Why don't RS485 hubs include termination resistors?

[5U4GB]

I sometimes have to work with modbus (so RS-485) setups that include hubs and repeaters, typically to allow localised star configurations in locations where the standard multidrop setup isn't practical.  However of the several hubs I've now worked with, none have built-in termination resistors, which makes them a huge pain to wire up since you can no longer run each wire pair directly into the hub but have to add termination resistors across each one.

Is there some technical reason why you'd build a hub, whose entire job is to constitute one end of an RS485 bus, without the termination resistor that this entails, or is it just a case of not-my-job?

[jwet]

They should and they should also included pull up/pull down resistors so the lines don't just flap in the breeze.

The technical reason that you will be told is that there are subtleties to terminations, not all nodes require it, some use AC terminations and in many cases you can use none.  The truth is that it adds some cost and complexity and the suppliers side step the issue.

RS-485 is a poorly written standard with gaping holes and ambiguity that are traps for those applying it.

My 2 cents- curious to see what others say. 

[Siwastaja]

Cheap-assiness of the manufacturer. Not just component cost, but the "someone else's problem" mindset.  I understand that not every random device (e.g. with RS485 being a rarely used extra feature) have those, but I would expect a specific freaking hub product to have configurable (e.g. by switches or jumpers) termination and biasing resistors for ease of install and use.

We have a controller product (basically always modbus master) with two RS485 buses and I opted for a three-way mechanical SMD switch where you can choose termination, termination+bias and none. Cost of components is negligible even when this is a relatively low-cost product.

A thousand have been sold and in 100% of installs the correct position has been T+B. Still, we are not even considering dropping the switch. It's such small expense and gives such boost in flexibility and ease of install that we think having it is a no-brainer. We are surprised how others want to push a relatively complex technical problem into poor electricians who almost never have competence in understanding pecularities of modbus biasing and RS485 termination.

[Salitronic]

The reason is simply that RS485 is meant for a multidrop configuration with short stubs to each device and with only two termination resistors one at each loose end of the multidrop cable.
Using RS485 in a star configuration (with or without termination resistors) is wrong, it does "work" in many cases but may be totally unreliable in other cases, there will inevitably be an impedance mismatch at the star point.

A permanent termination resistor built-in to the hardware therefore makes little sense for the hardware manufacturer unless the hardware is made for a specific application where the hardware will always be at one of the loose ends of the multidrop and the network is well defined. For example RS485 bus between boards within the same equipment. For a generic RS485 hardware it's not always the case that it is at the loose end of the multidrop.

In some cases the termination resistor is not just a simple resistor either but it could be an ac termination or split 60ohm with low pass filter.

I understand the usefulness of having a built-in switchable termination resistor it would surely cover the vast majority of applications, but quite frankly I see the cable termination as part of the wiring rather than the RS485 hardware itself. I've designed several products using RS485 and never included built-in termination (I sometimes do add a footprint for a termination resistor but never mounted it). It's not so much a BOM cost issue, I think it is more an issue in adding a method to enable/disable the termination in a product that it typically meant for rugged/industrial environments. If you add a switch it would often need to meet some IP rating, you need to handle ESD at that switch, etc... it adds quite a bit of complexity for something that in reality will only be used on a small percentage of the units. The industry also typically expects wire termination to be external so you'd often have the wiring installed with terminations by the wiring installer.

Edit: On re-reading the OP's post I realized that what he refers to as star topology is in fact using hubs and repeaters so that is not actually a star topology but rather a multitude of point-to-point topologies, in which case it is of course correct and requires terminations at both nodes in each point-to-point branch.

[langwadt]

. If you add a switch it would often need to meet some IP rating, you need to handle ESD at that switch, etc... it adds quite a bit of complexity for something that in reality will only be used on a small percentage of the units. The industry also typically expects wire termination to be external so you'd often have the wiring installed with terminations by the wiring installer.

one way of doing it is to add the termination on board and an extra pin so you only need to jumper two pins to terminate

[sw_guy]

....I sometimes do add a footprint for a termination resistor but never mounted it...

