After some thought analysis about what I’ve learned about transmission line reflections theory, I can re-affirm my explanation of transmission line bus construction. Changes in current flow are the primary cause of reflections. So, when the question comes down to how to tap out a T connection to your device connected to your bus wire, the solution is simple. You simply want your device on the T connection to be a “high impedance” connection, i.e. it does not draw much current from the line. This way, you are minimizing your interruptions to the line current when connecting.
But on second thoughts, there may be other considerations. You can have multiple receivers on the connection simultaneously, but only one sender can send at a time. So, although you can have a high-impedance receive, you must have a low-impedance send. And ultimately, you want to be able to electrically remove a device from the transmission line when it is not sending so that it does not provide a current flow path that would interfere with other senders. Ah, yes, looking at the Wikipedia article for “three-state logic” in more detail, this is precisely what it is: in addition to an output being able to assert two states, it can also assert itself to be electrically removed from the connection, creating a high-impedance state.
20191209/https://en.wikipedia.org/wiki/Three-state_logic
So, that’s the magic on control of the sender side. So, finally, your primary current path is pretty much just the “bus” line connection itself, and you finish that off by terminating the ends of the bus line with resistors equal to the characteristic impedance of the line. These resistors, of course, have lower resistance than the high impedance receive connections, which are designed to have a higher impedance so as to not interfere with the line.
So, another consideration. What kind of cable should you use for your own IoT network where you plan on connecting devices via RS-485? In a large scale construction, sure, you could use cheaper cable, but my recommendation would to be to simply use the same standard Cat6A cable that you are using for building out the Ethernet network, and likewise strive to wire it up in a star topology. This provides for easy upgrade to a high-speed Ethernet network should the need arise in the future. “Design for the future, because it will be here sooner than you think.” And especially for construction at the building scale, often times there are sources of delay and procrastination with making any changes at all that more than tilt the cards to building future-resistant structures that can still provide value even under procrastination and deferred maintenance.
Now, what if you want to put a single RS-485 receiver at the hub of a star topology with point-to-point connections? How can you do this that provides for proper termination at the characteristic impedance? Simply use line drivers for termination, this will provide for electrical isolation of the current loops of different wires. Then, you can wire-OR everything together on the board-side of your circuit, without worrying about creating interference in the lines.
Okay, so enough hypothesizing, let’s take a look at the data sheets for some actual RS-485 line drivers. Indeed, I can confirm my explanation in the data sheets. The input impedance determines the maximum number of transceivers that can be connected, such as 32. That’s also especially why you must use a purpose-built “receiver” circuit if operating on multiple access bus communications. Likewise, only one transmitter can be connected at a single time, and all other transmitters must be in the high-impedance state. High impedance input, low-impedance output.
20191210/DuckDuckGo digikey rs-485 line driver
20191210/https://www.digikey.com/catalog/en/partgroup/rs422-rs485-differential-line-driver/11774
20191210/https://www.digikey.com/catalog/en/partgroup/rs-422-rs-485-transceivers/9879
There are also some nice through-hole DIP components to boot.
20191210/https://www.digikey.com/products/en?mpart=ADM485JNZ&v=505
20191210/https://www.analog.com/media/en/technical-documentation/data-sheets/ADM485.pdf
20191210/https://www.digikey.com/products/en?mpart=ADM488ANZ&v=505
20191210/https://www.analog.com/media/en/technical-documentation/data-sheets/ADM488_489.pdf
This last one is probably more complicated than needed for practical applications, but it may be useful for experimentation. Typically you only use either full duplex or you use half-duplex with 3-state logic. It also has the slowest data rate at 250 kilobits per second, but that’s still plenty fast.
20191210/https://www.digikey.com/products/en/integrated-circuits-ics/interface-drivers-receivers-transceivers/710?k=ADM489&k=&pkeyword=ADM489&sv=0&pv16=96863&sf=0&quantity=&ColumnSort=0&page=1&pageSize=25
RS-422 and RS-485 are 5 V logic, easy as that. Use level shifters to interface the input/output of these line drivers with 3.3 V logic.
Whoa, this RS-485 line driver is rated for 30 megabits per second. Much faster than I could ever need for most of my projects.
20191210/https://www.digikey.com/products/en?mpart=ADM1485ANZ&v=505
20191210/https://www.analog.com/media/en/technical-documentation/data-sheets/ADM1485.pdf
Okay, but I still don’t know about the whole T connector thing. Well, if I had to throw in by best guess, it is that you want the branch on the T connector to be as short as possible, if such a design is unavoidable. And termination is agreed to be 120 ohm resistors for twisted pairs.
20191210/DuckDuckGo rs-485 t connector
20191210/https://eeci.com/rs-485.pdf
20191210/https://www.windmill.co.uk/rs485.html
On the other hand, Apple LocalTalk was very similar to this whole RS-485 thing, and they used transceiver modules that two connectors for the daisy chaining. And, they had automatic termination and some additional filter electronics inside the modules. Filter electronics such as an inductor, capacitor, and some resistors. Then there is a finite length of wire for connecting between the module and your Macintosh.
