# Opamp current noise

• posted

Opamp current noise.

I got a call the other day from someone seeing =91too much=92 warm up drift in the Johnson noise voltage from our Noise fundamentals instrument.

I had them send it back to me. (A few hundred dollar learning experience) After much head banging, it turns out that the voltage noise of the first stage opamp increases by ~3-5% in an exponential way with a ~10 =96 20 minute time constant. (I haven=92t had time for a carful measurement.)

Anyway, along the way I became worried about the current noise of the opamp (opa134). I=92m wondering if any one has an idea on how to measure this. I think Phil H. mumbled something about a correlation technique when I mentioned this a year or so ago. But now I can=92t see how to make that work.

My current idea (NPI) is to just switch in a second similar opamp. Like this,

switch | V |\ +-/.---|+\ | | >-+-->(output to nowhere) | +-|-/ | | | |/ | | +------+ |\ +---------------|+\ | | >-+-->(output to more gain, R +-|-/ | filter, multiplier) R | |/ | R +------+ | GND

And measure the increase in noise for different R=92s. Will this work? Any other ideas?

Thanks,

George H.

• posted

This AN has some useful stuff:-

• posted

t

I'm trying to remember what my 24 bit sound card's noise density is, but 17nV/rtHz comes to mind. If put a LT1115 preamp in there, you could get down below 2nV/rtHz and use that as an instrument to measure the noise density function.

Plus you get a free tracking generator output along with it.

• posted

at

Thanks Spehro, The current noise gets =91trouble some=92 out beyond 100kHz. See figure bottom of page 4.

Making a big R kills the bandwidth. I=92ve got switches and terminal blocks in the front end, which give about 7pF of input C. Making some sort of difference measurement would be ideal. (It's fairly easy to see 1% change in the noise.)

George H.

• posted

hat

text -

Thanks Robert. The issue isn't measureing the total noise, but getting a measure of the fraction that is due to the opamp current noise. Measuring the opamp voltage noise is a snap... just short the input.

George H.

• posted

Well, if the gain of the amplifier and its DC errors makes something saturate, that can fail. There's also opamp current noise at the inputs, and that responds to some kinds of thermal effects, which could be your major issue. Current noise doesn't diminish when you short an input.

Shot noise in a transistor base current (bipolar transistor) goes up with temperature, because current gain (beta) drops.

• posted

Well OK, but it works fine in this case.

There's also opamp current noise at the

Sure but the current noise contributes I^2*R^2 to the V^2 noise output.

There are Jfets on the input of opa134. They have a current noise that rises with frequency. (and temperature) I read somewhere what this is due to, but have now forgotten it. (too much beer.)

George H.

• posted

If you make the second amp a FET input one, it won't contribute current noise. Then the correlated part of the noise of the two amps will be just the current noise times R.

So bung the two of them into a multiplier, and lowpass filter the daylights out of the output.

Cheers

Phil Hobbs

```--
Dr Philip C D Hobbs
Principal Consultant```
• posted

Before 5 O'clock ?:-) ...Jim Thompson

```--
| James E.Thompson, CTO                            |    mens     |
• posted

I

that

ed text -

I didn't mean total noise, but true "noise density function". The soundcard easily finds you noise density. I use it all the time.

How do you do it now, since at the output you have both current and voltage noise.

Floundering around a bit here, but I perceive simultaneous measurement [soundcard has two channels at 192kS/s] measure at input terminal measure at output terminal. then compare. Voltage noise is new at the output, but current noise also appears at the input, so with correlation you can discern which is which.

• posted

hat

OK that kills the opamp voltage noise. But won't the Johnson noise from the resistor get into both channels? That's going to be a much bigger contribution. I think the above switched circuit work. I'm going on vacation tomorrow, so the measurement will have to wait.

George H.

t -

• posted

up

10

he

n I

ke that

e

oted text -

Well I've just ignored the current noise. If I keep the maximum frequency below 100kHz the current noise from the opa134 is below 1% of the total. If I believe the spec sheet! I guess that's what I wanted to confirm.

Yeah I think that's right. But the johnson noise from the resistor is correlated in both channels.

I must admit it's not obvious to me why the voltage noise of the opamp doesn't 'leak' out of the input.

George H.

- Hide quoted text -

• posted

m up

he

~10

l

the

hen I

make that

.
,

r)

,
u

ure

quoted text -

he

hmmmm...yes, it should divided by the Zinput ratio against the equivalent input impedance. But isn't the equiv input impedance usually much less than the Zinput? so the signal could get really small.

• posted

Yes, you'll see the Johnson noise, but you can jack up R until you start seeing the current noise clearly.

There are other things you can do with linear combinations of 3 or more amplifiers, e.g. putting two identical resistors in series, with two FET amps. The number of combinations goes up faster than the number of amps, so you rapidly get more equations than unknowns.

Cheers

Phil Hobbs

```--
Dr Philip C D Hobbs
Principal Consultant```
• posted

up

10

he

n I

ke that

nt

be

Thanks Phil, But it's a jfet input opamp. The current noise rises at high frequency. (Something like 100fA/rtHz at 100kHz and 1pA/rtHz at 1 MHz.) I'll only be able to see it at high frequency. On paper it shouldn't be a problem. But I started to worry that things might be worse than the spec sheet. And wanted to try and measure it.

George H.

• posted

A picoamp at 1 MHz is quite doable. You only need a few kilohms to make that bigger than the Johnson noise, and you can remove the Johnson noise by a first principles calculation.

Cheers

Phil Hobbs

```--
Dr Philip C D Hobbs
Principal Consultant```
• posted

George Herold a écrit :

Noise current in JFets mainly comes from capacitive coupling (jfet internal parasitics) of noise voltage, hence the first order frequency dependency.

Beware that it's not the only current noise coupling mechanism in opamps. You also have input clamp diodes with parasitic capacitance which translates power supply voltage noise into current noise into the inputs. If you have something like a linear voltage regulator (78xx or such) which are notoriously noisy, you may have your noise source: 2pF at

160kHz and 0.5uV/rtHz of noise (yep they are that bad) gives you 1pA/rtHz of current noise.

A simple RC bypass cell on your opmap will fix that.

```--
Thanks,
Fred.```
• posted

arm up

g

the

a ~10

ful

of the

when I

o make that

mp.

in,

ier)

urrent

ill be

ise

uch

t.

art

re

FET

at

e

Yeah, the bad news is that the bandwidth of the amp chain is 1-2 MHz. So its a bit tricky separating more noise 'amplitude' from bandwidth effects.

For the above (switched in amp) circuit, a man could worry about the extra input C of the second amp.

George H.

• posted

10

he

n I

that

ent

be

e
h

rt

e

ET

at

Ahh, so if I measure the voltage noise and assume the capacitive coupling is about the same from device to device, then I've got a measure of the high frequency current noise.

That's useful. Thanks

I've got a cap multiplier on the opamp power lines. Noise is something less than 1nV/ rtHz. (It's actually kinda hard to measure (the PS noise) without a lower noise opamp (Some BJT one.)

George H.

• posted

rises at

the

I had some idea of looking through the LTSpice jigs directory for noise measurement circuits but that may be a barren idea.

?-)

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.