Oops

..

There are exceptions to every generalisation, but I've still got fond memories of the time one of the other graduate students wanting to dismantle a vibrating reed electrometer head that I was looking after, because he wanted to take out the 1 gigohm resistor and use it to measure a low current by putting a multimeter across it.

I didn't know all that much about electronics back then, but I did know enough to be able to tell him that his multimeter movement had a resistance of around 20k ...

-- Bill Sloman, Nijmegen

Actually, modern electronics was invented (and a war won) by the folks at the MIT RadLab, and they were mostly physicists. That's embarassing, so don't tell anyone.

I've published a comment in Rev. Sci. Instrum. where even the abstract reads like that

Some physicists can do electronics superbly, but Rev. Sci. Instrum. does publish quite a bit of electronics designed by physicists who are considerably less competent.

-- Bill Sloman, Nijmegen

Yes. They love diffamps and current mirrors, whether they make sense or not, and their favorite single component is the trimpot.

This was designed by chemists:

It's a preamp for an IonSpec Fourier Transform Mass Spectrometer. Those are JFET opamps. It runs in UHV so the only cooling is radiation. They run at 130C or so, so the bias current is huge, so they needed 1M resistors to ground, so the Johnson noise was huge. They threw away about 30 dB of s/n. Bruker, their competitor, only threw away about 20.

Note the sophisticated layout: holes drilled in ceramic, soldered wires. This was a *production* unit.

Hmm..as i remember, coax i used ran in the 80-90% C, a lot faster than 60-70% cited for fiber. Signal loss and edge loss over a long line is another story, natch.

Foam dielectric coax can get up there, but it has a lot of drawbacks, e.g. when you bend it, the centre conductor tends to cut through the foam and cause all sorts of nasty impedance discontinuities, and in really bad cases, may even short to the shield.

Cheers

Phil Hobbs

The trimpot on a cable, please. We do have some standards.

Cheers

Phil Hobbs

The GoreTex "high-speed" coax we used in the '80s was pretty decent stuff, albeit expensive as hell.

or...

last

be

nds

er.

es

on

or,

ch.

And don't get it warm. The cable TV connection at our previous house ran across the cellar ceiling - the previous owner had been inclined to improvise - and when we first moved in the plumbers who tidied up the improvisations in the central heating got a bit of the cable warm enough that the inner ended up in contact with the shield, even though outside of the cable looked fine - polyethylene softens at a lower temperature than polyvinyl chloride.

I had to borrow a time-domain reflectometer from work to find the short (which hadn't actually killed the cable signal everywhere) and ended up chopping out about a foot of cable and splicing in new with cheap (video) coax connectors, leaving the system even more improvised than it had been before.

-- Bill Sloman, Nijmegen

ror...

last

ke

be

onds

ter.

kes

s on

tor,

x?

er

er

tch.

Gore do seem to specialise in exploiting the virtues of fluorocarbon- based plastics, though that's not all they do.

-- Bill Sloman, Nijmegen

The particular disaster I was commenting on was free from any of that, being purely digital, but did get enthusiastic about the virtues of MC10K ECL at a time when ECLinPS had already been commercially available for a few years. There was a particularly irritating passage where the authors made a fuss about 10K being four times faster than TTL, when ECLinPS was in turn roughly four times faster than 10K.

-- Bill Sloman, Nijmegen

Their best product is Gore-Tex

The gates inside a good $15 FPGA are another four times faster than EL.

There is some really fast logic around, Gigacomm at roughly $40 per gate, and the Hittite stuff for even more, but it doesn't make sense to do any serious amount of logic design with that stuff.

You mean to imply that a trimpot on a coax cable is a good substitute for a trombone assembly?

ent

ists

Obviously, though you've still got to drive transmission lines at the outputs, and FPGA's tend to get unhappy about driving 50R loads.

At the time I wrote my comment - 1996 - FPGA's weren't quite as quick. The late Peter Alfke was famous back then for using a Xilinx FPGA to implement a 1GHz divider (and it took a lot of clever fudging).

-- Bill Sloman, Nijmegen

No. John was teasing me for being a physicist who presumes to do circuit design occasionally, so I'm just entering into the spirit of the thing.

Cheers

Phil Hobbs

The fast way in/out of an FPGA is differential, LVDS or PECL. We just did a controller that stole the SERDES blocks to make 32 fast laser driver differential logic outputs, clocked at 1 GHz.

Many FPGAs can be programmed to source terminate a single-ended 50 ohm trace, with sub-ns edges. ECL is becoming a niche.

I don't find that scenario plausible at all. Having been on the scene = for the installation of over 300,000 feet (cable of many configurations) of fiber cables, jumpers, etc., Having 40 feet coiled up handily to cut and re-terminate does not normally happen.

The dirty connector story doesn't work at all either.

A more likely story is that validation measurements (OTDR) showed a 60 ns discrepancy (40 foot) in the total cable length.

?-)