USB Isolation

[Demon-Xanth]

Here's the deal, I work for a company that makes medical devices and we want to be able to use some USB peripherals. But we need to make sure it can handle some leakage current tests in upwards of 5kV iirc.

Is there any devices or ICs that'll do this or is there any optoisolators that can handle the speeds of USB?

 

[Mark R]

Could you not feed the USB into a microcontroller which translates it into an internal bus, and then optoisolate that bus? The microcontroller being powered from a seperate high-isolation PSU. This should give you more flexibility, and alleviate the need for the isolator to operate at the speed of USB.

Most opto-isolators should be fast enough for USB (at least in low-speed mode). 1Mb/s seems to be the standard, although the very high isolation devices (>10 kV) are often slightly slower at 0.5 Mb/s. Conversely, those with highest speeds tend to not have medical grade isolation. It depends what you want to transfer across the bus.

Hmm. I read your question as if you were making USB peripherals - rather than a USB host controller. If it's the latter then things might be a bit tricky. I'll think about it some more.

 

[Demon-Xanth]

Sorry for the late reply, but the entire system is isolated as medical grade, however the device we're choosing to plug into it is not.

So the basic deal is to make it so the problem device cannot affect the main device, which is going to be attached to the patient.

 

[Demon-Xanth]

A note: the critical spec on USB is the 35nS (iirc) rise/fall times. The optocouplers must be able to match that spec.

 

[Mark R]

I'm not entirely clear on what you are trying to do - I'm assuming that you are designing the main device, and want to use 'consumer grade' for want of a better term devices as peripherals.

Presumably this device attaches to the patient, and should ideally feature a USB port for monitoring etc. I'm not entirely sure why you specifically need to isolate the USB port - is it because the USB controller would have to share the same supply and communicate directly with the patient connections?

I've not seen any opto-isolators fast enough which I would feel confident would work at USB speeds. For this reason, the only solution I can come up with is to move the isolation point - at this point you have control over bus design and can use as many parallel lines as necessary to provide the required bandwidth whitout having to compromise the strength of isolation.

Would it not be acceptable to put the isolation barrier between your device and the USB controller, instead of between the USB controller and the bus? I'm not closely familiar with the regulations for medical devices, is there a clause that states that this would not be acceptable? Or would this option be to difficult to design in?

 

[Demon-Xanth]

The device that we're making is based off a PC using the i810 chipset. So seperating that from the USB controller is out of the question. The USB peripheral is a label printer.

The regulations state that there is a leakage current test of about 4kV that must be performed, and I highly doubt that the label printer's internal power supply could pass that. If the label printer is connnected it provides a path that current could leak through. However if we provide isolation on the USB ports this path is blocked.