[noob mistakes] importance of op amp decoupling

[exe]

Hey guys, I had an almost detective story today. I was experimenting charachterizing different opamps I have.
Essentially, I measure offset voltage (Vos), input current and Vos change with temperature (I use a heatgun set at ~100C, cheap thermocouple shows peak temperature at around 118C).

I recently bought a bunch of NE5532D from TI because it was cheap, and one expensive NE5532D from ON SEMI.
And I couldn't believe my eyes, the on semi's device was total crap all around.

I kept staring at both chips until I realized there was something wrong... Got it, I soldered decoupling cap to wrong pins. Damn!

So, once I fixed the cap all went good .

The final stats: TI showed offset 0.05mV and 0.1mV (both showed 0.19mV at ~100C) if you trust uni-t 61. ON SEMI showed 0.43mV and 0.51mV (0.15mV and 0.51mV at ~100-110C). I tested only one sample of each, so results are not statistically significant and no conclusions to draw. Both chips are well within spec for the parameters I measured. But I would prefer TI chip for DC stuff . Anyway, I have two more chips from TI, will see how they perform.

Now question. Do you think voltage offset will be better if I remove flux? I'm yet to try, didn't have spirit at hand.

PS Just realized that chips only specified to work up to 70C. Huh, I wanted to buy SE5532 that has extended range, but it was not on sale on tme.eu .

[danadak]

Possibly the flux could increase bias current which in turn would
generate offset from the source impedance.


http://www.ti.com/lit/an/slyt166/slyt166.pdf


Regards, Dana.

[David Hess]

The 5532 has a high input bias current anyway so it would take a lot of leakage to become a problem.  In audio applications there is an assumption that AC coupling at some point will remove DC errors so bias current and leakage are of secondary importance.

Low bias current is used where high impedance is needed to keep capacitor values reasonable.  In low cost applications, FET input operational amplifiers are used not for their low input bias current although this does make design easier but because of their AC performance advantage from lower input stage transconductance.  (1) Some people think they sound better but I am dubious.

(1) Do a search for transconductance or Gm reduction in operational amplifiers.  This is a good place to start and it is interesting to study how the 5532 achieved its AC performance.

[exe]

Thank you very much for the link.

I cleaned the flux but, unfortunately, I confused wires and applied to the inputs 5 volts below the ground. The result -- no real changes in offset voltage (it become slightly worse, but after thermal cycle it become slightly better, so, I guess, mechanical stress?). But the input current now just like a wild between 2 and 4uA. I think I blowed the input diodes (even though my PSU was set to source only a few mA a current, but it has 10uF output cap, so peak discharge current is high).

So I took another TI opamp and... it has Vos 1.26mV and 0.68mV (still within spec). Way much worse than the previous sample from TI. Damn, I destroyed my best opamp

One thing surprised me is that thermal cycle may change Vos after cool down. On the picture below you can see I managed to get zero offset (according to my uni-t 61, opamp connected in unity gain configuration, supply voltage 17.5V unipolar, input voltage 5V). Although I'm not sure if this is a permanent state.

[David Hess]

It can be a tricky to measure the input offset voltage and bias current directly from the input terminals; the capacitance of a long lead at the inverting input can cause oscillation or RF can get in and be rectified.

Usually offset and bias current are measured at the output in a configuration where the operational amplifier amplifies its own offset by 100, 1000, or any convenient fixed amount.  Bias current is measured by adding a known resistance in series with one input at a time and measuring the change in offset.

[exe]

Thank you very much for suggestion with the resistor. I actually only measured input current only at positive input (the sum of both inputs, I connected positive inputs together). So, perhaps, it's time to repeat measurements again, but in a proper way.

I used unity-gain configuration because I don't have precision resistors. But now I think I may want to try it. I will not get accurate numbers, but I can compare opamps between themselves (provided I use the same resistors).

[danadak]

One approach to measurement -


http://www.analog.com/en/analog-dialogue/articles/simple-op-amp-measurements.html


Regards, Dana.

[Zero999]

In an op-amp with a BJT input stage, the bias current has a negative temperature coefficient. As the temperature is increased, the Hfe of the input transistors also increases, so they sip less base current in order to maintain the same collector current, which is normally fixed by a current source/sink inside the op-amp.

Whilst it's very good to understand this, you shouldn't lose sight of the fact that the NE5532 was never designed to be a precision op-amp.  It was primarily designed for used in audio amplifiers where the DC component is removed with AC coupling capacitors. Offsets due to the bias currents can also be minimised by making sure both inputs see the same impedance, with respect to 0V.

[David Hess]

That is called bias current compensation and operational amplifiers intended for audio applications rarely include this feature for the reason you identify, AC coupling makes DC errors irrelevant, and because the bias current compensation both increases noise and usually increases distortion.

There are ways to implement bias current compensation externally if necessary.

[Zero999]

That is called bias current compensation and operational amplifiers intended for audio applications rarely include this feature for the reason you identify, AC coupling makes DC errors irrelevant, and because the bias current compensation both increases noise and usually increases distortion.

There are ways to implement bias current compensation externally if necessary.

I think you've misunderstood. I made no reference at all to bias current compensation. I just stated that: for BJT amplifiers the bias currents have an NTC, how to mitigate them and briefly explained why.