RJ-45 / RJ-11 / Premade Cables

I see my idea of switching to RJ-45 was met with a bit of resistance. I still think its the right way to go, but there are definitely some ways we could do things:

the main reason i see for this is the ability to use pre-made cables that will save time, energy, and frustration. there will be no more posts on the forums like "check to make sure your cables were made right". we can effectively eliminate one more thing that humans have to do, and are error prone at doing.

here are the options:

1. switch to straight RJ-45.

mouser header: $0.57 / ea
[www.mouser.com]
cable cost: $1.00 / ea from the RRRF.
[www.cat5ecableguy.com]

2. switch to straight RJ-11.

mouser header: $0.57 / ea
[www.mouser.com]
cable cost: $0.50 /ea from the RRRF.
[www.deepsurplus.com]

3. put spots on the PCB for RJ-45 (or 11, whichever we decide) and a .100" header. the advanced user can use the .100 if they really want to.

4. continue to use .100" headers

mouser header: $0.29
mouser connector: $0.22
wire (x3 feet): ???
crimp-on connectors x 2: $ 0.14
time to make cables
(part #'s in BOM)

^^^ note that once you add that up, making your own cables is definitely not as cheap as using premade ones. if they are, we're talking like a couple bucks. ^^^

now that i've looked up those parts, i'd say that RJ11 is looking like the real winner here. i really, really dont think we have to worry about someone plugging their RepRap machine into a phone jack. We have to assume at least *some* level of intelligence here.

I cant be the only one who thinks this is a good idea here...

As I said in the other thread, I agree with you that the RJ11 is the best option. And I wouldn't be worried about someone plugging the RepRap into a phone jack (although troubleshooting that issue would be quite entertaining!) as much as a built-in PC modem - all dem durn plugs in the back look the same anyway...

Kyle

Adrian Bowyer Wrote: (quoted from [forums.reprap.org], moved here for pertinence)
-------------------------------------------------------
> > * switch to RJ-45 plugs for Rx/Tx
>
> Why? Much more complicated, and no advantage that
> I can see. Remember that _none_ of the signals go
> to and from the same place; they go in a ring
> (i.e. just one wire between each board and the
> next). And, of course, it only takes ten seconds
> to solder the end of a single wire to make a fully
> insulated pin socket for an 2.54mm pin using a
> shell socket and a bit of heat shrink.

I don't see this change as making things more complicated, but less so, from both a building and a support standpoint. RJ11 (or 45) cables are a standard solution that works.

It may take only 10 seconds to solder a wire for a whiz who's done it a hundred times before, but the people involved in this project have a wide range of skillsets and thus very variable experience levels with soldering/electronics. For many, assembling the RepRap boards may be the first soldering work they've ever done. Making all aspects of the project as easy as possible should be a priority IMO.

Kyle

If you do go with RJ-11/12, keep in mind that the "standard" phone cable swaps conductors (1234 -> 4321) when designing the PCB. I've also come across straight-through cables, but I don't have a clue where they came from, hopefully they're not common. Most phones work fine either way so I haven't paid much attention.

Yes come to think of it I have had problems with RJ-11 model cables. You would think they were interchangeable between different modems but they are not.

The biggest problem I see is being able to get standard cables the right length. If you look at Darwin the boards are quite close together. I have never seen RJ-11 cables less than a couple of metres. You can make them up yourself but you need a special tool and is a lot easier to get it wrong than a simple crimp is,

Also I still can't see the logic of somebody having no trouble assembling a PCB with dozens of soldered joints having trouble making a 1 wire link. Perhaps we should remove the ground and connect both pins together. That way you can't wire it up wrong other than plugging it in the wrong place and you can do that with RJ-11 anyway.

---

[www.hydraraptor.blogspot.com]

> Also I still can't see the logic of somebody having no trouble assembling a PCB
> with dozens of soldered joints having trouble making a 1 wire link.

