another hack

Where's that link you were going to post, about skin loss not mattering?

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

is

ng,

h t

We had a six layer board with FR4 between the four inner layers, and isocya nate bonded Teflon (PTFE) cloth for the two out layers. The people who made it warned us that the outer layers were a bit soft, so the board might be a bit bendy - it was an extended triple Eurocard, so quite big, but that di dn't give us any trouble, and certainly did bow in any direction.

We had a lot of trouble with the first prototype because some idiot at the board shop didn't like the layer order and scrambled it - all the controlle d impedance microstrips on the surface had the wrong impedance - but the la ter versions worked fine. It wasn't strictly necessary to get Gigabit Logic GaAs parts to work right, but both the boards and the logic we put on they were very expensive, so we did go in for a bit of over-kill.

--
Bill Sloman, Sydney

Huh, cool, thanks. I've only done manual machining, there are a few mill marks on the vise jaws of the student shop of my alma mater that should have my name on them. (I have given them money.) Oops, George H. in

[snip]

On a recent high speed board, we specified that the board be rotated 8 degrees with respect the the weave, so that differential pairs would be better matched. Otherwise one half of the pair might be in an impedance "trough" and the other might be on a "peak". The figure of 8 degrees was a compromise to do with area efficiency at the PCB manufacturer. Most of our traces are 0, +/-45 or 90 degrees with respect to the board edges.

We also separated some differential pairs (into two uncoupled single ended 50 ohm traces) to match the weave pitch, if they had to travel a long distance.

Regards, Allan

On Fri, 11 Dec 2015 15:54:03 -0800, John Larkin Gave us:

"mattering"? About as much as your concentration on spelling does. Funny you should use that term talking about skin effect.

Just a smattering of humor since there isn't much physical presence involved... just a smattering. But sure... I knew you meant the present participle of matter.

gold plating, it rides on the plate. Silver is better.

Gold oxidizes less (near never), but takes solder well. But it conducts even worse than copper.

But the substrate medium matters too, and you just blow shit off you don't understand worse than Donald Trump, in fact.

So tell us, genius. What frequency is Litz wire good for. Why is nobody using it in high frequency applications, since the entire reason it exists is skin effect?

You posted:

Skin effect is insignificant. Do I need to send you a link?

So, send me the link.

Cool. Do you thing the trough effect is real, or are you just being cautious?

I often see a small wiggle in the TDR impedance of a trace, and it looks like the scale of the fiberglas weave. But it's very small. Other things, like slower Z variations and attenuation/dispersion, are worse.

Thank Goodness for equalization chips. Fast differential data over any distance wouldn't work without them.

Anything conductive gets very lossy to HF magnetism. Metglas is just less conductive, is all. Nickel-loaded conductive paint is good enough shielding that the first-generation Macintosh needed no metal case.

"Taking the consultant too literally" might be closer to the truth.

I did my first 10Gb/s design in 2003. That seemed fast at the time. Now, for 10Gb/s, I just do the schematic and let the layout artist do his thing. As you noted, the equalisation in contemporary transceivers is amazing and can correct for a variety of PCB sins.

This design (the one with the oriented weave) is our first design to do

28Gb/s on diff pairs. That sounded scary at first, so we contacted a consultant. Based on the consultant's advice, we made the following changes:

- low surface roughness layers to reduce loss and dispersion

- a better via design

- oriented weave

- "long distance" diff pairs used a wider spacing to match the weave.

Hopefully the prototype will work ok. However, there is no chance to do a "what if" that would allow us to determine whether it would have worked as well without (e.g.) the oriented weave.

I've seen the weave in RG58 on a TDR. I never knew braid shielded coax was so bad until I compared RG58 against a piece of semi-rigid (which had a perfectly smooth trace).

Allan

e

s

Ever tried "conformable coax"? It uses solder-soaked braid rather than copp er tube, which is what makes more conformable than semi-rigid, if less smoo th-surfaced. The local funny connector distributor (for the Cambridge area) rather liked it - it was good enough for most high-frequency stuff, and ea sier to work with. Worked fine for us back around 1990, but back then we we re happy with 0.5nsec transition times.

--
Bill Sloman, Sydney

We did a gen1 PCIe interface, in an Altera FPGA, and we were all worried about the layout. Turns out that you can practically do gen1 on barbed wire.

