John Curl's Blowtorch preamplifier part II

[Steve Eddy]

Originally Posted by Wavebourn

Very probably! Like our cat still can't figure out that she brings us wrong food!

Or our dog who thinks the cats' litter box is an hors d'oeuvres tray.

se

[jacco vermeulen]

Rubbing some snake oil on usually helps.

[Bonsai]

Originally Posted by Steve Eddy

Or our dog who thinks the cats' litter box is an hors d'oeuvres tray.

se

Ha ha we have the same problem.

[brianco]

Originally Posted by Steve Eddy

AH HA! So all these years the aliens have only been trying to FEED US! You'd think by now they'd have figured out that food goes at the OTHER end.

se

Not from what one hears national representatives say.....irrespective of country!

[gpapag]

I googled for IEEE Std.1050 and I found the 1989 edition in pdf.
Well written and carefully worded document. Thanks jneutron for pointing to it.

In there, I read something (one of the things at least that) I didn’t know. That low – impedance circuits are less susceptible to capacitive coupled noise and crosstalk . This may answer my own question:

Originally Posted by gpapag

As I read it, you target the problem at the bonding wires and pads. That’s in agreement with Scott words.
This I find strange. I mean how all the routing within each layer and the stacked layers between them end up less troublesome than the bonding wires.

The circuits inside audio ICs are relative low impedance (Vac/Iac). Furthermore, the“de” part of the capacitive coupling equation I=C de/dt is fairly low (*).

On the other hand these low impedance circuits are more susceptible to inductive coupling and the “di” part of the inductive coupling equation E=M di/dt becomes significant, thus your emphasis specifically on inductive coupling

Originally Posted by jneutron

For old style chips with aluminum bonding pads on the top surface, the pads are connected to the leadframe terminals via a wirebond, either aluminum or gold. This wirebond is looped a bit so as to not touch the edge of the substrate in it's travel to the lead. Back in the 80's (IIRC) it was determined that the parasitic inductance of the wires (14 nH for some reason, comes to memory) significantly affected the speed that could be attained for all I/O lines of the chip. Also, two parallel wires can also crosstalk.

The basic issue was the slew rate of the current, how it coupled to neighbors, how it affected the speed of the signal, and for hot and ground runs, it can introduce unwanted ground or rail bounce.

At lower frequencies, the slews are not as bad, so the parasitic inductance doesn't have as great an effect.

Nowadays, the problems I see in things like the one chip amps, is the proximity the rail currents are to the input nodes. Any proximity coupling from either rail to that node can introduce error signals at the input pair, where gain is full.

j

makes sense to me now.


But for this question I haven’t been honored an answer yet

Originally Posted by gpapag

Scott, I hadn’t noticed your response, sorry.

Now, to exercise your patience, one more step on the ladder:
Why these “conventional methods” brake up above say 100MHz?
Is it that the conception of L/R/C can not model the reality anymore, and if so, why?
(I guess the next floor is a few more steps higher, so prepare yourself )

George

(*) May be this is the reason that high package density modern CPUs are designed for lower voltage supply than the previous generation CPUs which were less densely packaged.

[bcarso]

The methods don't "break", but I've found it instructive to consider resonances and note just how rapidly the L's and C's of significance get very small indeed. And this is before abandoning lumped circuit elements.

We tend (or used to tend) to think of a picofarad as a small capacitance, and maybe a tenth of a microhenry a fairly small inductance. But as a resonator those two values have a frequency of resonance of about 500MHz, and if constituting the distributed L and C of a lossless transmission line, a characteristic impedance of about 316 ohms.

I recall an experimental "opamp" described some years ago that was reasonably well-behaved out to about 10GHz. The developers remarked that, despite their fears, it behaved in accordance with an analysis that didn't change in character from ones appropriate to far lower frequencies. But of course the dimensions were very small.

[DF96]

"Conventional methods" break above the frequency when the size of the circuit begins to be a non-negligible fraction of the wavelength. The lumped LC circuit model is always an approximation of reality, but a very good approximation for sufficiently low frequencies. The wave model is always true, but unnecessarily complicated when the lumped model is good enough.

For pulse circuits substitute (approximately) 1/(rise time) for frequency - maybe there should be a 2 pi in there somewhere too.

[gpapag]

Originally Posted by bcarso

The methods don't "break", but I've found it instructive to consider resonances and note just how rapidly the L's and C's of significance get very small indeed.
And this is before abandoning lumped circuit elements.

Ouch!

Originally Posted by DF96

"Conventional methods" break above the frequency when the size of the circuit begins to be a non-negligible fraction of the wavelength.

Largest dimension ~0.15 of wavelength (IEEE Std.1050)

Originally Posted by DF96

The lumped LC circuit model is always an approximation of reality…
The wave model is always true….

I’ll keep these two. They match my understanding too.

Thank you both for answering (*)

George

(*) In lew of Scott. Most probably he is messing around with SY, excusing themselves they are doing some live recordings again.
Hedonists

[hitsware]

>>where there appears to be diode action at very
>>low levels on a mechanical contact.

If it's ' diode ' , I would suggest it's like 2 diodes
opposent polarity in parrallel ....
It may not be diodic, but arc-over @ nano-gaps

[Bonsai]

". . . inductance of the wires (14 nH for some reason . . . "

That seems incredibly high. Bonds even on the old DIP packages are a fraction of this.