Extraction of parasitic from trace

Hi

I am working on a high frequency design in which we in principle is using a very high frequency switch mode converter

I need low loss, so need to minimize capacitive parasitics, and also minimize inductances. They conflict of course, smaller trace has lower capacitance but higher inductance.

We use Altium for layout and have used ADS for parasitic extraction. But we cannot afford the license.

Is there a low cost alternative that is ok in terms of amount of time to get it done?

I was thinking maybe somebody has made a tool, maybe even excel, that could import the trace and image plane /traces to do a finite element analysis /model.

Any hints?

Reply to
Klaus Kragelund
Loading thread data ...

The Saturn software does the basics for standard trace geometries.

Sonnet Lite does 2.5-D em analysis for free.

A switcher should be small enough that the speed of light is inconsequential. How high a frequncy did you have in mind?

Reply to
John Larkin

Thanks, that looks quite nice.

It's not a standard switcher :-)

Operating at 100MHz, and with 1ns switch node rise time. So even laminate loss is important.

Reply to
Klaus Vestergaard Kragelund

Sounds like fun. GaN?

One problem we encounter is driving laser diodes fast. At 5 amps per ns, it gets hard to wire things up. Eveny nH of inductance drops 5 volts. Tx line impedances need to be a fraction of an ohm. Most laser diode makers don't seem to understand this and sure don't furnish IBIS files. The laser physics can help, or hurt, and that's not specified.

Driving Pockels Cells can have similar issues, but kilovolts.

A proper sim, device and package models, Spice, IBIS, 3D e/m sim, thermals, PCB, is probably impossible, or would take years and megabucks. I use instinct and Dremels.

If you crowd parts to keep the electricals tight, then they get hot. Yet another dimension in the optimization space.

Reply to
John Larkin

OpenEMS - full 3D. Don't let the cost (£0) fool you, it'll give the same results as the mid 5 figure commercial offerings.

Reply to
John May

Yes, EPC GaN.

Excactly what we are looking into also. The layout did a mistake, and we are fixing it.

We have used ADS so far, but no more license.

The power we have is low, so we pack it very tight.

Reply to
Klaus Vestergaard Kragelund

Thanks, I will check that out.

Reply to
Klaus Vestergaard Kragelund

I love those parts electrically, but they are fragile and close to impossible to rework. I have a pulse generator output stage design that's great, way better than possible with mosfets, but we did it as throw-away baby boards to make rework possible. We build, test, and glob-top the baby boards.

formatting link
formatting link
That pulse is lowpass filtered for beauty. The unfiltered pulse is faster but a little ratty looking.

I used the tiny 4-ball parts, which fracture or break off the board if they are touched. The bigger ones with fat rectangular contacts are tougher, but the corners will still fracture if tapped.

Reply to
John Larkin

Sounds like VHF transmitter design :-)

At those frequencies you also need to consider the skin dept, i.e. the current flows only at the surface of the traces. Thus the resistive losses of the trace is larger than on DC and you may have to use wider traces if large currants are involved.

Reply to
upsidedown

More important, copper that's shiny on the bottom. Most of the current is on the dielectric side.

Regular PCB:

formatting link
Isola, medium-range microwave laminate

formatting link
A serious microwave laminate is full shiny and peels with a toothpick.

Reply to
John Larkin

There is one good thing about FR4 at high frequencies: The high losses mean that reflections are quickly damped, so impedance matching is not quite so critical. This is quite useful on 10Gbit/s data tracks where attenuation is not necessarily a problem but multiple reflections causing notches in the frequency response would be.

John

Reply to
John Walliker

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.