Low leakage parts
Tom Bruhns
years about the self-discharge rate of polyester and polypropylene
caps. The time constants I saw were on the order of a few years for
the polyester and over 50 years for the polyprops.
A month or so ago, someone asked about making a simple toggle circuit
for turning 12V lights on and off, and I posted a couple solutions.
One of them was a "this is really simple, but it probably won't work
very well" circuit using just a capacitor to hold the voltage on the
gate of a power mosfet. Well, I built that ckt, using an 0.01uF cap
across the gate-source, and toggled it "on", so the cap was charged to
about 12V. Then I disconnected the power and went on a holiday trip.
Just came back, and it's still in the "on" state, three weeks later.
Sooo...I'd say that modern power mosfets also have pretty low gate
leakage current. To hold the voltage above the nom. 3V required to
turn the mosfet on, the average leakage must have been less than 50fA,
assuming 21 days, 0.01uF and 9V delta (and no arithmetic errors). (It
was cool, about 18C, and likely wouldn't do quite so well inside a car
with the windows rolled up in Phoenix in the summer...)
Cheers,
Tom
Phil Allison
"Tom Bruhns" <k7...@aol.com> wrote in message
news:3200347.03010...@posting.google.com...
** 21 days is 1.8 exp 6 seconds, divided by 0.01 exp -6 gives 1.8
exp 14 ohms as the resistance in the cap, fet and wiring etc. I don't
believe it - do you ??
I suspect there was a DC leakage path HOLDING the 0.01 uF cap voltage
up that was the real story here.
......... Phil
John Larkin
Tom,
I did a post a while back: a 2N7000 with a 9-volt battery and an LED
in its drain, gate hanging open. This can be set on or off and stay in
that state for many days. You can also perch the gate voltage just at
the turnon point and get the LED to stay on dimly, with no perceptable
brightness change over many hours. This is cool... if you wave a
charged object near the gate lead, you can modulate the LED brightness
by inducing voltage into the gate. Maybe I'll repeat this with a power
fet.
I figured that the gate leakage was in the 100 electron/second range,
maybe much less... attoamps. After all, eproms and flash store bits as
a few thousand electrons each, for many years.
John
Winfield Hill
>
> Tom Bruhns wrote
>> Some of you may recall some postings I've made ... [ snip ]
>> Sooo...I'd say that modern power mosfets also have pretty low gate
>> leakage current. To hold the voltage above the nom. 3V required to
>> turn the mosfet on, the average leakage must have been less than 50fA,
>> assuming 21 days, 0.01uF and 9V delta (and no arithmetic errors).
>> (It was cool, about 18C, and likely wouldn't do quite so well inside
>> a car with the windows rolled up in Phoenix in the summer...)
>
> ** 21 days is 1.8 exp 6 seconds, divided by 0.01 exp -6 gives
> 1.8 exp 14 ohms as the resistance in the cap, fet and wiring etc.
> I don't believe it - do you ??
>
> I suspect there was a DC leakage path HOLDING the 0.01 uF cap voltage
> up that was the real story here.
Phil, having made similar measurements and calculations myself, I'd
say that you could well be wrong. As Tom and I have both observed in
the past, leakage resistances in excess of 100T-ohm are not all that
unusual. Moreover, we know that sub-5fA leakage currents have been
routinely measured for NSC mosfet opamps, 10x below Tom's value.
On the other hand, FET gate leakages often tend to turn the part on.
Tom can easily expand his experiment to test such a hypothesis simply
by completely discharging the capacitor and waiting to see how long
the FET stays off. If the off state exceeds 21 days, I'd say he's
made a good point.
Thanks,
- Win
Phil Allison
"John Larkin" <jjla...@highSNIPlandTHIStechPLEASEnology.com> wrote in
message news:3gcc1vsedgol8lbll...@4ax.com...
** Not quite the same thing as components mounted on a PCB.
Ever measure resistance between close spaced tracks on a glass or
phenolic PCB ?
I attemped to make condenser mic pre-amps once and this issure
became the major one - the slightest surface contamination or moisture
reduces the resistance to less than 100 Mohms at room temp and WAY less of
the PCB is heated.
