LMC6001 drift

Hello Bela,

Please define "drift"....How long is the drift? Seconds? Minutes? Approximately how many volts/second?

I this a single or dual supply application? Are there guard traces?

A drifting power supply can affect the input leakages. Are the supply voltages stable? DC power supply rejection ratio goes out the window at high very high impedances due to the ESD diode capacitance and leakages affecting the input directly and bypassing the amplifiers inherent PSRR.

Does the problem follow a particular datecode?

Are the "good" and "bad" devices starting out at the same temperatures? The same input current level? How much output load is there?

Temperature would have the greatest effect on bias current. Bias current for this device tends to increase at 1.9x for every 10°C. Would something be warming the circuit a few degrees after power-up? Most likely the devices you see as "good" have bias currents already close to zero so they don't have far to drift.

Circuits in the T-Ohm range do not stabilize quickly. How quickly do you expect it to stabilize? Do you zero-check (ground) the input during power off?

Are you sure it is the 6001? Does the problem follow the 6001 if you remove the "bad" 6001 and place it in another circuit? Don't rule out the possibility the problem is elsewhere in the circuit.

Are you sure the ion detectors supply voltage is stable? How high is the HV? What happens at power-off? Where does the HV go? If the HV is blead-off by the 6001's ESD diodes, the ESD diodes could be stressed and may be damaged over time. Since the "bias current" of a CMOS amp is the difference in the leakages of the two ESD diodes, if one diode is being degraded, they will not longer track and the leakages will be worse (and more sensitive to temperature).

I am also assuming you have properly cleaned the circuit beyond normal PCB cleaning procedures, which should involve an alcohol or acetone flush followed by a blow-dry and baking.

Regards,

Hallo Paul,

sorry for the late answer, I had to do other tasks as well. I try my best to answer all your questions, but let me write a few words about the overall construction as well:

The ions (a gas) are coming from an ionisation chamber into a sensor assembly driven by a little ventillator. High voltage cannot come in, only the charged particles, but during test, both HV and the ventillator are off. The sensor assembly is hermetically encapsulated into a grounded alu shield. When the lid of this shield is off, the whole measurement worth nothing, it is completely wrong. When the lid is on, I can't access the circuit inside.

The sensor is formed of a sensor material that catches the ions, this way producing some current. It's internal impedance is some 10^15 Ohm, it can be considered as a current source. The current range is from a few tenths of fA to a few pA. We'd like to measure as low a current as possible. LMC6001 is connected as a transimpedance amplifier, pin3 grounded, pin2 receives the input current through a 5M6 resistor and the feedback 1 TOhm. There's no guarding, the pin2 is mounted in the air, both resistors are coming direct to the pin, none touches anything. LMC6001 has a load of 100kOhm, an inverting, unity gain buffer formed of AD8638. Then comes two more AD8638, they create differential signal out of this ground referenced one. This signal can be measured outside of the sensor box, it drives a 24 bit ADC (ADS1251 from TI) and then comes the processor and so on.

I've got a reference sensor, it functions good, and we also have 4 others that are performing well. But: we have some tens more that is not stabil.

When an LMC6001 is replaced, the whole circuit is cleaned up first by some aceton-alcohol mixture (a flux remover material) and then by clean ethanol. Than it goes to a dry chamber and it is there at 50°C for an hour. One other hour is left him to cool down, and than can it be mounted into the equipment.

Drift means that the output has some offset other than null and it is moving, during a few tens of second. Therefore the measurement gets invalid.

It has dual supply, but, interestingly, asymmetric: +7V / -6V. It doesn't drift, stabilized.

Date codes that performing good:  JC87AB (2 piece) ; XR78AC (1 piece) ; XC93AB (2 piece)

The ones that having offset and drift:  JR05ABE0 ; JR17ABE3 ; JR18ABE3 ; XC93AB (1 piece)

Unfortunately, I cannot decode these date-codes, nor have I any effect, wich ones to be purchased. The ICs above came from different distributors.

Turn-on happens always at room temperature, on the same desk in out lab.

The sensor module has a heating inside, I cannot turn it off. It also works when the pieces are good or when they're bad.

We expect a stabil behaviour right at turn-on. Up to 5 second can be tolerated, but it is already a compromissum.

Once the "bad" 6001 is removed it is cut-out from the pcb. It cannot be built into other circuits anymore. We save the pcb, not the IC. It is a DIP version but mounted on SMD pads of the proper size.

Well, I'm not sure it is the 6001 (I'm even not sure of my own name already ... ) but the unity gain buffer and the difference circuit cannot made such a drifting error.

Hi Paul, just another info:

I fed back LMC6001 by 1MOhm instead of 1 TOhm.

Offset is negligible, no drift at all. I read out the display of the whole instrument. I'm already pretty sure that this error comes due to the LMC6001.

The bad news is: this way I lost all the sensitivity, I cannot use this modification. We do need 1V / pA transimpedance. It's been just a test.

Dear Paul,

during the last few month, I could manage to create some long-time tests.

If I could share the results with you, you could better understand what our problem is.

How could I send you a few excel sheets, about the "good" amplifiers and about the "bad" ones as well? They're pretty large, above 10MByte each.

I've got a lot, but I would send only a few ones that can illustrate our problem.

Additionally: yes, this type of error comes with specific date codes. The ones we could purchase today, are wrong, I've got a few older pieces, they're the good ones.

How could you receive these excel diagrams?

Regards:

Bela Istvanfi

Hi Bela,

It sounds like you are doing things correctly.

What happens if you disconnect the sensor from the amplifier (snip the resistor wire) - and just let the 1Meg float in the air?

If it does not drift, then that would rule out the amplifier circuit.

Cleaning must also involve scrubbing between the pins with an acid brush, not just dousing the board with cleaner. Just like brushing your teeth...you have to get the chunks out before the cleaner can do it's job.

I assume you are rinsing the entire IC package, including underneath. Even with the leg in the air, flux can wick down the leg and across the pins or "spray" flux across the pins during soldering.

I have seen drips "hide" under the package and evaporate during baking, leaving a conductive path of burnt flux behind. Use a air gun or air can to immediately (within 5-10 seconds) blow all the ethanol/acetone residue off first before baking. Blow under and between the pins and into all the nooks-and-crannies. Avoid flux removers that leave an oily residue behind...they are difficult to clean completely for these circuits.

You can send the files to paul.grohe@ti.com if you do not wish to post publicly. If you are using Excel 2007 or greater - save it as an Excel binary file (.xlsb) -  that will knock the filesize down significantly - better than zipping..

Also include full schematics (input stage and at least the next stage, as well as biasing and regulator circuits), as well as a photo of your setup if possible.

Regards,