About link test pulse

That is correct; link test pulses are not frames, and have no significance beyond the link on which they are sent/received.

Normal link pulses are used by 10BASE-T devices. Fast Link Pulse bursts are used by devices that perform Auto-Negotiation, and are present only during Auto-Negotiation.

-- Rich Seifert Networks and Communications Consulting 21885 Bear Creek Way (408) 395-5700 Los Gatos, CA 95033 (408) 228-0803 FAX

Send replies to: usenet at richseifert dot com

Reply to
Rich Seifert
Loading thread data ...

Hi,

I try with my poor head to understand the integrity link test pulse.

(1) The link test pulse (e.g. for 10BaseT) is not repeated on the other ports of the hub when this one receives it. Is it correct?

(2) Is someone can explain me in a few words why there is sometimes NLP (normal link pulse) and why there is sometimes FLP (fast link pulse)? Is NLP still exists?

Regards, Michelot

Reply to
Michelot

Correct, although you should understand that, in 10BASE-T, the "TP_IDL" signal is transmitted as a lack of any signal altogether; "TP_IDL" is "silence" in 10BASE-T. If the transmit link is idle for 16 ms (nominal), then the device sends a link test pulse, just to tell the receiver that the link is still up and functioning, even if frames are not being sent.

The TP_IDL signal does not have any "shape"; a TP_IDL of 2.5 BT min simply means that the line must go quiet for at least 2.5 bit times.

No, auto-polarity (as it is sometimes called) is outside the standard, but commonly implemented in commercial transceivers because it is so useful.

-- Rich Seifert Networks and Communications Consulting 21885 Bear Creek Way (408) 395-5700 Los Gatos, CA 95033 (408) 228-0803 FAX

Send replies to: usenet at richseifert dot com

Reply to
Rich Seifert

Bonsoir Rich,

Thanks for your clear replies, I understand better. Can I ask you two more questions?

(1) If I understand correctly 802.3-2002, the TP_IDL signal is sended just after the last frame data bit (last FCS bit I suppose). And, 16 ms after TP_IDL, comes LTP pulse if TD is always idle. Is it correct? I note that the minimum TP_IDL signal duration is greater (2.5 BT min.) than the one of LTP (0.85 BT min.), but perhaps, in real circuits they have the same shape.

(2) Is the automatic detection of the cable twisted pair connection (straight or crossover) is also an effect of the autonegotiation? I don't see that in 802.3 but perhaps, I haven't read correctly.

Regards, Michelot

Reply to
Michelot

(snip)

And what does a 10baseT device built before auto-negotiation do with a fast link pulse?

I once had some Asante (for Macs) NICs that would blink the link light when connected to a 10/100 hub. The book didn't say anything about a blinking link light (maybe 1Hz or so). It just didn't work on that hub, but worked fine on ordinary 10baseT hubs.

-- glen

Reply to
glen herrmannsfeldt

Correct. The transceiver ensures that the last transition is positive-going (either because the data worked out that way, or by forcing it that way), and then lets the signal naturally descend to zero. The waveshape of the return-to-zero is a function of the cable impedance, the transformer magnetics, etc.

You can detect the polarity by looking at the link test pulse; it should be unipolar and positive (seen differentially). If it shows up as a negative differential, just invert the output of the receiver.

-- Rich Seifert Networks and Communications Consulting 21885 Bear Creek Way (408) 395-5700 Los Gatos, CA 95033 (408) 228-0803 FAX

Send replies to: usenet at richseifert dot com

Reply to
Rich Seifert

Bonjour Glen,

Probably Rich will add some good comments to our remarks.

Normally, NLP is compatible with FLP: FLP can act the role of NLP.

You're talking a 10baseT NIC without auto-negotiation connected to a 10/100 hub. To detect correctly the 10baseT, I think the hub has to treat the parallel detection mecanism. I heard that sometimes, there are some bugs with that.

