Shunt Reference Problems 2

Feb 22, 2020 1 Replies

Rate this thread:

A few days ago I posted Shunt Reference Problems re. issues I was having with ADR530 shunt regulators. Thanks for many suggestions, prize to Phil Hobbs who pointed out two issues:

1) most (maybe all) suppliers incorrectly/misleadingly talk about ppm/C when that isn't what they mean. Some of them do explain clearly that they are taking the pk -pk deviation and dividing it by the full device working temperature range. Maxim and ADI data sheets for modern parts describe this as the box method. Next time I'm stopped for speeding I'll expalin to the officier that it's OK because I use the box method to calculate my speed :-)

2) Other things are going on which are probably die stress as suggested by Phil, or maybe humidity but I doubt it because some of the worst effects are when devices are hot and the Env chamber is pretty dry.

I tested two each of Microchip LM4040, ADR530, LTC6652A, ADR525 over the range -40 to +85C.

Plots are here:

formatting link

This isn't strictly correct for the LTC6652A which should go up to 125C but it isn't a shunt regualtor ans is included to show that may measuring system is up to the task. (I know it doesn't prove that beyond all possible doubt but it helps.)

The Microchip parts are dead cheap and meet thier own spec. The response on on of them is very lumpy and I've seen this effect in a real circuit.

The ADR530s are very similar, pretty wll behaved and out of spec.

The LT6652A are well in spec, obviously compensated in rather more complicated way than the other parts. They probably have the worst ratio of dV/dT to "box method" ppm/C of all the parts. The shape of the V/T curve is quite like the typical examples in the data sheet.

The two ADR525s are amazingly similar (there are two lines on that plot) and in spec.

If all the ADR5XX parts are as nicely behaved (apart from the little matter of not quite meeting the spec) they would be easy to calibrate.

The actual "box method" ppm for the 8 parts are:

LM4040 51, 18 ADR530 46, 51 LTC6652 3.9, 2.9 ADR525 30, 30

Thanks again for suggestions, I hope some of this is useful.

Yup. One must always bear in mind that datasheets are marketing documents.

Video is a classical example--a display with N by M colour pixels is quoted as NM pixels. A camera with the same layout is quoted as 4NM, because each colour pixel takes four filtered monochrome pixels. I call this "marketing megapixels".

(This despite the presence of a crossed walkoff plate that smears out the image by one pixel in each direction to eliminate colour jaggies at the edges of bright regions.)

In the voltage reference case there's some justification, because higher-order temperature compensation produces Chebyshev-like error curves--tighter limits but maybe not much difference in max |dV/dT|.

Cheers

Phil Hobbs

Shunt Reference Problems 2

Join the Discussion

Didn't find your answer?