TAPR TICC boxed
On Thu, 6 Apr 2017 22:23:43 -0400, you wrote:
David wrote:
I know one thing to watch out for if you are looking for low
leakage is gold doping
Anything that increases carrier mobility increases leakage current (all
else equal -- i.e., for each particular device geometry). This accounts
for the much higher leakage of Schottky and germanium junctions.
I mentioned this in connection with some manufacturers using gold
doping in transistors which would not normally be expected to have
gold doping. So you end up with a bunch of lessor named 2N3904s which
meet the 2N3904 specifications but are useless if you were looking for
low leakage diodes.
And I have another question if you know. How is rb'Cc measured?
One way is to drive the transistor with a medium-high frequency (well
down the 1/f portion of its current gain curve -- typically 10-50MHz for
small-signal BJTs) and measure the base-collector phase shift. It can
also be calculated from fT and Cc-b. There is a JEDEC standard for
measuring rb'Cc, but I'm not finding my copy at the moment. It may be
posted on the JEDEC web site.
I thought there was a more sophisticated way but that sure sounds like
something Tektronix would have done for grading parts.
The JEDEC standard is probably what I need to find or at least start
with. Thank you for the tip.
The advantage of the 4117/4118/4119 is that the leakage is already
tested to a given specification so no qualification or testing is
necessary.
That may be true, but there is nothing in the data published by Vishay,
Fairchild, Calogic, or InterFET to indicate this. Spot-checking, along
with the part design, should be sufficient to guarantee meeting the
spec. I'll try to remember to ask the Vishay process engineer next time
I talk to her.
Best regards,
Charles
If they are not being tested, then where is the maximum specified
leakage number coming from? For a small signal bipolar transistor it
will typically be 25nA, 50nA, or 100nA, but the InterFET datasheet (1)
shows 10pA maximum and 1pA maximum for the A versions.
When this discussion of low leakage input protection started, I did a
quick search for inexpensive alternatives to the 4117/4118/4119 JFETs
and came up with nothing; all of the inexpensive JFETs are much worse
until you get to premium devices.
(1) I only picked the InterFET datasheet because it was the most
readily available of the ones you mentioned. The current Fairchild
and Linear Systems datasheets show the same thing.
On Fri, 7 Apr 2017 01:06:17 -0400, you wrote:
Another thing to watch out for if you need very low leakage, is if the
package is transparent. All junctions are photodiodes.
Maybe it's less of a problem now with SMTs, than it was with glass body
diodes or translucent transistor packages.
Andy
I got caught by this once. We had a design which had to use hermetic
parts and this happened with the diodes used for input protection
during development and testing. Luckily I noticed within a few
minutes that the apparent drift coincided with the angle that I was
observing the circuit leading to the discovery that my desk lamp was
controlling the offset voltage.
We ended up painting the diodes black after soldering.
I have also heard of it happening with metal TO-18 packages through
the lead interface under the package.
On Fri, 7 Apr 2017 04:09:38 -0400, you wrote:
David wrote:
what doping is used for PNP RF transistors and saturated switches
if it is not gold? Does it also increase leakage?
I replied:
Gold doping doesn't affect the speed of BJTs in the active region very
much -- its purpose is to reduce minority carrier lifetime and, thereby,
to reduce storage time when a transistor recovers from saturation. I'm
not sure how manufacturers deal with this in the case of PNPs.
After I posted, I recalled learning in a long-ago device physics course
that both Gold and Platinum doping were used to reduce minority carrier
lifetime in PNP saturated switches. According to Motorola, the
MPS3639/3640, 2N4209, and 2N5771 were gold-doped PNP saturated switches
(all are now obsolete, although SMD versions of the 3640 and 5771 appear
to still be available).
And yes, doping PNPs with either Gold or Platinum does increase reverse
leakage current (Platinum less so than Gold).