You can find an exemple in the reverse situation (10/100 NIC face to 10 Hub) here (dated on 1996)

formatting link

Minimum frequency should be 62 Hz (1000/16) and the light seems stable for our eyes.

Difficulties with the parallel detection?

Regards, Michelot

Reply to
Michelot

Bonjour Rich,

I understood that it was a real signal added, confirmed by the case when the last cell was CD0.

"If the last bit transmitted was a CD0, the PLS will generate an additional transition at the bit cell boundary following the CD0. After the zero crossing of the last transition, the differential voltage shall remain within the shaded area of Figure 14-10".

Are you saying in your reply that after the last positive transition (normal if CD1, or added if CD0 is the last bit), the signal returns freely to the quiet (conforming to the standard template)? So, I can understand now the term "start" in the expression "transmitter waveform for start of TP_IDL".

Do you know how is the mecanism? I suppose the polarity is detected before any frame is sent.

Regards, Michelot

Reply to
Michelot

Bonsoir Rich,

Thanks for that talking which helps me.

(1) I should be curious to know why 802.3 uses sometimes the term repeater instead of hub. Is there a difference of meaning?

(2) The NLP pulses (or FLP) don't carry the synchronization. Can we say that these test pulses are asynchronous of the frame bit timing? Perhaps, 2 consecutive frames issue from the same end TD could be even asynchronous. Is it correct?

(3) When the AUI isn't separated from the DTE, can we still talk about the CO (control out) et CI (control in) signals? I'm wondering now if the test pulses come from DO or CO!

Regards, Michelot

Reply to
Michelot

Bonsoir Rich,

Thanks for that talking which helps me.

(1) I should be curious to know why 802.3 uses sometimes the term repeater instead of hub. Is there a difference of meaning?

(2) The NLP pulses (or FLP) don't carry the synchronization. Can we say that these test pulses are asynchronous of the frame bit timing? Perhaps, 2 consecutive frames issue from the same end TD could be even asynchronous. Is it correct?

(3) When the AUI isn't separated from the DTE, can we still talk about the CO (control out) et CI (control in) signals? I'm wondering now if the test pulses come from DO or CO!

Regards, Michelot

Reply to
Michelot

(snip)

I would expect it to always use repeater.

In the beginning there was coaxial ethernet. Multiple stations could tap into a long coaxial cable. The electronics for building a repeater was expensive, so they were only used when needed. The cable could be 500m long, so a fairly large net could be built without any repeater.

Coaxial ethernet is inconvenient in that the cable can't have any branches in it. AUI cables up to 50m long reach from the host to the nearest tap on the cable.

As electronics got cheaper, twisted pair ethernet began. This allowed the more convenient hub and spoke topology, where a hub in a wiring closet is connected to each host, and a separate cable is needed for each host. Electrically such hubs were ethernet repeaters, as the term had been used since the coaxial ethernet days, but, as all twisted pair hubs were repeaters, the term hub became commonly used.

As electronics became even cheaper, the ethernet bridge, as previously defined in the ethernet standards, became affordable. To emphasize the fact that the signal was not sent out every port, marketing departments started to use the term switch, related to the telephone system device which connects individual phone subscribers to the phone network.

So, technically there are repeating hubs and bridging, or switching, hubs. The word hub alone is often taken to mean repeating hub, and the word switch to mean switching hub. Hub is the topology, repeater is the function.

Does that help any?

-- glen

Reply to
glen herrmannsfeldt

In marketing-speak generally if it says "hub" on the label one may assume a repeater and if it says "switch" one may assume a bridge, but this has become complicated as multiport bridge ICs have become cheaper, the ability to support multiple speeds has become necessary, and as routing features have been added to switching hubs.

For the rest of this when I use "hub" and "switch" assume I mean a device labelled as such, regardless of its actual function.