I agree; unless there is some strange design flaw we do not yet know about, which didn't bite Adrian, Vik, Simon, Zach, or me, but did bite several others recently (seems unlikely) this looks to me more and more like a documentation issue of some kind, not a directly technical design flaw that requires changing our design to use different interconnect wiring.

It would (IMO) help us get to the bottom of the underlying issue if those experiencing it continue to post daily updates about what they are trying and what happens. Some threads seem to be turning into discussion and speculation among those trying to help, instead of dialogue with those asking for help. While that discussion may be useful in the longer term, we need to keep on hearing from those with the boards/wiring that are apparently not functioning correctly.

Jonathan

Yes it would useful if everybody that has had problems give a quick post recapping what problems they have had, or still have, regarding UCB PCOM interconnect and testing. Also if people who had no issues could also post we get an idea of the scale of the problems.

We should also be able to put together a debugging flowchart that differentiates each of the problem people have encountered so far. With such a simple system there can only be a small number of possible faults even with a large number of people building it.

---

[www.hydraraptor.blogspot.com]

Hi

I am completely new to this but have built a comms board and 1 universal controller.

I had a problem but it was only because the build documentation seemed ambiguous.

My problem was the reference to connectors being the wrong way round. It was not clear to me that this only referred to the mark one board. I see this confused some one else as well.

I would suggest a few edits in the wiki stating which board revision any text referred to would clear this up and ideally a picture of the latest board revision would help. With regard to the cables a couple of pictures clearly showing the 2 headers and the tab position with coloured wires would solve the wiring problem. I think changing to new connectors is only going to add to any confusion as you will now have three styles of board to deal with plus the possibility of straight through and cross over cables.

---

Regards

Ian

i guess i'm not going to be able to convince you guys. personally, i hate the crimp-on connectors. i think they are a waste of time. we say that we want this project to take off, and be as easy as possible for someone to use, but then completely shoot down my suggestion to eliminate a large problem with our current wiring problems. its really quite frustrating, and i can definitely see why forrest decided to go off on his own direction. mind, i'm not going to do that, but it is definitely a temptation.

heres what i'm going to do: i'm going to place both a RJ11 jack, as well as the .100 headers on the v1.3 circuit board. if you are a masochist and you want to micromanage every single facet of your reprap machine, knock yourself out. personally, i would much rather plug a known-good cable into the machine and be done with it. if you dont like that, modify the boards and get them manufactured yourself

Good on you for sticking with it. I hate making cables and this will definitely save some time and frustration. I'm not too enthusiastic about registered jacks, but I don't have any better ideas.

I really doubt that many of the new user problems are poorly-made cables. I have a feeling that when a lot of people say "maybe there's something wrong with the cables I made" they're just looking for a polite way to say "your @!#% hardware doesn't work". Still, if we can make assembly substantially easier just by switching connectors, I'd say it's worth doing.

Is there any way we can make this more palatable by putting the sync lines on the same cable? If we use four-conductor cable, one is ground, one is data, and two are unused. Assign those two to the sync connectors. Connect the Sync 1 pin in the TX jack to the Sync 1 pin of the RX jack on each of the boards. Do the same with Sync 2, so all boards see the same signals on these pins. Stepper boards would route those pins to the PIC for use as sync signals. Extruder boards (and any others) would have to disconnect them from the PIC, because those pins are also used for heater/thermistor. The only way I see to do this would be to require two shorting links on each stepper board, which is pretty ugly. Aside from that, it doesn't sound like that bad an idea.

Well it makes a bit more sense using a 4 pin connector to handle three signals but how do you buy them at the right length? The link Zach gave is 7m long. Making them yourself is no longer an easy option because you need an expensive tool and it is quite tricky to use.

I made a video of how to put on a crimp and it only took me 1 minute 40 seconds even while taking time to explain how to do it.

