LMC6001 Thermal Stabilzation Time

Hello David,

All non-chopper amplifiers exhibit some power-on drift as they 'warm up". Of course, this is influenced by many factors - Ambient Temp, internal quiescent current dissipation, thermal conductivity (PCB area), package type (theta-Ja), output load (pdiss).

You will see some precision devices actually specify their offset after 10 seconds, or some as high as 20 minutes, after power on. Some precision devices will even show a typical power-on drift plot in their datasheet...

How much drift are you seeing? How long are you waiting? What is the total gain? What is the load on the output? Is the drift identical each time?

The "drift" can also be caused by thermal effects external to the amplifier. Is there a regulator or other heat-generating device near, or below, the amplifier circuit? Just the junction of the device leads and the solder and trace can generate several nanovolts each (depending on the metal combination). Multiply that by the gain and you can have significant drift at the output.

Regards,

You asked some good questions that are tough to answer. The first part is that the circuit is an integrator used to measure the E field due to an electret. So, it's far from a normal application. As a matter of fact it's one of those that when taking electronics the prof said would never work in a real world!

Just as an FYI if it helps any, back in 2007 I had worked with Merrill Johnson, then at National (Merrill.Johnson@nsc.com) on some problems with this circuit. At that time we were seeing oscillations and he was able to help. If he happens to be around any more and he hung on to them he has the schematics, assembly, etc.

The onservation about an offset is coming from the customer who observed that the readings seemed to be more stable after the unit had been on "a while". In most applications,the user turns on the unit makes a few (10 or so)  readings and allows the unit to turn itself off after a 30 second inactivity timeout. In this case, the customer has a need to make many readings over an extended time, so it never reaches the timeout. So the diea of stability is a percieved one.

What I was hoping for was some idea to try. For example, if we turned on the unit and didn't allow readings for 30 seconds (inhibit the timeout) would we expect any better performance?

To answer the other questions, the load is into a 20K input resistor of an MCP6273 OpAmp used to buffer and add offset. The lead from the LMC6001 to the 20K resistor is approximately 1" long. The regulators for the +/- 5V are just over 3/4" away. They only power the LMC6001 and a few ancilliary circuits - no significant power dissipation.

Any thoughts welcome!

Hi David,

Merril retired several years ago...and the Texas CRC was closed a few years after that..so tracking down his old files may be a challenge...

It sounds more like stored charges or "soakage" is creeping back into your signal path. It can take a while for these charges to "bleed" off. This "charge" can be stored in the insulation materials around the input.

Is there any physical shock or vibrations to the element? Piezoelectric crystals can generate large voltages (>10V to 100'sV) when physically shocked. This can charge the input stray capacitance and "bleed" back into your signal path (See "capacitive soakage").

Coax cable on the input can also cause similar problems.

Try grounding the input while the power is off. See if the "drift" goes away.

The problem with circuits like this (Gohm high impedance circuits) is that the schematic is only part of the story. Layout and materials used also contribute to the circuit behavior.

Have a look through some articles I wrote in EDN about low current techniques...see if anything rings some bells and make sure you are not violating any of the suggestions...

http://www.edn.com/design/analog/4368681/Design-femtoampere-circuits-with-low-leakage-part-one

http://www.edn.com/design/analog/4375459/Design-femtoampere-circuits-with-low-leakage---Part-2--Component-selection

http://www.edn.com/design/analog/4395651/Design-femtoampere-circuits-with-low-leakage---Part-3--Low-current-design-techniques

And the "bible" of low current techniques from Keithley (particularly chapter 2):

 http://www.keithley.com/knowledgecenter/knowledgecenter_pdf/LowLevMsHandbk_1.pdf

If you wish, post the schematic AND a picture of the layout. If you do not want to post publicly, you can send it to me directly at paul<DOT>grohe<AT>ti.com.

Regards,

I recall reading the EDN series with interest as it was related to this project and I found the Keithly "teaser" to be interesting. However, Kewithly seems to be having problems with their web site right now and I've not been able to establish a login, so I'll need to wait and try again later.

