logic levels over long cable runs

It's called "word salad."

John

You're an idiot. No is not the answer.

In your case, it is commonly referred to as 'shit for brains'.

You're extra grumpy today. Rheumatism acting up?

Series termination allows a point-to-point link that preserves the full signal amplitude and speed, is insensitive to reflections, and in many cases saves a boatload of power. But naturally Real Engineers don't care about any of those things. (Reminds me of the old joke about the merchant who lost money on every sale, but made it up in volume.)

What exactly don't you like about it, and why?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

And, in your case, it is commonly referred to as 'My family tree has no branches!'

Can you prove that? If not, you're the idiot.

Even 1 Gb is pretty complicated.

full duplex on all 4 pairs at 125 megabaud

5 level signaling echo cancelation trellis coding

--
These are my opinions, not necessarily my employer's.  I hate spam.

Yes, I can. You are the idiot, but that was a known before you posted this tripe.

Series (what I call source) termination works beautifully. In real life. We do it on pc boards and with cables all the time. You can use a driver's internal impedance as part of the termination (say, a cmos output and an 86 ohm resistor to drive a 100 ohm trace) and magically get the full Vcc swing at the far end.

I think - I've seen it done, but I'd I'd have to simulate it to get the details - you can shunt the source termination with a mostly capacitive network to compensate for moderate skin losses in the cable, resulting in a clean, full-swing step at the load, without introducing much in the way of reflections.

JT is starting to dislike you, which shuts off his higher reasoning centers, such as remain.

John

And that's commonly referred to as "TheDimmerMan's scat fetish."

John

It doesn't matter how good your ancestors are. What matters is how good your descendants are.

John

I don't have rheumatism... I'm just mad as hell and I won't take it any more ;-)

It's the clipping I didn't like. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

      Remember: Once you go over the hill, you pick up speed

Dimmie fell from the top of his family tree, and hit his head on every branch on the way down. :(

--
You can't fix stupid. You can't even put a Band-Aid? on it, because it's
Teflon coated.

You can also use LVDS, there are a raft of chips available for this. They are lower power dissipation and a LOT more available than ECL. Probably faster, too.

Jon

You spout some of the dumbest "jokes" I have ever seen in my entire life.

The only thing that could possibly be amusing about them is that *you*

*think* they are amusing. That makes *you* the joke. You billboard you lack of maturity every time you post one. Generally, in your case, that is every time you post.

Well, the impedance of the series clipper will go from some low value to just a small nonlinear capacitance. That will cause a distorted reflection, which will disappear into the series termination resistor.

Dual diodes in SOT23 are pretty well matched, so the static offset would be small. If there's any significant dissipation in the diodes, it might cause a very small thermal tail, but I doubt it would be anything to worry about. The only question in my mind would be whether the improvement would be enough over the range of cable lengths anticipated. I don't know the answer to that, but I could find out easily enough.

Given months of lead time and an unlimited budget, it's often possible to do things better with a custom IC, or for that matter with a bunch of automatic tweaking in the FPGA. Application circuit design is a different skill, dealing with different constraints and possibilities (as you know perfectly well).

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

[with adaptive amplification, which is NOT time-preserving]

IMHO, the wire 'limit' at high frequencies for logic level (baseband) signalling is in the wire resistance, due to skin effect. So, one approach is to use a mass-produced cable/receiver assembly, with the receiver gain (say, 10 dB) high enough for attenuation in the cable length (which I'm thinking would be about 12 meters ==40 feet). Once the standard assembly is characterized, you just use the power supply and receiver to restore the slew rates at each stage, and cascade as many stages as you need to do the long line.

Optical cables go miles between repeaters; I'm thinking for copper, it'd be shorter than the 150 foot length under consideration here, for nanosecond timing and pulse shape integrity.

Here is a crazy idea...

Use 2 fibers. Send pulses that don't overlap down each fiber. That's to keep the AGC happy. When you want to send your signal, force both lasers on. The output signal is the AND of both receivers.

If the pulse you want to send is longer than the AGC time constant you will have to do something like send both for a short time, then switch to sending pulses that do overlap. The receiver switches to an OR. At the end of your signal, turn both lasers off.

You may be able to do something similer with copper. Use two pairs or coax, one for the start of the pulse and one for the end of the pulse.

--
These are my opinions, not necessarily my employer's.  I hate spam.

It looks like we may just do sixteen single-fiber links, each with a connectorized VCSEL laser, fiber cable, and a connectorized PIN diode. That's klunky but safe. The customer is risk-adverse, so reliable and klunky wins.

John