Looking for isolated switches

[mribble]

I'm working on a project that has to quickly switch on 3-12V lines with 0.1 to 20 mA current (more voltage and current capacity on the switch is a bonus, but not a requirement).  The control for the switch needs to be a 3.3V line from a microcontroller.  Today I'm using this: http://www.digikey.com/product-detail/en/LTV-827S-TA1/160-2040-2-ND/388468

It works pretty well.  I like that it is isolated for my use case.  I also like the 2 channels in one chip since that is ideal for my pcb layout.  I also like the low cost since I need 16 of these switches per device and it's targeted at a low cost device.  The main problem is switching speed.  The switch speed seems to depend on current and I've seen over 10 microseconds to switch at the lower current range.  3.5 us is what I see for my typical currents.  Ideally I'd have something that switches in the 0.1 us or faster range, but I'd be willing to consider changing for anything around 1 us or better.

I'm not sure I'd want to go with just a switching transistor because its not isolated.  My exact use case is triggering camera flashes and cameras which can be kind of noisy and attaching them to the same ground could be a problem.  That said, I'm willing to consider any alternatives you experts can think up.  Let's call this a brainstorming session.  I don't have enough electronic's experience to know very many options for this sort of case.

[Christopher]

If collector current is causing the delay, can you not drive an external transistor e.g 2n7002 with the opto?

[wraper]

You can use H11L1 optocouplers. Please note that detector needs 3-16V supply voltage and have open collector output.
https://www.fairchildsemi.com/datasheets/H1/H11L1M.pdf
For lower currents you can use:
https://www.fairchildsemi.com/datasheets/6N/6N136.pdf
You can forget about any usual phototransistor optocouplers without additional detector voltage supply because all of them are slow. Downside of all fast optocouplers is they are expensive.

[Zero999]

How about a pulse transformer driving a MOSFET such as the 2N7002? That will work but only for AC. It won't be possible to keep the switch on for long periods of time.

[mribble]

Christopher, it is not the collector current that is causing the delay.  It's the current on the transistor on the other side of isolation.  That said, I did oversimplify the problem before.  I don't directly drive the collector.  I use a buffer chip to prevent too much current being drawn from the micro's IO lines.  The buffer chip is fast and not a source of the delay.

Wraper, these chips do seem a little better.  The 6N136  says it has a propagation delay of around 1 us.  This is the latency from when it goes high on the collector until the switch opens on the isolated side, right?  The H11L1M says it has a rise time of 0.1 us and a turn on time of 1 us.  Could you clarify what those times are exactly to me?  I also only have 2 lines on the flash side of the circuit.  Both these chips seem to expect a vcc.  So let's say the flash has a 4V and a isolated gnd.  How would I hook up Vcc, Vb,  Vo, and Gnd for the 6n136?  The down side of these chips are they are more expensive and the H11L1M uses 6 pins for a single channel (not sure if it will fit on my circuit board.

These optoisolotor chips seem to allow large currents on the cathode side to drive the LED.  I've tried smaller currents and they seem to work just as fast even at 2 mA instead of the max 30 mA currents.  I imagine they just list the max current the LED can handle.  Does anyone know a general rule on how much current you should drive these devices with?

Hero999, I don't think a pulse transformer would work.  I'm not using AC and the switch needs to stay on for long periods (in some cases many seconds).

[wraper]

Wraper, these chips do seem a little better.  The 6N136  says it has a propagation delay of around 1 us.  This is the latency from when it goes high on the collector until the switch opens on the isolated side, right?  The H11L1M says it has a rise time of 0.1 us and a turn on time of 1 us.  Could you clarify what those times are exactly to me?  I also only have 2 lines on the flash side of the circuit.  Both these chips seem to expect a vcc.  So let's say the flash has a 4V and a isolated gnd.  How would I hook up Vcc, Vb,  Vo, and Gnd for the 6n136?  The down side of these chips are they are more expensive and the H11L1M uses 6 pins for a single channel (not sure if it will fit on my circuit board.

That means that rise time is fast because of the Schmidt trigger but actual turn on delay is 10 times higher than rise time. There is faster 6N137 and derivatives but they are lower currrent and up to 7V. You need power rail which never goes low. Therefore if you have only 2 wires at flash side, you would need some low power isolated DC/DC converter too.

[wraper]

BTW why do you need such speed as it seems excessive for flash applications. H11L1 and 6N136 optocouplers can pass 1MHz signal. I cannot imagine any flash or camera that could work closely as fast. And Considering that you have MCU there, my guess it will give more delay than optocoupler itself.

[mribble]

Thanks Wrapper!  I don't think I'm willing to add an isolated power supply to improve timings so I'd probably stick with what I have.  Unless someone can suggest a better option...

That's a good question Wrapper.  I'll assure you I'm not one to optimize a part of the system that isn't the bottleneck.  It is probably true with standard xenon flashes 4 us doesn't matter since I've measured a latency in Canon/Nikon speedlite flashes of around 50 us.  However, I use an airgap flash which only has a latency of 3 us.  The sensor/microcontroller of my new trigger system should be under 1 us.  This means with the airgap flash you are doubling latency and it's worse for some newly emerging LED flashes which have less than 1 us latency.  How can 4 more microseconds of latency matter in photography?  When I'm photographing a 3000 ft/sec bullet that equates to 0.15 inches.  I can live with that, but reducing it would be nice.  I'm also worried it could increase with certain models of flashes if they start to reduce the trigger current on flashes.

[wraper]

I don't think there are any options to make it faster without any additional isolated power. As of faster options, you could use fast 10+ mbit low voltage/current optocouplers or Analog devices ADUM series digital isolators + transistor switches to increase voltage/current handling. Neither option is very cheap, especially ADUM.

[Jeroen3]

Have you looked at isolated gate drivers?