I am designing a troubleshooting lab for a team of networking technicians and I have a need to purposely make cables that will only negotiate at 10 mbps. Wiring 1-2 & 3-6 pins forces 100 Mbps as we all know; which is also what 10 Mbps uses. So its not a matter of pairs.

I've seen plenty of bad cables in the wild negotiating at 10 mbps and we usually re-tip or re-run as required. But I've never been sure what causes this and how to purposely induce this? My research is coming up empty on the physical circumstances that would force a 10 mbps negotiation.


Anybody know?

Thanks.

I would try putting two RJ45's on a 4 wire not twisted cable.

Or you can try mixing the wires of the pairs on a normal cable. (pins 1 and 2, TX+ and TX- should normally go to a pair, and as well 3 and 6, RX+ and RX- ,should go to a pair, if you exchange two wires so that they go through different pairs it should generate enough issues to downgrade speed)
The needed length to degrade enough the signal is to be found experimentally, I would expect something like 10 meters.

Wouldn't it be easier to find an old 10 Mbps switch?

Yes it would be much easier to plug into a 10 mbps switch; but for this particular lab the cable needs to be the bottleneck that the technicians need to discover during their troubleshooting. I'll play with it as you mentioned; I'll even try putting some resistors in there and see what happens.

I don't think that resistors would do anything effective.
What you need (in theory) is a "low pass filter".
This can be electronic but it needs a handful of components, so it will be visually detectable (a little box on the cable).
As well an alternative could be a ferrite bead (that would behave as a high frequency choke) but very easily detected.
And besides, I believe that both are not trivial to calculate/build/find.

a little box on the cable

Chokes and caps are bulky, but we aren’t building power supplies here. Properly engineered for the tiny currents involved, they might be surprisingly small. We live in a world where Thunderbolt cables look identical to USB-C cables despite one having a complicated IC embedded and the other not.

Whether anyone here is willing and able to do that task-specific engineering, and the OP able to do the micro-soldering work is a separate question.

ferrite bead

Cheap and easy to try. Good thought!

very easily detected.

Naah. We’re already talking about crappy “Cat 3” or telephone wiring hiding inside a Cat-5 jacket. There’s room for a small bead in there. They make them with ID sized for insulated 26 ga wire, and they can be staggered down the length of each conductor. Multiples, if need be.

Whether that’s enough to get the effect wanted, though….

not trivial to calculate

The equations for passive filters are nearly as trivial (and subtle!) as Ohm’s Law. (See equation (1) here.) Finding the right terms to put in is the tricky bit. The parasitic capacitance of the cable is measurable with a good cap meter, sensitive down into the pF range. The corner frequency you can look up from the 10Base-T specs.

Don't make it complicated, just miswiring the cable (not the correct color sequence, but of course the same on both ends) will do it.

miswiring the cable (not the correct color sequence, but of course the same on both ends) will do it.

I held off suggesting this because I have successfully run gigabit over "rainbow-wired" cables. (e.g. Orange, white-orange, green, green-white, blue, blue-white, brown, brown-white.) Purposeful misuse of the twisted pairs isn't guaranteed to get the effect the OP wants. It's worth a shot, but it might not work as hoped.

Moreover, the suggestion of using untwisted wire (e.g. telephone cable) inside the Cat-5 jacket gets the same effect, but with the same lack of guaranteed result.

Ethernet PHYs differ in how tolerant they are of this type of gaming.

The low-pass idea is at least based in solid engineering principles. Whether it can be done inside the volume of a Cat-5 cable jacket is an open question to me, but it can in principle work.

Don't make it complicated, just miswiring the cable (not the correct color sequence, but of course the same on both ends) will do it.

Yep, that is the original idea, with the advantage that it can be done using a normal Cat5e cable, with nothing added to it.

The question is more how long needs the miswired cable need to be, as said I would expect some 10 meters length needed at least, maybe I am too pessimistic and a shorter length of cable would do as well, but it is also possible that a much longer cable is needed (and that would look also suspicious).

Not the same technology we are using, but 100 Mbps are possible on barbed wire , so ...
https://www.sigcon.com/Pubs/edn/SoGoodBarbedWire.htm

@tangent
I don't think that such a miniaturized low pass filter is something that can be done at home, unless there is some available integrated chip, maybe *something like* this:
https://markimicrowave.com/products/bar ... datasheet/
would do, but really cannot say, and in any case it would probably cost an arm and a leg.

I don't think that such a miniaturized low pass filter is something that can be done at home

I would not reach for active filters or IC passives here unless you had to have something that worked every time, across all manufacturers and devices.

