• State TI Thinks Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 1 answer
  • Subscribers 34 subscribers
  • Views 1548 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

SN75178B: RS-485 based uni-directional star / repeater design

Christoph Hesse
Prodigy 150 points
Part Number: SN75178B

Hello,

I am designing a unidirectional bus with RS-485 standard in an extended tree topology and have several question about the design. The signals should simply be flooded through the complete tree, starting from the root node. Acknowledgement is not necessary. The aim is to use few low-cost components to keep component cost very low per module, still realizing a throughput in the range of 5-50 Mbps, the higher, the better.

In the sketch below, 4 connected modules are displayed - ideally I need 7.  I am thinking about hundreds of cascaded modules. The Z comes from Visio and does not mean anything.

I would like to ask:

- Can I reach unlimited cascading with this design, or what would limit the cascading, and how could I improve it?

- if for short distances I can simply use a star-design at the output side of the repeater as illustrated, or if I need some unidirectional hub for implementing the star-type design. I know that it is not recommended to use a star design, but it seems that for the short distances involved, a star design would probably work until roughly 20 Mbps. The following repeaters in the connected modules would then need to re-condition the signal to get rid of any problems caused by reflexions, would that work? I understand that I cannot use terminating resistances, as this would overload the repeater's output, correct? Is there a workaround to that?

- Should the star branches / stubs be all connected as close to the repeater as possible, or as far away as possible?

- if RS-485, which is defined for twisted pair, can be entirely realized on PCB (I do not plan to use any cabling)

- In what way I have to take into account the connectors, should I expect additional reflections for instance and therefore lower bandwidth?  

- Is the order of the slave / repeater ok, or should it be vice versa?

- if a different kind of bus, with low-cost interfacing microcontrollers available for max. USD1-2 (therefore Ethernet seems to be too expensive), is more suitable for this design? 

 

Best regards,

Christoph 

  • 0
    TI__Mastermind 25705 points
    Theoretically the cascades can move on forever. But without termination, you may need to take the reflection into account. In some applications, the system is designed without terminations. But usually for short distance and low speed. I think you have to live without them in star topology. The performance would depend on the media characteristic (cable or PCB) and the data rate. The stubs should be close to the repeater as possible to reduce reflection. If it’s possible, please choose the connectors to make them have minimum impact to your system. You have so many I am afraid they could hurt the communication when everything adds up together. I think it's better to put the slave before the repeater like you did. The signal after the repeater might not be good as before it. To be honest, I didn't quite get what you mean 'low-cost interfacing microcontrollers for mux'. It looks like USB can implement the function you need.

    Regards,
  • 0 in reply to Hao L
    Prodigy 150 points

    Dear Hao,

    thanks for your reply, that answers already many of my questions. A few remain though:

    Is the SN75178B a viable chip to use for this application, or are there other chips that I should look at?
    Is there a better solution to keep the signal good, for instance using a repeating multi-port hub instead of the simple star connection? Do multi-port repeater hub ICs, circuits or possibly IP exist for serial buses?
    Are there other solutions to ensure minimum signal degradation in this application?

    For the "low-cost interfacing" question: Sorry, this might not have been formulated clearly. The design will have many modules and each of them should have chip component costs of around < USD2 in mass production (10000s prices). An Arm processor without Ethernet is available for around USD1. That means, additional interface components should not cost more than roughly USD1-2, which would be the case with the very simple design of my sketches. On the other hand, it means that Ethernet is too expensive for the application, according to my price research.

    You suggested to think about USB which is interesting to me too.
    However, I did not see an advantage of USB so far. What could be the advantage and how would I have to change my design? As I understand, USB is a point to point design, correct?

    I could change the current design using RS-485 to point-to-point as well. To my understanding, point-to-point connections would eliminate the signal deterioration problems, as each slave would recreate clean signals, correct?

    I would then use an ARM uC slave with two serial bus ports, one for the point to point connection towards the root module and one towards the 7 point-to-point connections to the next modules, instead of using the bus transceiver cell of my sketches above. All the data would need to be transferred somehow from the input to the output port of the slave.
    It is unclear to me how I can realize this data transfer through the slave sufficiently fast while checking for messages that are addressed to it. Maybe there are standard solutions or ICs to do this (DMA?), if you have any information, this would be very helpful to me.

    Thank you and best regards,

    Christoph

  • 0 in reply to Christoph Hesse
    TI__Mastermind 25705 points
    Christoph,
    If you follow this reference design, the transceivers can be any general 8pin RS485 devices, like HVD82, HVD3082.
    www.ti.com/.../TIDA-01365
    TI has USB hubs. www.ti.com/.../hubs-products.page
    But I’m not familiar with USB spec. That’s just another idea I thought you could try.
    It would definitely be helpful for signaling quality if you change the structure to point to point. I’m not sure about the impact on the cost though. Speed would be another concern. The bottleneck would be uC. Sorry that’s not the area I’m familiar with. If you have other questions about transceivers, I would be happy to answer them.
    Regards,