Copy that. I "always" add a footprint for a termination resistor... and they are "never" used.
Speaking of AC termination, there are tons of theoretical information available. However, is someone able to summarize what are the use cases in practice. I am interested when it is possible to use AC termination instead of a termination resistor.

sw guy

[Siwastaja]

but quite frankly I see the cable termination as part of the wiring rather than the RS485 hardware itself. I've designed several products ...

Of course you see it that way because it tremendously helps you with your design work, pushing the problem forward 

And sometimes it's the right solution, it depends on the target audience of your product. Ignoring designers and looking at your customers (for whom the product should be designed!), there are three sorts:
1) Those who have maybe heard the term "RS485" once or twice, and never carry any resistors with them. Normal electricians.
2) Those who have wired up many RS485 buses in their life and know there must be 120-ohm termination resistor, maybe even carry a few with them or at least know where to buy them.
3) Those who know what "failsafe biasing" is and carry other resistor values with them, are able to measure (e.g., with handheld oscilloscope) and debug RS485/modbus buses. They build complex industrial control system out of modules.

For example, out of our installer base basically everybody belongs to 1. They wire up energy meters, solar inverters, battery inverters, EV chargers and other building automation stuff to our controller. And significant % of these devices are actually hard-wired to have 120-ohm termination making them point-to-point only devices. Others have jumpers or switches to enable termination, just like our box. These are slave devices and possibly take care of termination of one end, but leave biasing as "someone else's problem".

And we are that someone else. If we try to push it to installers, all we get are very unhappy customers.

Maybe you deal with mostly group 3, maybe 2. Then it might be indeed a great idea to leave the problem for the installer to solve. They can apply whatever AC termination scheme they wish.

[Salitronic]

but quite frankly I see the cable termination as part of the wiring rather than the RS485 hardware itself. I've designed several products ...

Of course you see it that way because it tremendously helps you with your design work, pushing the problem forward 

I wouldn't say this is 'pushing the problem', signal reflection and hence the need for proper termination is fundamentally a cable problem not an RS485 problem.
This is like blaming a singer for the echo and expecting them to bring sound absorber panels with them!

If it is a device meant for a specific system or pre-defined panel installation, I agree there is a strong argument in favor of built-in (or built-in jumper connectable) termination. The AB DH485 is a typical example of this. However for a generic RS485 device the argument is much weaker especially considering that all of the connector and cabling aspects in RS485 are completely non-standardized. With just the basic A and B connections there is already widespread inconsistency about labelling and polarities. I would imagine having some devices with an extra terminal for termination jumper other devices with fixed built-in termination, others with switch controlled termination, ...would just add more confusion, especially for the group 1 that you mentioned.

[5U4GB]

They should and they should also included pull up/pull down resistors so the lines don't just flap in the breeze.

I thought pullups/pulldowns were mostly optional and sometimes even discouraged, with only the termination resistors being necessary?  I've only ever wired in the termination/EOL resistor.

Which is actually extremely useful as a quick go/no-go check, 120 ohms on the line means you've most likely got the right cable pair, 60 ohms once you add the resistor at your end as a double check.

And for @Siwastaja, there's a (2a) type, "carries a bandolier of 120 ohm resistors in their tool belt".

[5U4GB]

Cheap-assiness of the manufacturer. Not just component cost, but the "someone else's problem" mindset.  I understand that not every random device (e.g. with RS485 being a rarely used extra feature) have those, but I would expect a specific freaking hub product to have configurable (e.g. by switches or jumpers) termination and biasing resistors for ease of install and use.

That was my feeling as well, the hub constitutes one end of the bus (or at least the downstream ports do since they're isolated bus segments) so they should have termination resistors.  I've worked with gear from the same vendor where device A has it jumper-settable and documented, device B has it but it's undocumented, and device C doesn't have it.

[ejeffrey]

They should and they should also included pull up/pull down resistors so the lines don't just flap in the breeze.

I thought pullups/pulldowns were mostly optional and sometimes even discouraged, with only the termination resistors being necessary?  I've only ever wired in the termination/EOL resistor.

