Lost a Bet

Jun 15, 2017 13 Replies

I was figuring that if you made a really tight meander delay trace, charge would prefer to jump between the loops, rather than go the long way around. That would make a precursor on a voltage step, and raise the impedance. So one of my guys tried it.



formatting link



formatting link



formatting link



formatting link



That last meander is about an inch of total length. PADS does that on demand.



The TDR and TDT (reflected and transmitted step responses) are:



formatting link



In my defense, there is a tiny precursor on the through-pulse, and the impedance changes a little. But not enough to declare victory.



The total one-way delay is about 1.5 ns, and the output risetime is



100 ps. You can clearly see the launch SMA connector (big capacitive spike at 3.5 cm, the three vias, the meander from 7 to 9, and finally the 50 ohms of the second scope channel.

The meander trace is 16 mils wide, 22 mils pitch, and it has is 9 mils of FR4 above a ground plane. Total length between connectors is 9".



So, tight meanders seem to work.


John Larkin Highland Technology, Inc lunatic fringe electronics

Nifty. Can you take a shot of the whole trace, including the SMA at the far end?

-- john, KE5FX

Corollary to that: coupling between adjacent microstrip is s***ty. Good news for parallel buses, bad news* for differential traces.

The drool means it's pretty bad for certain RF purposes (say, filters and matching networks?), and longer analog delays (maybe with some compensation though?). When you don't care about any of that because your logic thresholds are 30-70%, it's no problem.

*Not actually bad news at all. It just means you should treat the two traces as individual microstrip traces, which instantly gives you an idea what to do about common mode: terminate it!

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

formatting link

The TDR includes the whole trace, up to about cm 9. The losses in the traces blur the TDR resolution such that the end SMA doesn't show up much. TDR is only approximately a graph of impedance vs distance; losses mess things up.

I should have done another TDR starting at the other end. I'll do that.

John Larkin Highland Technology, Inc lunatic fringe electronics

With 16 mil traces and 6 mil spaces on 9 mil dielectric, coupling is small.

The step response is fine for logic. It is a pretty long trace.

100 ohm differential is usually just two 50 ohm traces of about the same length.
John Larkin Highland Technology, Inc lunatic fringe electronics

What Phil said, and also, stripline can be made with much more beneficial geometry -- face-to-face rather than side-by-side, for example.

There's also lots more stackups than a standard 4 layer build: a pair of 30 mil traces, on 62 mil core, with 7 mil separation (a standard 2-layer board), are coupled more closely than a pair of 10 mil traces, 7 mil separation, 10 mil prepreg (standard 4-layer board).

Or, for the 4-layer build, simply removing the ground plane beneath the trace, so you get a half trench stripline geometry.

You could even make a slotline geometry with even higher coupling (but also fairly high impedance, unless you use a high-k substrate).

I think it would be difficult to make wideband (>> 1 octave) directional couplers over, say, -20dB, in microstrip.

Tim

Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com

Here's the TDR starting from the tight meander end. Again, the opposite SMA connector (at the right-most cursor dot) is blurred by the trace losses.

formatting link

The vias are less distinct, too.

John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
[...]

Well I hope your colleague is extremely patient, 'cos I'm still waiting for the burger & fries you owe me from our wager back in 1998!!

Isn't there a statute of limitations on newsgroup bets?

What was that about anyhow?

John Larkin Highland Technology, Inc lunatic fringe electronics

Yes. 6 years. After that it's only binding in honour only.

F***ed if I can remember after all this time, John. But I do recall my prize was a burger & fries and a beer in a biker bar somewhere dangerous (Oakland, in all probability). Oh - and I of course regard you as a highly honourable man.

Well, we could go to Zeitgeist. It's a biker bar, but the kind of bikes with pedals. 44 beers on tap. Al fresco dining, sticky picnic tables under a freeway ramp. Joerg loved it. Free matches.

formatting link

The food is actually good. Burgers, home fries, grilled cheese sandwiches.

John Larkin Highland Technology, Inc lunatic fringe electronics

"Warm beer, cold women"????? N..i..c..e! Since you're picking up the tab, I shall require something a bit more classy (well, classy for Oakland, anyway:)). Somewhere not totally burnt- out, for example. It's not too much to ask after a 19 year wait, surely?

Oakland? I don't know much about Oakland, except that it's generally dangerous.

My not-yet-then-wife used to live in Oakland. I decided that the world's biggest impedimet to romance is the Bay Bridge. So, she moved.

Zuni and Absinthe both have high-end burgers, $16 ballpark and worth it. Trick Dog is amazing, both the burger (long and skinny, on a hot dog bun) and the fries (hot and gooey inside, crispy as broken glass on the outside.) Dang, I'm hungry.

John Larkin Highland Technology, Inc lunatic fringe electronics

That's all part of the appeal, though, don't you think? I mean, that's why we decided to meet up in a biker-bar honky-tonk burger-joint there. Have you lost your sense of adventure?

Very sensible. All that fog and whatnot. No good.

After a near 20 year wait, so am I! :P

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required