John Curl's Blowtorch preamplifier part II

[jcx]

fasthenry, fastcap are free, opensource quasistatic "field solvers" can give parasitics, frequency dependent R,L, mutual L
take awkward text descriptions of elements, must descretize "manually" for frequency range

I've known of their existence for a while - never really tried to enter a model before

so 1st try here for a U shaped 100 mil wide trace 1" inboard from edges of a 10"x10" ~1 oz gnd plane, connecting the upper, lower centerline points, 0.062" layer separation

resulting impedance output:
Computed matrices (R+jL)
Row 0: n4 to nout
Freq = 20
Row 0: 0.10945+3.61561e-007j
Freq = 200
Row 0: 0.109462+3.60903e-007j
Freq = 2000
Row 0: 0.110282+3.19306e-007j
Freq = 20000
Row 0: 0.116489+1.95738e-007j


so from 20 Hz to 2 kHz the inductance has dropped <20% due to gnd plane "current centroid" return path shifting towards the trace
by 20 kHz the effective loop area (~=inductance) has been reduced by <half due the proximity effect gnd plane current path crowding towards the trace

2 oz copper's lower sheet resistance reduces the energy "cost" of the current path increasing length - so the proximity effect is slightly stronger:

Computed matrices (R+jL)
Row 0: n4 to nout
Freq = 20
Row 0: 0.109029+3.61529e-007j
Freq = 200
Row 0: 0.109053+3.58945e-007j
Freq = 2000
Row 0: 0.10998+2.76665e-007j
Freq = 20000
Row 0: 0.114031+1.72646e-007j

ran in the demo dist from: Fast Field Solvers


my Uloop fasthenry file:


*Ground Plane

.unit mils
g1 x1=0 y1=0 z1=0
+ x2=10000 y2=0 z2=0
+ x3=10000 y3=10000 z3=0
+ thick=1.4
+ seg1=20 seg2=20
+ nin (5000,1000,0) nout (5000,9000,0)
.default nwinc=8 nhinc=1

*U track 1000 from edge
N1 x=5000 y=1000 z=62
N2 x=9000 y=1000 z=62
N3 x=9000 y=9000 z=62
N4 x=5000 y=9000 z=62

E1 N1 N2 w=100 h=1
E2 N2 N3 w=100 h=1
E3 N3 N4 w=100 h=1

.equiv nin N1

.external N4 nout

.freq fmin=2e1 fmax=2e4 ndec=1

.end

 

[gpapag]

pfft. Clearly, that leaves me out..

j

That was for Scott. He does not see it as a restriction, as he rarely (during full moon maybe) exceeds that number.

You2k are not Scott20. Go ahead

The observation and the understanding came before and outweigh the mathematical descriptors. The math may allow other things to be seen or understood but cannot stand in place for personal observation and logic applied to analysis.
.....

Within the issue of complex unknowns, the matter of eventuality is what can loom. The notion of logic and observation... is to correctly make those distinctions.

Valid observations and a good reminder KBK.
Thanks for putting the human nature-uncertainty and variance- into focus.

When I was young, my then girlfriend bewitched me with a curse: “I wish, you always question your conclusions and your decisions”

You’ve met her too?

George

 

[SY]

And its just as likely that the odd result was just that- a mistake, like faster than light.

This was a much less remarkable finding- when you're measuring things at really, really tiny energy levels (in this case, smaller than the energy of an ant doing a pushup), stuff that's easy to neglect at normal power levels (e.g., phonon-photon coupling) starts to matter. The spin (if you'll forgive the pun) is from the publicists- the science here is far more mundane than the superluminal stuff that later turned out to be wrong, albeit it's still interesting to solid state physics geeks like me.

But when one wants to hype frauds and cranks, any straw within grasp will do.

 

[vacuphile]

This is probably best setup as a wiki...

The current centroid is the average center of the current. Think of a frisbee balancing on a finger. If you have the (mass) centroid of the frisbee directly above your finger, the frisbee is balanced. If the frisbee is off center, it will tip. If you glue a quarter on the edge, you have changed where the centroid is, and it will have a different balance point which you can find by trial and error.

Now think of the current in a uniform conductive cylindrical wire (with the return path for this current at infinity so there are no proximity effects). At DC, the current will flow equally through the entire cross section of the wire. The current centroid will be exactly in the center of the wire. As you increase the frequency of the current, the current will begin to skin (go to the outside of the wire as a result of the lenz effect (law). The current centroid will remain exactly in the center of the wire.

A braid is exactly this...a cylindrical conductor where the current is confined to the outer "surface" of the wire, in this case the outer surface being the wire. Any current flowing in this braid (again, no proximity) will flow uniformly, and the current centroid is exactly in the geometric center of the braid.