Semi-rigid is beautiful. Once people order custom cables for a project, and something goes wrong, buckets of the stuff show up on ebay.

I sometimes cut one in half, let the inner poke out, and use it as a scope probe. Solder an 0603 resistor on the tip to make a 500 ohm

10:1.

You can also lap solder it onto a PCB and make a pretty good transition.

(I suspect that those wiggled traces on that board are silly.)

Second from right has one less wiggle. What's it supposed to be, some sort of delay matching or anti-reflection gimmick?

Mark L. Fergerson

PCIe_Pickoff3.JPG

Length or delay matching. The inside track is shorter when a differential pair goes around a bend. The wiggles add extra length to compensate. It's better to wiggle near the bend than to save it up for one big wiggle at the end of the trace.

The edge coupling of the differential pair is a minor part of the overall impedance (it's typically more like two 50 ohm traces side by side) so the wiggle doesn't affect the impedance much.

Allan

The idea is to make both traces exactly the same length overall, to avoid skew between the conductors of the pair. The wiggle thing is naiive, because a simple wiggle like that is not the same as a longer trace, and has goofy side effects. There's no point to matching trace lengths to way below the signal risetime, which is something like 100 ps on the board I showed.

There's not a lot if interaction across the wiggle. Two 45 degree bends are better than one 90 degree bend, so they aren't great wiggles.

Name one.

20psec is 4mm on FR4 and materials with similar dielectric constants. Those wiggles were in that ball-park.

--
Bill Sloman, Sydney

Also Indium? Indium-Tin Oxide is not only conductive but also optically transparent, so it's used for contacts on glass e.g. for LCD displays. I think it's an alloy but I'm not sure in what ratio and what is the phase diagram. Perhaps the conductivity is entirely from tin and indium is just there for something like mechanical or anti-corrosion properties? Anybody knows?

Indium oxide is a metal as well, but by itself it's much poorer than tin oxide. ITO is better than either--iirc you want about 10-15% indium oxide.

Cheers

Phil Hobbs

? 2015?12?9???? UTC

+8??1:43:39?John Larkin???

Jiangmen ABQ Electronic Material Co.,LTD was estimated in the year of 2006.

specialized in the production of high-end photoimageable PCB ink.

We passed the ISO9001 quality standard identification.and ISO14001.

Our Main products:

PISM white colour RS-2000 (W series)for LED ( patent products ),

PISM white colour matt for automative (we have TS16949 certificate , in Eur ope our RS-2000 WDM already be appointed to use in Automative industry,so t here are a lot of PCB factory from china ,from korea ,from Thailand are buy ing RS-2000 WDM for their Europe automotive order).

Electronica spray ink RS-2000 SP(our electronica spray ink is sucessful in Korea market already)

PISM green colour RS-2000GL

PISM other colours RS-2000 BL,K,Y,R ,and others.

$UV cured solder mask UV-1000 G green colour

UV cured solder mask UV-1000 W white colour

Etching ink RS-1200,RS-1280,RS-8580

UV etching ink UV-680,UV 690.

Conduction current carbon ink RF-1860

Low pressure coating ink:RS-2000 BDHF

Marking ink.

All of our products are approved by UL ,(UL filenumber :E-310593).

Main customers in china :Founder technology ,Tiger builder ,Red board ,KB

,mankun and so on.

Due to we are setting up a new factory ,the total produce capacity will

reach 800tons per month.

In order to meet the new capacity we must develop oversea market largely

with very competitive price.

Our oversea market is developing very fast on the basis of our competitive

price ,stable quality ,excellent service.

Now our oversea market is :korea ,russia ,india ,parkistan,saudi,brazil

,usa,vietnam,turkey,Iran,malaysia and so on.

Consindering the cost ,now more and more PCB manufacturers choose china

mainland PCB ink supplier.

Now we are developing your market,so if we can get cooperation chance from

you then we will be very much appreciated.

Looking forward to hearing from you and best regards.

Mrs. Wang

****************************************************************

Export Manager

Jiangmen ABQ Electronic Material CO.,LTD

Facebook:

Video:

Web:

E-MAIL: snipped-for-privacy@hotmail.com | snipped-for-privacy@gmail.com

snipped-for-privacy@jmbq.com.cn

SKYPE: snipped-for-privacy@hotmail.com

MP: 0086-18501626226

(Whatsapp,Viber,Line,Kakao,Wechat)