The major contributor to thermal runaway in OP valves is
anode/screen to grid 1 leakage on the plastic base - remember these get
pretty hot. I have measured as low as 20 M ohms between adjacecent pins on
a *clean* octal base with the base at 180 degrees C.
If one pin is at 500 volts and the other adjacent one at -42 with
270 kohms in series........
.............. Phil
Phil Allison
"Winfield Hill" <wh...@picovolt.com> wrote in message
news:av5d8...@drn.newsguy.com...
> Phil Allison wrote...
> >
> > Tom Bruhns wrote
> >> Some of you may recall some postings I've made ... [ snip ]
> >> Sooo...I'd say that modern power mosfets also have pretty low gate
> >> leakage current. To hold the voltage above the nom. 3V required to
> >> turn the mosfet on, the average leakage must have been less than 50fA,
> >> assuming 21 days, 0.01uF and 9V delta (and no arithmetic errors).
> >> (It was cool, about 18C, and likely wouldn't do quite so well inside
> >> a car with the windows rolled up in Phoenix in the summer...)
> >
> > ** 21 days is 1.8 exp 6 seconds, divided by 0.01 exp -6 gives
> > 1.8 exp 14 ohms as the resistance in the cap, fet and wiring etc.
> > I don't believe it - do you ??
> >
> > I suspect there was a DC leakage path HOLDING the 0.01 uF cap voltage
> > up that was the real story here.
>
> Phil, having made similar measurements and calculations myself, I'd
> say that you could well be wrong.
** The issue is NOT the resistance inside the semiconductor or cap
di-electric (they are super high as you say ) but *stray* resistance and
therefore leakage on the surface of the plastic package, PBC material and
any coatings used for insulation of wires etc.
This resistance carries leakage current between any and all points in
a circuit. A super high impedance like a fet gate will float at the leakage
network's value at that point unless a deliberate resistance swamps the
network of all possible strays.
It is always good design to have a real resistor of at least one
order of magnitude LESS value than the stray resistances at the highest
operating temp and humidity expected.
I have seen the specific problem of a leaky PCB in poorly designed
audio gear ( both valve and SS) many times - it is a nightmare to fix.
................ Phil
Spehro Pefhany
<bi...@bigpond.com> wrote:
>
> I suspect there was a DC leakage path HOLDING the 0.01 uF cap voltage
>up that was the real story here.
He did say he removed power for the 21 days.
Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
sp...@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
Phil Allison
"Spehro Pefhany" <sp...@interlog.com> wrote in message
news:g3kc1v8nfssf4733m...@4ax.com...
> On Sat, 4 Jan 2003 11:02:34 +1100, the renowned "Phil Allison"
> <bi...@bigpond.com> wrote:
> >
> > I suspect there was a DC leakage path HOLDING the 0.01 uF cap
voltage
> >up that was the real story here.
>
> He did say he removed power for the 21 days.
** Does not prove there was *no* leakage path in his design.
BTW I, for one, never take poster's words literally. Having done
electronics servicing for 30 years I know to disbelieve EVERYTHING I am
old - especially when it is only in writing.
................ Phil
Spehro Pefhany
<bi...@bigpond.com> wrote:
>
>"Spehro Pefhany" <sp...@interlog.com> wrote in message
>news:g3kc1v8nfssf4733m...@4ax.com...
>> On Sat, 4 Jan 2003 11:02:34 +1100, the renowned "Phil Allison"
>> <bi...@bigpond.com> wrote:
>> >
>> > I suspect there was a DC leakage path HOLDING the 0.01 uF cap
>voltage
>> >up that was the real story here.
>>
>> He did say he removed power for the 21 days.
>
>
> ** Does not prove there was *no* leakage path in his design.
No, but no visible voltage source is a darn good start. Where do
you think the DC could come from? AC sources are a dime a dozen, but
mystery DC leakage is not so common. Electrolytic action, perhaps,
if there was a lot of humidity. Regular electrolytics won't hold
a charge for very long.
> BTW I, for one, never take poster's words literally. Having done
>electronics servicing for 30 years I know to disbelieve EVERYTHING I am
>old - especially when it is only in writing.
