Surface mount alternative to LMC6001

Hi Louis,

Note that the LMP7721 has a non-standard pinout - so it is not a "drop-in" replacement for other SMT devices. It also has a servo-guarded ESD input stage, which gives it the fA bias current across the lower VCM range.

The LMC6081AM would be an ideal SMT replacement for the LMC6001.

 www.national.com/mpf/LM/LMC6081.html

Why? The dirty little secret is that the LMC6001 is based on the LMC681 and undergoes a full room temp Ibias test to make sure the bias current is under 25fA - which can take up to a minute - which is also why it is so much more expensive.

Other notable devices are the LMC662 and LMC6482 series. They do not have singles, but I recommend using the second channel of a dual for the best layout and leakages (use CH1 as a guard driver). The LMC6482 adds rail-to-rail input with almost no bias current penalty.

Most any of the LMC6xxx series will have less than 30fA at room temperature, many less than 10fA. We do not guarantee the fA bias currents on the standard LMC series - but they tend to be very good "right out of the box". Just be sure they are gently handled and kept clean/dry.

The LMC6041, LMC6482 & LMC662 are my  femtoamp favorites, and I have yet to see a fresh one with more than 20fA bias current. If you need a single, check out the LPC661 (a single micro-power 662).

FYI: Board cleaning will become more critical with SOIC devices. Your lowest leakages will be when using DIP's. The DIP lead spacing is wider and there is more room for guards and spaces between the pins. I would not go any smaller than SOIC, since it is very difficult to run guards between TSSOP/SC-70 pins.

Regards,

Paul Grohe

SVA Precision Applications

Thanks for the dirty details.  Looks like I should have searched harder.  The 6041 looks great because we don't particularly need the rail-to-rail inputs, but I'll keep the 6482 in mind.  Since I'm using this stage basically for impedance translation (voltage follower), is there an advantage to buffering the gaurd ring with a second amplifier?  The app note suggestion indicates that simply tying the guard ring to the feedback net is sufficient.  Also, is there a silicon reason why the second channel in the dual amp package is superior, or is it simply that the pins are in a better position for layout on that channel?

Are any of these devices in a position for long-term availability?  The LMC6482 seems to be a little better stocked by distributors, but is either one significantly more popular?

Sorry if these are obvious questions.  I'm new to working with signals at this kind of sensitivity and want to try an get a wholistic understanding to cut down on respins.

Thanks,
Louis

Hi Louis,

If it is just a High-Z buffer, then you want the guard potential near, or juuuust slightly below your average input voltage. You could do that with a slightly attenuated version of the buffer output. You do not want  a gain of 1 or more since it will lead to oscillations (positive feedback) since the guard is essentially cap-coupled back into the input.

Normally you would connect the guard to the summing node potential (inverting input), but since this is a buffer, the inverting input is connected to the output. But you do not want to apply the output directly to the guard because of the reasons stated above. So you can either attenuate slightly with a resistor divider (~0.97 division ratio) or buffer the attenuated signal with the second amp to have a low impedance driver for the guard. In short - you can drive the guard off the resistor divider, or, buffer it with the unused amp. Beware of AC peaking!

If the guard is a small, local trace to the sensor, then the resistive tap is probably good enough. But if you are driving large guard planes, parallel guard traces or triax cable, then a guard buffer should be used (with appropriate output current limiting resistor).

The amp is there - basically for free - so you might as well use it.

There is no "silicon" reason to use the second amp - it is more because of the package layout of the second channel. For the second channel, the inputs are in the lower corner away from the power supply pins. On CH1, the +IN pin is right next to V- with little room to guard. For the buffer app, the CH2 +IN pin is in the bottom right, "guarded" by the inverting pin above with plenty of room to slip a guard or shield between the V- pin on the other side. For inverting or transimpedance amps, use CH1.

The LMC6482/4 has always very popular part for 18 years. It's unique "sliding" single-stage rail-to-rail input eliminates the usual abrupt common mode offset step. The single input stage also reduces internal leakages (lower bias currents) and has the input capacitance of a single stage (~3pF). This all makes it hard to replace...

TI has committed to keeping the National parts and keeping them in their original National fabs, which will keep the quality consistent. I do not see the LMC6482 disappearing anytime soon. So you need not worry...

Regards,

Paul Grohe

SVA Precision Applications

Hi Paul,

Sorry for the delay in the reply.  This project goes up and down in priority pretty quickly.  The LMC6482 sounds like the way to go.

My background is as a digital engineer, and I'm finding that as I go, the layout is coupled much more closely tied to the schematics in order to get the expected performance then with low/mid-speed digital designs.  I was wondering if you had any suggestions for resources to study analog layout?  I've had The Circuit Designer's Companion recommended to me, though that seems more of a generic text.