Best regards,
Charles
So gold doping does work with PNP devices. Previously when I brought
it up, I was told gold doping only applied to NPN devices leading to
my confusion.
The Siliconix PAD1 at 1pA and 0.8pF is still available :
http://www.micross.com/pdf/LSM_PAD1_TO-72.pdf
On Sat, Apr 8, 2017 at 4:52 PM, David davidwhess@gmail.com wrote:
On Thu, 6 Apr 2017 22:23:43 -0400, you wrote:
David wrote:
I know one thing to watch out for if you are looking for low
leakage is gold doping
Anything that increases carrier mobility increases leakage current (all
else equal -- i.e., for each particular device geometry). This accounts
for the much higher leakage of Schottky and germanium junctions.
I mentioned this in connection with some manufacturers using gold
doping in transistors which would not normally be expected to have
gold doping. So you end up with a bunch of lessor named 2N3904s which
meet the 2N3904 specifications but are useless if you were looking for
low leakage diodes.
And I have another question if you know. How is rb'Cc measured?
One way is to drive the transistor with a medium-high frequency (well
down the 1/f portion of its current gain curve -- typically 10-50MHz for
small-signal BJTs) and measure the base-collector phase shift. It can
also be calculated from fT and Cc-b. There is a JEDEC standard for
measuring rb'Cc, but I'm not finding my copy at the moment. It may be
posted on the JEDEC web site.
I thought there was a more sophisticated way but that sure sounds like
something Tektronix would have done for grading parts.
The JEDEC standard is probably what I need to find or at least start
with. Thank you for the tip.
The advantage of the 4117/4118/4119 is that the leakage is already
tested to a given specification so no qualification or testing is
necessary.
That may be true, but there is nothing in the data published by Vishay,
Fairchild, Calogic, or InterFET to indicate this. Spot-checking, along
with the part design, should be sufficient to guarantee meeting the
spec. I'll try to remember to ask the Vishay process engineer next time
I talk to her.
Best regards,
Charles
If they are not being tested, then where is the maximum specified
leakage number coming from? For a small signal bipolar transistor it
will typically be 25nA, 50nA, or 100nA, but the InterFET datasheet (1)
shows 10pA maximum and 1pA maximum for the A versions.
When this discussion of low leakage input protection started, I did a
quick search for inexpensive alternatives to the 4117/4118/4119 JFETs
and came up with nothing; all of the inexpensive JFETs are much worse
until you get to premium devices.
(1) I only picked the InterFET datasheet because it was the most
readily available of the ones you mentioned. The current Fairchild
and Linear Systems datasheets show the same thing.
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/
mailman/listinfo/time-nuts
and follow the instructions there.
You need to be careful how you paint the package black. My first electronics job was in a place that made, among other things, mass spectrometers. We made very high input impedance electrometers for the mass specs using TO-5 can mosfet transistors. One batch was found to be very photo sensitive through the glass/ceramic lead interface. Someone had the idea to spray paint the bottom of the package with black paint. Not a good idea. The black paint, likely loaded with carbon, decreased the electrometer input impedance by many orders of magnitude. Considering that our electrometers had an input impedance of 1E-12 to 10E-15, even a fingerprint made a huge difference. The carbon filled black paint was practically a short.
Maybe an overcoat with silicone or some other type of low leakage sealant, then the black paint?
Tom
From: David <davidwhess@gmail.com>
To: Discussion of precise time and frequency measurement time-nuts@febo.com
Sent: Saturday, April 8, 2017 10:00 AM
Subject: Re: [time-nuts] TAPR TICC boxed (input protection)
On Fri, 7 Apr 2017 01:06:17 -0400, you wrote:
....controlling the offset voltage.
We ended up painting the diodes black after soldering.
I have also heard of it happening with metal TO-18 packages through
the lead interface under the package.
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
David wrote:
So gold doping does work with PNP devices. Previously when I brought
it up, I was told gold doping only applied to NPN devices leading to
my confusion.
