What is the advantage of servoing instead of directly measure?

Question

I was looking at the LT6118 datasheet, since the TI amps (like te INA381) I am used to are not available. They mostly do the same thing, and probably are intended as direct competitors, IMHO. The job is moderately simple, amplify the signal of a current shunt and offer a trip comparator.

However it seems that the architecture is completely different:

• TI uses a (probably chopper stabilized) instrumentation amplifier to directly amplify the voltage drop on the shunt (gain is factory selected) and that's it, they output the voltage; they probably use laser trimmed dividers inside to achieve the required common mode range;

• Linear/Analog instead servoes a current source to balance the shunt with an external resistor, making in fact some kind of automatic bridge; the same current needs to be developed externally into an output signal; gain is user controllable by resistor; the CM range is, frankly, a mess.

What would be the advantage of such an 'indirect' measurement technique? It's only a design choice or philosophy? My first guess is that the current servo affords more bandwidth (the zero drift chopper restrict it) and maybe there is some self cancellation in the amp errors. But on the other hand gain accuracy is mostly delegated to the external resistor (the 'second arm' of the bridge). Also there is a significant power dissipation involved with the current source.

Maybe it's just that TI does better chopper amps while Analog is better on some other kind of topology (they to the over the top amp so I don't think that would be a good reason)

Just curious, am I missing something?

Answer 1

The LT (voltage-controlled current source) approach is a great way to get very high common mode rejection without exotic techniques. I've implemented this method with RR-in op-amps rather than using boutique parts (in that particular case I wanted a part that was rad tested). I would say it's the default method. Accuracy is first order dependent on resistor values. Usually we're not all that concerned about current accuracy (and the shut accuracy is usually the limiting factor) so this is fine. The nominal rail voltage is typically known and fixed.

We are not shown what methods are used to implement the in-amp in the TI part. Just a mysterious block. If a conventional 3-amplifier in-amp was used (with some method of dealing with common mode range, forget that for now) the accuracy becomes dependent on the ratio of rail voltage to sense voltage times the resistor tolerance, which is a much bigger sensitivity (due to common-mode rejection). We usually want the sense voltage to be as small as practical, of course, so that ratio will be as large as practical. There would still be a "servo" with currents related to the sense voltage, of course, just hidden inside that in-amp block. Of course I suspect they are doing something else in there, probably involving capacitors and switches. It's not easy to do this outside of a chip. If you can use that part and if you can get it.. .it seems to be aimed at volume markets so price/performance is excellent.

As to cost difference- anything starting with "LT" will be premium-priced- that's what they do. If a function becomes commoditized they are not interested. The LT part, being 60V, is useful for 48V systems, for example.