High speed data via optocoupler

[kd6oji]

I am working on a project that requires serial data (RS-232) to be transmitted via an optocoupler for isolation. I am not worried about the level shifting, as I am able to use a MAX232 style driver after the opto for that purpose. What I am concerned with is the reliability of my design as such, and asking for a quick review just to be sure I am on the right track. I have a fully isolated 5vdc supply on the remote side, so power isn't an issue, but I will of course need to make sure I don't trash my budget on current available with my current part, but that is a different subject

Attached is an eagle schematic of what I have worked up in Multisim that seems to work, so please review and give me a thumbs up or constructive responses!

Thanks in advance,

Shawn

[JohnnyBerg]

Funny, I tried the exact same thing. I ended up using a ADUM1201 for the RX/TX lines. I used optocouplers for the RTS, DTR lines. I found the optocouplers not reliable enough at 115200 baud.

[digsys]

If you want guaranteed speed / isolation, you need the Silabs Si8442 or similar in that series !! They go from 1Mb and 150Mb, 1KV-5KV isolation.
... and fixed low current draw, irrespective of baud rate. Awesome beggers :-)

http://www.silabs.com/Marcom%20Documents/Resources/isolation-selector-guide.pdf

[kd6oji]

Johnny,
     That looks like a great solution, thanks for the input. I was noticing some glitched output off the phototransistor in multisim, as well as occasional missing pulses when running over 500khz 50% duty cycle clock input, which is what prompted my question. With data, those missing bits would be troublesome. I had added the additional logic in an attempt to overcome that error, but it wasn't super reliable. Seriously with a 1mbd rated part, missing pulses at half the frequency? just didn't make sense.

The analog devices part isn't as cheap, but compared with the additional gates required, isn't much of an offset in pricing overall, and is far cheaper than the ADM3252 isolated line drivers which would have required a substantial redesign of that portion of my circuit.

[kd6oji]

Digisys,
   Thanks, I have apparently learned something new today, glad there are resources for dummies like me
Looking at digital isolators turned out to be the ticket, optos are good for their purpose, but not so good for really fast stuff per my sims.

thanks again guys, this has helped me a lot, and helped traverse a hurdle I had become hung up on in my design!

[T3sl4co1l]

6N136 (or SFH6345 if you need more noise isolation) and 6N137 range optos are more than sufficient for serial like this.  The resistor values shown seem a little off, but if you've checked them against minimum CTR, then that's fine.

The magnetic isolators are good too, but a little more expensive.  Maybe worth it if you're isolating a few more channels, like SPI or something.  Beware that the cheaper ones aren't guaranteed DC stable, so they can randomly flip state and transmit incorrect data in the presence of strong impulses (ESD?).

Tim

[f5r5e5d]

there are plenty of options for logic gate digital optoisolators with guaranteed speeds http://www.avagotech.com/pages/en/optocouplers_plastic/plastic_digital_optocoupler/

there are higher transient immunity versions with internal shields

and considerable innovation in various newer isolation techs as mentioned already

[kd6oji]

Yes, I am finding lots of good options. My main requirement is to separate the grounds from the digital/computer half and amateur radio/remote half mainly for RF ground loop mitigation. Absolute worst case for isolation would be a lightning strike I'd imagine, with antennas outdoors, so higher withstanding voltages are a good thing. Normal day to day operation would be serial communications to control various aspects of the amateur gear, so missing data could cause unintended operations. I've tried the 6n136 and 137 in the simulations, the 137 behaved a bit better, but still had the occasional missing pulse, and lots of switching glitch, which may or may not occur, as it is a simulation. The 136 misbehaved more often than not, with low voltage response at best, or very strange glitched signals.

[f5r5e5d]

the logic gate optos have built in amplification and hysteresis - much simpler than rolling your own and then having to characterize/debug/harden

surviving lightning strike is a rather tough requirement - various surge standards exist for lightning induced spikes

still may require layers - spark gaps, gas tubes, transorbs, low Z grounds, limiting impedances with requisite V ratings

preventing RF ingress/disruption should be a lot easier

[T3sl4co1l]

I've tried the 6n136 and 137 in the simulations

Opto models are notoriously bad, measure it on the breadboard!

[paulie]

Opto models are notoriously bad, measure it on the breadboard!

More likely it is "notoriously bad" weekend engineers who don't understand some basic principles. Saturated output transistor may be biggest limit to throughput. Easily solved with proper choice of output pull-up and limiting resistor in the LED circuit. I've had no trouble with 230kbaud transmission to regular PC serial USB adapters using old tech 4n37.

Noise problems can be a little tricky. When dealing with microvolt/1ppm measuring circuits capacitive crosstalk became a problem. The generic optoisolator parts did sustain spikes but only a caused trouble at high data rates. My solution was building a DIY discrete isolator using IR LED and phototransistor (harvested from a 10 cent TC5000 detector) poked into 1" shrink tube. Even high end multi-thousand dollar lab equipment use a similar approach as seen in some of Dave Jones breakdown videos.

[T3sl4co1l]

Yeah, if you need better performance or isolation, you'll quickly find you're spending more board area or money on patching it up, at which point you reach for the fancier kinds of isolators.

One such example; t_r is around 300ns, pretty remarkable for a 4N35, but you're spending another four transistors plus assorted resistors on it, too.
http://seventransistorlabs.com/Images/4N35_TIA.png
Some of this is inevitable, say if you have a gate driver circuit that needs opto isolation.  You need the level shift, sharpening and amplification to drive the gate.  But even doing better than 3 transistors and a handful of resistors is tough, at which point you might as well go for the integrated version.

Tim

[electrodacus]

If all that you need is 115200 then that 6n135 is more than good enough.  I was even able to use the LTV-357T at 115200 (that wan was at the limit there) but your choice is 10x faster.

[kd6oji]

Well, I have samples enroute for a few different options to breadboard and do some live testing.
As I said, the sim was giving erratic results, which prompted the question, and surprisingly in some 30 years of hobby electronics, I've never moved data through an opto.

Thanks for all the pointers everyone, I've enjoyed the info and learning experience.