emi hardened opamps

Seen a few of 'em before. Not much data, and haven't seen a test setup for said data. Haven't gone looking for it though.

Guessing the mechanism is input capacitance, ESD diodes, and series resistance; it's not a bug, it's a feature. :)

Tim

The LT1124 is a bipolar op amp. If the nose at the input kicks one side of the long tailed pair more than 50mV out of balance - no matter how briefly - the nonlinearity of the response gets significant, and you can see a DC offset.

The quick and dirty solution is to replace the bipolar amp with something with FET inputs. They don't start showing significant non-linearity for excursions of less than about a volt.

Joerg has posted here on the subject, amongst others.

No, but Joerg has said that fet input's are less susceptible to RF rectification. I have no experience if that's true or not, I assume it's true. I using this fet opamp that has kinda good offset and noise for a fet. ~100uV and 5nV/rtH(?) something like that. (I was thinking opa129 but that's not right, an opa something. :^)

Can you filter the inputs? (Are you sure it's the input? Maybe ground bouncing around.)

George H.

OPA2192... I knew there were, twos and ones and nines.

Just use CMOS opamps.

I need +- 16V supplies and really good DC and noise specs, which knocks out a lot of candidates.

I have a minion researching amps, but I thought I'd ask here too.

AFAIK, you aren't going to be saving a buck, here. Not that that's much of a priority to you--so that should be alright.

OPA171, and 192 if you need the inputs unequal sometimes (rectifier, comparator, etc. service), are the two I know of.

Tim

When I was doing electronics for space, we more or less only had the LT1014, Rad Hard version. But that is probably extremely expensive and it is bipolar, so does perhaps not fit your design

Cheers

Klaus

Keep your opamps. Get rid of the noise.

Get a good spectrum analyzer and find out what frequency the noise is at.

Then design a proper LC input filter to attenuate it by 20 to 40 dB as needed. Measure the response to make sure you are not picking up some ground loop noise or introducing overshoot and ringing.

Check also the supply voltages to the opamps for spikes.

Why? I could understand this thirty years ago but things change. You're really limiting your options all around.

We had a problem with 2.4GHz TDM noise (200Hz) at my PPoE. Filters were only marginally useful. We had at least five spins of the board and added conductive paint (also used for ESD) on the inside of the case but the problem was never really "solved". We tried EMI "hardened" opamps, too. The solution was CMOS but they were too cheap.

The problem cropped up after a "process improvement" increased the Ft of the opamps transistors and then was lot specific (made it hard to chase down the problem).

Make sure he/she looks at the opax192. I haven't really looked closely at it, (tried to measure noise over a wide freq. range) but in general I like it.

George h.

Right, if you have to deal with microwaves, they kind of just get everywhere. Traces, components, in through holes and slots in shields...

You must be very methodical about the construction of the shield, and how signals are allowed to penetrate it.

Peppering low-pF caps everywhere on a board could help. But that's a PITA, and not feasible in many circuits.

Tim

if its a new PWB layout, you can include series Rs (1k or so) directly at the op amp input pins.

keeps the evil spirits out

It's the front end of a 16-channel differential-input data acquisition module, and I need accuracy at high common-mode voltages. I boosted the +-12V VME supplies to +-16.

Some fet opamps have shockingly low PSRR, which has consequences. "Always Invert" is good advice, except when you can't.

The dual version, the LT1013, shares the weird rotated-die pinout of the LT1124, so that would be a drop-in replacement. The front end is very different from the 1124, too, so it might be less RF sensitive.

We'll probably hack a board to try different things on the 16 channels, and compare them. I'll need to come up with a way to blast RF at a VME module in the crate.

It's a VME module, and some of the noise comes from VMEbus digital activity (which we thought we fixed) and some seems to come from adjacent VME modules at customer sites. It's not the sort of paths that show up on schematics.

We might have to hack some channels in various ways and send that module to the customer to test in his system.

TDMA and 200 Hz sounds about GSM, but GSM is at 800/900 MHz and at some frequencies below 2 GHz

When evaluating a prototype RTD interface circuit board (and different opamps to use on it) for sensitivity to GHz interference, I just hooked up a RF signal generator to a cable with a SMA connector on the end, with a ~5pF through hole ceramic capacitor soldered to the centre pin. Then I put a microvoltmeter on the output of the amplifier, and went around the input connector of the circuit board probing each pin with the capacitor lead, and turning the ~1GHz RF on and off a few times to see whether there was any DC shift. For a rough experiment I did not bother connecting anything to the sheath of the coax from the RF generator. I was able to make a lot of improvement that way, even though it is not a very well defined measurement condition.