Hardening a simple op-amp circuit against EMI

"Russell Warren" schreef in bericht news: snipped-for-privacy@13g2000yql.googlegroups.com...

Wiith any luck, the relevant datasheet will tell you.

shows the output impedance of the part peaking at about 200R at around 5MHz. 100R might be enough.

Exactly.

Symmetry. If you are using small resistors right up against the inputs to linit RF pick-up, it makes sense to use identical resistors.

I got caught by a Texas Instruments CMOS input op amp - I think it was the TLC2201. No input capacitance on the data sheet, guessed about 3pF - found about 15pF (based on the amount of feedback capaitance it took to stop it oscillating). Fourteen years ago now, and it still rankles.

The OP727 won't have any trouble driving the 12k7, and won't notice

10nF at the other end of the resistor

--
Bill Sloman, Nijmegen

I've used "the cell phone test" quite a bit as an extra test on input stages in various designs. Hold cell phone an inch away and see what happens :->

Often it's the phone speaker magnetics that cause the trouble though - a two for one test, bargain!

Very common to get the "rat-tat-tat" effect in your car radio when your mobile is sitting nearby.

Dave.

Thanks for the reminder. I am trying to finish my big ticket work while at least some small portion of it is covered. (working stiff) .

You didn't know?? .

work

a 1

ntal

n

ould

ch pin

issues,

on

ed

m a

t some

ng

d

or

ou

d , e

nd

t

amp

ur

the

t it

the

her

the

to

puts,

such

ot.

t

This begs the question of how it gets rectified. If the RF current being injected is big enough to sweep out the stored charge in the BE junction, then you can get rectification.

Otherwise the RF goes through to the collector and gets lost in the parallel capacitance of the node.

My point is that small RF currents won't get rectified; if they are big enough to get rectified they are going to show up as a more- than-26mV voltage excursion between the base connection to the input transistors.

True, but the whole point about the advantage of FET-input op amps over bipolar input op amps is that you need a sufficient voltage excursion for the RF to get rectified (and its a lot higher for FET- input amps).

There's a rational arguement that says that the threshold lies close to 26mV for bipokar op amps, and you seem to want to revise this down to the microvolt level without providing any explicit justification for this revision.

I didn't want to disagree with you about this - I didn't want to give John Larkin more justification for describing me as cranky - but I think that you've got it wrong on this particular point (even though you very rarely get things wrong).

-- Bill Sloman, Nijmegen

We have a new dentist in the neighborhood offering 15% discount for cash or check.

I'm hearing more and more tales of good deals for direct pay.

Also a doctor here offers "executive medical care"... flat fee of $79/month, immediate service, covers testing, too !!

...Jim Thompson

--
| James E.Thompson, P.E.                           |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
 I love to cook with wine     Sometimes I even put it in the food

[ RF into BJT-opamp inputs ]

Oh, oh. Watch the goons from the language police swooping in now ...

(and I still believe that wording is correct so we'll probably share a holding cell soon)

It doesn't have to be big if what you are processing is in the microvolt range. Whenever they call me out to cases like that this is almost always so.

Yeah, 99% does. But the remaining 1% is what causes all the grief, frantic searches for the next flight out there, and so on ;-)

Ok, I can't show you schematics for reasons of confidentiality. All I can tell you is that it does get rectified and I did measure RF levels. Well below 26mV, yet rectification occurred. Enough of it to show up in the data signal. Heck, we could tell whether the person coming down the stairway next door was wearing an AT&T cell phone (GSM network) or not.

Sure. But sometimes they don't offer the low frequency noise performance you need.

Sorry, I can't share the data because it was work under contract for clients, but it's right here. Typically went like this: RF amplitude measured at the inputs, usually in the hundreds of uV. Bursts coincide with the dirt in the data. Rip out BJT amp, replace with FET amp, problem gone, if the baseband noise performance is still good enough plus margin the job is done, call airline, try to get an earlier flight back, wife happy.

