What BAUD rate is considered "high speed"?

[oldfortune_]

Hey guys,

I am currently in the process of designing PCB to re-route some RS-485 signals, as well as convert them to TTL in order to communicate with a microprocessor. One of the design requirements is that the system has to work at 1MBaud, and while I'm vaguely aware that at "high-speed" some different layout principles have to be implemented, I can't seem to find what constitutes high-speed...

Is 1MBaud considered high-speed? And if so what design considerations should I be mindful of in terms of layout?

Thanks in advance

This is my first post on the forum! woo

[Psi]

1M is fast but not so fast as to have serious design considerations.

Important info: How far does the signal need to travel?

If you're going 2 inches i doubt you could stuff it up if you tried 

[tggzzz]

Speed read this to find the answer to your specific question
http://www.edn.com/collections/4435129/Bogatin-s-Rules-of-Thumb

At your leisure, read all of them. They are pleasingly terse; they consider the reader's time is valuable.

[oldfortune_]

On the PCB the signal will only have to travel 50mm at a maximum, but the overall system will be around 60 modules daisychained with the total cable distance of approximately 40m. My other concern was to do with the connector and cable choice, as each module requires both power and communication. Is twisted pair cable absolutely compulsory for this kind of distance?

[T3sl4co1l]

Well, 40m is more than enough that you'll need to terminate it.  But you should terminate RS-485 anyway.

And you'll want to have EMI filtering and transient protection, but that goes without saying for any cable leaving the PCB.

Tim

[mikeselectricstuff]

1Mbaud won't be a problem, either for teh 485 side or the TTL output. Try to minimise stub length on the 485 side though.
485 will happily do 2mbaud over 50m+ of cat5 cable, and I regularly use it at 4 and 6Mbaud.

This is nowhere near "high speed" in the context of PCB layout. That would start at several tens of megahertz.

[oldfortune_]

Okay great thanks for your help everybody , that sounds very reassuring!

One of the other issues that I'm struggling to solve with the design requirement is using a single cable to carry power as well as data for a run of say 20 modules in series. Each module requires approximately 500mA at 12V, in a daisychained configuration. So this would require a 4 core cable and connector assembly able to handle loads of up, and ideally above 10A. Originally I was thinking of using XLR type connector but I could not find a right angle PCB mount type that satisfied this requirement. At the moment the best solution I could come up with was a 4 pole speakon connector and cable.

Could there be any issues with using speakon connectors and cables to carry data and power? My concern would be that the cables are not really designed to carry data... Not sure if this is just irrelevant.

PS. Love your videos and installation work Mike!

1Mbaud won't be a problem, either for teh 485 side or the TTL output. Try to minimise stub length on the 485 side though.
485 will happily do 2mbaud over 50m+ of cat5 cable, and I regularly use it at 4 and 6Mbaud.

This is nowhere near "high speed" in the context of PCB layout. That would start at several tens of megahertz.

[mikeselectricstuff]

If you can increase supply voltage that would be the way to go.
Note that RS485 transceivers are typically only protected to 12v so you should add extra protection - against shorts and out-of-sequence plug insertion.
I use a 100mA polyfuse, then a 6v8 zener to ground. If you may have significant ground shift, which you probably will if doing power over the same cable, then two back-back zeners would be better.

[T3sl4co1l]

If you can increase supply voltage that would be the way to go.
Note that RS485 transceivers are typically only protected to 12v so you should add extra protection - against shorts and out-of-sequence plug insertion.
I use a 100mA polyfuse, then a 6v8 zener to ground. If you may have significant ground shift, which you probably will if doing power over the same cable, then two back-back zeners would be better.

Yeah, back-to-back zeners (or bidirectional TVS -- beefier than the average zener) is preferred: you aren't burning common mode range that way.

Since after all, if you didn't need to accommodate the input voltage being at +12/-7V some times (or a bit less with a modest TVS), you wouldn't have needed an RS485 receiver!

Speaking of plugs, I like to put in extra grounds, and place them towards the ends of the connector (first and last pins).  This tends to make ground the first to connect, especially if the connector is rather tilted as it goes in (but doesn't help much if it's going in straight; in that case, any random pins sticking out will still mate first).  Bonus: more grounds means better signal quality!

Tim

[brucehoult]

Hey guys,

I am currently in the process of designing PCB to re-route some RS-485 signals, as well as convert them to TTL in order to communicate with a microprocessor. One of the design requirements is that the system has to work at 1MBaud, and while I'm vaguely aware that at "high-speed" some different layout principles have to be implemented, I can't seem to find what constitutes high-speed...

Is 1MBaud considered high-speed? And if so what design considerations should I be mindful of in terms of layout?