[www.youtube.com]

---

[www.hydraraptor.blogspot.com]

Cables are the main failure point of any system. I've used RJ 45 comms cables before. Yeah, they work, but it's really annoying if you manage to cold work one wire inside. It doesn't bother me because I have a spool of 1000' of wire and a crimper. Most don't. Buying spare cables works, but then you have to keep track of where you put them.

A somewhat more expensive option that I favor is going the terminal block route. We use these extensively for analog and digital signals in significant excess of that the Reprap operates at. My suggestion is to look at Phoenix Contact's Combicon line. You can get screw terminals, spring terminals, and removable terminals (thus avoiding the permanancy issues of RJ45 AND the crimp issues of the current solution without adding in the problem of mixing up your wires).

So, let's go with a 4 wire system (2 sync, 2 serial) as an example.

Mouser pricing, single quantitiy, removable socket, 2.5MM spacing
MCV 0.5/4-G-2.5 ($1.15, 4 pole 4A socket )
FK-MC 0.5/4-ST-2.5 ($3.86, 4 pole 4A spring terminal block)

Mouser pricing, single quantitiy, removable socket, 3.5MM spacing "economy" model
PT 1.5/4-PVH-3.5 ($0.59, 4 pole 10A header, screw terminal)
FK-MPT 1.5/4-ST-3.5 ($1.15, 4 pole 10A header, spring terminal, alternate part)
PST 1.0/4-3.5 ($0.23, 4 pole 3.5mm header)


Mouser pricing, single quantitiy, fixed screw terminal socket
MPT 0.5/4-2.54($2.09, 2.54mm spacing, 4 spot)


Obviously the most expensive option if you're trying to keep to 2.5/2.54mm spacing, but you never have to solder nor do you lose the terminal block if you need to rewire. 3.5mm and 3.81mm are significantly more common than 2.54mm in this format. This would also be compatible (electrically) with V 1.2 boards. Some variants (they have a 1 position modular spring terminal block style, too) may be able to be retrofit back into the V1.2 boards to provide similar solderless/crimpless capability.

nophead Wrote:
-------------------------------------------------------
> Well it makes a bit more sense using a 4 pin
> connector to handle three signals but how do you
> buy them at the right length? The link Zach gave
> is 7m long. Making them yourself is no longer an
> easy option because you need an expensive tool and
> it is quite tricky to use.

true, there are a few easy solutions:

1. use the long length. a couple zip ties and you'll be fine.

or

2. cut the excess length from the middle, splice each color back to itself, then tape back up.

> I made a video of how to put on a crimp and it
> only took me 1 minute 40 seconds even while
> taking time to explain how to do it.
>

i understand that its possible, and also pretty easy to make these cables... but the whole point is that it wastes time, introduces potential for error, and distracts from the main goal here: making a self replictating 3D printer.

thanks for the video tho =)

SOI Sentinel Wrote:
-------------------------------------------------------
> Cables are the main failure point of any system.
> I've used RJ 45 comms cables before. Yeah, they
> work, but it's really annoying if you manage to
> cold work one wire inside. It doesn't bother me
> because I have a spool of 1000' of wire and a
> crimper. Most don't. Buying spare cables works,
> but then you have to keep track of where you put
> them.

true, but its very rare to have a pre-made cable go bad on you, and at these prices its not too bad. plus, RJ-11 phone cables are EVERYWHERE.

>
> A somewhat more expensive option that I favor is
> going the terminal block route. We use these
> extensively for analog and digital signals in
> significant excess of that the Reprap operates at.
> My suggestion is to look at Phoenix Contact's
> Combicon line. You can get screw terminals,
> spring terminals, and removable terminals (thus
> avoiding the permanancy issues of RJ45 AND the
> crimp issues of the current solution without
> adding in the problem of mixing up your wires).
>

correct me if i'm wrong, but i see these as being unsuitable for main-stream use for 2 reasons:

1. price
2. polarity. you have to know exactly which wire goes where. with a pre-wired cable, you just plug it in, and the pins are all connected correctly. this is big deal when people who are not electronics experts are working with the board. (and thats the goal, isnt it? to get these into the hands of artists, mechanical people, kids, etc)

1. If you coil up 7m or wire and place it anywhere near a stepper motor or DC motor I would not be surprised if you got comms errors due to it making a parasitic transformer.