From the little I could read without downloading the full text is that our circuit would fall in the coulombmeter category. After I have a chance to review the Keithly material, I'd like to follow up in private e-mails. There are some minor problems with our circuit that I'd like to pursue. Since I've been working with this circuit for almost 25 years, I think I've started to understand the nuances and I feel we can address some of them rather than just living with them.

Thanks

Hi David,

I use LMC6001 since about a half year, trying to make the performace a bit better.

It is in a transimpedance amplifiet topology.

My experience is: the output gets stable some 15 - 20 minutes after turn-on. In our application, the electronics is heated up to 35 Celsius, but this temperature is stable after 5 minutes.

After this 20 minute period, LMC6001 is stable through a whole week, but only when not turned off.

Temperature does effect the output voltage. In my circuit the input bias current procuces the significant portion of the error signal.

Our application is battery powered, so time is important. I was wondering how much "drift" you are seeing from the time of turn on until it becomes stable? Are we talking micro- or milli-volts?

Thanks

I talk about millivolts. Max 5 - 10 mV, depending on device. Please don't forget: this drift is not only due to the offset, it includes the variation of bias current, multiplied by a teraohm-resistor also. It looks like an exponential curve, when a capacitor discharges. Unfortunately it is not a capacitance somewhere, it is made by the amplifier itself.

LMP7721 stabilizes much more quickly, in a few second, but in longer term (days long) the overall stability is somewhat worse than that of the LMC6601. I say 1-2 mV change is possible. LMC6001 was one year before also better.

I also have to say that none of the devices are going outside the specified parameters, but in our application, it is not enough.

In our case, the circuit is an integrator which is held in reset most of the time. I'll have to investigate the effect of a few millivolts.

Thanks for the input.

Hallo David,

please consider, my application is a transimpedance amplifier with 1TOhm feedback. It is never held in reset. The error on the output is basically due to the bias current flowing through the feedback resistor.

Once I tried to measure the offset of LMC6001, only one piece, a few months before. As I can remember, it was some 70 microvolts or so. I didn't made a longer "offset only" test against time and against temperature.

Hi Paul,

I wanted to get back to you and thank you for the reference to the Keithly "bible". I had more or less forgotten about it until it arrived at my dorrstep yesterday. Most important for me was the discussion on Johnson noise (section 2.6.5) and the realted grpahs which appear at various places in the book. By using the rate of decay, I have been able measure the resistance of our circuit as being around 10^12 Ohms. So, I started to focus on those regions in the charts. I started to realize that we are operating near the "theoretical limit" shown there!

Using the information given in that section, I was able to show that we really can't expect any higher precision that we are getting now. I have been able to put to rest the quest for "more digits" that has consumed far too much time over the years. You read about Johnson noise but tend to forget about it in working with "everyday" electronics.

As far as the possible drift that started this thread, I'm not sure I have any answers but your input has helped narrow the focus.

Thanks again for your help, Dave

Hi David,

Glad I could help.

I mentioned in the articles that the resistor noise will limit the ultimate sensitivity (even had a graph). Most people are just not familiar with these large value resistors, and don't realize that there can be millivolts of noise from these resistors alone.

To get below the resistor noise limit, you have to switch to an integrator. But the integrator brings along it's own set of problems (resetting, charge injection, slow throughput).

Regards,

In the light of the previous comment by Paul Grohe it is perhaps worth reminding readers that though the Johnson noise of the feedback resistor becomes dominant as the resistance increases the signal-to-noise ratio improves as the resistance increases. This arises since the gain increases proportionally to the resistance while the noise only increases as the square root. One of the few (text)books that considers the design of such high gain transimpedance amplifiers is Operational Amplifiers by Jiri Dostal  (ISBN 0-7506-9317-7, Butterworth-Heinemann, 1993). However the limiting factor eventually becomes the shot noise arising from the amplifier input bias current.

Regards, Scott Hamilton.