Yes, I've seen a *lot* of that stuff too and tend to discount the
words until I get a measure of the person. But Tom Bruhns seems to
know what he's doing.
Phil Allison
> On Sat, 4 Jan 2003 14:21:33 +1100, the renowned "Phil Allison"
> <bi...@bigpond.com> wrote:
>
> >
> >"Spehro Pefhany" <sp...@interlog.com> wrote in message
> >news:g3kc1v8nfssf4733m...@4ax.com...
> >> On Sat, 4 Jan 2003 11:02:34 +1100, the renowned "Phil Allison"
> >> <bi...@bigpond.com> wrote:
> >> >
> >> > I suspect there was a DC leakage path HOLDING the 0.01 uF cap
> >voltage> >up that was the real story here.
> >>
> >> He did say he removed power for the 21 days.
> >
> >
> > ** Does not prove there was *no* leakage path in his design.
>
> No, but no visible voltage source is a darn good start.
** That is a massive assumption. There may be a switch, PCB tracks or
simply plastic coated wires with leakage current of the order of 1 exp -13
amps.
Hey we *are* talking 1 exp 14 ohms here !!!!! Air is that leaky.
> > BTW I, for one, never take poster's words literally. Having done
> >electronics servicing for 30 years I know to disbelieve EVERYTHING I am
> >old - especially when it is only in writing.
>
> Yes, I've seen a *lot* of that stuff too and tend to discount the
> words until I get a measure of the person. But Tom Bruhns seems to
> know what he's doing.
** OK - then let him explain how he got 1.8 exp 14 ohms across
little bits of plastic insulation that were exposed to the atmosphere.
Most designs battle to get 1 exp 10.
................ Phil
Winfield Hill
>
> John Larkin wrote
>>
>>> A month or so ago, someone asked about making a simple toggle circuit
>>> for turning 12V lights on and off, and I posted a couple solutions.
>>> One of them was a "this is really simple, but it probably won't work
>>> very well" circuit using just a capacitor to hold the voltage on the
>>> gate of a power mosfet. Well, I built that ckt, using an 0.01uF cap
>>> across the gate-source, and toggled it "on", so the cap was charged to
>>> about 12V. Then I disconnected the power and went on a holiday trip.
>>> Just came back, and it's still in the "on" state, three weeks later.
>>> Sooo...I'd say that modern power mosfets also have pretty low gate
>>> leakage current. To hold the voltage above the nom. 3V required to
>>> turn the mosfet on, the average leakage must have been less than 50fA,
>>> assuming 21 days, 0.01uF and 9V delta (and no arithmetic errors). (It
>>> was cool, about 18C, and likely wouldn't do quite so well inside a car
>>> with the windows rolled up in Phoenix in the summer...)
>>
>> I did a post a while back: a 2N7000 with a 9-volt battery and an LED
>> in its drain, gate hanging open. This can be set on or off and stay in
>> that state for many days. You can also perch the gate voltage just at
>> the turnon point and get the LED to stay on dimly, with no perceptable
>> brightness change over many hours. This is cool... if you wave a
>> charged object near the gate lead, you can modulate the LED brightness
>> by inducing voltage into the gate. Maybe I'll repeat this with a power
>> fet.
>>
>> I figured that the gate leakage was in the 100 electron/second range,
>> maybe much less... attoamps. After all, eproms and flash store bits
>> as a few thousand electrons each, for many years.
>
> ** Not quite the same thing as components mounted on a PCB.
>
> Ever measure resistance between close spaced tracks on a glass or
> phenolic PCB ?
Hey, Phil, those of us who work in the area of very high-Z circuits
always take standard precautions: guard traces for moderate-impedance
wiring, and lifted-leg wiring in air for serious low leakage. This
is my preferred approach, and isn't as painful as it sounds; usually
only one node is involved. If several nodes are necessary, one can
use a teflon standoff. These circuits can work reliably for decades.
We've added a special component to our PCB library, with only one pin.