For most async serial applications of RS485 (including modbus and similar protocols) pull up / down bias resistors are mandatory.  Unfortunately a lot of people believe nonsense like this and it is one of the things that makes RS485 a disaster.

Also, transceivers that claim to have "failsafe biasing" which is a terrible and misleading name can usually not be trusted unless you are staying in the same chassis.  You need real bias resistors if you run cables between devices.

Leaving out the bias resistors creates the worst possible system: one that always works on the bench and usually works in the field.

[tszaboo]

Is there some technical reason why you'd build a hub, whose entire job is to constitute one end of an RS485 bus, without the termination resistor that this entails, or is it just a case of not-my-job?

Because the placement of the termination resistor is based on the wiring, not on the role.

[Siwastaja]

I wouldn't say this is 'pushing the problem', signal reflection and hence the need for proper termination is fundamentally a cable problem not an RS485 problem.
This is like blaming...

Talking about blame is interesting because it has built-in assumption that things go wrong and unhappiness ensues

I mean, as designers we are the best experts possible on stuff like signal reflection. If we can do anything to help our customers with it, we need to do it. For example in products that are often placed at the end of the bus, that means adding easily configurable termination. In products that are often modbus masters, that means adding configurable bias resistors - I know it's not related to master role per se, just "once in the system somewhere", but at master is a logical place to do it because the condition of having exactly one (no less, no more) master matches the condition of having exactly one (no less, no more) set of bias resistors.

And then of course, documentation is paramount, because our customers usually are not signal integrity experts. What type of cable to use, when to enable our integrated resistors or when to add external resistors.

With Chinese modbus slave devices it is very usual to not see any mention of termination resistors at all. We then measure with a multimeter if there is a permanent built-in resistor, making the device easier to install but point-to-point two-device bus only. Quite often there is.

[Siwastaja]

For most async serial applications of RS485 (including modbus and similar protocols) pull up / down bias resistors are mandatory.  Unfortunately a lot of people believe nonsense like this and it is one of the things that makes RS485 a disaster.

Also, transceivers that claim to have "failsafe biasing" which is a terrible and misleading name can usually not be trusted unless you are staying in the same chassis.  You need real bias resistors if you run cables between devices.

Leaving out the bias resistors creates the worst possible system: one that always works on the bench and usually works in the field.

This post should be cross stitched and framed and put on the wall of the coffee room with spotlights aimed at it at every company designing or installing RS485/modbus stuff.

[5U4GB]

With Chinese modbus slave devices it is very usual to not see any mention of termination resistors at all. We then measure with a multimeter if there is a permanent built-in resistor, making the device easier to install but point-to-point two-device bus only. Quite often there is.

I do that too, which is how I found out about the present-but-undocumented ones I mentioned.

On a vaguely related note, I finally found the application note on termination types that I remembered reading years ago but lost track of... which was also a chance to refresh my understanding of the different types.

[jwet]

Speaking of AC termination, there are tons of theoretical information available. However, is someone able to summarize what are the use cases in practice. I am interested when it is possible to use AC termination instead of a termination resistor.

sw guy

A few posts back- SW guy asked about AC terminations and it fell aside.  Another poorly understood deal...

The idea of an AC termination, a cap in series with the termination resistor, is a power saving technique that usually doesn't apply to 485.  Since the feflected waves are steps with fast rise times, AC terminations solve the termination problem (the cap is a short to the steps) but don't draw any DC power with the bus in a fixed on state.  RS-485 in its usual form is a "tri-state" kind of bus for non talkers so this shouldn't be an issue.  A 485 node is high impedance when receiving and only drives the but when its enable to talk.  AC terminations are more applicable to RS-423 (used in Apple talk, etc).which is very similar to 485 but doesn't use the tristate feature.  The idea is to not have DC current just heating up termination resistor and wasting power.  RS-485 is sometimes run in a non-duplex mode where the bus is never floated- applications like PIN pads in retail often are run to a back room hub.