The result is that multiple-speed "hubs" are often several multiport repeaters connected by two-port bridges to allow multiple speeds to be handled, sometimes inexpensive "hubs" are implemented as bridges because it was cheaper to implement using bridge chips (a 5-port gigabit bridge which also supports 100baseTX and 10baseT is a single not terribly expensive chip for example, and in a few more years they'll be 50 cent parts) than as a conventional repeater, and a "switch" may in fine print say "layer 3" which makes it a router. Unfortunately spec sheets these days seldom contain enough detail for one to be able to determine how the particular device is actually implemented--you have to take one apart and refer to chip vendors' data sheets in many cases, _if_ the identifying information hasn't been sanded off of the chip or covered with a blob of epoxy.

I suggested a while back that the Committee look into finding some means of beating the marketroids into submission with regard to nomenclature but I think I worded my suggestion poorly as it was taken as criticism of the terminology used in the spec and not a request to recognize that the marketroids are creating chaos by making up their own nomenclature and by redefining standard nomeclature to suit themselves.

Reply to
J. Clarke

Bonjour Glen,

Yes, it's wonderful! As I always say, I have some difficulties talking about telecom technicals with my dear wife, or with my neighbours, on Sundays during aperitif! Thanks for that interesting synthesis.

We always talk about AUI in 10Base-T, and MII or GMII. Are these interfaces really used? Perhaps it is used in specific applications as banking terminal... But, nowadays, I don't see the advantages.

Yes, it's commonly used by the users, not by the Standard. In the IEEE, I note that the term hub is used 2 times in the sections relative to twisted pair or optical 10 Mbit/s and, on the contrary, the term repeater is used more than 300 times.

Great! I just trying to look for a generic word for naming hubs (in the sense of repeatind hubs) and bridges (either a full equipment, either embeded in routers, domestic modems, industrial modems as LMDS CPE, stations...). So, as you say, we can define the generic hub word which can be qualified of repeating or switching feature. In France, hubs is opposed to switches (or bridges), but I can all the same to be clear on a generic word for gathering both device concepts.

Regards, Michelot

Reply to
Michelot

As others have noted, "hub" carries two meanings:

(1) Marketeers/end-users generally use "hub" as a synonym for "repeater".

(2) The standards try to avoid the use of the word "hub" altogether, although if pushed, most network architects would define hub as a device at the center of a star-wired topology. Thus, a hub could be a repeater, a bridge/switch, or a router. The distinction is in the higher-layer functionality; the commonality is in the physical topology.

NLPs are completely asynchronous with respect to any data frames sent over the same pairs. FLPs are also asynchronous with respect to data frames, but since FLPs are NOT interspersed between data frames, it is strictly correct only to speak of the timing relationship between the FLP burst and the *first* data frame. In any case, there is no strict timing relationship; the two signals are asynchronous.

It is possible that two 10BASE-T frames emitted from the same station will have a precise timing relation between them, synchronized to the same clock. However, the receiver cannot depend on this relationship, and resynchronizes its receive clock independently for each received frame. The reason is that it is possible (on shared media systems such as coaxial cable) that sequentially received frames were sent by different transmitting stations, each with its own independent clock. Since the receiver must deal with this (worst-case) situation anyway, there is nothing to be gained by trying to take advantage of some potential synchronization between received frames in the case of dedicated media systems.

In 100BASE-T, the line signaling is synchronous; there is an active-idle signal on the line at all times, maintaining receiver clock synchronization even in the absence of data frames. This is, in part, why 100BASE-T cannot support shared-media wiring systems.

Yes, we can still speak of the signals, although they may be present only in a chip-to-chip connection, or even inside a single chip. Also, virtually no devices even implemented the CO (Control Out) signals, regardless of whether the AUI was exposed or not. (CI--Control In is just another name for the Collision-Presence pair of the AUI.)

Both NLPs and FLP bursts are sent on the DO (Data Out) pair; they emanate from the device regardless of whether there is an exposed AUI/MII or not.