Thanks in advance

This is my first post on the forum! woo

In the late 80s Farallon sold passive little transformer boxes that plugged on to the RS422 ports on Mac computers and officially gave 4000 ft (1200 m) range for 230.4 k Baud on a single 22 AWG unshielded twisted pair (i.e. phone line). They could also do 1 M Baud for shorter distances, but still hundreds of meters.

You're not going to have a problem on a PCB :-)

[Feynman]

Whether a design is "high speed" basically depends on how sharp your edges and how long your transmission lines are.

One common rule of thumb:
A design where the transmission line is longer than 1/4 of an edge is considered "high speed".

This could be a many-GHz-link between two cores of an IC or a few-Hz-link across a big country.

[T3sl4co1l]

Emphasis on edge rate!  That 1Hz link is fine if it's only 1Hz, but if the transition (edge) is a few nanoseconds, don't expect it to arrive at the other end looking the same.

Common question: "I'm running SPI at ~kHz and it's still getting garbled! Why?!"
Common answer: "Because the MCU is still driving the same 5ns edge..." 

Tim

[tggzzz]

Too right.

I'm getting tired of trying to correct the "necessary bandwidth is 5*clock rate" misunderstood and misapplied rule-of-thumb.

[T3sl4co1l]

It's only "necessary" if you use that much.  If you just leave it wide open and unfiltered, well...

As I've said many times... I always put filtering and protection on any line that touches the dirty outside world.

Tim

[oldfortune_]

Ok cool, so it seems like it would be a good idea to use a 24 V bus and on the board have use DC buck converters to convert to both 12 V and 5 V. Is this what was meant by using back-to-back zener diodes? As I understand this will clip the level, protecting the 5V components against both positive and negative voltage spikes? See the attached schematic.

Thank you for the responses

[mikeselectricstuff]

Is this what was meant by using back-to-back zener diodes?

On the data lines, to protect against shorts to the supply or an open ground. You also need polyfuses to protect the zeners.

[oldfortune_]

Is this what was meant by using back-to-back zener diodes?

On the data lines, to protect against shorts to the supply or an open ground. You also need polyfuses to protect the zeners.

Ah right. my mistake. So something more like this, where D+ and D- are the 485 data lines.

Please excuse my newbieness, I have been thrown in the deep end for this project. Trying to learn as much as much possible as fast as possible, to make sure the tech works!

Thanks once again.

[mikeselectricstuff]

..and obviously your RS485 transceiver connects between the fuses and the zeners

[oldfortune_]

haha yup obviously.

[T3sl4co1l]

What's great about that connection is, if the zeners heat up the polyfuse, the polyfuse can go open before things get de-magic-smoke-ified.  This isn't practical on board, but there are polyfuse-TVS devices available, if you're interested.  (Probably pricier and less available?  YMMV.)

Tim

[oldfortune_]

Alright, here is the schematic, and layout of the board if anyone feels like trying to find my error... pull it to pieces if you feel so inclined, I'm sure there are plenty of errors

I tried to use a polygon pour for the 24V BUS, but I'm not sure if the ground stitching is sufficient (can't really see in this layout image). Also I wasn't 100% sure which filtering caps I should use for the DC-DC switching converters. There seem to be a few datasheets flying around with different Application circuit recommendations. Anyone know anything about Recom-78 series?

Best,
S.

[Richard Crowley]

Each module requires approximately 500mA at 12V, in a daisychained configuration. So this would require a 4 core cable and connector assembly able to handle loads of up, and ideally above 10A. Originally I was thinking of using XLR type connector but I could not find a right angle PCB mount type that satisfied this requirement.

Even the cheapest plastic XLR connector is rated for 50 VAC and 10A contacts in the 4-pin variety.  500mA would be a walk in the park.

Ref: http://www.neutrik.us/zoolu-website/media/download/13540/Product+Guide+-+Section+XLR  p. 40

[mikeselectricstuff]

How much current do you need at 5V? If less than 50mA I'd use a 78L05 from the 12v rail unless you really want to squeeze the last drop of efficiency

[oldfortune_]

It's about 150mA - 200mA at 5V unfortunately , A TTL laser and the micro. 12V is for 2 dynamixel servos. I was contemplating running the 5V regulator off the 12V rail haha but then I thought maybe it was not such a great idea for some reason.

How much current do you need at 5V? If less than 50mA I'd use a 78L05 from the 12v rail unless you really want to squeeze the last drop of efficiency

Hey Richard, from what I can see the some of the 4 pin XLR connectors (horizontal PCB mount) are only rated to 6A, and are far more expensive than a 4 pole speakon (horizontal PCB mount). Also its looking like it will be reasonably cheap to get the speakon cables made to length here in the UK! 

Fixed errors mentioned. Cheers evb149.