Low-C input protection diodes

[Joerg]

So what you're saying is that, with 1.3pF in the diode pair (at ZERO
bias), you have 2pF in the single pad, 0.8x0.9mm? So the ground plane
is less than 0.02mm (less than a mil) under that pad? You obviously
don't have a diode problem, you have a layout problem. You're right,
cutting the diode capacity to ZERO isn't all that much help, and it's
also not very realistic. On the other hand, it's also obvious you
didn't look very closely at the package for the part I suggested; on
the same scale (same ground plane less than a mil away), its pads
would be around 0.3pF each. Even with your (apparently) ridiculously
thin dielectric, its pads can be integrated into a 50 ohm line with
very little effect. Even with a SOT23, though, protecting a
microstrip on FR4 with 20 mil dielectric should not introduce any
discontinuity; the pad "capacitance" (well under 2pF anyway) is just
integrated into the line impedance. An 0.8mm trace width (the width
of a SOT23 pad) 0.5mm (20 mils) above a plane at Er=4.7 gives you
about a 50 ohm line, at 0.12pF/mm. If the SOT23 straddles the line,
you can neck the line down a bit where it passes between the pins 1
and 3 to maintain 50 ohms. Then it becomes an issue only of diode
capacitance, and it does seem to me that getting rid of nearly half
the capacitance is worthwhile.

Even if it WERE 3pF across 50 ohms, that's only a quarter dB or so at
500MHz. With the suggested diode-pair, at zero bias, you can get that
down to well under 0.1dB, and unless the design is remarkably clean or
simple, the rest of it won't be nearly that good.

That number came from datasheets and app notes and includes the SOT23
package as well. But as I said if the diode has a better package it
would lower the total. The HMPP in the MiniPak you suggested would
certainly have a better package. However, the other downside is that
it's around $0.40. A bit steep in this application versus the usual
$0.02 for a dual.

 

[Joerg]

Le Tue, 23 Oct 2007 14:40:09 +0000, Joerg a écrit:


BAV99T in SC75/SOT523 package, from infineon (yeah, I know), Diode INc,
and MMC.

I try to stay away from Infineon unless Digikey has stock. Got burned
one too many times. Should ideally be from ON Semi or another major US
or Asian mfg, or ST. The BAV199 is also pretty cool, also SC75 and
sometimes half a penny less in quantities.

 

[Joerg]

BTW, it isn't yet known what is the cause of the fault: is it ESD or the
excess RF power at the input or something else. Depending on that, the
protection measures should be quite different.

It's pretty similar for protection. Basically you have to make sure the
voltage and current excursion doesn't exceed certain limits, no matter
what causes them. Then it'll survive. With RF there is the added risk of
causing harmonics with such diodes and suddenly the 3rd, 5th or 7th
harmonic of some other powerful source swamps your RF link. That would
not be cool.

RF chips often seem a bit wimpy. 0dBm abs max, I wouldn't even dare to
release something like that.

There is a great deal of small size low-C protection devices intended
for the fast USB and FireWire.

Yes. John Larkin had pointed out a few really nice ones.

 

[Joerg]

None, actually.

We do use a lot of Skyworks low-c schottky diodes, 0.2 pF range, but
not for esd protection.

That would probably be like using a Lamborghini to haul firewood ;-)

 

[Vladimir Vassilevsky]

Joerg wrote:

It's pretty similar for protection.

I have to disagree. The RF clamping is a different business. There are
the special diodes for that. Stray L matters. BAV99 are not good at
stray L at all.

Basically you have to make sure the
voltage and current excursion doesn't exceed certain limits, no matter
what causes them. Then it'll survive. With RF there is the added risk of
causing harmonics with such diodes and suddenly the 3rd, 5th or 7th
harmonic of some other powerful source swamps your RF link. That would
not be cool.

Aha. Now it is clear: you have the RF communication device. The failure
of the preamp FET due to the ESD is a well known problem.

