LMC6001: lmc6001

Hello Craig,

What type of 10pF capacitor are you using? How are you switching ranges? How is the signal getting delivered to the amplifier (Coax? Connectors?).

What happens to the settling time if just the capacitor and resistor are connected to the input pin? What if you remove the capacitor?

Have you looked at the amplifier output with a scope to make sure it is not oscillating?

Does it eventually settle at the correct reading? Does the step always go in the same direction?

Avoid 'railing" the output for extended periods as it puts a large voltage across the capacitor and aggravates the settling time due to capacitor "soakage".

Ideally, there is a "zero check" switch around the resistor to discharge the capacitor and keep the system at zero until the measurement is ready. You don't want the output "flopping" around as you change ranges.

Generally, measurements in the pA or below range are done by integration. Have a look at the IVC102.

Can you provide a schematic and a *photo* of the layout?

Regards,

to TI

Hi Paul, Thanks for helping out.



First let me tell you I a m working with a nuclear instrument electrical engineer that has 50 years design experience. This is his design and I am helping him.

The answer to the questions are as follows.




The feedback capacitor currently is a Corning glass capacitor of 10pF and may change to a 5pF. Jim (the engineer) has tried other types including polystyrene and ceramic. The results are the same.




The input connection is made with coax cable.




There is no difference in the settling time if you disconnect the capacitor from the feedback resistor.




There is no oscillations on the output with virtually no noise or hum.




Yes, it does finally settle to the final value. After power on, the output goes to +4.1 volts and decays from there. It appears to be a single rc response (single pole). If you cycle power after the required settling time, the output performs the same. It immediately to to +4.1 volts and decays in the same manner. Always goes toward the positive rail. The settling time is far longer than the circuit’s rc time constant .5 sec or 1 sec depending on capacitor value.




Jim is relative sure that it not capacitor soaking. He is thinking about looking at the noise contribution of the 100G feedback resistor.




JIm’s feeling is that is it a phenomena internal to the chip during application of power like building a charge on some gates that some how get discharge over time.




A 4.4e_14 current at the input give a 1volt full scale output.




Jim looked at the data sheet of IVC102 and has discounted it for 2 major reasons for this application: 1) the input bias current is about 4 times higher than the LMC6001A version, and the internal capacitors total to a value much higher than the 5-10pf he is using at our required lowest current range and therefore the time constant will be much longer.




Please let me know your thoughts with this added information. We are open to all ideas and suggestions.

Since I tried to send this reply, we disconnected all the circuitry from the LMC6001A with the exception of the 100G resistor and the 10pF capacitor and it made no difference. Therefore it is our belief that it must be related to a charge build up in the chip structure itself when the bipolar voltages are getting distributed at turn on time. It take a good 10 minutes to get to .1% settling time and much longer to completely settle out.

Thank you for helping out.

Craig Shull

Hi Craig,

Yes. At turn-on there will be a large transient. This is unavoidable. There are ESD structures from the inputs to the supplies - common to almost all CMOS amplifiers. Each of these structures has a few pF of capacitance that will inject some charge as the supplies move. You will find that other CMOS amps will do the same (older JFET based amps lacking ESD structures may not have this issue - but will still see some turn-on transients).

I have used the LMC6001 and it's LMC60xx siblings down to the fA levels, so I know they can do it. The DIP with a lifted-leg is about as clean as you can get for 1pA. The metal package does have longer legs and is easier to manually wire - but you probably won't see much difference between the two *as long as* you properly clean the soldered leg. I always poked the leg through a piece of paper to protect against any flux/smoke residue/splatter, and to soak up any flux that tried to go up the leg. Then tear it off when done soldering then clean.

In integrator and electrometer applications, the amplifier needs to be held in a buffer or low-gain configuration until the measurement is ready to be made. This is commonly known as 'zero check'.

The simplest way to do this is to short the large feedback resistor with a low value (say 1kohm to 100kohm - but not zero) until the measurement is ready to be made. This essentially makes it a follower and keeps the amplifier in a controlled state during power-up.

Switching in the resistor is the fun part. Ideally this is done with a low leakage reed relay (as Keithely & HP do it), a mechanically operated switch with a grounded/guarded rod., or JFET's or MOSFETS.

JFET and MOSFETs can have gate charge that will aggravate the settling time. The trick with the MOSFET/JFET - and even the coil on the relay - is to not use a fast edge on the control signal. The 100ns risetime of a logic signal will blast through the few pF gate capacitance. Use a R-C network to slow it down.

As for the capacitor - Glass and Ceramic can be piezoelectric. Avoid any stress or vibrations as they will add charge. Some poly capacitors are worse than others. The best are polypropylene and PTFE (Teflon).

For low value capacitors, my favorite low leakage capacitor is a length of Teflon coax (RG-188) or flexible microwave hardline. For the RG-188, it only takes a few cm to get 10pF (you can coil it up). The coax is self guarding (connect the shield to the amp output side), and the inner conductor to the summing node (you can also have the coax double-duty as an input cable, too). Using a good cap meter, you can pre-trim the value to exactly what you want by cutting it slightly long and trimming the end until the value is correct.

In a pinch (pun intended), you can twist two pieces of Teflon wire together (preferably solid core).

If you also put a switch in series with the 100G resistor, then you could choose either trans-impedance or integration mode. Integration is commonly done at <10pA due to the noise involved with the required large value resistors. We use the IVC in our testers and can easily resolve 1pA given enough time.

You will always have a baseline leakage. It takes a lot of work to get 0.000pA!

Regards,

Hi Paul,

Jim Larsen and I would like to thank you for your help. Jim feels that you have confirmed what he thought was the issue. Basically his design consideration and layout approach was very close to your comments. He like the idea you mentioned about using the coax trick and will consider that approach in the future. We will look at a polypropylene type cap around the 100G. We believe, all things being equal for this application we will go with what we have and live with a little longer warm up settling time. He still thinks using the LMC6001A is the better choice because of the much lower input bias current compared to the IVC chip when going down to 4.4e-14 current and will not have to deal with a much longer RC time of the internal capacitors.

Once again, thank you for your time and experience in these type of applications.

Craig Shull