Thanks,
- Win
Winfield Hill
>
> ** OK - then let him explain how he got 1.8 exp 14 ohms across
> little bits of plastic insulation that were exposed to the atmosphere.
> Most designs battle to get 1 exp 10.
WRONG.
Thanks,
- Win
Phil Allison
** Screw you Win, that is just posting abuse.
......... Phil
Phil Allison
** That is lovely but of no relevance to the either to OP or my
remarks.
.............. Phil
Spehro Pefhany
<bi...@bigpond.com> wrote:
>
> ** That is lovely but of no relevance to the either to OP or my
>remarks.
I didn't get the impression that Tom used a PCB, so your assumptions
may fit the facts less closely than Win's. A haywired circuit will
have similar performance to Win's doggy-style low leakage methods.
Tom's been working with, and thinking about, this low-leakage stuff
for a *long* time. Here's an interesting thread he started back in
July of 1999:
Incidentally, the conversation also includes David Klassen alias
Dangerdav who passed away just over two years ago.
Phil Allison
> On Sun, 5 Jan 2003 00:06:30 +1100, the renowned "Phil Allison"
> <bi...@bigpond.com> wrote:
> >
> > ** That is lovely but of no relevance to the either to OP or my
> >remarks.
>
> I didn't get the impression that Tom used a PCB, so your assumptions.....
** What assumptions ?? You have assumed I made some - I believe I
made none.
I am still waiting for the OP to come back with all the info about the
construction details of his magic cct.
The first thing I imagined was that he built a rats nest - since
anything else has more leakage, but even then the plastic pack of the mosfet
will have leakage. Whatever he built was exposed to the atmosphere - day
and night for 21 days. Was it in a permanently controlled air environment
or in a home workshop. Was it a period of very low humidity ? etc
180 thousand gigohms is a lot of ohms.
.................... Phil
Tony Williams
Spehro Pefhany <sp...@interlog.com> wrote:
> Incidentally, the conversation also includes David Klassen alias
> Dangerdav who passed away just over two years ago.
Good grief..... and how do you find out these
things Speff?
--
Tony Williams.
Spehro Pefhany
Word gets around. In this case, it was reported by Robert Paterno.
Winfield Hill
>
> I did a post a while back: a 2N7000 with a 9-volt battery and an
> LED in its drain, gate hanging open. This can be set on or off
> and stay in that state for many days. You can also perch the
> gate voltage just at the turnon point and get the LED to stay
> on dimly, with no perceptable brightness change over many hours.
> range, maybe much less... attoamps.
At low currents, such as 1 to 5mA, the transconductance of a
2n7000 approaches that of a transistor, where an 18mV change
makes a 2x current difference. If we assume your FET stayed
within 2x current over 5 hours, that's dV/dt < 1uV/s. Given
a 40pF gate capacitance, we get I = C dV/dt = 4E-17A, or 40aA,
which is 250 e-/sec.
While impressive that's not unreasonable, and it demonstrates
that under good conditions both air and semiconductor plastics
have much better insulation capabilities than some would think.
You don't mention the FET's gate-drain voltage, but assuming
2V, we get R > 5E16 or 50000000G, or 50M Gig-ohms, another
interesting value. The high-quality plastics used by most
semiconductor manufacturers must be impressive, especially
considering that the gate leakage is likely in the gate oxide
rather than the plastic.
Thanks,
- Win
Mike Monett
[...]
> While impressive that's not unreasonable, and it demonstrates
> that under good conditions both air and semiconductor plastics
> have much better insulation capabilities than some would think.
[...]
> Thanks,
> - Win
Just to put some numbers on it, here's snips from Bob Pease:
"What's All This Femtoampere Stuff, Anyhow?"
http://www.national.com/rap/Story/0,1562,5,00.html
"The other material we like to use for an insulator is the body of
the plastic-DIP ICs. This has an impedance up near 10^16 ohms, which
is about as good as you need."
"Ceramic DIP packages, on the other hand, aren't nearly as low in
leakage, more like 10^14 ohms - not very good for precision work at
subpicoampere levels."