2. More time consuming than making a few crimps and a much messier result.

Feel free to put the video of the RepRap site if you think it will help anybody.

---

[www.hydraraptor.blogspot.com]

Zach:

Premade cables are office environment hardware. They aren't moving constantly as in a RepRap.

As for my Combicon solution, the 2.5mm models ARE expensive. There are other manufacturers. They're also keyed so you can't reverse them. We'd still need to make the cables first, but once made they will work fine. Moving to the 3.5mm models are actually quite simple. The "economy" models are NOT keyed, however they're made to hang over the side of the PCB. This is a simple and effective method of "keying". I do not know if the economy (which if you look at the math cost in the same ballpark as any other cable option) is a high friction hold, but its an option.

So, cost is a bit higher, but most people can follow a simple "A to B" connection diagram. You can change the wiring if you make a mistake. You can splice in other existing hardware simply (say, if thery want to put an Adurino board in for some reason). Wire length is simple to change and keep short. Time and testing will tell.

Also, if you think this is expensive, wait till you see the price for ethernet magnetics. That will range from $2.50 to over $8 depending on application, and that still doesn't include the silicon.

Electronic parts are always much cheaper in volume. My company gets Ethernet sockets with magnetics for 1$ / 1000 off and PHY chips for less than $2. I think once the projects reaches critical mass the RRRF will be able to take advantage of volume discounts to sell remarkably cheap kits fo electronics parts.

---

[www.hydraraptor.blogspot.com]

Sorry my comments earlier about 7m should have said 7ft.

---

[www.hydraraptor.blogspot.com]

No one to my knowledge has addressed transmission reflection and the need for buffers and a termination resistor. I have to check the rise and fall times for the Pic I/O in nanoseconds, then divide by 4 and multiply by 6inches/nanosecond. This will give the minimum length for which no termination resistor is needed. I will try to find the characteristic impedance of the twist pair I'm using. I know the pic can only output 20ma max. Buffers might be necessary. I will do some calculations and get back later, unless someone else has done this already?

I think you will find that at these sorts of baud rates all the reflections will have died down long before the mid bit position where it matters.

---

[www.hydraraptor.blogspot.com]

The smallest of the rise or fall time is the proper variable to calculate the length for safe unterminated signals. Since the voltage has to be zero the reflection is 180 degrees out of phase. I thought this was already taken care of, but this could explain some comm errors. I need to get the rise and fall times for the I/O and check this out. I sure know you have to terminate PLC networks correctly, (the RS486 blue hose network)

That is if your logic is sensitive to ringing on the edges, asynchronous comms is not.

What will happen is that the signal will bounce up and down the cable a few times because it gets reflections from each end that is not matched to the characteristic impedance of the cable. Traveling at around half the speed of light it can bounce up and down this length of cable many times and will dissipate long before the UART samples it. Simple UARTs sample at mid bit time so at 19200 baud that is about 25us so you are talking three orders of magnitude longer than the cable transit time. All that will happen is that if you look with a fast scope you will see ringing on the edges but when you slow it down so that you can see the bit times it will be insignificant.

True if you use RS485 down long cables, and / or at high data rates (it can do 10m bits) then you either need termination or drivers with controlled edge rates. However, with cables a few meters long and you can get away without even at 100KB.

---

[www.hydraraptor.blogspot.com]

Another vote for RJ-45!

Last year I looked through a ton of stuff, for a similar cabling problem. Ended up making my own since I needed 30 ga magnet wire with no overcoat.

But so many advantages for RJ-45.