Yeah, really interesting that the LocalTalk article does not mention RS-485, the similarities are many.
20191210/https://en.wikipedia.org/wiki/LocalTalk
20191210/https://en.wikipedia.org/wiki/Choke_(electronics)
But, looking elsewhere, the magic trick is revealed. Here is a datasheet that shows how you can connect an RS-485 transceiver to a LocalTalk network. So, what’s actually in that LocalTalk interface box? A transformer is used to electrically separate the transceiver from the LocalTalk line bus network. This effectively allows you to isolate the impedance effects of your “T connector” from the LocalTalk bus wire, keeping its electrical path as straight as possible. However, it also imposes different electrical standards, such as the inability to pull the line high or low and the need for proper communications frequency matching. The receiver circuit is connected to with two 1 K resistors, thus providing the required high-impedance input. Also, each transceiver input/output is filtered with a 100 pF capacitor to ground surrounded by two 22 ohm resistors. The LocalTalk bus wire is terminated with 120 ohm resistors.
Additionally, it is also possible to directly connect to an LocalTalk network without the transformer, and using full duplex with separate twisted pairs for transmit and receive.
20191210/DuckDuckGo localtalk rs-485
20191210/https://html.alldatasheet.com/html-pdf/547355/LINER/RS485/2150/14/RS485.html
20191210/https://html.alldatasheet.com/html-pdf/547355/LINER/RS485/2304/15/RS485.html
Okay, enough with the nonsense on that site, let’s get it straight from the Digi-Key source.
20191210/DuckDuckGo digikey ltc1321
20191210/https://www.digikey.com/product-detail/en/linear-technology-analog-devices/LTC1321CSW-PBF/LTC1321CSW-PBF-ND/891269
20191210/https://www.analog.com/media/en/technical-documentation/data-sheets/lt1321.pdf
One more point in hand. LocalTalk is also limited to 32 devices per network, just like RS-485. So indeed, the input impedance configuration is not somehow superior to that used on RS-485.
20191210/https://en.wikipedia.org/wiki/AppleTalk
Again, I reiterate, because this is important!
Now this really gets me thinking. Isolating transformers on a RS-422-style differential communications through LocalTalk? Well if that works just fine, now I’m thinking power over RS-422, similar to Power over Ethernet (PoE). As I’ve learned on the Wikipedia article on Power over Ethernet, PoE is pretty simple to implement due to the center tap transformer design of Ethernet transceivers. Now I see that clearly for RS-422/485 communications too, thanks to the LocalTalk schematics. In the case of RS-485, you’d use a spare wire pair to carry the other common-mode power or ground, whichever is the opposite of what is carried on the data wires.
Likewise, the schematics of the Raspberry Pi 3 Ethernet jack are also very insightful for the particular design of Power over Ethernet. Note that the center tap on the Raspberry Pi side of the circuit is used for the power. Notably, there are also ferrite core chokes used to suppress any common-mode noise common across all Ethernet wires on the outside world input/output side.
20191210/https://en.wikipedia.org/wiki/Power_over_Ethernet
20191210/https://en.wikipedia.org/wiki/Transformer_types#Pulse_transformer
20191210/https://hackaday.com/2018/06/01/raspberry-pis-power-over-ethernet-hardware-sparks-false-spying-hubub/
The particular types of transformers used in Ethernet jacks are called pulse transformers. These are transformers that are optimized to work with square wave signals. Also, a duty cycle near 50% is pretty much required. Okay, so thinking about this, sure that may work well for Ethernet and LocalTalk, but it may encounter problems with RS-422 and RS-485 serial communications. But surely, there is technical info on how to make it work for that application? Indeed there is.
So, here’s what looks like the trick is to putting RS-485 through isolation transformers. You might add an additional oscillator for signal modulation so that your flat runs of high signal value will register as the presence of high frequency oscillation. But, I guess I don’t fully understand exactly how the pulse transformer is matched with the serial communications frequency and what the frequency response curve looks like.
20191213/DuckDuckGo rs-485 pulse transformer
20191213/https://www.ti.com/lit/ds/symlink/iso35t.pdf
20191213/https://www.digikey.com/catalog/en/partgroup/iso35t-series/58979
Also, thinking more about PoE, if you extract your power on your device’s side of the transformer, the common mode power source must also be an alternating current signal in order to make it through the transformer. Ah, yes, so that’s why the article on the Raspberry Pi’s Ethernet jack noted that the transformer needed to pass the alternating current power through. You don’t want to use a direct current common mode signal because then you must tap out away from the device side, and you render in the risk of ground loops and limiting the common mode voltage range.
Yeah, the whole concept of these transformers in longer data transmissions lines is pretty much the same as that of the isolation transformers in single-wire earth return.