Another related idea is "unit loads".  RS-485 was initially designed to have a maximum of 32 drops on a bus.  This is related to the fan-out of the drivers to the fan-in of the receivers.  Maxim (my old employer) and others made 1/4 unit load and 1/8 unit load devices that maintained the fan-out of the drivers but raised the input impedance thus allowing up to 256 nodes on bus.

This topic is near and dear to my heart, I managed the standard products new product machine at Maxim where we proliferated the original CMOS MAX485 for the 75176 type devices in every way imaginable.  I spent a lot of time talking to 485 users about their issues and we designed parts to try to address these needs.  I came to really despise 485- though it made a lot of money for the semi companies chasing problems.

 

[Siwastaja]

This topic is near and dear to my heart, I managed the standard products new product machine at Maxim where we proliferated the original CMOS MAX485 for the 75176 type devices in every way imaginable.  I spent a lot of time talking to 485 users about their issues and we designed parts to try to address these needs.  I came to really despise 485- though it made a lot of money for the semi companies chasing problems.

The problem is, as semiconductor designer you can only do so much. Sure features like slew rate control for controlling EMI are really useful, or you can reduce power consumption, increase ESD ratings, make sure your IC does not blow up at short circuit or accidental application of say +12V, useful stuff like that -  but you really can't do much on the worst real world problems that appear on the field, namely failure to bias and failure to provide ground lead.

As one recent train wreck thread on RS485 biasing showed, even people working in this field do not fully understand the consequences of the bus idle level on UART-based RS485 protocols like modbus. Bias-resistor-less transceivers are suggested from time to time and chip manufacturers of course like to market them, but they are no help at all if you need interoperability, as is almost always the case with MODBUS. If you can design all bus participant devices in-house you can as well drop MODBUS completely as unnecessary complication which it is. But if you need to interoperate, most others just keep using the traditional transceivers with traditional hysteresis / levels. As long as even one uses them, you need the bias resistors, or those slaves keep seeing start bits on any random noise and become unresponsive, or worst case, shout replies on the bus.

[langwadt]

https://mx-en.rs-online.com/product/siemens/6sl32550vc000ha0/71560658/

[Siwastaja]

https://mx-en.rs-online.com/product/siemens/6sl32550vc000ha0/71560658/

Broken website. "You are blocked". Fuck you rs-online.

[John B]

 It's very common for DMX devices to have switchable termination resistors, where you'd only select it on the last device. Obviously a relay and the associated logic/control takes a bit of space and costs money.

The other way is to have a termination resistor in an XLR housing that you manually plug into the last device. Ironically the "not my problem" type approach is more expensive for the consumer as the XLR termination resistors cost more than a relay.

The RS485 transcievers are usually rated for an output load of ~55 ohms, ie either 2x 120 ohm cable lengths in parallel, or a single cable in parallel with a termination resistor.

[Siwastaja]

It's very common for DMX devices to have switchable termination resistors, where you'd only select it on the last device. Obviously a relay and the associated logic/control takes a bit of space and costs money.

Yeah. If it has to be software controllable a smaller and more affordable (than a mechanical signal relay) photo-mosfet can be used. A simple DIP switch is even lower cost, especially if you are going to have a DIP switch there anyway to configure something else, then an extra contact plus a 120-ohm 1206 resistor would cost only a few cents. If you assign any cost for customer dissatisfaction and them trying to source a resistor or your special terminated end-plug, the switch+resistor is cheaper. Similarly, if you have to ship a special end plug, it's at least ten times more expensive than the switch+resistor on PCB, and this is not even assigning any cost for the packaging of the extra part. And then customer loses it so it's of no value anyway.

[ejeffrey]

For most async serial applications of RS485 (including modbus and similar protocols) pull up / down bias resistors are mandatory.  Unfortunately a lot of people believe nonsense like this and it is one of the things that makes RS485 a disaster.

Also, transceivers that claim to have "failsafe biasing" which is a terrible and misleading name can usually not be trusted unless you are staying in the same chassis.  You need real bias resistors if you run cables between devices.

Leaving out the bias resistors creates the worst possible system: one that always works on the bench and usually works in the field.

