optical isolation (fiber??)

[nazardo]

Hello
I have a board carrying a power section and digital control (functionally it is a sort of inverter). Despite the fact that I kept nets separate and galvanic insulation, in some conditions I have interference problems, due to high currents (10A) and high voltage switching (the output is going up to 2000V).
I think I need to separate more the two sections, and I am looking at some alternative to the optocouplers. For example, I might place the components on two different boards and then use some insulated coupling to transmit the signals. I do not need speed, just 7-8 lines (4 in one and 4 in the other direction). Is there any commercial component that does the task? I mean something like fiber coupler. I have no experience on that.

[m98]

Whats wrong with optocouplers? But yes, there are fiber optic transceivers: http://www.mouser.de/search/refine.aspx?Ntk=P_MarCom&Ntt=140260866
Bit expensive, though.

[f5r5e5d]

do you have Ott? http://www.amazon.com/Electromagnetic-Compatibility-Engineering-Henry-Ott/dp/0470189304

understanding the problem is good  - could be EMI which could still reach around optos

electrostatic shielding is trivial - any self supporting metal sheet will do

mag fields need iron to divert but fall off with distance faster than E fields so use both space and iron if high magnetic field coupling is the problem (use low carbon steel - Mu metal is useless at high field levels)

may need better supply bypassing, grounding common impedance coupling is often a problem

I/O lines can be hardened with clamps, filters - ferrite beads - you do need to know about common mode vs differential - which is the problem



fiber optics is a big step up in cost compared to packaged optocouplers - the internal Faraday shielded optocouplers are a step up in performance for not much cost

[Berni]

There is no reason a good optocoupler wouldn't do the job when designed in correctly. You can get interference in all sorts of other ways like radio waves, capacitive coupling, magnetic fields etc. Also it is a good idea to not have any high impedance nodes in your sensitive circuit, make sure things have pullups/pulldowns, filter the power rails etc.

Where i have seen fiberoptic used for isolation is with tesla coils. The ones that can be modulated with music need the signal sent to them and a fiberoptic cable going in to a shielded box is about the best way to do that.

[Marco]

http://i-fiberoptics.com/summary-info.php?id=1678

Painful amount of money for a bunch of LEDs/Phototransistors and plastic though.

In addition to what fiver said the return currents of the high current might be influencing the ground of the control electronic if you have a shared ground plane, a slit in the ground plane might help in that case.

[Circlotron]

Toslink cables,transmitters and receivers are pretty cheap. Would something like that do?

[nazardo]

Toslink cables,transmitters and receivers are pretty cheap. Would something like that do?

I do not know them. Is there any application note? Would I need a SIPO on the driver to send out 4 bits? And maybe a PISO to receive the 4 inputs?

[nazardo]

Whats wrong with optocouplers? But yes, there are fiber optic transceivers: http://www.mouser.de/search/refine.aspx?Ntk=P_MarCom&Ntt=140260866
Bit expensive, though.

Suppose I use optocouplers. Then I will have a flat connector in between the two boards: 4 wires plus a ground will exit from the mainboard and will go to the diode on the mezzanine optocouplers, and viceversa. In this way I will have a cable with the two grounds that will couple together inductively. If I need bidirectional communication (just 4 bits GPIO in both directions) I do not see how to physically separate the two boards with optocouplers . If you have a solution I will be happy with using them of course.

[DanielS]

I do not know them.

Toslink and SPDIF are standards for digital serial audio over coax and optical cables. Optical modules and the plastic optical fiber they require are available from several manufacturers for a few bucks a piece. They are little more than an optocoupler with fiber between them instead of direct coupling. While their intended application is audio, there is nothing preventing you from repurposing the transmitters and receivers to carry arbitrary protocols or just plain trigger signals.

[Marco]

I do not know them. Is there any application note? Would I need a SIPO on the driver to send out 4 bits? And maybe a PISO to receive the 4 inputs?

You won't find them much cheaper than the low end stuff from Industrial Fiber Optics I linked (when I said painfully expensive I meant ~5$ per transmitter/receiver in singles). Which have simple screw on connectors for plastic fiber which will be easier to deal with than TOSLINK.

[Marco]

If I need bidirectional communication (just 4 bits GPIO in both directions) I do not see how to physically separate the two boards with optocouplers .

Simply having the increased distance between the circuits, increased inductance of the connections and common mode filtering could help.

Of course this is assuming your guess that the interference is being transmitted through those traces and not through EMI or magnetic coupling is correct. You could put 100 meters of fiber in between the boards, but if it's RF/magnetic energy coupling into your circuit outside those connections it would obviously not help if the circuits stay in the same relative position with the same shielding.

What I would try is to take a saw to the existing PCB, make the connections through ribbon cable (with signal/ground/signal/ground/etc) and put a clamp on ferrite on the cable near the control board (the clamp on ferrite could be replaced by a PCB mounted common mode filters later on).

[Fungus]

Separate boards and a longer wire to the optocoupler should do the same job.

[actuallyjaseg]

Apparently, HP has used optical isolation via fiber on some 8.5 digit multimeters in the past. See page 12 of this pdf: http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1989-04.pdf