4-Layer boards

That's a delay-line experiment. It turned into an arbitrary waveform generator that didn't work very well.

The linear one did work:

On thinking about it, it's probably as much reflex as anything else -- I mostly do 2-layer boards where I fill anything that's left over with copper at ground potential.

I'll try it the "usual" way next time & see...

My point is that the other ultra-low cost ones all do the same thing.

I am talking ~$14 for 10 double sided through-hole PCBS, solder resist, silkscreened, everything, including shipping.

Yes OK you may have a point there. I have never used eagle so do not have that perspective.

Yes I suppose, although there is something to be said for having a consistent workflow and particularly for using something the online visualizers/checkers can work with.

The inhomogeneous dielectric (part air, part something else) also causes dispersion, if I remember correctly.

Pere

Right, but on FR4, other losses kill fast edges as much as dispersion.

OK thanks, I wasn't thinking about transmission through the air... I was thinking about different E-field strengths in the dielectric.. (which may not even be important.)

George H.

It's an asymmetric dielectric. I don't have any tools to analyze dispersion, but as I noted, other losses probably swamp it.

ATLC can calculate the impedance and capacitance and inductance of weirdly shaped txlines and mixed dielectrics, but not dispersion.

Fun stuff.

Sonnet Lite might do dispersion... not sure.

No download left here, (no pics) How about a layer of fiberglass over those few traces you really need fast? (Maybe that would make it worse?)

I always think of dispersion as the phase shift from some nearby (in frequency space) absorption/resonance. I don't have a picture of two dielectric layers... Well if I send in a pulse, I get a little pulse first and then the big one.

George H.

IIUC, microstrip dispersion arises from the EM boundary conditions. Assuming a perfectly conducting ground plane, the field has to go roughly like sin(k_z z) inside the board. (Z is the vertical coordinate.)

The higher the frequency, the more curvature you get in the sinusoid, so at higher frequency dE/dz at the surface is smaller, which means that more of the field is in the dielectric compared with the air.

That makes the mode slow down.

Cheers

Phil Hobbs

Yeah, I've got this image of polarization charge at the surface.

Oh... high frequencies are confined more. Thanks, (There are these pics by Howard Johnson (?) showing ground return current at different frequencies.)

It's kinda like the picture I have of a ground shield* (danish butter cookies perhaps.) At low frequencies the current flows everywhere, at HF the two sides decouple, there's all this electro-static crud on the outside, and my signal on the inside.

George H.

HoJo is an idiot. I started highlighting the gross errors in his Black Magic book but got tired less than halfway through.

He's a terrible writer, too.

Nice!

The above link gave me "server not found"

Have to agree.

Four or five years ago, he came out with this absolutely hilarious video purporting to show that ground currents follow traces. Brownie points for anyone whose Google-fu can bring it up--mine didn't.

Thing was, he and his colleague were using a big electrostatic generator and an array of metal squares to mimic the ground plane. The idea was to watch the HV arc over between squares to demonstrate the preferred current path.

Surprise surprise, he found that the return current wanted to travel under the trace....um, oops, well as long as it was on a 45 degree diagonal...i.e. the field concentration at the corners of the squares caused corona breakdown, so the current wanted to go diagonally. The current wanted to go diagonally even if the trace didn't.

He never seems to have thought about the fact that it's magnetic effects that steer ground currents, and that in order for his demo to mimic normal circuit board reality, he'd need the trace inductance to be in the tens or hundreds of henrys.

Now I see he's selling a one time license to his "collected works' for a mere $10,000 per company.

Nice work if you can get it, and aren't too worried about the collateral damage due to spreading misinformation.

Cheers

Phil Hobbs

I suspect he's made a fortune from his seminars. His obsession with return currents is bizarre.

He is capitalizing on the fear that many engineers have of the n-word (nanosecond) or the greater fear of the p-word (picosecond.)

rote:

te:

Having spent quite a lot of time getting junior engineers to pay attention to return currents, I can see the pedagogic point. I can remember Win Hill recommending Howard Johnson's opus when it first came and how peeved I was when I finally got my hands on it. Sadly, if there is an alternative, I hav en't heard of it. I've got a couple of highly technical microwave transmiss ion texts that go further, but aren't really accessible to junior engineers .

Ralph Morrison is much better at organising information for the reader, but the sections on transmission lines in the later editions of "Grounding and Shielding Techniques" aren't all that comprehensive.

Not unjustified. People who grew up with TTL mostly found it hard to get to grips with transmission line effects. When I was talking about making a ch unk of circuit super-quick by realising it in ECL (back in 1985), another e ngineer commented that in his experience it had been hard to get ECL to del iver the performance it promised.

When I said that I'd not had any problem getting it to deliver, his respons e was that I was an analog engineer (which was true, though I was also perf ectly happy with digital logic) and used to finnicky side-effects.

Works for me. I was just being lazy, there are some tables of their default gerber extensions like:

G1, G2, etc Mid-layer 1, 2, etc

GBL Bottom Layer

GBO Bottom Overlay

GBP Bottom Paste Mask

GBS Bottom Solder Mask

GD1, GD2, etc Drill Drawing (assignment based on order of drill pairs appearing in the Drill-Pair Manager dialog)

GG1, GG2, etc Drill Guide (assignment based on order of drill pairs appearing in the Drill-Pair Manager dialog)

GKO Keep Out Layer

GM1, GM2, etc Mechanical Layer 1, 2, etc

GP1, GP2, etc Internal Plane Layer 1, 2, etc

GPB Pad Master Bottom

GPT Pad Master Top

GTL Top Layer

GTO Top Overlay

GTP Top Paste Mask

GTS Top Solder Mask

(I have no idea what a "pad master" is by the way...!)

I never read any of his books, but just to be clear about the physics. (And not some crummy demo that he did. I recall a link to it.. he probably took it down.)

At high frequencies the ground currents do tend to gather closer to the trace. To me this is exactly what you were talking about with the E-fields... I mean after all the E-fields start and stop on charges. (OK that last is an electrostatic approximation... and once we are talking about stripline it's somewhat suspect.)

George H.

Hmm... I found it useful. Is there something "wrong" in his physics? (But as I said I haven't read him.)

George H.