resetting a filter

Oh that's cheating.. :)

Jamie

But it kept people employed, mainly the IT guys.. :)

Jamie

So much for diplomacy in engineering.

Nope. Gigabit ethernet is 4 parallel signals, at 125Mbits/sec, with 2 bits per baud. 4 * 125 * 2 = 1000 Mbits/sec. The 4 signals are sent simultaneously on the 4 pairs of wires in the CAT6 bundle. If they arrive not so simultaneously, there will be problems with errors.

Doing the math, 125Mbits/sec = 8nsec per baud. Without de-skewing, if datta arrives on one pair 8nsec late, you get garbage. Looking at the specs:

CAT5, 5E, and CAT6 are speced at 50ns at 100 meters maximum skew, so obviously, the transceiver is doing something to reduce the sensitivity to data skew.

Ethernet is CSMA/CD (collision detection). Wi-fi is CSMA/CA (collision avoidance) with 802.3 ethernet packets encapsulated inside

802.11 wireless packets. Your description is how CSMA/CD works. CSMA/CA works by having the transmitter announce its intention to send some data before actually doing so.

--
Jeff Liebermann     jeffl@cruzio.com
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

OK, now you want me to add differential drivers and receivers to the

42-ga twisted-pair delay line. All to replace four 0603 parts.

I haven't done necl in ages. All our fast stuff runs in pecl mode. Works fine. It's better to not drop chunks of metal onto live circuits.

PCI Express is LVDS. Gen1 runs 2.5 gbps per lane with roughly 80 ps edges. Gens 2 and 3 are faster. We use a lot of LVDS in and out of FPGAs. Great stuff.

I still haven't figured out what sets the LVDS common-mode voltage. The cartoons in data sheets and appnotes show dueling current sources. Anybody know?

--

John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

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

It likely deserializes each pair, FIFOs the data, and merges the FIFO outputs. PCI Express does that explicitly. PCIe allow 20 ns skew, which is 50 bits worth at Gen1 speed.

--

John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

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

Great idea. All I need to do is double the number of overpriced CAT6 connectors in an installation plus a double female CAT6 connector at mid span. I get to stock crossover cables in a multitude of common lengths. (Right now, I use crossover adapters and standard EIA568B wired patch cables). When I use a crossover cable, I'll need to remember which end to plug into the cable and the device. Of course, nobody is going to insert it in the middle of a cable run. More likely, there will be a mysterious extra coil of wire on the floor with strange connector wiring. My wiring nightmare come true.

Instead, how about just a reversing adapter to go at mid span and another at the equipment end? In effect, it would convert 1/2 the cable run into a crossover cable. Much easier to install (and explain).

Patent search?

--
Jeff Liebermann     jeffl@cruzio.com
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

Most Ethernet gear these days doesn't care if the pairs cross or not.

--

John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

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

Right. At my PPoE they had a product in the field for over a year before we figured out that the wires on the PCB were crossed. Someone eventually found a switch that cared. It took a while to replicate the problem. ;-)

No need to. We stock about twenty different configurations. Across about five different brands.

Work with them on a daily basis. At 10GbE speeds.

We also use them for serial control.

So does Cisco.

Good point. That solves one item on my complaint list. I still think that standard cables and adapters is the way to do it, especially since crossover adapters already exists:

Unfortunately, I can't tell if this one is wired for 1000baseT (all 4 pairs crossover), or 1000baseTX (2 pairs crossover and 2 pairs straight through). Checking around, I find that most are 1000BaseTX, which will NOT do the trick. Caveat Emptor.

1000base-T has been somewhat of a problem with some of the hardware that plagues my life. I've seen auto-MDIX (auto crossover) fail a few times. Also NWAY negotiation failure, which is another story. It doesn't happen often, but when it does, it requires an equipment power cycle. If it's a managed switch, I can usually reset it remotely. If it's a cheap junk switch, and the box is in a server farm or corporate hindquarter, I jump in the car, drive 50-100 miles, argue with the security people, power cycle the box, and drive back home. For some unknown reason, this only happens at 2AM, on weekends, during vacations, etc. So, in self defense, I lock down the interface speed and protocol on most boxes, and use a crossover cable as if auto-MDIX didn't exist. No such problems for many years.

--
Jeff Liebermann     jeffl@cruzio.com
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

ds

n t

so

e

es

s

That's a fairly spectacular variation from the original concept.

suggests that for at least 1000BASE-T, this is what's going on, but it does negate the basic idea underlying all the original local area networks.

Would you care to specify the relevant level in terms of some comprehensive layered model? The original ISO seven-layer model never looked all that convincing so I imagine it has been superseded.

-- Bill Sloman, Nijmegen

s o s

You were good at what you were interested in, and interested in enough of what you needed to know that you could get away with ignoring the stuff that you didn't find interesting. From the persistent holes in your general knowledge that you are still exhibiting here, you seem to have ignored quite a bit of stuff.

How can you be sure?

A bit too easy, it would seem.

Sure. But it was a strictly serial system, so skew between cables didn't matter. Dispersion could shrink the eye pattern until you couldn't reliably distinguish between 0 and 1 (intersymbol interference) but differential timing was a non-issue.

Even today it only seems to be an issue for 1000BASE-T, which is a decidedly freakish variation.

-- Bill Sloman, Nijmegen

In the late 80s ethernet had collision detection, when a collision was seen all nodes would echo the jamming signal, and the conflicting senders would detect this and wait for a (hopefully different) random time before retransmitting these days that feature isn't exercised much on wired networks unless the switch you're connected to is overloaded.

--
?? 100% natural

--- Posted via news://freenews.netfront.net/ - Complaints to news@netfront.net

Oh, man, 10Base2 coax ethernet, you had to remind me. What a nightmare. All those BNC tee connectors--one goes flaky, the whole subnet goes down. Back in the day, token ring was a _huge_ improvement. (Of course even 10Base2 was a lot better than 10Base5 thickwire.)

Then the Ethernet guys jacked up the name and slid a new network technology under it, and life got a lot better.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

That's why each signal (on each pair) has an embedded clock. so threy can be re-aligned.

--
?? 100% natural

--- Posted via news://freenews.netfront.net/ - Complaints to news@netfront.net

Well, maybe he hid a time machine somewhere in one of those cabinets.

--

John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

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

Make up your mind, Johnny. You had me cleaning toilets.

Pretty much makes anything you say about me a lie, and pretty much of anything you say about anything else as well.

"Word salad"? Explain your failure to understand vapor phase cleaning processes again. You seem to think full immersion is part of the deal.

You lose. again.

You guys never conceived of the idea that they might "handshake" when attached or "powered up" and all gets determined at that point?

Get a clue, guys. It ain't a modem, but there is certainly a handshake, or it would not be stepping down to 10Mb/s when you put a low grade patch cable in the path.

Wrong again. I never said that. I think you're a tech who doesn't really understand electronics very well.

--

John Larkin Highland Technology Inc

jlarkin at highlandtechnology dot com

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