NPN signal transistors make pretty good low-capacitance signal and MOSFET-gate protection zeners, etc. For example, the emitter-base junction of the common 2N3904 breaks down near 7.25 volts, and has 6 to 7pF of capacitance at 0V, drops to about 3.7pF at 5 volts. Has anybody taken pulsed high-current voltage-drop measurements on these guys?
--
Thanks,
- Win
I would guess, that a reasonable limit would be a ten microsecond pulse at 10mA.
Would that protcet the gate against ESD though ? ( body model)
-- This email has been checked for viruses by Avast antivirus software. https://www.avast.com/antivirus
Bah, humbug, I'll bet it can do amps. Hope so anyway.
--
Thanks,
- Win
I have used BJT e-b as gate protection (combined with duty as an emitter follower) and brief 1A reverse pulses were clamped nicely.
I may get some bench time next week to try higher currents.
piglet
It's limited by Rbb. I have a book around here somewhere that has a table of transistor Rbb's.
An alternate would be to put a schottky diode in series with a charged-up zener diode maybe paralleled with a cap.
There might be a FOM for schottkies, which would be Cjo * Rs. That's a time constant, and we want it to be low. It would be interesting to scatter-plot that. BAT15 is about 1 picosecond. 1N5819 is around 6.
Zeners could have thye same FOM.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Schottkys are bad, by their own admission so to speak. Under surge conditions, the guard ring activates, and you're just using another dumb PN diode. I don't think the schottky barrier itself actually gains you anything.
But, BAT54S and such continue to be used (including by myself), so it seems a diode of sufficient size isn't bothered by it, regardless which part of it is doing the work.
Wideband ESD devices (for protecting PCIe and HDMI for instance) use a PN diode as small as possible, into a common rail with a TVS onboard.
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website: https://www.seventransistorlabs.com/
Yes, Rbb' Aha, good point, Table 8.1a, page 501. We measured 110 ohms for the 2N3904, but that'd be an extra 1.1V at 10mA, now I gotta go check.
The table reveals much better choices, low Rbb' with low capacitance. Time to do some browsing.
--
Thanks,
- Win
If we're trying to protect a mosfet gate from a spike, activating the guard ring doesn't hurt... unless we want to recover fast.
I wonder if the really fast low-barrier schottkies, SMS7621 and BAT15 and such, have guard rings. They are only rated for 2 or 4 volts reverse.
I guess that an initial spike would have to charge up the TVS capacitance. One could bleed in a little current to keep it charged, on a multi-section part.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
The ones for Ethernet are precharged by the signal itself, kind of annoying I suppose but not destructive for that kind of data. The others are usually used with supply bias (e.g., super easy for USB), but can be alone as well.
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website: https://www.seventransistorlabs.com/
OK, it appears Rbb' isn't an issue, at least the parameter we measured with our noise measurements. I measure ON-Semi and Fairchild parts, with both C-B tied together, and with B alone.
I found, first, the V_EB breakdown is very sharp, changing by only about 5mV/decade from 1uA up, increasing some above 1mA. E.g., 7.413 volts at 1uA, 7.447V at 1mA, and 7.555V at 10mA. No extra Rbb' voltage visible at 10mA.
Then it rises faster, to 7.985V at 50mA, 8.39V at 100mA, and 9.23V at 500mA. That's Rs = 3 ohms. I was quickly switching the power on and off, grabbing a reading. At 1A for 1 or 2 seconds, it shorted. Sorry, I didn't setup my SMU machine for sub-ms pulsing. After the book is in, I'll take some 10A measurements at 20us.
The E-to-B and E-to-CB connections gave the same answer, SFAICT, except for 100mA and up, where the CB form had lower voltages. Maybe it was a little inverted-mode gain going on in that case? At any rate, that's the way I like to
--
Thanks,
- Win
Also please reconsider the injustice you did to my favorite gumdrop, the BCX70. It came in dead last.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
There's package limits, too; I've seen the emitter wire explode on a TO-18 transistor where the plastic-package version just kept going. Surges can generate hypersonic stresses, and wires don't just fail by thermal shocks.