Since I posted, I dug through my books and found a few more references
on point:
A couple of textbooks say commercial PNP fast switches are (were) Pt
doped.
Motorola and National Semiconductor say they use(d) gold doping for fast
PNP switches. Section 6 of National's 1978 Discrete Databook ("Process
Characteristics") lists 10 gold-doped processes (6 NPN and 4 PNP) and no
Pt-doped processes.
Raytheon's Transistor Dice Catalog indicates that the company used gold
doping for NPN switches (Processes CJ and CK) and Pt doping for PNP
switches (Processes GJ, GK, and GR).
Raytheon sourced small-signal, high-speed PNP switches 2N2409, 2N2894,
2N3640, 2N4208, 2N5771, and 2N5910 from its platinum-doped Process GR.
National sourced these parts from its gold-doped Processes 64 and 65.
Best regards,
Charles
Am 08.04.2017 um 17:52 schrieb David:
If they are not being tested, then where is the maximum specified
leakage number coming from? For a small signal bipolar transistor it
will typically be 25nA, 50nA, or 100nA, but the InterFET datasheet (1)
shows 10pA maximum and 1pA maximum for the A versions.
The large print giveth and the small print taketh away.
Usually there are footnotes and weasel words like "sample tested",
"by characterisation" or "not production tested".
The time such a small device sits on the wafer tester costs much more
than the silicon. For 100 msec.
At 1 pA it takes an eternity until the capacitances in the setup
are charged. Just the waiting time makes such a diode or FET
a premium part.
When this discussion of low leakage input protection started, I did a
quick search for inexpensive alternatives to the 4117/4118/4119 JFETs
and came up with nothing; all of the inexpensive JFETs are much worse
until you get to premium devices.
(1) I only picked the InterFET datasheet because it was the most
readily available of the ones you mentioned. The current Fairchild
and Linear Systems datasheets show the same thing.
Ouch, Interfet and data sheet in one sentence! But then they could
condense it further and just give the abs.max. ratings. I have
checked out my first 7 pairs of IF3602. Some have > 100 mA
at Vgs=-0.5V, others don't have any drain current at all. I wanted
to parallel 4 pairs for noise reasons, found just 2 pairs that are
reasonably similar. At €50 a pop finding another matching 2 will cost
a pretty penny probably.
The noise spec also seems "optimistic" and there was troubling gate
current with the 2 pairs, even at Vdd=2V. The 1/f corner seems to be
OK at 30 Hz.
Back to input protection:
Someone in the sci.electronics.design group mentioned these
< https://www.digikey.de/products/de?keywords=cmpd6001s >
but, as usual, typical values, and watch the plot with the temperature
as parameter. At least they are cheap.
Also interesting, while not exactly low leakage diodes, are these
USB3 lightning arrestors:
< https://www.digikey.de/products/de?keywords=296-25509-1-nd >
Looks like they don't spoil the timing.
regards, Gerhard
David wrote:
I mentioned this in connection with some manufacturers using gold
doping in transistors which would not normally be expected to have
gold doping. So you end up with a bunch of lessor named 2N3904s which
meet the 2N3904 specifications but are useless if you were looking for
low leakage diodes.
I believe all 2N3904s and 2N3906s are gold doped. National's certainly
were (Processes 23 and 66), and TI's and Fairchild's are. Not heavily
doped, like 2N2369s (with storage times of ~20nS), but just enough to
bring the storage time down to ~100nS. 2N2219s, 2N2222s, and 2N4401s
are also lightly gold doped.
If [4117 leakage is] not being tested, then where is the maximum specified
leakage number coming from? For a small signal bipolar transistor it
will typically be 25nA, 50nA, or 100nA, but the InterFET datasheet (1)
shows 10pA maximum and 1pA maximum for the A versions.