This post should be cross stitched and framed and put on the wall of the coffee room with spotlights aimed at it at every company designing or installing RS485/modbus stuff.

Well, I never really thought about the problem or worried about it until I had a system that was deployed in the field with a few hundred systems with ~15 nodes each where I started getting a few systems with random but data dependent corruption (e.g., on each system specific communication patterns were much more likely to cause a problem). 

[John B]

They should and they should also included pull up/pull down resistors so the lines don't just flap in the breeze.

I thought pullups/pulldowns were mostly optional and sometimes even discouraged, with only the termination resistors being necessary?  I've only ever wired in the termination/EOL resistor.

Which is actually extremely useful as a quick go/no-go check, 120 ohms on the line means you've most likely got the right cable pair, 60 ohms once you add the resistor at your end as a double check.

And for @Siwastaja, there's a (2a) type, "carries a bandolier of 120 ohm resistors in their tool belt".

If you're talking about a multi-drop setup, each transceiver should be in a receiver/high(ish) impedance state until they need to drive the bus. So the idle state leaves it more susceptible to noise, at least in my experience the transceivers weakly set the lines to a logic high, but don't drive the lines to the sames voltage as you'd expect if they're in transmit mode.

The added biasing resistors just give you more margin for noise/error. It's all dependent on whether your intended environment is noisy, like the worst case industrial environments.

Just to add as well, I've seen quite a few diagram from Texas Instruments et al, showing fairly low values like 330 ohms to 5V and GND, but I've never used anything that low. 10k or 4.7k sets the lines just fine for my needs, but I'm also not doing anything industrial. It's more just to overcome any potential manufacturing variation in transceivers as I'm not sure if any and every transceiver is guaranteed to result in the bus being a logic high when all units are in an idle state. Also, the idle state resistors are only on one device, ie the master, or the client if you're talking MODBUS.

Edit again: Using 330 ohms on each line to 5V and GND respectively, then 150 across the lines should give a termination close to 120 ohms?

ie ((2x330)^-1 + 150^1)^-1 ~= 122 ohms

So I guess you could hard-wire that combination of resistors on your master device.

[PGPG]

in my experience the transceivers weakly set the lines to a logic high

Receivers that weakly set lines to logic high are specified as being fail-save with lines open, but lines are open only if you have not terminated bus.

So I guess you could hard-wire that combination of resistors on your master device.

Provided that you force your master device being at one bus end what not always is the best solution. And to ensure the correct voltage along the whole bus you have the same polarization being done at the other bus end.
In 90s we have tried with such polarization but:
- switching on termination is more complicated/understandable for users compared to single resistor,
- we didn't found any solution allowing the bus to function normally with any one slave device having its supply switched off (when this 'any one' happens to be that one ensuring bus polarization).

At the end of 90s we found transceiver being specified as fail-save for open and short bus and decided to use them instead of polarizing the bus. For bus shorted (0V at bus set by terminal resistors) they read logic high (with 50mV noise margin, if remember well).
If the device at which terminal resistor is switched on have its supply switched off the terminal resistor still works and all other devices can communicate without any problem.

[Siwastaja]

Well, I never really thought about the problem or worried about it until I had a system that was deployed in the field with a few hundred systems with ~15 nodes each where I started getting a few systems with random but data dependent corruption (e.g., on each system specific communication patterns were much more likely to cause a problem).

The problem is that people think RS485/modbus is an "industrial quality" "very robust" "against noise" and "long wiring" and whatever marketing bullshit. In real systems with real misunderstandings (e.g. lack of biasing, grounding through parasitics) it's closer to a Arduino tutti frutti dupont cable contraption: works on the lab table and often works on the field with near-zero noise margin.

People overestimate how demanding industrial environments are. Often crappy hobbyist kludges would work well in industrial environments if they were allowed. RS485/modbus has social acceptance in industry, but in practice is often on the same level as those kludges, people just don't know it. Until it fails, and then RS485/modbus specialists are brought in. Or problem sidestepped without them, doing random changes until it seemingly works again, hobby way.