CA3240 Advantage over TL082?

Most specs favor the JFET '082, but some favor the '3240. If he needs operation to the negative rail, either IN or OUT, the CA3240 gets the nod. Being a mosfet, it also has less leakage at high ambient temps. But many would consider the lack of a second source to be a strong mark against it.

view in monospace font:

item CA3240A TL082A input: mosfet jfet

Vcm input range: to -rail: yes no to +rail: no yes

output to rail? Vee: yes no Vos max: 5 6 mV Ib max: 0.64 7 nA at +70C e_n typ: 40 18 nV slew typ: 9 13 V/us f_T typ: 4.5 3 MHz Iq (both) 8.0 2.8 mA cost: $1.18 $0.43 for 100 pieces

2nd source: no yes

--
 Thanks,
    - Win

Thanks for your effort

Frank-Stefan

"Winfield Hill"

** Examination of the EIN curves for both devices shows big differences in the way noise varies with frequency.

The MOSFET input device has a constantly falling curve which is quite unlike most JFET and BJT input amplifiers which have curves that remain level above a certain break frequency - usually at a few hundred Hz.

See fig 18 on page 13.

See fig 21 on page 14.

Means the MOSFET device is significantly *less* noisy above about 5 kHz than the JFET one.

Could be useful to know....

..... Phil

n

ike

ove

han

Interesting, Thanks Phil. Say I noticed the other day that the opa134 has current noise that starts to rise dramatically with frequency above 2kHz or so.

See the graph at the bottom of page 4.

Is this common for all jfet opamps? I've never seen other spec sheets that plot current noise vs. frequency for FET opamps.

George H.

The current noise is so low, 3fA/rt-Hz, that it's scarcely significant. But it's still interesting to analyze.

Often this type of very low spectral-density noise, rising proportional to frequency, is due to a capacitively-coupled signal from a spectrally-flat voltage-noise source.

We can calculate, i_n = e_n 2pi f Cx. For example, assume the JFETs have 60k drain resistors, which gives about 10nV of Johnson noise. If this is coupled to the input via some capacitance, etc., we can calculate, Cx = I_n / 2pi f e_n. For noise at some frequency on the plot, we get Cx = 0.06pF. Is that due to Cdg, or is it some other small capacitance?

If the opamp had a cascode input stage for the JFETs, their drains should not be exposed to high voltage noise. It's possible the opa134 doesn't have a cascode input. TI says, "The p-channel JFETs in the FET input stage exhibit a varying input capacitance with applied common-mode input voltage."

--
 Thanks,
    - Win

4

Thanks Win, The capacitive coupling makes sense. The 3 fA/rtHz is no worry. But at 100kHz the nosie has risen to 100fA/rtHz.. If my source impedance is 100k ohm.. that looks like 10nV/rtHz of voltage noise. That's starting to look significant.

any better way to measure current noise than something like this?

+--100k ohm--+ | |\ | +-- \ | | >-------+---out +-+ / | |/ | GND

Thanks,

George H.

Thanks Win, The capacitive coupling makes sense. The 3 fA/rtHz is no worry. But at 100kHz the nosie has risen to 100fA/rtHz.. If my source impedance is 100k ohm.. that looks like 10nV/rtHz of voltage noise. That's starting to look significant.

any better way to measure current noise than something like this?

+--100k ohm--+ | |\ | +-- \ | | >-------+---out +-+ / | |/ | GND

Thanks,

George H.

---------------------------------------------------- Some low noise circuits by VLF listeners here

I've tested some simple pre-amp/voltage amps over the years, found out 2 key points.

1. Use a discrete JFET with careful bias set by low noise resistors.
2. Then find a place FAR FAR away from any grid power lines.

100k is a bit on the low side.

--
 Thanks,
    - Win

That setup has a huge noise gain at high frequencies.

I'd set up the opamp as a follower and dump its + input into a grounded resistor. Then amplify that voltage with another opamp. Measure the amplified noise with and without the DUT and do the math.

John

the opa134

.pdf

pec

assume

a some

.