Cheap. Only the absolutely most crude other solutions will be cheaper, and have their disadvantages.
Reliable with most mass marketed cables.
Extra capacity. You're not up on real world experience if you think extra wires are a disadvantage when you don't pay extra for them. Separating power/ground out onto standard pins so things don't blow if something is plugged in to the wrong place is an extreme advantage.
Well engineered solution, no real chance of plugging in backwards.
NOT RJ-11. RJ-11 is spotty as heck for this type application, many are only a thin foil ribbon wrapped around fiber core for strength. You know something about what you're getting with CAT-5.


Disadvantages..
Terminal blocks: Easy as heck to mis-wire. Have to screw around manually with each wire. Unwieldy as heck for more than a few wires. Have to have bare end wire laying around for problems. Anyone that can pick up a spare LAN cable and plug it in is miles ahead of you.

CD-ROM cables: Only a few wires at best, and EASY as heck to plug in backwards. Wire to the wrong board where it's used for different things and easy to blow something out since you can't separate signals.

hey Alan,

thanks for the support! I've decided to let this issue drop for the near future, but rest assured I wont forget about it, and I'll be pushing hard for RJ-45 for v1.4/v2.0 of the boards. I have a feeling that when we're ready to release those boards, we'll be very close to prime-time and that an easy cabling solution will be extremely useful.

Hey Zach,

I've been reconsidering the RepRap cabling issue for future use over the last month. I can't say what you want to do will be optimal, but let's look at an example from my projects.

I started looking into CAN networks for highly deterministic communications between my modules (ack I need a design team). There are two standards of connection for this, DB9 (Devicenet) and RJ45. Both tend to provide the two wire differential CAN signal and a 5V power pair (2A max per bus). I don't need additional connections for what I need, but I also worried about someone plugging in a network cable and something going POOF. Especially since I'm going to be providing 5V logic supply to my boards via the same cable.

I've recently been dealing with doing some network cables at work and looked at the EIA 568 standard. At 10/100 baseT speeds we have half the wire pairs used. While from the standard I do not think I can daisy chain ethernet, I can set things up differently. My boards are going to have two RJ45 connectors, one will have the wiring for ethernet and CAN (labeled), the other will have only the CAN connections. This will prevent any damage from a mistaken Ethernet connection and still allow me to daisy chain CAN or even switch over to Ethernet wherever I wish.

Something similar could be done with RepRap in V2.0 of the boards, allowing for a protected upgrade path. Given that Ethernet and Serial will not coexist on the same cable at the same time (unless I can verify that I can convert back to a daisy chain Ethernet setup) you should not have issues with signal interaction. I personally believe the signals should still be converted to/from balanced differential, but that's me. RS422 would be a direct signal transformation, but I'm considering putting the CAN transciever on a daughterboard and providing a swapout option for RS485 (and connected to the serial instead of the CAN controller) on my own designs.

I was originally against this idea, but I now see that what we needed was a better implementation vision. Have you put more thought into this?

Hey,

I've also been reconsidering things. One of the things I've reconsidered is that I don't really like the token ring network at all. It seems like overkill to design a new network when modern computers have many output ports (not to mention USB hubs) I've recently fallen in love with Arduino, an open source microcontroller project that is powerful, dead simple, and awesome. I havent fully thought it through, but I definitely think we should switch to arduino as a base for v2.0.

I dont think it has enough pins to run an extruder + 3 axes on one board, but it would be easy to attach multiple arduinos via USB and then address them individually in the host software.

Then it just becomes a matter of attaching each modular board (stepper driver, extruder driver, etc,) either directly, or via simple, solder-on pin headers that then plug in to the arduino. In the software you specify what board and what pins each module is connected to, and the software takes care of the rest.

bingo! no non-standard uses of cables needed, and the hardest cables you have to make are soldering pin headers to wires. you could probably even get by with using those off the shelf .100" cables that were found before.