A coax puts a wire in the exact same location as the braid's centroid. For this case, both the central core wire and the braid have centroids which are on the exact same line in space, coaxial... So we say this structure has a common centroid.

If the current is flowing equal and opposite for the core and braid, the magnetic fields external to the structure will cancel. This applies to all structures which have a common centroid.

Loop area is defined as the area which exists between the centroids of the "send" wire and the "return" wire. This loop area is what will trap time varying magnetic flux and as a result, a voltage can be induced between the conductors. A coax has no loop area.

What jcx and I have touched on, is the fact that the loop area can be current dependent. A simple example would be a pc board with a long trace that starts and ends with a source and load rca jack an inch apart on the board, with the "hot" trace on the top surface of the board travelling the entire length and back on the top surface, with the current return via a ground plane. At DC, the "hot" current will flow via the entire top trace from the source jack to the load jack, and the current will return via the shortest resistive path, a straight line jack to jack within the ground plane.

Inductively, this loop has the largest area at DC because the send or hot current goes the entire top trace, and the return current takes a much different path, direct. Because current always takes the path of lowest impedance, at DC the ground return will be resistance dominated. As the frequency increases, the inductance along the straight line starts to come into play, and the path of least reactance within the ground plane starts to migrate towards the top trace. The ground current is attempting to bring it's current centroid as close to the centroid of the hot current. If you measure this system's inductance, you will find it is very high at low frequency, and drops as frequency rises.

In the limit, this setup will act like a stripline, high frequency currents will ignore the entire ground plane and travel directly under the top trace.

With wires and coils and electromagnetics, this effect is referred to as the proximity effect. The biggest problem with coils can be seen in the previous example...at dc, the resistance of the ground plane is very small because it's a honkin conductor. But at high frequency, only a very small portion of the copper is actually conducting, most of it does nothing anymore. In coils, this means that the effective resistance is much higher than the dc resistance of the wire.

This will also occur in a chassis.

The important thing to realize is that as the frequency goes up, the current centroids will attempt to get closer to each other lowering the system inductance. Many physical equipment designs do not worry about this effect. I believe JC et al has been successful because they do not ignore it. Perhaps not fully understand it, but use their ears in liu of maxwells equations in entirety.

My apologies for the length...I'll stop here, it's too long already.

j
ps.. going to have to stop editing for clarity, my 15 minutes is gonna run out.

J,

Thanks a million for this lucid explanation, it should be a wiki indeed. Took me longer to fully internalize than for you to write it down . This is really helpful knowledge for what I am doing now (trying to push up the frequency where my power supply goes over 1 mOhm output impedance, around 2Khz at the moment).

high regards,
vac

 

[bcarso]

Thanks for those suggestions and examples jcx.

Brad

 

[jneutron]

My question is: Why did the distortion pattern remain relatively constant, even when I changed the oscillator frequency at least 10:1?

Lets review: From what you've told us in the past..

1. You have a test setup which is notoriously dependent on the cleanliness of the contacts. Resistances in the 10 milliohm (give or take) range affect the outcome of your test, even with the same cable. That is what we call a RED FLAG. STOP...figure out why your setup is so problematic. You have a ground loop and must remove it before you can believe your test results.
2. You received the 15 khz horizontal scanning frequency from various two axis linear electron accelerators in your building in your test results. This means that you are susceptible to external time varying electromagnetic fields. At that frequency, it is near field coupling. Because your test loop is low impedance, (less than 377 ohms), time varying magnetic fields are typically the culprit. Picking up these signals is an indicator of a ground loop problem. Another RED FLAG. You must fix this ground loop problem before you can believe your test results.

3. The relationship between the ground loop, the cable shield resistance, the contact resistance, the cable impedance, all conspire to create a physical structure which will have a frequency dependent centroid loop area. In any test setup, when you have a frequency dependent EMC problem which will affect the outcome of the test, you have a RED FLAG.. Before you can trust the results of your test, you need to eliminate the loop.

In all of this, there is one consistency. You have RED FLAGS, and must fix it first. If you can't fix it, you MUST consider the results as suspect, and you cannot draw conclusions from the results. Especially when the conclusions derived require abandonment of known science. (condensed matter physics as it were..)

Belly up to the bar and put it on the table. Design and produce world class gear.

I do. Well, actually, it's more interesting than just "world class gear". I built some stuff back in '98 which will become your state of the art in another 10 years. I pray that it does by that time, as a subset of my work is necessary for a carbon 12 medical beam delivery system, and I've the genetic history to need it by then..

And my current work...very interesting indeed.

To not bend and understand another one's terminology means you will likely miss the translation, the data..and come to an incorrect conclusion.

first, you've provided no data. second, you've used orthogonal buzzwords to elucidate a phenomena not within your domain of understanding. That is why I stated that I do not understand what you said.