Thanks,
Louis

Hello Louis,

I was incorrect...In reviewing the LMC6001 binder, the LMC6001 is based on the LMC6081, with a slight metal routing tweak for better ibias performance.

I have corrected my original post above to keep the records straight.

My apologies...

Regards,

Paul Grohe

SVA Precision Applications

Thanks for following up with those corrections!

I have two questions:

1) Why is buffering the guard trace and improvement?  Shouldn't the output from the op amp be low impedance already?  If we're worried about stability, wouldn't you still need to have a slight attenuation in the buffer before it's output is used as a guard trace?

2) I am not wedded to using the LMC6482 only as a buffer.  Would there be stability/accuracy benefits of using it as an inverter, or as a small amplifier (Ag = 2 or 3)?  The input is A/C coupled in order to level shift it between the rails, so I could use an inverter and not need to use such large resistors in biasing (since the input signal has a very high source impedance).  All of these configurations simulate similarly, but I'm not certain which is superior in a real PCB.

I'm currently reading through http://www.analog.com/library/analogdialogue/archives/43-09/EDch%2012%20pc%20issues.pdf.

Thanks,
Louis

Hello Louis,

Driving the guard with the output of the signal amplifier is adding more capacitive load to the output. This extra capacitance can cause output peaking if it is over a few pF. Since you have a unity gain buffer, you have the configuration most susceptible to output peaking due to output capacitive loading.

Using an output derived guard (again, assuming a "unity" gain stage), a small amount of attenuation is recommended. But even with a little attenuation, the peaking can still cause the guard to go over unity.

To stop the peaking, the loop usually needs to be compensated or slowed down (by capacitor across RF) - which means you now sacrifice signal BW to gain output stability - OR you can increase the values of the divider string so that the "top" divider resistor effectively isolates the guard capacitance.

However, this also increases the driving impedance of the guard and weakens the "guarding" ability at anything over DC. You could add a bypass capacitor across the guard to lower AC impedance to ground, but that can kill the "bootstrapping" effect at AC by introducing a large lag.

By using an external buffer the main amplifier is saved from having to drive the extra capacitance. The buffer can then be properly compensated to drive the guard without peaking.

Remember that the guard also works both ways. The "captured" guard noise has to go somewhere..

If you use a guard derived from the signal amp output - any AC noise or large currents the guard is "protecting" the input traces from will now appear on the amp output. Remember that the output impedance of an op-amp goes up with frequency - so you will not always have an "ideal" low impedance over all frequencies. HF "junk" can get into your signal chain. Most low-leakage applications are fairly low frequency - so "junk" like 50-100KHz switcher noise can travel along the guard with little attenuation. If your guard "wiggles", your input will "wiggle" (and integrate the noise).

Again, there is no fixed rule. It all depends on the particular circuit, layout and application. If you are just driving a thin trace around a photodiode, then the output derived guard would be fine - you may not even need to attenuate.

So:

IF you are guarding a few mm worth of traces between the amp input and the sensor - than the output derived guard should be fine. It may not even need the attenuation.

IF you are driving more than 1-2 cm of traces, then it should be attenuated/resistor isolated from the output.

IF you are driving long guard traces, planes, cables or shields, or going off-board, then the guard should be buffered.

As to #2...By configuring the guard driver as a inverter, or with a positive gain, you are slowing down the loop to gain stability by increasing the closed loop gain - basically the same as throwing a cap across a feedback resistor of a buffer.

Sections 12.16-12.20, 12.27-12.28 and 12.31 in the ADI text you referenced are applicable.

Regards,

Paul Grohe

SVA Precision Applications

Hi Paul,

Can we power LMC6482AI from +5 to -5V dual-supply? I could not find any specifications recommending powering it from dual-supply in the datasheet, other than the specified maximum supply voltage (V+ − V−) = 16V.

Kindly suggest.

Regards,

Sinoj

Hello Sinoj,

Sure. The LMC6482 will work just fine at ±5V (10V). It is specified up to 15V (±7.5V).

There has not been a "ground" pin on an op-amp in 40 years. The op-amp only cares what happens between the V+ and V- pins - so it does not care if it is run single supply or split supply. There really is no such thing as a "split supply" or "single supply" op-amp - those are just marketing terms.

99.5% of the standard, 5-terminal op-amps out there can be run single or split supply - they do not care.

However, the datasheets can be confusing because as they may present the specifications in single or split supply terms. You have to look at the power supply specifications to see if the datasheet is in "split" or "single" supply terms.

Regards,

Hi Paul,

Thanks for the clarification.

Regards | Sinoj

Dear Paul,

In the LMP7721 Evm Users manual it is mentioned that 10T resistors are available in 1206 form. Can you indicate a source for these please?

Regards, Scott Hamilton.