op amp questions

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

tmbg

Well-known member
Joined
Nov 7, 2004
Messages
438
Location
Atlanta, GA
When is it appropriate to use a BJT-input opamp, and when is it appropriate to use a FET input amp?

I was reading through the other concurrent thread about the 60dB 990 stage, and I'm playing with opamp circuits on the simulator, and I realized there's some stuff I don't really know well enough!

FETs have lower bias currents, is that correct?

How do you observe the effects of bias current in a simulator?
 
If your simulator is accurate it will reveal the effects of op amp bias current. Put some R's in series with the input pins and look at the voltage that is dropped across them, both the magnitude and the polarity (or if you have a current probe tool, just look at the currents directly). Do this for what is already a stable configuration with feedback, so the equilibrium conditions are reasonable.

For an LM833 for example, the input stage is a PNP pair, and you should see a voltage drop that reflects this---the amp inputs source current, so the input pin will be more positive than the other end of the resistor.

A 990 uses an NPN input stage, so (if there were a model in a given simulator) you would see the input pin lower in voltage than the other end of the resistor.

Another way: Set up a circuit as an integrator, with a shorting switch across the feedback capacitor and leave the input open. Open the switch and see which way the output voltage ramps; the rate of change in volts per second times the capacitance is equal to the input current, and since the input is open the current has to come from the op amp itself. There's an error due to the op amp's gain/phase behavior, but if you use a microfarad as your feedback C it should be small.

Or, just have the n.i. input grounded, put a feedback resistor from output to input, and measure the difference between the input voltage and the output voltage. Divide by the resistance to get the input current.

This input current won't necessarily be constant with input common-mode voltage, but most simulators probably ignore these effects.

JFETs have a very small gate current under normal bias, when the part is below its saturation drain current. This current behaves much like the leakage current of a reverse-biased diode operated below the breakdown voltage. It doubles for roughly every 10-12 degrees C, so it can get quite a bit larger at high temperatures.

Bipolars OTOH have significant base current, the collector current divided by "beta", which is typically in the range of 50 to several hundred. Low current low-bandwidth op amps have low input stage currents and low base current. The 990 has high input stage current to get low voltage noise and has consequently relatively high base current.

The current in both bipolars and FETs has at least full shot noise, proportional to the square root of the magnitude of the current. When this current flows in external resistors it generates a noise voltage which adds its associated power to the system noise power. Since FETs have so much less current they are appropriate for higher source impedances (and if you are constrained to use them, higher Z feedback network components).

Typically, FET voltage noise for garden-variety FET op amps is higher than bipolar counterparts. There are some very good modern large-area FETs that come close to good bipolars in voltage noise, although they have high input capacitance which may be a bother in other ways.

At sufficiently high frequencies, generally above the audio range, typical silicon JFETs have an induced gate noise that is a current noise, and is partially correlated with the voltage noise. If you have a high impedance source this contribution starts to get important after a while.

Anothe nice thing about FETs is the tempco of the leakage current means you can improve the high-Z low-medium frequency performance greatly by cooling. Carrier freezeout sets a limit here to about 100 Kelvin (so liquid nitrogen is too cold). The carrier mobility goes up as well so the voltage noise goes down too, although not nearly as fast as the leakage and current noise.
 
> FETs have lower bias currents, is that correct?

Maybe.

> How do you observe the effects of bias current in a simulator?

Same way as in real life: set up a simple but non-trivial circuit and see how bad it screws-up.

What is "non-trivial"? Well, we can always reduce the effect of input bias current to Zero by shorting the inputs to ground. Problem solved. Except we also like our amplifiers to amplify, which means that we gotta put a signal in somewhere.

The most basic reason we care about bias current is when we C-R couple. We want to block DC and maybe bass from a source, yet still have a path for Base/Gate current to reach ground and set a DC level on the input pin. And with simulator accuracy, we can wire the amplifer for unity gain (remembering that whatever we get at unity gain may be worse at higher gains).

Three op-amps with 1 Meg input resistors. The AC source and capacitors are not needed to check DC bias currents, just to remind us why we don't just short-out the inputs and their annoying currents.

bias-L.gif


Top amp is "ideal", and shows zero volts across the 1Meg, zero bias current.
Middle amp is the popular LM324 with BJT inputs, showing 44 milliVolts.
Bottom amp is a BiFET similar to TL071, showing 30 microVolts.
(If I'd had a 5532 or 990 to throw in there, they would show most of a Volt of offset in a 1Meg resistor.)

"Clearly" the FET input beats the BJT input.

But that was 27 deg C. Try 85 deg C:

bias-I-85.gif


BJT bias tends to decline a bit when hot, FET leakage increases exponentially. It just happens these parts come to similar values at 85 deg C. It is true that in very wide-temp applications, you can sometimes do better with a tight BJT design, or at least it won't go wild above a certain temperature.

Yes, I agree there seems to be some funny-stuff here. The output voltages "should" be the same as the input voltages, since the feedback wire has no bias-current error. I suppose the op-amp models have some offset voltage in them. Op-amp models should be treated with EXTREME distrust: full modeling would clog your CPU, most models attempt only the gross behavior, and possibly not all the embarassing details. While I don't believe the exact values shown, certainly not the exactness shown (point oh two microvolts?), I do think the trend is valid. An LM324 will show measurable (on a simple VTVM) offset with a 1 Meg input resistor, a TL071 won't, not in the shop (maybe down in the oil-well). As Brad says: "input current won't necessarily be constant with input common-mode voltage, but most simulators probably ignore these effects."

> Set up a circuit as an integrator

Show-off. Sure, that's how you measure super-low DC current. But we simple audio people figure: if we don't see much DC across the grid resistor, that's all we need to know.
 

Latest posts

Back
Top