AD620 (instrument amp.) input current

OK I guess I just need to drive the inputs with a stiff source.

GH

Fig 36 of the data sheet suggests that, once the feedback loop rails, the transistor BE junctions, or the paralleled protection diodes, will conduct hard. Then you'll see roughly 800 ohms + Rg plus a couple of diode drops between input pins.

Lots of diffamps and conventional opamps draw lots of input current when they run open-loop. Back-to-back diodes between the inputs are common, to protect the front-end bipolars from zener damage.

I've even seen a comparator with antiparallel diodes between the inputs! Guess who would be dumb enough to do that. Starts with "M".

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

So when the output rails the data is useless (other than knowing what direction it railed), right? And when the signal is within range again, the output works properly, right? So You're good to go.

--
 Thanks, 
    - Win

Precision Op-Amp situations usually try to avoid output saturation and inpu ts going too close to the rails. Even though the specifications may say th is is permissible, the ultimate performance will often be degraded. Satura ted output stages usually take time to recover, and this can be highly asym metrical depending if sourcing/sinking and loading. Input bias current is t rue only over the common-mode input range and differential maximum. Limit you input so that your op-amp does saturate and keep your inputs pre-biased within common-mode input range. Although some op-amp input structures may allow over-the-top rail and or under-the-bottom rail, there can be input p rotection diodes that come close to being forward biased. These situations might be possible, but there can be some reduced EMC immunity. It's impos sible to define every situation on a datasheet, so I would encourage you ac tually test your various candidate op-amps over the full range of condition s that you expect it to operate. The more you can "confine" the possible v ariables/combinations/ranges to test for, the easier it is to find an op-am p for a given application.

Assuming it unsticks cleanly. If not, it seems as though you could have hys teresis there.

Cheers

Phil Hobbs

(who vividly remembers a long-ago outing with a FB loop containing two MC34

084 op amps in cascade [noninverting]--their commendably wide output swing exceeded their input CM range, and the whole works locked up.)

No, I send that signal out as a set point (SP) monitor. Which is then wacko. Another opamp cures the problem.

The 1 k ohm resistor is there because I wanted a voltage limit on the SP. I used Speff's circuit, (I can't spell sphero, unless I look it up.) to make sure the SP stayed above ~1V, |\

---Vref=1V----|+\ | >--+ +--|-/ | | |/ | +--K------+ | diode, I hope the right way. R1 (10 ohms, may not be needed.) | +---> in(-) AD620 | +--> to SP monitor | R2 1k ohm | SP in-----+ from opamp (opa2134)

Some better/other way to clamp a voltage? I could put the SP monitor on the other side of the

1 k ohm resistor (R2) but then you wouldn't see the voltage clamp.

George H.

Right, is that the one with integrator opamps feeding the diff pair? (I didn't understand the intergator's.... are the pseudo-inductors?) But once the output railed it took about 2V (differential) for me to start seeing current flow.

puts going too close to the rails. Even though the specifications may say this is permissible, the ultimate performance will often be degraded. Satu rated output stages usually take time to recover, and this can be highly as ymmetrical depending if sourcing/sinking and loading. Input bias current is true only over the common-mode input range and differential maximum. Limi t you input so that your op-amp does saturate and keep your inputs pre-bias ed within common-mode input range. Although some op-amp input structures m ay allow over-the-top rail and or under-the-bottom rail, there can be input protection diodes that come close to being forward biased. These situatio ns might be possible, but there can be some reduced EMC immunity. It's imp ossible to define every situation on a datasheet, so I would encourage you actually test your various candidate op-amps over the full range of conditi ons that you expect it to operate. The more you can "confine" the possible variables/combinations/ranges to test for, the easier it is to find an op- amp for a given application.

Thanks Terry, I make stuff to sell to physics students. They will explore the entire parameter space, I often only find the edges of my design during testing.

George H.

ysteresis there.

34084 op amps in cascade [noninverting]--their commendably wide output swin g exceeded their input CM range, and the whole works locked up.)

Another int amp (classic 3-opamp) would fix it too. (I think) I use the AD620 lots, it survives my abuses. It's not all that quiet.

George H.

inputs going too close to the rails. Even though the specifications may sa y this is permissible, the ultimate performance will often be degraded. Sa turated output stages usually take time to recover, and this can be highly asymmetrical depending if sourcing/sinking and loading. Input bias current is true only over the common-mode input range and differential maximum. Li mit you input so that your op-amp does saturate and keep your inputs pre-bi ased within common-mode input range. Although some op-amp input structures may allow over-the-top rail and or under-the-bottom rail, there can be inp ut protection diodes that come close to being forward biased. These situat ions might be possible, but there can be some reduced EMC immunity. It's i mpossible to define every situation on a datasheet, so I would encourage yo u actually test your various candidate op-amps over the full range of condi tions that you expect it to operate. The more you can "confine" the possib le variables/combinations/ranges to test for, the easier it is to find an o p-amp for a given application.

Graduate students are worse. They will "improve" the design.

You aren't paranoid enough.

--
Bill Sloman, Sydney