resetting a filter

e:

@fonz.dk"

te:

duction staff

and solder 42

ible. And

C board?

oned RM core

re compact.

surface-

delay line,

ld you that

ternal fields?

ive it

s,

he fields

n lines)

same

mission

A point that I made when I originally mentioned the Filotex cable - it is available from Farnell in Europe, but not - as I pointed out that post - from Newark in the US. I did give you credit for knowing what your local distributors do stock, and indeed Joy Signal Technology, LLC1020 Marauder Street Chico, CA 95973 USA does appear to be tolerably close to your patch.

se

ECL is quick and was always designed to drive transmission lines. The negative power supplies are a pain, but when I was doing it they were also handy to provide more or less +/-5V power supplies for fast analog amplifiers, comparators and the occasional more exotic part.

The ECL I was using was not at all picky about it's negative supply rail. The 100k ECL was specified for -4.5V, but the GaAs wanted -5.2V and running the 100k on -5.2V just meant that it ran 16% warmer (and a bit faster). We could cope with the extra heat.

RTL - on the other hand - has nothing to recommend it.

One wouldn't avoid PECL just because it's easy to blow up, but it's certainly a reason to find another solution, if you can.

-- Bill Sloman, Nijmegen

Squirm, worm.

Hey, the offer goes for you, too.

--

John Larkin Highland Technology Inc

Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom timing and laser controllers Photonics and fiberoptic TTL data links VME analog, thermocouple, LVDT, synchro, tachometer Multichannel arbitrary waveform generators

" snipped-for-privacy@fonz.dk"

production staff

solder 42

invisible. And

board?

RM core

compact.

surface-

delay line,

you that

external fields?

it

fields

lines)

Hey, Maxim is in the fast logic business now.

That's encouraging.

Hittite and Inphi make really fast flops, but some are $500 each.

These guys are interesting...

but their logic isn't terribly fast.

If you couldn't follow John Fields' logic, you've got a comprehension problem.

Admittedly, my wife has got much too much sense to get involved in the petty bickering here, and your wife is probably equally sensible, but since it's your wife whom you think has been insulted (rather than you personally), she really should have the opportunity to call Jim Thompson something unpleasant on her own account, rather than being stuck with you as the sole - not all that impressive - defender of her reputation.

-- Bill Sloman, Nijmegen

It's still hairball logic. Tangled, asynchronous, glitchy. But he sure does use a lot of big words.

Since nobody here knows her name, except for the few who have met her, her reputation isn't an issue. She'd just laugh at the stupid old geezer, as she laughed at other inarticulate MIT geeks in a later day.

Something else I thought I knew that wasn't so. Thanks for the education.

Crimped BNC's were a whole lot better. At Cambridge Instruments we eventually banned everything else. You only had to spend a couple of hours fault-finding on a million dollar machine and solve the problem by tightening a BNC nut to get the idea that crimped connectors were a very good idea. The crimp tools were expensive, but fault-finding time was very expensive.

-- Bill Sloman, Nijmegen

rote:

...@fonz.dk"

production staff

ip and solder 42

visible. And

l PC board?

ntioned RM core

more compact.

an surface-

ir delay line,

told you that

external fields?

drive it

ends,

, the fields

sion lines)

he same

ansmission

the

ires

but

you

lier

ils

TL

cause

ot

Maxim is well-known for producing interesting data sheets, less well known for shipping the products that the data sheets describe.

Nor is it up to driving transmission lines, or traces that look anything like transmission lines.

-- Bill Sloman, Nijmegen

Impotent John Larkin and (you fill in the blanks) wife Mo ;-) ...Jim Thompson

On Sun, 12 Aug 2012 17:27:26 -0700, John Larkin

ooops, can't reset a hard wired ego.

Ah, topic drift and nostalgia time...

Tightening the UG-88/u nut was not a common problem. Having the shield braid fall apart inside the connector and short everything together was about equal to having the center conductor retract back into the connector to the point where it didn't make contact. If the nut looked loose, I would always take apart the connector and inspect the guts. It was not unusual to find some of the rubber components missing.

During the age of CheaperNet, I only used crimped BNC connectors. When I found a UG-88/u, I would try to replace it with a crimp type BNC male. Unfortunately, I ended up with a mix of RG-58a/u (solid center conductor) and RG-58c/u (stranded center conductor), which used different connectors. I also had an assortment of "waterproof" BNC connectors left over from a USCG direction finder project. After mixing these, I successfully reproduced the UG-88/u connector failures using crimp type connectors. There were also multiple attempts at solderless BNC male connectors, all of which I consider to be bad ideas. That was the 50 ohm part of the puzzle. I was also doing ARCNET (93 ohm) and some video (75 ohm) at the time. The cable jungle in my palatial office and service truck was barely manageable. Where everything came together was also umm.... interesting:

There were plenty of ways to do CheaperNet wrong. I cleaned up several networks, where the BNC tee connectors were in the wall, and someone had simply added a single coax cable extension cord between the card in the computah, and the tee in the wall. I installed far too many double BNC connector wall plates to fix such creativity.

One thing I really liked about Cheapernet was the ability to sniff ALL the network traffic at any point along the cable bus. I retained that advnatage when Cheapernet was replaced by 10baseT using hubs (repeaters), but lost it when ethernet bridges and switches took over. Oh well, network paradise doesn't last.

I don't recall having problems with the connectors or the tees. It was the cable topology, reflections and losses, that were the problem.

BNCs are excellent connectors, and work fine to at least 3 GHz.

" snipped-for-privacy@fonz.dk"

production staff

and solder 42

invisible. And

board?

mentioned RM core

more compact.

surface-

delay line,

told you that

external fields?

drive it

fields

lines)

same

transmission

What's encouraging is that there's enough interest in Eclips Lite style glue logic that a third vendor has joined the game. I worry about that sort of logic going away, but it seems to be thriving.

The problem with cheapernet was collisions. Not "reflections and losses".

Until somebody yanks the cable - as in tripping over it.

That's nonsense. Ethernet relied on collision detection for traffic control; collisions were expected, and and the system handled them gracefully, by waiting for a randomly varied time before attempting to retransmit.

-- Bill Sloman, Nijmegen

But you aren't brave enough to post it, and risk revealing that it's actually pretty straightforward.

Not the adjectives usually used to criticise a rhetorical construction.

None of them outside my active vocabulary. I'm not sure that he used "quixotic" entirely correctly, but the implicit reference to tilting at windmills probably justifies him.

-- Bill Sloman, Nijmegen

Only if it was oversubscribed, and it was no different to 10-base-5 or hubbed 10-base-T in that respect.

What I remember is changing cable lengths, or passing through various PCs in a different order, and having things work or not work. Or connecting/disconnecting one network card and having consequences elsewhere. Collisions are resolved by the protocol, and only cost a little time. Electrical problems would lock up the whole system.

On a tapped transmission line, every tap makes a small reflection, and the reflections propagate to all the other taps and interefere. The complexity is impressive.

A good BNC is tougher than an RG45.