It's

says,

Yeah a TIA with no 'external' current. A pf of capacitance to ground gives me a noise gain of 2 out near 1MHz or so. I could roll the gian.

I've been using the follower circuit. (Well I usually put a bit of gain in the first stage. 200 ohms to ground and a 1k feedback resistor.) I see no sign of excess noise out at higher frequencies. (I can only measure things with the needed accuracy out to 100kHz.) So I was trying to think of some way to measure the current noise of the opamp.

But the TIA configuration looks like it gives the same answer as the follower. (Well with the TIA, capacitance to ground gives noise gain and more noise at higher frequencies. Whereas with the follower the C to ground rolls off the signal from the DUT. Both effects tend to throw a monkey wrench into the measurement. Sigh. I should try Win's suggestion of a bigger resistor. I could tyr asking Burr Brown (TI) how they measured the current noise.

Thanks,

George H.

Thanks, There's a lot of stuff on those links! I haven't done very much design a the transistor level and I think it would take me a while to do better than what I can buy in an opamp. But thanks again for the input.

George H.

"George Hairoil"

Thanks Win, The capacitive coupling makes sense. The 3 fA/rtHz is no worry. But at 100kHz the nosie has risen to 100fA/rtHz.. If my source impedance is 100k ohm.. that looks like 10nV/rtHz of voltage noise. That's starting to look significant.

** Not compared to the self noise of a 100 kohm resistor it ain't.

My god you are a tedious wanker.

..... Phil

The OPA134 will need a couple of megs of input or feedback resistance to get the current-induced noise up to the level of its voltage noise, where you can resolve it: 8 nv/rthz. But the Johnson noise of that resistor will be around 20x as much. This doesn't look good.

There must be a trick somewhere. Maybe something with a capacitor.

The easy was is to ask them.

Maybe they calculate it, and don't measure it, based on the shot noise of the bias current. They claim 5 pA bias current, 3 fA/rthz noise as typs. The shot noise of 5 pA is 1.3 fA/rthz. Assume the bias current is contributed by partially canceling ESD diode leakages, and it's in the ballpark.

John

Maybe make a follower but leave the + input open. The bias current, a few pA, will charge the stray capacitance, a few pF, at roughly 1 volt per second. Measure quickly.

Add a Phil Hobbsian trick: cancel the bias current, in whichever direction it turns out to be, with a 1 pF silicon photodiode illuminated as needed. You can measure the noise, which is caused by the opamp noise and the shot noise of the photocurrent. It will need a tiny photodiode. At least the Johnson noise is gone.

The capacitance will of course integrate the noise current. 5 fA/rthz dumped into 5 pF is only around 160 uV/rthz at 1 Hz.

Nasty problem. Ask TI!

John

Well, the current noise is a real current that comes out of the input, whereas the voltage noise is purely internal to the device. Thus if you took two followers connected like George's, you could use correlation, like this:

1. Leaving them separate, so all their intrinsic noise sources are independent, cross-correlate their output noise, and integrate for long enough that the correlation product decays below the level you want to measure.
2. Connect their inputs together, so that the current noise is now 100% correlated in both outputs, and re-do the cross-correlation with the same integration time. The voltage noise remains uncorrelated, so the result will be a measurement of the RMS sum of the current noise of the two amplfiers.

Cheers,

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net

Needs one more turn of the crank--use normal followers, connected to ground via large resistors. Then proceed as above.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net

o

Sorry to be tedious Phil, you don't have to answer. I'm trying to measure the noise of the 100k resistor to 1% (or so). That tells me Boltzmann's constant if I know the temperature. So the 10nV/rtHz is a few percent of the ~40nV of thermal noise from the resistor. ( a few percent of the noise power.)

George H.

a134

is no

urce

d n C

in's

u %

t -

Wow, Phil Hobbs to the rescue! Thanks. (I think I get the above and I think I can do it.)

Can I send you a bottle of my favorite Belgian style abbey ale? (are you allowed to ship alcohol via the mail?)

George H.

opa134

y

is no

urce

C

n's

Yeah I thought I was missing something obvious.