* * *
When this discussion of low leakage input protection started, I did a
quick search for inexpensive alternatives to the 4117/4118/4119 JFETs
and came up with nothing; all of the inexpensive JFETs are much worse
Same as any "guaranteed by design" spec -- by the device design. The
4117 series is unlike any other JFET -- the geometry is TINY, and the
4117 Idss is only 30-90uA (hundreds of times lower than other low-Idss
JFETs). [BTW, lowest Idss is why I recommend the 4117 over the 4118 and
4119 for use as a low-leakage diode. The 4118 and 4119 have higher Idss
-- up to 240uA for the 4118 and 600uA for the 4119 -- and tend to have
higher gate leakage, as well.]
Best regards,
Charles
I have run across the conductive carbon filled plastic problem before.
We did not actually use just paint. We took black mastic electrically
insulating tape, dissolved it in thinner, and painted the parts with
it. It dried to form a pliable black coating.
On Sat, 8 Apr 2017 17:49:01 +0000 (UTC), you wrote:
You need to be careful how you paint the package black. My first electronics job was in a place that made, among other things, mass spectrometers. We made very high input impedance electrometers for the mass specs using TO-5 can mosfet transistors. One batch was found to be very photo sensitive through the glass/ceramic lead interface. Someone had the idea to spray paint the bottom of the package with black paint. Not a good idea. The black paint, likely loaded with carbon, decreased the electrometer input impedance by many orders of magnitude. Considering that our electrometers had an input impedance of 1E-12 to 10E-15, even a fingerprint made a huge difference. The carbon filled black paint was practically a short.
Maybe an overcoat with silicone or some other type of low leakage sealant, then the black paint?
Tom
From: David davidwhess@gmail.com
To: Discussion of precise time and frequency measurement time-nuts@febo.com
Sent: Saturday, April 8, 2017 10:00 AM
Subject: Re: [time-nuts] TAPR TICC boxed (input protection)
On Fri, 7 Apr 2017 01:06:17 -0400, you wrote:
....controlling the offset voltage.
We ended up painting the diodes black after soldering.
I have also heard of it happening with metal TO-18 packages through
the lead interface under the package.
On Sat, 8 Apr 2017 21:43:31 +0200, you wrote:
Am 08.04.2017 um 17:52 schrieb David:
If they are not being tested, then where is the maximum specified
leakage number coming from? For a small signal bipolar transistor it
will typically be 25nA, 50nA, or 100nA, but the InterFET datasheet (1)
shows 10pA maximum and 1pA maximum for the A versions.
The large print giveth and the small print taketh away.
Usually there are footnotes and weasel words like "sample tested",
"by characterisation" or "not production tested".
The time such a small device sits on the wafer tester costs much more
than the silicon. For 100 msec.
At 1 pA it takes an eternity until the capacitances in the setup
are charged. Just the waiting time makes such a diode or FET
a premium part.
Low leakage is the defining characteristic of these JFETs so they
better be testing them.
The Calogic datasheet was the only one I checked which said anything
like "For design reference only, not 100% tested" and it did not apply
to the leakage current.
The non-A parts are only tested down to 10pA.
Back to input protection:
Someone in the sci.electronics.design group mentioned these
< https://www.digikey.de/products/de?keywords=cmpd6001s >
but, as usual, typical values, and watch the plot with the temperature
as parameter. At least they are cheap.
I think these were pointed out to me before. Since I would have to
test them to guaranty leakage below 500pA, I might as well test a
cheap small signal transistor.
If you want a laugh, take a look at NXP's various "low leakage
diodes"; they only specify and test them down to nanoamps. But I
assume for most new EEs that is low leakage.
Also interesting, while not exactly low leakage diodes, are these
USB3 lightning arrestors:
< https://www.digikey.de/products/de?keywords=296-25509-1-nd >
Looks like they don't spoil the timing.
regards, Gerhard
USB is not leakage sensitive. It looks like these were only tested to
the same 100nA standard as many transistors which makes sense; they
just need to weed out bad parts.