RS485 chip LTC491; optocoupler MOC5007

If you read the fine datasheet at

Figure 12 suggests that - connected directly to the wires - you should be able to get a bit over 100k bps at a bit over 300 ft (i.e. 1000 m).

The effect of the optocouplers are an extra confounding factor. Can you define a nearly-worst-case combination of conditions to do a long bit-error-rate test?

--------------------------------------- Posted through

Eh? 300 m is approx 1000 ft. It looks like your 10x too short.

Peter

Em 17/2/2011 12:57, RCIngham escreveu: [snipped]

Not really, in fact ten percent of that distance in meters: 91,44 meters, for a 12 inches foot more exactly.

--
Cesar Rabak
GNU/Linux User 52247.
Get counted: http://counter.li.org/

"Correct termination" is a can of worms. Almost all current RS485 devices have the A/B circuits reversed re the spec. Also, simply terminating across the A/B lines will result in an indeterminate state. You need some bleed current in there to ensure a MARK state when idle. Hint: AC-coupling the termination often helps with the bleed current.

Without termination, you're at the mercy of cable capacitance and transmission-line reflections. From experience, you can probably get away with 4800baud over that distance, if you're lucky.

Thanks. The tests I did were with termination - a 150 ohm resistor across the A B terminals, plus a 47K resistor from A to VCC (I think this was for biasing, whatever that is).

The optocoupler datasheet says "high data rate 1MHz typical NRZ".

We use the 68302 DDCMP mode. The error rate at 250 Kbps over 1000 M non twisted cable was very small / negligible.

You Know they Make RS-485 Transceivers with the Isolation built-in.

at

fa.pdf, Figure 12 suggests that -

0k

Thanks. I guess you meant 3000 feet as that's what Figure 12 shows. I ran a test at 250 Kbps for several hours over 1000M with very few errors but I don't have a worst case optocoupler or a worst case RS485 chip.

I thought that speed / distance rating thing was for when there are multiple devices on the wire. We have only two devices on the wire.

A single 47k is not enough to guarantee a mark signal on an idle line. You need neough bleed current from both lines, one to voltage supply and another to ground, to create a guaranteed differential signal across the termination.

For the speeds you're aiming at, the wavelength is below the

1000 m cable length, and you have to have a decent transmission line. Using a non-twisted pair and/or unknown insulation is inviting trouble.

--

Tauno Voipio

[...]

That's why I consider EIA-485 broken by design.

Do you have a working real world example?

I can't imagine that AC termination works if the bit rate is not way beyond the propagation time. Either it's time constant is too short for long cables, or it's too long for very biasing patterns.

Oliver

--
Oliver Betz, Munich
despammed.com is broken, use Reply-To:

The non twisted cable was just what I was testing with because we have

1000 M of it here. We can specify whatever kind of cable we like for use in a real system.

Yes.

I'm sure there are more I just googled "rs485 isolated transceiver" and found that one.

True. A 680 to Vcc on 'A' a 680 to Vss on 'B' and a 150 terminating is a lot better.

Nothing wrong with the spec. Plenty wrong with the devices which followed.

Although - I presume your point is that, with termination and no bleeds, there's that indeterminate state. Yes, the specs could have been clearer about this. But it's not *that* hard. I've been using RS485 without problems (apart from those danged chips) for nearly 30 years now... (Eeeeep.)

I fear you misunderstand me. Given the lowish value of the terminator, the bleed to/from rails results in fairly high DC current. A transmission line is only an issue with AC ;). So, AC-coupling the terminator resolves the termination issue, while resolving the DC current issue.

There was an app note from my old company detailing this online, but they're rebuilding their site, and it's now "under construction"... gah...

Steve

--
http://www.fivetrees.com

I disagree. Requesting a DC bias was the wrong way IMNSHO. The better design choice is to make 0V a defined state, as CAN does.

Then I can use passive terminations at the ends of the bus and I don't need to select a bus node to provide DC bias.

[...]

That's what I understood. I think it doesn't work under the conditions I pointed out.

Oliver

--
Oliver Betz, Muenchen (oliverbetz.de)