Why is it people call unsupported verbage "data"??

Perhaps, in person, we'd not be too far apart...

Scary thought.

Probably closer than you could imagine. And not scary at all..

And when someone approaches me with a room temperature superconductor if there is some element of validity I'll support it even if very unlikely (as I have done).

I would drop the superconductor into one of our test setups and measure it's critical boundaries....magnetic, current, and temperature.

fasthenry, fastcap are free, opensource quasistatic "field solvers" can give parasitics, frequency dependent R,L, mutual L
take awkward text descriptions of elements, must descretize "manually" for frequency range

I've known of their existence for a while - never really tried to enter a model before

so 1st try here for a U shaped 100 mil wide trace 1" inboard from edges of a 10"x10" ~1 oz gnd plane, connecting the upper, lower centerline points, 0.062" layer separation......

....so from 20 Hz to 2 kHz the inductance has dropped <20% due to gnd plane "current centroid" return path shifting towards the trace
by 20 kHz the effective loop area (~=inductance) has been reduced by <half due the proximity effect gnd plane current path crowding towards the trace

2 oz copper's lower sheet resistance reduces the energy "cost" of the current path increasing length - so the proximity effect is slightly stronger:

What an excellent thing to do. Thank you very much for this effort. Your simulation is exactly what is seen during actual measurements. I fear that the solvers will not be able to do 3-D easily. Does it have the ability to model magnetic materials with hysteresis and remnant fields?

J,

Thanks a million for this lucid explanation, it should be a wiki indeed. Took me longer to fully internalize than for you to write it down

It would have taken me longer, but I ran out of editing time to fix all my spelling errors and phrase bumblings..

j

 

[jneutron]

Believe it or not, I don't. No insult intended. I'm talking with a bunch of 'guys' on line. The end.

If I was rude, I'll take it back in a heartbeat. If I was stupid, I'll take it in stride, nothing else to do, after that.

To me, if one makes a misstep in public, they fix it in the same place, in public.

(actually I might have a clue. I recall something about 'big science',and that's about it. Unless I want to get into half remembered things.)

None taken.

j

 

[jneutron]

2 oz copper's lower sheet resistance reduces the energy "cost" of the current path increasing length - so the proximity effect is slightly stronger:

This effect is much more evident when dealing with superconductors. When I use a supeconducting cable comprised of 6 spirally wrapped around 1, the 1 sees an entirely different magnetic environment than the outer 6. When I ramp the current above one half ampere per second, the center conductor sees no current and the cable will quench at 6/7th of the cable's dc capability, this due to skinning. Conversly, I can ramp the cable at 4000 amps per second as long as I do not exceed 6/7th the cable capability.

When I bend a super, the inner edge sees a higher magnetic field, and that in itself causes proximity effects. It gets worse if the bend radii is on the order of the cable twist pitch..

Supers are really fun to use, but boy they can do things you would not expect at first glance.

j

 

[scott wurcer]

This is EXACTLY what the free energy experimenters have claimed for nearly the past 100 years.

No it's not, they routinely claim energy from nowhere (or usually the zero point, not local thermal energy). Still nothing of substance from any of them. As SY said this stuff makes good press and I seriously doubt it scales to useful proportions. Same as the Casimir force "engine", the ironic fact is that things like this validate quantum mechanics even at totally non-intuitive levels.

 

[gpapag]

This was a much less remarkable finding- when you're measuring things at really, really tiny energy levels (in this case, smaller than the energy of an ant doing a pushup), stuff that's easy to neglect at normal power levels (e.g., phonon-photon coupling) starts to matter. The spin (if you'll forgive the pun) is from the publicists- the science here is far more mundane than the superluminal stuff that later turned out to be wrong, albeit it's still interesting to solid state physics geeks like me.

But when one wants to hype frauds and cranks, any straw within grasp will do.

SY
I would love to have your intended meaning of "mundane" there.
"Earthy" (= well supported by observation)on one side and "boring" on the other would give a quite different meaning on your phrase.

Nevertheless, I can reason that there must be "something" that still attracts you to this superluminal stuff, something more rational than nostalgia.
I am confident that you know more on the importance of aether in the establishment of physics than your public image of you admits.
Thinking over a theoretical construct onto which some properties are attributed to, is not a fatal mistake.
If the Maxwellian equations still hold true, you don’t even have to arrange for a confession (*).

