.
re
But it is likely that what he was actually talking about was this
which comes from MIT and involves predictions for the period 2036 to
2065, which is still some way off.It seems to be close enough to keep John Larkin happy, but he's tends to stop thinking when he sees an answer he likes.
-- Bill Sloman, Nijmegen
Oh, that. I thought you were talking about the board-edge stitching. I ran their numbers on input-to-output. Musta been put together by the same people who dreamt up AGW. MY board isn't that big and the isolated areas are less than 10% of what it would take. I'll have stitching caps at the ends of the packages.
Computer? No computer in this box. ;-)
It's a gazinta-to-twelve-gazouta at 201Hz. ;-)
I need around 50mA for each 422 driver. I could use individual isolated supplies but then I have 100KHz, and up, that I need to filter out. Second source, as long as #1 is ADI, isn't important. Cost, for this box, isn't a biggie, either. If we make a hundred of them, I'll be surprised.
Haven't built it yet (just started in layout this afternoon), so no failing EMI testing is not my problem. Yet. ;-) I am worried about it, though.
I was more interested in your take on the theories and conclusions about EMI in app note. Is the board edge fence useful? As I mentioned to Joerg in another post, the numbers for the buried input-to-output cap just don't make sense. I figure a 390pF cap on both ends should work (360MHz). A 0603 won't, though.
Which brings up a related question... Does anyone have a formula for board spacing vs. isolation voltage? There are 0805 500V caps.
Not even close...
-- Les Cargill
50mA is not a problem. I never had to filter much there, sailed through EMC no sweat. Good transformers don't radiate and then there's a smoothing choke right after the rectifiers.
But if you make only 100/year it won't matter.
-- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
A formula usually isn't accepted by notified bodies. You'll have to get the standard that applies to your market and then look up creepage and clearance. I only have the standards for medical and aerospace.
For example, in the med standard basic insulation between parts of opposite polarity for 500V operation would be 4mm for creepage, basic/supplemental is 8mm and double or reinforced is 16mm.
You'll have to get special parts that are longer. And don't forget the hi-ohms bleeder resistor which is usually required so no static can accumulate up. Vishay has fairly decent prices for those.
-- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
Well, it's 100 * 13 per board. ...and likely not "per year". ;-)
I doubt there is any standard in our market.
Opposite polarity? So that's double the V-G voltage? 16mm kinda lets out
0805 capacitors. ;-)
I'll have a look at Vishay. Thanks. None of our stuff has bleeders now, though I'd always wondered about that.
1300 still falls under "miniscule quantity" :-)
However, if this is something that repeats itself as sort of an IP block in other products then the more discrete solution could make sense.
-- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
It does, sorta. This is the simplest version. This application is only unidirectional (one in, twelve out). The others either require it to supply or accept power sorta like POE. We're looking for a discrete solution for that box but it'll be far too complicated, and large, for this box.
won't,
I don't know what your market is but there's a standard for pretty much everything. Medical, industrial, residential, aircraft, spacecraft, extraterrestrial restaurants, UFOs ...
No, that is standards committee bureaucrat speak for voltage difference. Reversing polarity would be ok, too.
If you find that your barrier needs to be wider than an 1812 or
20-something cap can bridge I'd use leaded parts. Larger ceramic caps can suffer stress fractures over time.-- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
I'm skeptical of the fence thing. Looking at the ege of a pc board, any dipole made from copper layers is tiny compared to the 1-meter wavelength at 300 MHz.
They seem to be comparing a 2-layer board to a multilayer. Not fair.
I'd guess that the major emitter is antennas caused by traces, not board-edge emission. What I'd do is make sure the power to the isolators is well bypassed close to the chips. The dipole effect, their fig 3, is probably serious. The stitching caps should help, if you can tolerate the side effects.
Ferrite beads in the signal leads that leave the board can work very well. If you're worried about EMI, I'd do that.
We have one board with 12 of the ADUM things and 12 DC/DC converters... 12 isolated 4-20 mA channels. The DC/DC sips caused a lot more trouble at 60 KHz than the isolators at 300 MHz.
This is all fuzzy stuff. Lots of opinions.
John
won't,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^ How did you know? ;-)
I have never seen a real "standard" for anything.
Gotcha. Clarity in standards-speak has never been a priority (e.g. "grounded conductor" and "grounding conductor").
Leaded parts would seriously piss off the powers-that-be.
I was skeptical, too, but the polarity of the emissions suggest that's where it comes from. I should have (just) enough space to do it, so I'll try.
They also compared a more "standard" stackup to a tightly coupled G-V pair.
Side effects of the intrusion into the no-man's-land?
Yes, I've already put some 2.5K ferrites in the signal (and ground lines). There are a couple of other lines that already have 1K and 10K resistors in them, so I figured a ferrite was wasted. I also have a 390pF cap on the outbound side of the ferrite. That's something like .1ohm at 360MHz.
The problem is that these ADuM things have a power regulator built in. One, though I've decided not to use it for other reasons, can supply 100mA. It draws 290mA (34% efficiency) and has switching currents of .5A @ 180MHz!
Exactly. The boss has already acknowledged an extra spin in the board, though not because of anticipated EMI difficulty. Any port in a storm, though. ;-)
Yikes. We have done several boards where we made our own isolated supplies, using cheap ISDN-type transformers, driven with 60 KHz square waves with softened edges. That's pretty quiet, and certainly more efficient than 34%.
I wonder how much of that 66% loss is heat, and how much is radiated RF!
John
EMI
make
won't,
Actually the ones for aerospace are pretty good. Very down to earth, easy to understand. Meaning they were obviously written by real engineers.
Believe me, there is a standard for just about anything that has civilian, military, medical or commercial use.
[...]
Field failures down the raod probably even more so. Although a micro-fractured cap across the barrier probably won't be noticed. For a while ...
-- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
Do you have a favorite cheap XFMR type there that has 2nd source? In the past I've used European ones at times. ISDN never caught on too much in the US. The XFMRs are cheap alright but I've had purchasers complaining that the mfgs can sometimes be "difficult to deal with".
Probably mostly switching losses. Even if you got the transitions down to under 2nsec that's not very good at 180MHz.
-- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
They are sometimes long-lead. I think ISDN was a mediocre idea, lacking insight as to data rates to come.
The Talemas are 400 piece min, 16 weeks lately.
We buy custom ISDN-drop-in transformers from Minntronics. They wound us some with less inter-winding capacitance then the usual ones, and we can always get them.
Minntronics is great. Ask for Butch.
John
400 would be ok, 16wks not so much. Tamura is usually not in stock and Pulse wants a lot of money for theirs. Although, they do have 8-packs which you can get for around $4 in qties:
Thanks. Aren't they the folks who packed homebaked cookies with some shipments?
-- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
One,
Yeah, sometimes. They bake a batch of some sort of Swedish sugar cookies now and then.
John
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.