-- Rich Seifert Networks and Communications Consulting 21885 Bear Creek Way (408) 395-5700 Los Gatos, CA 95033 (408) 228-0803 FAX

Send replies to: usenet at richseifert dot com

Reply to
Rich Seifert

Historically, there were advantages, such as being able to build a workstation with an Ethernet interface without having to commit to a coaxial cable port vs. a twisted pair port vs. a fiber port. This approach was taken both in 10 Mb/s systems, and in early 100 Mb/s systems. (Sun workstations used to have an MII port, and used an external transceiver.)

Today, the interface is more useful as an abstraction for system designers. The interface may exist physically as well, for chip-to-chip connections (e.g., when connecting to discrete 1000BASE-T transceiver chips).

-- Rich Seifert Networks and Communications Consulting 21885 Bear Creek Way (408) 395-5700 Los Gatos, CA 95033 (408) 228-0803 FAX

Send replies to: usenet at richseifert dot com

Reply to
Rich Seifert

(snip)

I hadn't thought of this one before...

100BASE-T media can't support shared media wiring, but if one wanted to implement a 100BASE-?? at the MII interface, presumably whatever conversion needed could be done in the transceiver.

I haven't looked at the MII signals for a long time, but I suppose it would be synchronous all the way through the MII. Some part has to supply the clock and the other has to synchronize to that clock, I will guess that transmitters supply it and receivers synchronize, then the receiver should be able to synchronize to the incoming clock on shared media, presumably with a PLL.

It seems that we discussed here before that 100BASE-?? still have a preamble, even though it isn't needed for synchronous signaling. Is that preamble long enough to lock a PLL on the incoming bit stream?

It doesn't seem likely that a coaxial 100 Mbit ethernet will ever be created. I might wonder how well gigabit or

10 gigabit would do on coax, though. For wiring closet to end stations, would it be cheaper than fiber?

-- glen

Reply to
glen herrmannsfeldt

Bonjour Rich,

It's really interesting to read that. Thanks very much. Michelot

Reply to
Michelot

In theory, yes.

Actually, some of the control signals (e.g., Collision Detect) are defined as completely asynchronous, but you can always run them through a dual-rank synchronizer to align the edges.

Correct; the receive data stream is synchronous to a PLL-derived Receive Clock.

It should be long enough for that purpose, yes.

As a shared medium, I agree. The problem isn't so much the real-time synchronization of the receive clock as it is the sensitivity to reflections caused by tap capacitances. This was a serious enough problem at 10 Mb/s; at 1000 Mb/s it may be intractable (or at least, not worth the effort to try).

Perhaps, although not nearly as cheap as UTP, which currently supports gigabit (and perhaps 10 gigabit in the relatively-near future).

-- Rich Seifert Networks and Communications Consulting 21885 Bear Creek Way (408) 395-5700 Los Gatos, CA 95033 (408) 228-0803 FAX

Send replies to: usenet at richseifert dot com

Reply to
Rich Seifert

So which "manufacturers and private companies, the telecom professionals", made up and worded the "Pre 802.11N" standard with which certain equipment currently on the shelf at CompUSA purports to be compliant?

Yes, the marketroids are employees of some company somewhere. Being an employee of a company somewhere does not make one responsible or sensible or smart or even sane.

Every company that puts boxes on the shelves in stores is not run by "telecom professionals" or professionals at anything else except money-grubbing. The simple fact is that you can buy a chip from Intel or SiS or AMD or Marvell or whoever and stick it in a circuit board and sell it as a hub or router or switch or gonkalator armature or whatever without knowing much more than how to implement the reference circuit described in the datasheet. And the claims that are made by whoever builds and sells the box often bear little resemblance to the claims made by the true "telecom professionals" who designed the chip.

I don't know why you're defending these twits. The folks who write the standard are not the problem, the ones who act as if there is no standard or the standard doesn't mean anything or the standard doesn't apply to _them_ are the problem.

Reply to
J. Clarke

Bonsoir John,

Thanks for your advice.

Standards are often made up and worded by manufacturers and private companies, the telecom professionals. And the marketoids you're talking are members of these companies, or from their distributing system.

Regards, Michelot

Reply to
Michelot

Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.