Differences in packages

[Kontakr]

I'm putting together parts to build a uCurrent, and I noticed that the price of the MAX4239AUT+ on the BOM, which is in a SOT23-6 packaging, costs nearly twice as much as the MAX4239ASA+ in SOIC-8. My understanding is that packaging has negligible effect on the electrical characteristics of the chip, given that dies and pinouts are identical. Is this correct?

[Nauris]

Yes effects are small and generally bigger package gives better dc performance as die stresses are more uniform and stable.

[rstofer]

Dave just did a "Murphy's Law" on this op amp.  He shipped a bunch of units using a 'slightly' different part number and the gadget went into oscillation.  Best to order EXACTLY what is on the BOM.  Down to the last letter...

[ovnr]

It should be noted that there can be performance differences between packages - for instance, the MAX6126 voltage reference has roughly twice as good performance (tempco and initial accuracy) in SO8 compared to SSOP-8.

[Dave]

Dave just did a "Murphy's Law" on this op amp.  He shipped a bunch of units using a 'slightly' different part number and the gadget went into oscillation.  Best to order EXACTLY what is on the BOM.  Down to the last letter...

The circuit oscillated because the design was flawed, choosing a part from a different manufacturer just exposed the problem.

[Kontakr]

It should be noted that there can be performance differences between packages - for instance, the MAX6126 voltage reference has roughly twice as good performance (tempco and initial accuracy) in SO8 compared to SSOP-8.

Generally that would be noted in the datasheet like they do for that part, right?

[David Hess]

The circuit oscillated because the design was flawed, choosing a part from a different manufacturer just exposed the problem.

I agree.  This should have been a warning that there was a problem with the circuit or possibly layout.

The circuit itself has poor power supply decoupling and especially so with the rail splitter providing the virtual ground without decoupling so this sort of problem does not surprise me at all.  Tests would likely have shown marginal stability.

[SiliconWizard]

Dave just did a "Murphy's Law" on this op amp.  He shipped a bunch of units using a 'slightly' different part number and the gadget went into oscillation.  Best to order EXACTLY what is on the BOM.  Down to the last letter...

The circuit oscillated because the design was flawed, choosing a part from a different manufacturer just exposed the problem.

I also agree on a general level, and the design would certainly not be right for a full-fledged, high-production volume lab instrument, but I would still cut Dave some slack on this.
When designing a decent precision device with such a low-cost, low-count part, low-power design, you have to make some compromises, which he did. Relying on very specific parts in that case may make sense, even though that can backfire at some point, and that is exactly what happened.

That said, the fact his testing never showed the problem (until he specifically worked on it) while testing the final units that he was shipping (I think he does 100% testing before shipping), obviously shows that the test setups were lacking in representativeness.

That warrants a suggestion for future video topics for Dave: one about the specific topic of power rail split circuits (and their potential pitfalls), and one about testing in general and designing test rigs that take reasonable end-user scenarios into account.

[Croy9000]

So as a tl;dr. Is the problem fixable with some tweaks to the design? Or should I just stick with the original design, and source the identical parts?

If the design is fixable, has anyone taken a crack at it?

Steve

[mbless]

Dave just did a "Murphy's Law" on this op amp.  He shipped a bunch of units using a 'slightly' different part number and the gadget went into oscillation.  Best to order EXACTLY what is on the BOM.  Down to the last letter...

No, it was the virtual ground op amp that caused the issues.

[Dave]

When designing a decent precision device with such a low-cost, low-count part, low-power design, you have to make some compromises, which he did. Relying on very specific parts in that case may make sense, even though that can backfire at some point, and that is exactly what happened.

You can't possibly argue that the decoupling was screwed because he was saving on components. You don't compromise a design to save the cost of one capacitor.

If the design is fixable, has anyone taken a crack at it?

See this post. Eliminate one of the two 1nF capacitors on the ground splitter resistor divider. The post below mine explains why a single capacitor would be a better solution.

[David Hess]

So as a tl;dr. Is the problem fixable with some tweaks to the design? Or should I just stick with the original design, and source the identical parts?

Sure it is fixable; add a couple of bulk decoupling capacitors and short out the series resistor at the output of the rail splitter unless it is required for current limiting.  The series resistance is usually not needed to prevent oscillation of the operational amplifier if dominant pole compensation is used with a bulk decoupling capacitor.  You sometimes find this done to comparators to use them as operational amplifiers but it applies to operational amplifier as well.