Micpre of Graham

OK, simplest case first, an emitter follower:

+10 | | | c +5----------------b e | | 1K metal film | | | GND

Does Ib or Ic have full shot noise? Ie sure doesn't.

John

John Larkin a écrit :

Actually that's pretty easy to check with an LF SA (1M input)

15V >---. | .-. Rc | | | |1K '-' | | || Adjust +---||--> SA | || for Ve=5V | ___ |/ 10.7V >----|___|--+-----| | |>

Rb --- | --- | || | +---||--> SA SA

But metal film resistors don't have shot noise current. So, if a transistor's bias current is mostly set by the voltage drop across an emitter resistor, emitter current won't have much shot noise either. So Ic certainly won't. It seems to me that base current shouldn't either, but I'm not 100% sure about that.

John

I thought you misspelt that for a minute...

Regards Ian

;-)

Oh dear...

Ahmmm. The voltage being set across a noiseless resister makes no difference whatsoever to the internal shot noise of a transistor. Shot noise is the statistical effect of carriers across the junctions.

The shot noise is

icn = sqrt(2.q.Icdc) acrross the emitter juction transfering through the collecter ibn = sqrt(2.q.Ibdc) acrross the emitter juction transfering through the base.

Thats it. Nothing you can do will eliminate this shot noise.

Why dont you try it in spice? Hint, it dosnt model have a shot noise in resisters!

Ic and Ib are both inherent. End of story.

--
Kevin Aylward
ka@anasoftEXTRACT.co.uk
www.anasoft.co.uk
SuperSpice

Which is reasonable.

It is *not* the end of story.

If the metal-film resistor jams zero shot noise current into the emitter, how can full shot noise emerge from the collector?

I'm not sure about the base current in the case where Ie has no shot noise.

John

Right, thanks the gods. Kevin needs to take some measurements.

Surely the base-current shot noise is reduced. Worth checking.

I checked the datasheet input current noise of a few "simple" opamps (without bias current cancellation) as compared to input bias current, but they all had much more than shot noise, so that was inconclusive. Besides, I'd guess that internal to a linear IC, it's hard or impossible to make a sub-shot-noise current source.

I've been interested for some time in the shot noise reduction in metal resistors. I'd expect things like carbon comps, with a short mean free electron path, to be bad, but I can't find decent references and my measurements indicated well below full-shot. Some day when things are slow, I'll have to do better experiments.

John

I'd have expected the opposite (full base shot noise), so I just took time to check it.

Added 1K at the emitter for obvious stability issues (about 150p+1M load).

15V >---. | .-. Rc | | | |1K '-' | | || Adjust +---||--> SA | || for Ve=5V | ___ |/ 10.7V >----|___|--+-----| 2N3904 | |>

Rb --- | --- | || ___ | +---||-|___|-> SA SA

Fred Bartoli a écrit :

A small mistake was made there. Since we have identical voltage drop across Re and Rb, we have Rb=beta.Re This makes a current divider for the base noise current.

Then we'll see half that predicted noise voltage at the base point, and half too at the emitter point (and obviously too at the collector).

--
Thanks,
Fred.

It is as far as internal shot noise goes.

This statement makes absolutely no sense. Internal shot noise has absolutely nothing to do with any ability "to jam" noise free current in from an external source. The simplified model of shot noise assumes from the outset that external sources are noise free. So, its simply irrelevant whether this is via a metal film resister or teletubbies.

There is a shot noise generator associated with an ideal, otherwise noise free, DC current source feeding the emitter, irrespective of whence the current originates. This generator is due to the statistical nature of charge carriers crossing a junction. This noise generator is directly across the dynamic impedance re. This drops a voltage right at the base-emitter junction.

Well, actually, rather than stating that there is a collector noise dropped across re as one simplified convenient model, it is better to use the output noise equivalent model, that is:

icn_shotout = (icn_in.re).gm = icn_in

That is, the collecter shot noise is directly at the internal ce nodes. This internal noise current is a constant set by ICDC, irespective of any external circuitary. Now..if there is an emitter resister, a simple calculation on the equivelent circuit will show that the noise actually generated into the external collecter load is:

in_load = icn_shotout/(1+Re/re), because the controlled source swips some of the current away.

