RF galvanic isolator ...

On a sunny day (Tue, 8 Nov 2011 06:41:54 -0800 (PST)) it happened halong wrote in :

Just up the power, and use a folded dipole as antenna ;-)

Fiber optics!

Or capacitively couple into a tuned circuit. This is RF!

John

Hi Jan, it's medical environment where any extra EM is regulated.., few winding coils may work better than the dipoles

John: capacitive coupling... will it add more crying in hospital ?

Optics sounds feasible...but cost more than the air core transformer for sure...

What's the distance and geometry?

Does phase shift matter?

John

Have you tried widening your search to just specify a 200MHz signal frequency and galvanic isolation? I think the 200MHz part isn't a big deal, but the galvanic isolation may be.

If EMI is a concern then an air core transformer may not be a good choice: it's going to leak RF like a sieve, and the spacing you'll need for 5k isolation won't help.

If your production volumes are low, anything that works right off the bat is way less expensive than you think -- engineering time costs money, after all. So you may want to at least price out your fiber optics solution.

--
www.wescottdesign.com

Fiber optics? Definitely best for a clock! ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.

t

Phase shift is no problem. The clock needs to cross the two "islands" on same PCB...5KV is the different voltage...

btw, I think this would be good candidate

uct.html

but what I worry about is the pulse with distortion, the datasheet rates max distortion =3D 2ns... Does that means either reference clock- edge (rise/fall) will be 2 ns jitter ?...

Yup, optics, that would be my last plan

I hadn't mentioned anything, but presumably he wants to get data back from wherever his clock is going -- fiber would be good for that, too.

--
www.wescottdesign.com

g

You bet :-)

The clock has the timing reference information, that's why we need it...otherwise we can populate a crystal on the other side :-)

Hmmm, speaking out of this I used to dream about a wireless clock reference, and it's available everywhere in a closed system :-)))))

Hi whit3rd,

Thanks for suggestion, divide to lower frequency then multiply it back to overcome the 200MHz is a good idea...

What my concern about is the distortion/ jitter we adding up when the clock going through such stages...

Medical, defibrillator-proof? That's pretty much my home turf. I often use ferrite toroids, mostly #43 material and sometimes #61. Then 2-3 turns, bifilar winding. Bifilar is key here, without that it's not going to work.

One of the wires needs to be at least MIL-STD 22759 amendment 1 (5kV tested). Aircraft suppliers have that. Then the usual, strict training, assembly and storage requirements, all ECO'd, and yada yada.

If it's done right you can see amazing bandwidths. I did one interface (defibrillator-proof) where they needed 8nsec pulses and the slopes were less than 2nsec. Pretty much limited by what the driver circuitry could provide.

"Air core" becomes feasible at that frequency. You could possibly have loops on both sides of the board. But the compliance folks might have a hissy fit it you can't furnish the required documentation for the circuit board core material. A serious downside of air coupling is ratiated emissions which can hit you from behind during EMC cert.

--
Regards, Joerg

http://www.analogconsultants.com/

That part is only rated to 400 VRMS continuous, but it is OK at 5KV RMS for short-term use.

But it's not fast enough to pass a 200 MHz clock.

Some sort of capacitive or inductive coupling would probably be best, with a tuned circuit and comparator in the receiver. It would be fun to use the pcb itself as the capacitors or as the transformer, with the entire thickness of the board as the insulator.

John

For fixed frequency, you wouldn't need bifilar. Just tune both sides and live with a lot of leakage inductance. A few turns of HV wire on opposite sides of a little Micrometals toroid maybe.

How about a couple of Mini-Circuits RF transformers, balanced on both ends, connected with a pair of low capacitance, high-voltage ceramic caps?

John

Tuning is very much poo-poo'ed on in a production environment. That's why I use bifilar even for fixed frequency clocks. And it's not problem, since you have to wind the transformer anyhow why not to it bifilar?

Radio Yerevan would say: Basically yes, but ... in a medical environment you have pretty strict UL60601 leakage current requirements. Microamp range and even when measured at 60Hz you can quickly reach the limit. Usually because this 200MHz clock isn't the only thing that needs to cross the barrier. There's typically also the need to send a few watts of juice across.

Then there is the defibrillator jolt itself. 5kV out of a 32uF cap. A mishap could send personnel into cardiac arrest. Especially when the paddles don't find the low impedance path into the patient's chest for some reason.

The main reason why such a series connection is tough and will likely be shot down at agency approvals is that there are no real guarantees how much voltage the Mini-Circuits XFMR and how much the high voltage cap will get. Unless you have bleeders across each or ditch the Mini-Cicuits part (because you don't really need it, might as well use an LVDS receiver).

--
Regards, Joerg

http://www.analogconsultants.com/

t

ng

r).

Bifilar coils sounds promising, unless I have to find such bifilar with 5KV insulation... Another thing is how to pass the EMI regulation... how about bifilar inside a Faraday cage ? but how can I split and isolate the cage now ?

Right now I'm leaning to this: "divide the clock 1/2, or even 1/4, send it over the above Analog device, then use PLL to get it back on the other side"... how's that?

Thanks,

You want a transmission line transformer, made with coaxial cable that can stand 5kV between inner and braid. Many can - Farnell lists an RG58 cable good for 15kV (5mm diameter) and an RG174A got to 6kV (2.54mm O.D.).

-- Bill Sloman, Nijmegen

As I said, got to buy certified wire for that. There is no way around MIL-STD 22759 here.

I never needed a Faraday cage even though some of my patient interface designs are housed in ABS plastics. But Faraday is easy as well:

Make the bifilar transformer, twist the MIL-STD 22759 wire, leave it about 1" long. Place the transformer on the system side ground, put the shield around it (no sharp edges). Make sure the shield never cuts the twisted wire. Run the twisted wire across the barrier, and that's it.

MIL-STD 22759 wire if teflon-insulated and stranded (which it usually is) tends to resist twisting so you'll have to secure that with something. The wire will want to unravel the twists. Whatever you use, make sure it has the required insulating properties. No shrink tubing from dubious sources and stuff like that.

Depends on how clean the clock needs to be. It'll also add cost, complexity, size. But it does save you from having to use a custom-made part. In my cases most clocks had to be super clean and I didn't want to take chances. Plus there was also a power transfer which required a custom transformer anyhow, and whether peoduction makes one per unit or five doesn't matter much.

If you go that route _first_ make sure that the AD part has the right paperwork. For example, if you are building to category CF it must support that. IME manufacturers often wave oft or their folks get that deer in the headlight look if I mention CF. I guess they smell lawsuits or whatever :-)

--
Regards, Joerg

http://www.analogconsultants.com/

A bifilar winding is a twisted pair-transmission line, so what Joerg is recommending is - in fact - a transmission line transformer, which is what you need for 200MHz.

But don't make the winding length longer than a quarter-wavelength -

20cm in this context - and terminate the transmission line with its characteriistic impedance, unless you can exploit a reflection from the end of the line.

The core doesn't do much in a 200MHz transmission line transformer, more in broad-band transmission-line transformers that can handle components at 200MHz and above.

-- Bill Sloman, Nijmegen