ATLC2

I guess it depends on the image you dropped in there.

The DPI factor is going to show a horse shoe into it if it isn't correct.

You may want to investigate the image format a little closer, it is very possible the image data in the header is incorrect.

I've seen a good many programs really screw up the PPI in the header of a BMP. Most programs don't even bother to use that information, for the most part.

Jamie

I'm beginning to understand this. My real-life problem is to come up with a PCB footprint and pad stack for this edge-launch SMA connector

I'm going to do a 4-layer board with L2 ground plane, and I'll have to cut a window into the L2 ground to get the impedance up to 50 ohms. Just the connector pins, in free space, have an impedance of 96 ohms...

Here's the PCB stackup model, created with Paint. ATLC2 actually inputs a bitmap file.

Here's the solution. This ran for 20 minutes or so.

This problem is actually 3D, but this will get me pretty close. I can't imagine what the runtimes would be for a full 3D sim.

ATLC2 is a little quirky, but so are all field solvers. It's really very cool. I haven't tried yet, but I think it can solve arbitrary sheet resistance problems, something I've wanted for a long time.

That reminds me of the program I constructed years ago to calculate radiation images from various types of low frequency antennas.

I did that on a C-64 in BASIC. Even though it worked, It was a bear application on the machine. For a good detailed image to the printer it took ~ 1 hour of processing with all the trig that was in there for the graphic plotting of the image. The actual math for the RF calculations were not bad, but add that to the geometry/trig needed to do the graphing was a problem. Especially when I had to get a floppy disk drive in the operation to cache data.

I remember doing a 6 element beam that resulted in 8 sheets of paper. That was a 4x4 image that I put on the wall for a detailed display of what the RF looks like. I did that for a show and tell one night at a ham radio club meeting. That run took over 2 hours, 4 floppies and a lot of my attention!

It's been years since I've written any software of that kind.

I have a close friend that works in a GOV lab with G,T+ Hz frequencies and he tells me they have made some discoveries, indicating that our current view on how R.F or AC in general works and it's effects as we understand them, maybe slightly incorrect. Apparently they have been using some ideas and theories of their own, allowing them to do things way up there that could never be possible if RF was believed to work as most of us believe. At least that is what he tells me. He indicated that not knowing the greater aspect of what is happening does not much matter at the lower frequencies but it does up there and when you find that out, it then explains a lot more about the lower frequency effects that many of us just ignore or pad off as parasitic, etc..

And his last comments were, "If I tell you any more, I'd have to shoot you", So it was left off at that part.

Jamie

Not even worth protecting. There isn't any part of "the greater aspect" that we don't know about. General and special relativity, derived from Maxwell's Equations, have been verified to impressive accuracy on all size scales, from particle accelerators to astronomical observations. Quantum electrodynamics (and by extension, QCD and the core of the Standard Model) directly incorporates Maxwell's Equations, and have made numerous, high accuracy predictions of everything from subatomic particles unknown to science, to the structure of the universe (again, verified by astronomical observation). The predictive and practical power of these theories, over the last two centuries, completely overwhelms any heresay anyone might have.

More likely, their secrets are materials properties. Optics always apply, but are rarely used at radio frequencies because the physical size is prohibitive (a lens or mirror must be hundreds of wavelengths across to correctly approximate using geometry alone). At frequencies above UV, materials and individual atoms tend to absorb photons directly, and making mirrors, let alone lenses, is difficult (the Chandra x-ray observatory uses a very strange mirror design). The range from microwave to UV is special, because materials are available which reflect or transmit. A lens made of clear acrylic will work more-or-less the same from visible light (it absorbs UV) down to microwaves, except for a conspicuous IR gap (due to molecular absorption).

Engineered materials are very important, because up around light frequencies, conductors stop conducting as such. (Conduction is electron motion, required for electrostatic and electromagnetic induction which reflects or conducts electromagnetic waves. At small scales and high frequencies, even normally good conductors misbehave.) Copper is an excellent example of this: it looks pink because it absorbs blue light, rather than reflecting it like a good conductor. To carry a THz signal, no size wire may even be acceptable; fiber optics would then be required.

Tim

no decoherence.

I'm with you here. The science is good, but the materials may not be understood well.

Regarding the beam antenna, there is always NEC2.

And another:

At first glance, I see a drawing of a transmission line equivalent network having serial R and shunt G, then the statement "Zo=sqrt(L/C)". Simply not true for a line with either or both series or shunt loss.

Doesn't augur well for the software.

"heresay"

Hearsay or heresy?

Both seem appropriate ;-)

The software is a tad buggy, or at least non-obvious. But I'm not concerned with ohmic losses when I do a PCB transmission line calculation. If ohmic losses were enough to affect Zo, my signal would be truly trashed.

I eventually got it to do this:

which cross-checks against the "classic" Linux ATLC program to under half an ohm.

In ATLC2, you apparently have to do the "full solve" option to get the correct impedance in a mixed-dielectric case. The two selectable partial solutions, "Ls and Rs" and "C and vf" each report an impedance, but the numbers (76 and 32 ohms in my case) are, according to the author, not meaningful. ATLC does only an electrostatic solution, where ATLC2 does a full 2d em thing... I think. It runs hundreds of times slower than ATLC.

This program would be a lot better if it was properly documented. Maybe I'll volunteer.

It is fun to watch it work...

I think ATLC2 will give the correct impedance running only the electrostatic solution (like ATLC does) provided the only dielectric is vacuum. The 95 ohm result looks pretty good, checked against other methods.

I'm spending way too much time on optimizing one connector layout, but this is interesting. I do know that, if I just plop this edge-launch connector down over a layer-2 ground plane, it will trash my signal.

Anybody tried Sonnet Lite? I played with it some but it didn't immediately make sense to me.

Thanks for the link. Just starting to use it. Very powerful and appears accurate. I love it.

John S

It's much easier to use, and more fun to watch, than ATLC. You can change a material dielectric constant on the spot, without building a new bitmap or hassling with color values. It is a little quirky and buggy. If he'd tune it up a bit, and write a better manual, it would be a real winner.

Keep your bitmaps small or it will run for hours.

Ohmic losses result in a characteristic impedance having a reactive component. Probably small enough to ignore at *your* frequencies of interest. Pretty significant at LF. Nominal 50 ohm RG223 starts to look capacitive below about 100KHz.

Does anyone here have anything more to report on ATLC2 experience?

I am trying it for flexible circuit striplines with polyester coating in which the materials are so thin that I need to go to 0.0005 in/pixel to create bitmaps with reasonably accurate geometric proportions. My bitmaps are 2400 x 3200 pixels.

It does not show the entire bitmap, so I wonder if that is just a graphics display issue or if the solution domain is also being truncated/adulterated?

Anyone have experience with bitmaps that don't display fully?

Also, originally, ATLC2 was solving these in about 3 minutes, now it runs for hours without solution. Wondering if anyone has similar experience to this as well.

Thanks.

We occasionally use ATLC and ATLC2 to simulate PCB traces, like a microstrip transition to an edge-launch SMA connector.

Connector in free space:

Connector on PC board:

but we haven't tried any bitmaps as big as you are using.

ATLC and ATLC2 use different algorithms but produce pretty much the same results.

The actual PCB worked pretty well.

Also:

ATLC is a lot faster than ATLC2.

You might look at Sonnet Lite: professional, 2.5D, free.

Anyone have the email of the author (Kenneth L. Nist) (KQ6QV) or a link to his source code if published ?

It would be fine to port to linux and optimize it for SMP

Thanks

Nicolas PLANEL