Plug-package rectifiers are my idea of surge protectors. Or SiC grinding wheels between heavy plates.
Right, rbb' isn't a real resistance, it's a noise equivalent resistor. There is a real series resistance, in series with the b-e junction, that can be extracted from the DC e-i curve, but that's inherently less than rbb', a different number. Does that have a name?
-- John Larkin Highland Technology, Inc lunatic fringe electronics
This is the paragraph I'm adding, as a footnote to the graph in the x-Chapter section on protection devices:
"The reverse avalanche breakdown of the 2N3904 is very sharp, with no current conducted at all below 7 volts. Once it starts conducting, it has a low dynamic resistance, with the breakdown voltage increasing only 5mV / decade up to 1mA, where its resistance is about 5 ohms. The voltage increases by a volt at 100mA, and its still about 5 ohms. We get only one more volt by 0.5A, where the resistance is even lower. Above 50mA the breakdown voltage will be a little higher if you leave the collector pin open, you need it connected to benefit from its assistance with inverted-mode gain. The
2N3904 is fine carrying very high currents, protecting your fragile devices, if the time is short enough. It's worth noting that the r_bb' base-resistance we measured and reported in Table 8.1a, which was 110 ohms for the 2N3904, is not relevant to the breakdown."I'm hoping that by the time the book goes to print, I'll have replaced the graph with one that goes to 10A. I'm confidant the 2N3904, as with most other zener (actually avalanche) devices, can handle full 10A peak currents in the 8-20us surge tests, and pass. So it's a totally viable low-capacitance protection device.
If one gets it in the SC-70 SOT-323 package (at least four manufacturers*), it's not even that large (I still prefer diode-package parts). Fairchild offers a dual 2N3904 in an SC-70-6 package, FFB3904.** Huge stocking levels and low prices, too, for all of these. Lots of interest. Amazing, who knew?
** and the FMB3946, get one each 2n3904 and
2n3906 in SC-70-6, and other manuf as well.
--
Thanks,
- Win
BFT25 is a phenomenal diode:
The low end was limited by my measurement rig and may be even better.
Transistors in general make good diodes, better than the average diode.
Using b-e as a zener, there seems to be a small drift of zener voltage vs total charge conducted, but that's wouldn't effect its use as an ESD diode. Blow some up!
What happens to collector current when you zener the b-e junction? I've been meaning to investigate that for some decades now.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
600mV at 1uA. I'll keep it in mind next time I need a 100fA didoe.
Yes.
Hmm, I coulda taken a look yesterday.
--
Thanks,
- Win
Huh, can you say anything about the measurement rig. (610 electrometer?) I've measured current of about a pA with a ~10 pA bias current opamp (difference measurement...
I think I was looking at c-b junction of a ~200V npn. 'HV' diode with low leakage.
George H.
I built a thing:
After it was all built, I discovered that the crummy nylon banana terminals were leaky as heck, so I had to mill out the holes and add the polycarb insulator. That was a major nuisance.
I have a set of Pomona dual bannana plugs with various resistors from
1M to 1T ohms to use with that.I'm thinking that instead I could get a really good film cap and charge it with a device's leakage current. Let that sit for some hours or days and then measure the cap voltage. That would easily get below
1 fA. Some good film caps have self-discharge time constants of decades.1 nF would charge 1 uV per second at 1 fA. That gives 86 mV/day per fA.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Lots of JFET gates can make it well below 1pA. Our Table 5.3, Nine Low-input-current Op-amps, on AoE III, page 303, lists three CMOS op-amps. Try the LMP7721, rated 3fA typ, 20fA max. The ADA4530-1 has guard pins, and claims < 1fA.
--
Thanks,
- Win
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.