So, in an actual circuit, the output noise current, dropped across the load resister, is reduced by Re feedback. However, again, this noise is not a function of the nature of the driving source or resistance, in addition this reduced output noise does not imply a better S/N ratio, as the signal also gets reduced by Re. Of course, if Re generates additional noise, then this will have to be added into the calculation.

I hope the above explains what Re does, but draw the equivalent circuit and do the algebra by all means.

--
Kevin Aylward
ka@anasoftEXTRACT.co.uk
www.anasoft.co.uk
SuperSpice

Why?

I can only assume here Win, that you haven't actually read what I have wrote and that you are making statements not related to inherent transistor, junction caused, shot noise. To wit, internal base-current shot (and collector) noise cannot be reduced by external components. This is trivially obvious. The effect of such noise may well be modified by the external circuit, as I stated in my original post, and show in my later one to John.

--
Kevin Aylward
ka@anasoftEXTRACT.co.uk
www.anasoft.co.uk
SuperSpice

That was my point. If the emitter current source is noise-free, it reduces the shot noise in the collector load; it must, since Ie = Ic + Ib. Even if Ib has full shot noise, Ib is 1/b of Ie, so its "diversion" noise current is small. In the case of the differential mic preamp, if the transistors are biased by dc drops across big resistors, no-signal emitter shot noise is nil. The AC gain isn't correspondingly reduced, because AC-coupled emitter-emitter resistance sets AC gain, and that contributes no shot noise current (although it does add gain to the equivalent emitter noise.)

I suppose Ib has full shot noise because there's insufficient charge interaction path in the base of a transistor to get the smoothing effects seen in metallic conductors, and maybe because holes are clumsier than electrons, and maybe because the carriers are essentially thermally driven. Or something.

OK, next question, again for the simple emitter follower with a stiff base supply and Rbias from emitter to ground. Assume a forced Ie that is noiseless. Suppose we have some DC base current, say 1/100 of Ie. Now suppose that base current transiently increases by d above the normal current because of shot noise. That pumps 100*d current into the emitter. That raises emitter voltage by 100*d*Rbias, reducing b-e junction voltage, reducing base current. Since Rbias can be arbitrarily large, the effective gain here can be large. Since this is a feedback gain that reduces the base current, it sounds like we can reduce base shot current by something like beta, 100:1.

If not, why not?

John

This is so obvious that wouldn't, and I didn't, think for a moment that you are referring to this external to the transistor noise, as shot noise. Resisters don't have shot noise, so clearly, one is going to ignore them as shot noise generators.

Yeah, we really have a terminology problem here. "Emitter (or Collector) shot noise", just means the noise internal to the transiser to me.

I just don't understand what you are trying to say here.

All this transient forcing is all darkness to me. Look, draw the small signal equivalent circuit. Put in i_base_noise generated noise across r_pi (hfe.re), put i_collector across ce. put in the emitter resister. Add in any externally generated noise, and churn the handle. If you dont want to do that, do it is spice. You can even use ideal transistors, which still generate shot noise. You can even get spice to list out contributions from each component separately. To get thermally noiseless resisters, you can use a VCCC connected to itself.

Assuming a stiff base supply, i.e. Rs=o, and neglecting, rbb' as well, In is the base current noise, I get:

i_out_noise = In. Re.hfe/((1+hfe)(Re + re))

So, for the case of Re=0, we get i_out_noise = 0, as expected, to wit, ib is s/c directly on the input, so no vbe.gm. For the general case, with hfe >>1:

i_out_noise = In.Re/(Re +re)

So, for large Re, i_out_noise ~ In, where In is due to noise from sqrt(Ic/hfe.q)

Interestingly, Re allows for Ib to be feed directly into the output circuit, essentially because ib no longer has a s/c across it.

--
Kevin Aylward
ka@anasoftEXTRACT.co.uk
www.anasoft.co.uk
SuperSpice

In Graham's case its "Petty, Snide Donkey" ;-)

-- Service to my country? Been there, Done that, and I've got my DD214 to prove it. Member of DAV #85.

Michael A. Terrell Central Florida

No- it is Petty Simpleton's Demise...

My doctor told me I had "shotty lymph nodes". He pronounced it very carefully.

John