 volts and then the looser specified LM2904 is 26 volts.  Even more so, the LM358 and LM324 were deliberately designed for lowest cost and almost the first if not the first such operational amplifiers.

But a company like Linear Technology which deliberately built their business on premium parts at a higher price made only 44 volt versions.

[not1xor1]

so, since each op amp is different it may make sense to build a test circuit, i.e. a schmitt trigger oscillator and use a current limited supply to keep it within safe power dissipation and see if the circuit works and check when the PSU switches to current limit mode

unfortunately I do not have a suitable PSU so I cannot test if this may work 

anybody adventurous here ?   

[David Hess]

so, since each op amp is different it may make sense to build a test circuit, i.e. a schmitt trigger oscillator and use a current limited supply to keep it within safe power dissipation and see if the circuit works and check when the PSU switches to current limit mode

unfortunately I do not have a suitable PSU so I cannot test if this may work 

Most power supplies would not be suitable for this.  A source meter might be.  If you are going to build a test circuit, then you might as well build a high precision current limiter and test point as well.

For a part like the 741, I would do two tests:

1. Measure the quiescent current as the supply voltage is raised with the operational amplifier configured as a follower and the non-inverting input held at the midpoint of the supply voltage.  The quiescent current should rise linearly below the maximum voltage.

2. The NPN Vcb can be measured directly between the positive supply pin and either or both inputs.

[ebclr]

If you want +- 30 V, this one can do that

http://www.ti.com/lit/ds/symlink/opa551.pdf

[not1xor1]

so, since each op amp is different it may make sense to build a test circuit, i.e. a schmitt trigger oscillator and use a current limited supply to keep it within safe power dissipation and see if the circuit works and check when the PSU switches to current limit mode

unfortunately I do not have a suitable PSU so I cannot test if this may work 

Most power supplies would not be suitable for this.  A source meter might be.  If you are going to build a test circuit, then you might as well build a high precision current limiter and test point as well.

For a part like the 741, I would do two tests:

1. Measure the quiescent current as the supply voltage is raised with the operational amplifier configured as a follower and the non-inverting input held at the midpoint of the supply voltage.  The quiescent current should rise linearly below the maximum voltage.

2. The NPN Vcb can be measured directly between the positive supply pin and either or both inputs.

I'm (sloooooowly) building a dual output (i.e. positive/common/negative) PSU which should allow to test low voltage parts (max Vout is just +/-22V) since both positive and negative branches have tracking current limit and just 10µF on the output (i.e. less than 2.5mJ*2 capacitors at max Vout), so maybe in future I'll follow your suggestions and test some cheap LM358s...   

[not1xor1]

If you want +- 30 V, this one can do that

http://www.ti.com/lit/ds/symlink/opa551.pdf

come on... you can do better   
have a look at http://www.linear.com/LTC6090, +/-70V (140V) rail-to-rail output

but that's too easy... I just want to learn something new... 

[Zero999]

That makes sense for precision parts but for cheap ICs such as the LM741 and LM358? No way!

Actually it is the opposite.  Think instead of the standard 741 being a 44 volt part with 2 or 3 different grades based on bias current, offset voltage, and temperature range while the 36 volt grade (741C) is a way to recover all of the parts which function but with looser specifications.  The same thing occurs with most of the National LM series parts; the military and industrial grade parts are rated for 44 volts but the commercial part is 36 volts.  The LM358 (and LM324) is exactly one of these parts but with an extra 4th grade; the LM158, LM258, and LM358 are all 32 volts and then the looser specified LM2904 is 26 volts.  Even more so, the LM358 and LM324 were deliberately designed for lowest cost and almost the first if not the first such operational amplifiers.

But a company like Linear Technology which deliberately built their business on premium parts at a higher price made only 44 volt versions.

How much does the test actually cost to conduct? It wouldn't surprise me if it's close to the cost of producing an LM741 or LM358. I'm still not convinced they bin such cheap parts.