Have a look on electron effective mass calculations of Heaviside (Vol. III).
Variable is electron speed. Derivation through electromagnetic formulae.
See how they fit to recent experimental data.
The reasoning of the time was certainly not more limited by scientific dogma than the reasoning of our time

Thus only
the known can be accurately reasoned about. But this includes,
it will be observed, everything that can be deduced
from the known, without appeal to the unknown. The unknown
is not necessarily unknowable ; fresh knowns may make
the former unknowns become also known. The distinction is
a very important one. The limits of human knowledge are
ever shifting. But there must be an ultimate limit, because
we are a part of Nature, and cannot go beyond it. Beyond
this limit, the Unknown becomes the Unknowable, which it is
of little service to discuss, though it will always be a favourite
subject of speculation. But whatever is in this Universe can
be (or might be) found out, and therefore does not belong to
the unknowable. Thus the constitution of the middle of the
sun, or of the ether, or the ultimate nature of magnetisation,
or of universal gravitation, or of life, are not unknowable ; and
this statement is true, even though they should never be
discovered. There are no inscrutables in Nature. By Faith only
can we go beyond as far and where we please.

(Heaviside “Electromagnetic theory” Vol. I p.11)

George

(*)As far as I know, Einstein claimed and provided mathematical proof that aether as such a construct was not a neccesity. “Wrong” was not a word he used

 

[SY]

"Mundane" in the sense that there's no new physics there, this is an interesting effect from known and well-characterized factors, but not any kind of breakthrough or game-changer.

If the superluminal stuff had held up, we would have a LOT of interesting new physics to figure out. It didn't, ah well, we're a bit disappointed but that's how it goes. As for aether, I would quote Laplace: "Je n'ai pas besoin de cette hypothese."

 

[scott wurcer]

Still nothing of substance from any of them .

Looks like next stop Nigeria for the leader of the pack, http://www.cheniere.org/correspondence/020209.htm

 

[jneutron]

Scott
Thank you for the inside look.
If I would ask for your opinion as to why it is not performed on lower freq. Ics too? (Multiple choise question)
a. Negligible effects at these freq.
b. Effects are not that dramatic at these freq.
c. Effects are equally important but layout correction not feasible within IC size, due to large dimentions –length, height- required for lower freq.
d. None of the above. (Please elaborate in 20 words)


JC, KBK, jneutron, bcarso, simon, jcx, thanks for responding.

George

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.

For a one chip amp, I would make sure the supply rails have a common centroid...return, plus, and minus. See attached..a pic from 2006. Home made triaxial cable.
j

 

[scott wurcer]

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 numbers are 5-10x less on todays SMT packages OTOH we are trying to make 10G TIA's. If you check the AD8010 datasheet you will see a bypassing scheme that was necessitated by bond wire mutuals, that gets pretty amazing diff gain and phase with lots of parallel video loads.

 

[KSTR]

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.

For a one chip amp, I would make sure the supply rails have a common centroid...return, plus, and minus. See attached..a pic from 2006. Home made triaxial cable.
j

Well, for my favorite chip amp, the TDA7294 (and its cousins), the guys at ST got that bit right, it seems. Also it has split power and preamp/driver supplies, another nice feature given the rather bad PSRR

 

[simon7000]

See attached..a pic from 2006. Home made triaxial cable.
j

If you ever need any small quantities, let me know I still have scraps left over. Normally we get "Thick" triax on 3500' put ups for stadium use. Any length under 300' is scrap.

Although with the high price of scrap we actually are getting rid of some of it.

Today was a shipping day. Three crates for two speaker clusters went out. Attached picture!

Anyone have a contest for biggest loudspeaker?

 

[Wavebourn]

Anyone have a contest for biggest loudspeaker?

I have pair of line arrays made of capsules from dynamic microphones, 16 in each.

 

[SY]

I have pair of line arrays made of capsules from dynamic microphones, 16 in each.

Is that the ones I heard at your house? If so, they were surprisingly good.

 

[gpapag]

"Mundane" in the sense that there's no new physics there, this is an interesting effect from known and well-characterized factors, but not any kind of breakthrough or game-changer.

Thank you for explaining.

If the superluminal stuff had held up, we would have a LOT of interesting new physics to figure out. It didn't, ah well, we're a bit disappointed but that's how it goes.

Don’t give up hope (*). There may come one day…
Old stuff are like old men, stubborn. They refuse to give up.

http://maxwell.byu.edu/~spencerr/phys442/history.pdf

As for aether, I would quote Laplace: "Je n'ai pas besoin de cette hypothese."

Well Messier SY, here is Lagrange from the same episode (**): “Ah! c'est une belle hypothèse; ça explique beaucoup de choses.”

George

(*)I know you are dying for a new “tongue test”

(**)Talk about God existence (hypothesis)
Besides, both men were long dead by the time things were becoming hot with aether-based valid electromagnetic theories.

 

[SY]

Don’t give up hope (*). There may come one day…

I sure hope so. There's so much of the universe to explore and waiting millions of years for answers is not an attractive prospect.