Here's an interesting comment on reed relays:
"The interesting point is, that some standard, cheap and small
plastic dil reed relais by Siemens had a typical insulation
resistance of nearly 10 exp16 Ohms. Of course you could not put them
into the board, you had to wire the 'hot' contact directly and use
all the techniques described in Mr. Pease articles. But the relays
where damn good and we used them in some hundred devices. Maybe
their packages where made out of an equal low leakage material like
the LMC types or the BurrBrown INA116. The problem is, after six
years, they changed their packaging material and the wonder-relais
turned into ordinary ones with 10exp (12) or worse."
http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&safe=off&frame=right&th=6f21d2128856f40b&seekm=on18usskp2pl7o5c4eobi8hgpbiprtve8r%404ax.com
Mike
Tom Bruhns
hours. :-) Yes, you are right. The circuit was left with no voltage
applied, and since the FET is ON rather hard, it's safe to assume the
drain and source are at very close to the same potential. The gate
just has a capacitor, as noted, between it and the source. No PCB
involved. During my absence, there weren't any other significant
voltage sources nearby...things were all turned off. I could have put
the thing in a Faraday cage, but really didn't expect it to still be
charged after that long anyway. Now that I know it can hold a charge
for quite a while, I may try the complementary experiment, with it
"off" and power applied to the drain circuit. The pin layout on the
TO-220 mosfet puts the drain in the middle and the gate off to one
side, so surface leakage should be detected easily. (Of course, the
same surface leakage should have been/has been discharging the
gate/cap for the past three weeks, too.)
Perhaps Phil lives in an environment with an abnormally high number of
ions floating around in the atmosphere, but the roughly 0.1fA of
leakage I'm seeing in 0.1uF polyprop caps at 10V suggests about 10^17
ohms effective shunt resistance, and that includes the air in my
office as well as surface leakage over the outside of the caps and the
bulk resistance of the dielectric in the caps. I expect they'd
discharge rather faster in bright sunlight.
On the subject of FR4 PCB leakage: I've seen bad leakage on uncleaned
boards, but modern materials and manufacturing techniques yield some
pretty low leakages at room temperature, even in moderatly high
(non-condensing) humidity.
I am NOT (!!!) saying that it's good engineering practice to RELY on
such low leakages, I'm just making observations that modern materials
and methods are resulting in some pretty impressive (to me)
performance. BTW, Bob Pease related some similar results to me in
another area. I think the careful process control is reaching into a
whole lot of different areas.
Cheers,
Tom
Spehro Pefhany <sp...@interlog.com> wrote in message news:<eklc1v815rqjb2mfl...@4ax.com>...
Tom Bruhns
with a 2N7000 with relatively small gate area, but I thought the
IRF540 (I think that's what it is...) would be a lot worse.
In experimenting with my circuit, I also noted that you could easily
detect motion near the gate lead when it was floating and charged to a
value resulting in partial turn-on. Of course, it was also pretty
temperature sensitive!
Fun stuff.
Cheers,
Tom
By the way, the circuit I'm playing with is intended as a toggle
light-switch "toy" for low-voltage DC powered lights. The circuit is
just the FET with the lamp as a drain load to the (12V) power,
something like 1 megohm resistor from drain to a 4.7uF cap to ground,
an 0.01uF cap from gate to source/ground, and a NO momentary contact
switch from the 4.7uF cap to the gate. Then if the lamp has been on a
while, the 4.7uF is discharged and closing the switch for a short
(<1sec) time discharges the gate cap and turns the lamp off. After a
while the 4.7 is then charged, and closing the switch for a short time
turns it back on. It may actually be practical IF you turn the lamp
on and off often enough. The "better" version added a transistor to
make sure things stayed on or off, but the simple one has the property
of remembering its state over a power outage.
John Larkin <jjla...@highSNIPlandTHIStechPLEASEnology.com> wrote in message news:<3gcc1vsedgol8lbll...@4ax.com>...
...
Winfield Hill
>
>>> Incidentally, the conversation also includes David Klassen alias
>>> Dangerdav who passed away just over two years ago.
>
Another blast from the past. Robert seemed to often
uncover bits of obscure but useful information.
