Standard non-inverting opamp with a cap across + and - terminals...... Why?

great,

Yeah, probably closer to 4 for us. We don't drive that fast, or for that long... 8-)

Charlie

that

great,

I can easily make Riverside in 4.5 to 5 hours, not even half trying.

Of course, since going on Hytrin, I don't need to stop at every available bush ;-)

...Jim Thompson

On Apr 13, 3:53=A0pm, snipped-for-privacy@ieee.org wrote: [...]

BTW: Not all op-amp have the protection diodes.

I said that you were using it as a magic incantation. That is quite a different matter from me not knowing where the number comes from.

So it turns out that you really don't know about the effect. I pointed out the mistake of assuming that the variation in gain was the only thing at work. The bias current problem is real and is an added EMI issue. Before you claim to educate others, I suggest you study up a bit.

I'll even help you with a very simple spice model of just one transistor and a few other parts. Notice what happens to the average base current.

and apart =3D

on

example

everyone.

just wrote.

close

have no hands

really

may think neural

can only be found within you,

food,

teach anyone.

=20

who=20

mean=20

=20

outwards,=20

divine=20

left=20

=20

Alas, true. Est, rolf, and similar also suckered in a lot of wannabes as well. .

I suppose you could see it as a "variation in gain" problem, but it isn't a particularly useful way of looking at it.

With an NPN long-tailed pair, the point you need to appreciate is that the input transistor that sees appreciable fast voltage excursions ("fast" here means that the output stage is to slow to provide the negative feedback to keep the input transistor bases tracking within a fraction of a millivolt) responds fully to the positive excursions, and comes progressively close to turning off as the voltage between the bases moves out of the more-or-less linear region.

For by bipolar transistors, anything much above 52mV (at room temperature) leaves the input seeing only the positive half of the high frequency content.

MOSFETs eventaully go non-linear, but because they don't offer as much transconductance, it takes a lot more fast voltage swing to get inot this region.

Obviously, when the input transistors are spending almost half the time switched off, and the other half carrying twice the nominal emitter current, the bias current averaged over the cycle isn't what the designer expected, and the LT1028, which incorporates a bias current compensation circuit, could be expected to perform particularly badly, but this is still a side-show.

You might think about taking your own advice.

-- Bill Sloman, Nijmegen

n't

You've managed the simplicity okay, but lost all relevance in the process.

The discussion is all about the behaviour of the long-tailed pair at the input of an op amp.

I've been brooding about putting together a Spice model of a rudimentary op-amp - the 1986 National Semiconductor Linear Applications book included "The monolithic operational amplifier: a tutorial study" which was originally published in the IEEE Journal of Solid State Circuits as an invited paper in volume SC-9, issue 6, and its figure 1 - pretty much Widlar's LM101 - would be about as rudimentary as I would have thought that you could get away with.

-- Bill Sloman, Nijmegen

he

ldn't

ed

up

Do you now admit that the bias current changes? I made the model simple enough that everyone could see that the bias current will indeed change. There may be others who are less experienced reading along.

n't

It is time for you to stop trying to educate others until you study up a bit. Looking at it as a variation in gain with the input is a perfectly useful way to look at the effect of the gm curve.

We aren't arguing about the gm curve we are talking about whether you now admit that the bias current also changes.

The point you need to appreciate is that this is that I already know how an op-amp works and how the long tail pair responds. It is you who doesn't seem to know about the nonlinear change in bias current when a voltage difference appears on the input.

The RF I was talking about was only 5mVp-p you keep throwing your magic incantation of 52mV / 26mV in as though it matters even though I have pointed out that the effect I am talking about happens at a much smaller voltage than this. exp(X) doesn't have a sharp corner in it.

MOSFETs with the same gm curve as the bipolars still wouldn't have the bias current changes.

I never said anything about switched off and switched on. I told you to happened at 5mVp-p

Do you now admit that you were wrong about the bias current?

It is not a side-show at all. It is all about this bias current change issue that you denied existed. Do you now admit that the bias current changes?

I did and carefully. Now it is your turn. Do you now admit that the bias current changes?

And the bushes thank you!

As dry as it is around here, I'm not so sure ;-)

...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

he

ldn't

But not a signficant issue.

At 5mV the non-linearites involved aren't large, and there isn't all that much "rectification". John Larkin's examples would iterpolate to about 350uV of DC offset from 5mV of EMI.

I never said that bias currents didn't change. They don't change much, and the changes don't have a signficant effect on the amount of DC offset that you see at the amplifier output.

Since I've never denied that the bias currents are affected, this is asking me if I've stopped beating my wife. Since it doesn't seem to be an effect that is worth worrying about, I couldn't care less about it.

Carefully, but not intelligently. You've fixated on an irrelelvant detail.

-- Bill Sloman, Nijmegen

That would be adding insult, to injury. :(

On Apr 16, 1:56=A0am, snipped-for-privacy@ieee.org wrote: [.....]

"aren't large" is in the eye of the beholder. At 5mVp-p there was enough of a change in bias current to matter in the application where I ran into the problem.

"significant" is also in the eye of the beholder. It provides a method by which the changes in RF can cause a change in the bias current and hence a demodulation of the RF that depends on the elements that this changing bias current flows through.

Idiot. There are lots of situations where microvolts of offset, or nA of input bias current shift, would trash a signal. How can you glibly declare an effect "irrelelvant" and not "worth worrying about", "don't have a signficant effect?"

You can if you never design real electronics.

I'm currently working on a rather large and expensive analytical instrument, and just got the schematics of the existing stuff. It's ghastly, throwing away at least 20 dB of s/n ratio, on a gadget where extensive signal averaging - using many squirts of rare samples, on a megabuck machine - is the norm. 20 dB == 100x signal averaging.

The electronics was designed by physical chemists. Why are scientists, with a few rare exceptions, such notoriously bad circuit designers?

John

What's the book's title, Mr.Johnson?

RL

So I'd go a factor of 2 (or more) higher in frequency, just to be on the safe side.

...Jim Thompson

I've seen engineers go just a smidgen under where it goes berserk. Most of them were also motorcycle riders, folks who owned the more serious motorcycles ...