RF chips often seem a bit wimpy. 0dBm abs max, I wouldn't even dare to
release something like that.

0dBm @ 50 Ohm = ~ 600mVpp. It is not very trivial to clamp at this level.



Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

 

[Joerg]

Joerg wrote:



I have to disagree. The RF clamping is a different business. There are
the special diodes for that. Stray L matters. BAV99 are not good at
stray L at all.

Yes, but you can make sure that a simple filter takes care of anything
really high in frequency and then the BAV99 works pretty well. I am
using them since 20+ years. Now RF links run at higher frequencies and
this problem begins to nag, just as capacitance does.

Aha. Now it is clear: you have the RF communication device. The failure
of the preamp FET due to the ESD is a well known problem.

Yep, inside the chips and the datasheets are not very talkative about
what's inside the chip.

0dBm @ 50 Ohm = ~ 600mVpp. It is not very trivial to clamp at this level.

That's why I prefer chips like those from TI. Their abs max is +10dBm,
much easier. But often this wasn't my decision and by the time I get
called in it's all said and done. Since you are a consultant as well I
am sure you are familiar with that ;-)

 

[Joop]

john jardine wrote: ....

You mean the BAT18 PIN Diode? That one comes as a dual and is called
BAT18-04 for the series pair. But as with a lot of EU parts no stock at
Digikey. Also, its capacitance is about the same as that for the old BAV99.

I've killed a lot of RF stuff because sometimes I get involved in power
designs. Things like turning the ceramic of a resistor or cap into
bubbly green glass. Mostly they look kind of cool after the smoke has
wafted off.

I always thought PIN diodes have plenty of storage charge. This might
not make it the best input protection diode if you want it to be
"invisible" to your signal.

How about the dual BAT62? 0.35pF typical @ 0V. Its a Schottky, so at
0dBm signal you would need to clamp to the rails (or a biased zener).
That would take another 0.1pF off when both are biased to a couple of
volts.

Joop

 

[Joerg]

I always thought PIN diodes have plenty of storage charge. This might
not make it the best input protection diode if you want it to be
"invisible" to your signal.

How about the dual BAT62? 0.35pF typical @ 0V. Its a Schottky, so at
0dBm signal you would need to clamp to the rails (or a biased zener).
That would take another 0.1pF off when both are biased to a couple of
volts.

Thanks but it seems to be an Infineon-only part. As usual, zero stock at
Digikey and that's a red flag for me.

 

[john jardine]

[...]
You mean the BAT18 PIN Diode? That one comes as a dual and is called
BAT18-04 for the series pair. But as with a lot of EU parts no stock at
Digikey. Also, its capacitance is about the same as that for the old

BAV99.
[...]

Schottky BAT81.
Of no consequence though, as John L supplied an excellent list.

 

[Joerg]

[...]

You mean the BAT18 PIN Diode? That one comes as a dual and is called
BAT18-04 for the series pair. But as with a lot of EU parts no stock at
Digikey. Also, its capacitance is about the same as that for the old

BAV99.
[...]

Schottky BAT81.
Of no consequence though, as John L supplied an excellent list.

Ah, now I found it, at Vishay. That one is an oldie, can't use it
because it's thru-hole and also rather high in capacitance. Figure 4
doesn't look too great:

http://www.vishay.com/docs/85512/bat81s.pdf

 

[Winfield]

None, actually.

Well, then how'd you find so many good candidates so
quickly? E.g., if you search datasheet features lists,
lots of parts advertise themselves as "low capacitance"
but examining the data shows them to be 10 or 20pF, etc.

 

[John Larkin]

Well, then how'd you find so many good candidates so
quickly? E.g., if you search datasheet features lists,
lots of parts advertise themselves as "low capacitance"
but examining the data shows them to be 10 or 20pF, etc.

Well, it sounded like something I should know, so I googled <low
capacitance esd diode> and they popped right up.

So now I know.

But yeah, some people think 35 pF is low capacitance.

John