The thing is, @jaclaz's ferrite bead idea is a DIYable passive filter, one you can try at home. Moreover, it mimics actual problems you can see in the field, being too much inductance on the wire for the parasitic capacitance present. Kinking the wire and unkinking it creates a tiny parasitic inductance. It's a real effect your techs will deal with. All we're doing here is trying to figure out a way to get the effect reliably so the lesson doesn't fail due to clever electronics in the PHY compensating for the purposeful nonidealities in the hacked-up cable.

Simplified, equation (1) from my link above is L=X/(2πf). It gives you the inductance of the bead (or beads, in series) needed to get the effect you want. 2 and π are constants, and f we know from the 10Base-T specs, leaving the parasitic capacitance of the cable as our only mystery variable. I can't measure your cable for you, but I'd expect something on the order of 50-100pF per meter. (I could measure one here, but I'm lazy, and it wouldn't be your cable besides.)

According to Wikipedia, 10Base-T dates to the days where they used a simple MHz-per-Mbit encoding scheme, so f=10MHz. An online calculator I found gives X=160Ω, which means L needs to be somewhere around 2.5µH to get the desired corner frequency, presuming my 100pF-per-meter value is correct.

More inductance (more beads, more kinks and coils in the wire) pushes f down, so if one 2.5µH bead doesn't do it, try two in series, three, whatever it takes.

Also, don't neglect the "turns" through the beads, as that increases inductance. There are beads meant for passing the wire through multiple times. This adds bulk, but you might be able to stagger them down the length of the cable well enough without utterly wrecking the signal.

Degrading the cabling will never guarantee that the link will operate at 10Mbps. The endpoints transmit their capabilities in the regular fast link pulses and pick the best available speed and duplex option both are capable of, see https://en.wikipedia.org/wiki/Autonegotiation.

Depending on the ethernet driver software the outcome when the cabling does not support the negotiated speed the link may either just fail to operate due to packet corruption, or the driver may opt to advertise slower speeds and renegotiate when significant errors are detected.

If inserting a 10Mbps switch is not an option the only reliable method would be to construct something to remove the fast link pulse signalling, but still retain a single link integrity test pulse. This is likely to be more difficult than inserting a switch.

inserting a 10Mbps switch

Sneaky plan: buy a box of bulk cable, pull a few meters out from each end, terminate them, and pass the assembly off to your students as “a really long cable, portably packaged.” But inside, there are two cables joined by a PoE-powered CSS106 tucked into one corner. (Fallback to hEX based on parts availability.)

Ideally, what we want is a RouterOS version of the GPeR, but if it exists, I failed to find it. It’d be a great product regardless, useful anywhere you need to inject RouterOS logic in the middle of an existing cable run.

A simple mAP might be a cheaper alternative.
Can easily be battery powered too.

It has 2 Fast Ethernet ports and it runs ROS.
You can even be devious and change settings based on assignments/progress using Wifi.

use an old cable cat 3... ? and connect only 2 pairs

use an old cable cat 3... ? and connect only 2 pairs

Naah, CAT3 cable can do 100 Mb just fine in practice for relatively long stretches, like 30 meters or even more.
In any case only 2 pairs are used for both 10 and 100 connections, it's only 1000 that uses all four pairs.

Use Mikrotik racks https://mikrotik.com/product/rack_holder_sr_10u to set up test stands.
You can install RJ45 patch panels + FO panels and mix proper and bad wireing. Each stand could have different set of shortcuts, crosses, breaks etc. so each test team could actually face a different set of problems. Technicians would use standard cables as the trainers' "magical messing around with connections" would be done behinde the scenes
10U gives you a lot of place in a compact form to mount panel(s), switch(es) and other devices

A possibility is to split the pairs and run it over a few meters, it will give lots of crosstalk and poor performance, but
whether it is enough I don't know.

eg.

Pairs 1,4 2,3 5,8 6,7
or 1,3 2,4 5,7 6,8

(Cable terminated same at both ends)

• The cables will test as a simple straight cable (with a very basic tester)
• Will be pretty obviously miswired if you look at the terminations.
• Any tester that checks pairs will fail it immediately.

If the rack stand and patch panel is acceptable, it would be possible to conceal a Switchblox Nano (25x25 mm):
https://botblox.io/products/micro-ethernet-switch
which can now be programmed:
https://botblox.io/blog/botblox-aries-s ... -software/

If one is a real bastard, a tiny wi-fi uart module would allow to re-program the nano switch via wi-fi, even on-the-fly.