I just blew a $35 instrumentation amp (differnetial) by accidientally over-voltaging the inputs. Without degrading performance, how can I modify the inputs to ensure this will not happen again?
It was an AMP01 and I thought it had a degree of internal protection. Obviously needs something more.
The classic method is a series resistor followed by diodes to the power rails. Picking the diodes depends on what's important. Can you stand a bit of leakage current? I'm not familiar with the AMP01.
Note that most power supply regulators will let you lift the supply line through the didoes using this scheme. You need to insure there is some sort of protection device on the power supply rails to prevent this from happening. Generally regulated supplies like to source but not sink.
Oh that's interesting Miso, I never thought about that. Is the simple cure to just make sure the power supply is always sourcing current? (A resistor to ground on the supply rail perhaps.)
Hmm, The AD620 is not your 'classic' three opamp instrument amp. So maybe that is different. Is the AMP01 similar to the AD620? Do Zeners have much leakage current?
I was thinking about either, Schottky, PN, or part of a transistor or even a JFET. as a reversed biased diode to the supply rails... Section
7.4 in this is OK..
formatting link
.pdf
You've got to figure out how much leakage current you can tolerate, and then what over voltage you 'expect' and then how much series resistance you can tolerate... It's a bit of a juggling act as is ussual.
Sorry for being stupid John, but could you give me hint as to how this works? Are the gates tied to the source? Or do you need some 'programming' resistors?
I usually just tie gates to source and run them at Idss max, which is in the 1.4 mA ballpark for these parts. Most opamp ESD diodes will be happy with that. There is a tricky way to use one resistor to be effectively in both sources, to reduce max current, but that will add more series resistance and usually isn't needed.
This is the basic depletion-mosfet current limiter:
--------d s----+----s d------ g | g +------+------+
These parts are rated for 500 volts or some such. The opamp power rails have to sink any clamp current, of course.
You could do this with the bigger Supertex parts, for lower series resistance but higher currents; then the added resistor might make sense, to program the current limit.
Jfets would work too, but they usually have horrible parameter spreads and lower voltage ratings. These Supertex parts are slick.
Four series pairs in an MSOP-10 *with* clamp diodes to + and - rails. I'd buy buckets of them. It's shocking that nobody does this.
We're doing a 64-channel diff-input ADC board right now. That's 128 lines that need to be protected. Each line has a series resistor (1/4 of a pack) and two diodes, one to +12 and one to -12. That's 256 diodes! We used the BAV199, the pA leakage version with four diodes per can, but that's still a heap-o parts.
We searched hard for a multiple-diode clamp package, but no luck. There are ones with low-voltage zeners inside, or leaky schottkies.
ADI makes a signal protector, ADG465, but we'd need 128 of them!
I wanted to use a DG408 mux, just for its ESD diodes; that would have clamped 8 lines. But there turned out to be current sneak paths if you actually turn on those diodes, which could have caused channel interactions.
Cool, I thought it was long gone--mine are from about 1990.
I like to use inverse-parallel Schottky rectifiers in series with big (100 nF) bypass caps to protect diode lasers. That reduces the small-signal capacitance by about 100 times, which really helps the control bandwidth, while keeping them essentially immune to ESD damage during handling.
I just finished a laser locker for a client--it uses both temperature and current-tuning to lock a low-chirp DFB telecom laser to an etalon. The two-diode trick let me improve the locking bandwidth by a factor of about 5 without sacrificing protection. That was important, because they only had one of the magic $1k diodes, and the lead time was horrible...I had to debug this new board on the One True Golden Part, which is uncomfortable, but it did work...eventually.
I also always have NC relay contacts across diode lasers in instruments, preferably right on the diode mount. That makes them practically invulnerable.
Probably not what you need, but, way back - about ten years ago, I was working on a research project at the Nat'nl Severe Storms Lab. I was assembling and testing a data acq board that sat on top of the xmtr board. I was looking at the input that measures the statis charge across two aluminum spheres. I noticed there were two diodes on the input to the high gain op amp. They were connected to ground, in opposite polarities. "That's stupid", I thought, They just short the signal to ground". Then I realized the genius of the design: the diodes don't start conducting unil they reach 0.3 votls (or so - they may have been Ge diodes). This went into my bag of tricks to impress someone, if I ever need to protect a high gain, low input amp. I'm guessing you might be able to stack a series of these to allow a higher "window" before they conduct.
As Joerg has mentioned, typically you want to bias them so the input needs to be a solid, say, 1V or whatever before they start conducting (stacking them would be a crude approximation of this): Otherwise, even though the diodes aren't fully on, there's enough current flow that you can get some low-level distortion in your signals. In some cases that might not matter, but for something like an RF receiver, it does.
Icom is guilty of using the straight back-to-back diode approach in the IC-7000 transceiver... removing or biasing those diodes does wonders for the unit's performance -- particularly if you're near high-power commercial AM transmitters: ~1MHz is low enough that the diodes can start conducting significantly.
The resistor to ground works as long as you are not pulling more current out the diode than the resitor sinks. I think a transorb diode of sorts would be better for bulletproof protection.
If you ever tested a part for latch-up, this is a classic issue. If you are going to pull 100ma out the protection diode, you had better be pulling more than that amount of current out the power supply with a parallel resistor. Otherwise you over-voltage the device under test through the protection diode.
It's sometimes better to use separate devices for conducting the current to the clamp and the clamp itself. A TVS might have 1,000 x the capacitance of a switching diode. The clamp might be leakier too- with separate devices you can bias the clamp beyond the operating range.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.