Thanks,
- Win
John Woodgate
wrote (in <av6ub...@drn.newsguy.com>) about 'Low leakage parts', on
Sat, 4 Jan 2003:
>You don't mention the FET's gate-drain voltage, but assuming
> 2V, we get R > 5E16 or 50000000G, or 50M Gig-ohms, another
> interesting value.
50 Pet(a)ohms, I believe.
Incidentally, I see an unwelcome trend to render the 10^12 multiplier as
'terra', instead of 'tera'. Latin 'earth' instead of Greek 'monster'.
--
Regards, John Woodgate, OOO - Own Opinions Only. http://www.jmwa.demon.co.uk
Interested in professional sound reinforcement and distribution? Then go to
http://www.isce.org.uk
PLEASE do NOT copy news posts to me by E-MAIL!
John Larkin
wrote:
Hi, Win
Try this some time, it's fun. I have a 9-volt battery wired to a
resistor, an LED, and a pair of alligator clip leads, that I keep on
my desk for checking continuity and LED polarities and stuff like
that. So I just clipped it to the 2N7000 source and drain. With the
LED dim, Vd was likely around 6 volts.
You can set the gate voltage by touching the battery terminals with
two fingers, and touching the gate briefly with a finger from the
other hand... a human potentiometer. You can also charge-pump the gate
by moving a pencil point from b+ or b- to the gate many times. This
would be a very nice classroom experiment. If you blast the s-g air
gap with alphas from a smoke detector source, the LED dims; blast the
g-d gap, and it gets brighter.
Simple things like this amuse me disproportionately.
But you're not allowed to play with stuff like this until you finish
the Third Edition.
John
John Larkin
>Thanks for the note, John. Yeah, that wouldn't surprise me too much
>with a 2N7000 with relatively small gate area, but I thought the
>IRF540 (I think that's what it is...) would be a lot worse.
>
>In experimenting with my circuit, I also noted that you could easily
>detect motion near the gate lead when it was floating and charged to a
>value resulting in partial turn-on. Of course, it was also pretty
>temperature sensitive!
>
>Fun stuff.
>
>Cheers,
>Tom
>
>By the way, the circuit I'm playing with is intended as a toggle
>light-switch "toy" for low-voltage DC powered lights. The circuit is
>just the FET with the lamp as a drain load to the (12V) power,
>something like 1 megohm resistor from drain to a 4.7uF cap to ground,
>an 0.01uF cap from gate to source/ground, and a NO momentary contact
>switch from the 4.7uF cap to the gate. Then if the lamp has been on a
>while, the 4.7uF is discharged and closing the switch for a short
>(<1sec) time discharges the gate cap and turns the lamp off. After a
>while the 4.7 is then charged, and closing the switch for a short time
>turns it back on. It may actually be practical IF you turn the lamp
>on and off often enough. The "better" version added a transistor to
>make sure things stayed on or off, but the simple one has the property
>of remembering its state over a power outage.
>
Wow. A brief push toggles the lamp full on or full off; if you hold it
down longer, it will reach a stable partly-on equilibrium, which is
'lamp dim'. That would fool a lot of people into thinking there's a
PIC cpu inside!
Oops... once you reach the 'dim' equilibrium state, are you stuck
there? Yes!
John
Oh, when I was a kid I had a long-wire antenna on the roof. I could
hook it to my VTVM (with the 10M input resistor lifted) and watch the
induced voltage as clouds passed overhead.
John Larkin
<j...@jmwa.demon.contraspam.yuk> wrote:
>I read in sci.electronics.design that Winfield Hill <wh...@picovolt.com>
>wrote (in <av6ub...@drn.newsguy.com>) about 'Low leakage parts', on
>Sat, 4 Jan 2003:
>>You don't mention the FET's gate-drain voltage, but assuming
>> 2V, we get R > 5E16 or 50000000G, or 50M Gig-ohms, another
>> interesting value.
>
>50 Pet(a)ohms, I believe.
>
>Incidentally, I see an unwelcome trend to render the 10^12 multiplier as
>'terra', instead of 'tera'. Latin 'earth' instead of Greek 'monster'.
Then there's the Nissan Xterra SUV. Xterra is Greek for 'