2
\$\begingroup\$

I have been working on a system that requires data transmission between two devices powered by separate supplies. I have based my design on the MIDI-hardware Spec, as I'm comfortable working with that configuration and have successfully implemented it in other areas.

However, I am experiencing some strange timing issues with this latest design, and can't seem to track down the problem. The screenshot below shows the Data Input (blue) and Isolated output (red). The transmission frequency has been slowed from 200kHz to 30kHz to try and assist debugging:

Isolated Data Transmission

The high(1) and low(0) levels are as I expected, but I can't figure out where the timing issue or 'lag' is coming from. The falling part of the signal is always delayed by around 25uS.

I'm using an ACPL-247 opto-isolator. But I think the problem is more likely in the 2N3904 part of the circuit, as I've not encountered this problem in previous uses of the same opto-isolator IC. Example Circuit Below:

enter image description here

Thanks for reading, any advice or pointers to relevant literature will be greatly appreciated!

\$\endgroup\$
5
  • \$\begingroup\$ Do you have a trace of the signal on the base of the NPN ? try raise RC1 to 10k, and RI to 1k \$\endgroup\$
    – jp314
    Dec 22, 2015 at 3:46
  • \$\begingroup\$ Having separate power supplies does not necessarily mean you need isolation in your data signal. Is there a reason the two systems can't have a common ground? \$\endgroup\$
    – The Photon
    Dec 22, 2015 at 4:50
  • \$\begingroup\$ @ThePhoton, it's a mixed-signal system and I'm endeavoring to remove as much of the digital noise from the analogue ground plane as possible. \$\endgroup\$
    – Tom Wilson
    Dec 22, 2015 at 12:07
  • \$\begingroup\$ @TomWilson, your circuit still switches 5 mA (at least) through S_GND. If you want to reduce noise on analog ground then you should not be using PWM control. \$\endgroup\$
    – The Photon
    Dec 22, 2015 at 16:43
  • \$\begingroup\$ @ThePhoton, what alternatives might you suggest? Or are there any documents that you would recommend reading? Thanks, Tom. \$\endgroup\$
    – Tom Wilson
    Dec 23, 2015 at 0:31

2 Answers 2

3
\$\begingroup\$

Looking at the ACPL-247 datasheet, the off time is typically 40us. So the isolator takes 40us typically to turn off, in your setup it is faster and only taking 25us to trip the BJT. Note you do not see it on the on time because it is specified 2us, although at 200khz I would think it would show up... So you have a 200khz waveform?

Try another isolator, I'm 99% confident that a better one will solve your problem. If you really need 200khz, consider something with galvanic isolation or an isolator designed for digital logic signals.

See the screen capture below for datasheet summary...

enter image description here

\$\endgroup\$
2
  • \$\begingroup\$ I think you're right here. The Signal is at 30kHz, which would explain why the 'On-Time' looks relatively instant in comparison to the 40uS on-time. I will look into some capacitive or galvanic isolators that are designed for faster transmission. Could you recommend any ICs? Thanks, Tom. \$\endgroup\$
    – Tom Wilson
    Dec 22, 2015 at 4:25
  • \$\begingroup\$ Ecnerwal covered some options in his post, I have used high speed optical isolators and galvanic coupled ones (those get expensive). Looking at Digikey, they have ~1k parts classified as "digial isolators" with sub 1us Ton and Toff (digikey.com/product-search/…). Otherwise they have ~500 high speed "Optical Isolators" similar to what you are doing with open collector etc. \$\endgroup\$
    – MadHatter
    Dec 22, 2015 at 14:09
3
\$\begingroup\$

Mostly responding to comment: You don't have to use a non-optical isolator - you just need to use an opto that's high-speed, which will normally be one built on a photodiode and an amp, rather than a photo-transistor. Photo-transistors are slower than death when compared to photodiodes, in my experience.

A quick look at a few parts suggests that you might also need to accept inverting output; or possibly not, depending on what you really need. If I'm reading them right all these optos have turn-offs in the order of 20-150 nanoseconds. No particular recommendation, not associated in any way, just looking for how much better you could get while sticking with optos.

Looking for high speed optos many results come up. Here are a few. https://www.fairchildsemi.com/application-notes/AN/AN-3012.pdf or http://www.avagotech.com/products/optocouplers/industrial-plastic/digital-optocouplers/high-speed-cmos/hcpl-0723 or http://www.mouser.com/ds/2/408/TLP117_datasheet_en_20140901-739140.pdf

Edit, Add: Another approach that floats across my mind would be to use TOSlink plastic fiber optic transmitter/receiver, which are specified for a data rate of DC to 10Mb/s and would allow greater physical separation of the things you are trying to isolate without going crazy and pulling out SFPs and glass fiber. Not quite as fast as the fastest of the above optos, but possibly plenty fast for your actual application.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.