If the baseband/audio noise performance ain't good enough with FET amps it's back to square one. Usually meaning a relayout, better grounding, shielding, and so on. Call hotel instead of the airline, see if stay can be extended, wife not so happy.

Thanks for the kudos. But yeah, I am certainly wrong on some things, just not on this one ;-)

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

I subscribed to one of the dental plans for a family member at the suggestion of the dentists secretaries. At about $125 a year, saves considerable costs.

Just ask the dentist what insurance he accepts. I think I bought Avia. Saved thousands.

]greg

I knew. John's teasing me--I recently told him this. FWIW, Rumford brand has beaucoup more active stuff per $, but Clabber Girl cans and 0.25" copper tubing make 900MHz cavity filters with Q's of roughly 300.

Cheers, James Arthur

No, Joerg's right Bill. Operating, bipolars' junctions are forward biased, drawing bias current from their inputs, and they'll rectify the slightest signal.

Think of them as pre-biased detectors.

Cheers, James Arthur

Exactly. Bill, you might as well try it out: Take the most sluggish BJT opamp you find, hook up as 1000:1 amp or whatever, take a 100MHz or higher frequency source, attenuate down to a millivolt or so, feed into input. Hang a stereo amp on the output and then pulse the oscillator hard. You'll be surprised ...

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

[snip]

Wonder from which orifice "physicist" Slowman pulled 26mV ?:-)

(I know what he misunderstands ;-)

...Jim Thompson

--
| James E.Thompson, P.E.                           |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
       How severe can senility be?  Just check out Slowman.

Cranky isn't important if you're right. And you're wrong.

John

Jim should check out the Ebers-Moll equations or Gummel-Poon.

kT/q comes out at 26mV at 300K which is to say room temperature

i(v) =3D i(0).e^V/26mV

It is a bit of an over-simplification to say that the current is a linear function of voltage below 26mV, and exponential above it - what Jeorg is saying is that you may not get much non-linearity below 26mV, but you can get enough to rectify enough of an RF signal to get you into trouble - but as an arbitary place to set the boundary it isn't a bad choice.

In fact, for a long-tailed pair, the voltage difference gets divided over both transistors, so 52mv might be a better number.

-- Bill Sloman, Nijmegen

gs work

ity

ish

ing a 1

amental

l in

e

the

should

each pin

re:

ad

RR issues,

eason

ndred

I'm a

pent some

lly

y.

doing

tand

se or

e you

to

hat

fied

or

ing,

erse

s and

ight

d a

en

ed amp

ed

ge

ack

ny

a

your

hat the

what it

of the

other

ing the

ged to

inputs,

th such

r not.

k.

that

I don't think so. The relationship between voltage and current is still exponential for voltages below 26mV, so you can still get some rectification, even if it is then only a small proportion of the RF current that is getting rectified. In Jeorg's case, that small proportion was enough to create difficulties.

Your arbitary - and unsupported - claim that I'm wrong, really is cranky.

-- Bill Sloman, Nijmegen

Bill you're on the wrong part of the curve. dV(RFI) is small, but you've missed the fact that the input junctions are pre-biased forward, so Vf is 100's of mV, not 26mV.

Cheers, James Arthur

Slowman's ego, chutzpah and senility level are high enough that he could easily replace Barney Frank as Chief US Fairy ;-)

...Jim Thompson

--
| James E.Thompson, P.E.                           |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
       How severe can senility be?  Just check out Slowman.

No one could replace Barney Frank! Well, maybe Ted Kennedy.

Cheers, James Arthur

Actually, I'm not. The exponential relationship is exponential relative to whatever you take to be your reference point, and mine is where the two transistors in the long-tailed pair are both biased for equal currents in their collectors - needless to say, the equal current that biasses up the rest of the circuit so that all the active devices are in state to do something useful.

You can choose your own - equally arbitrary - reference point, but it isn't likely to be anything like as useful as a starting point.

-- Bill Sloman, Nijmegen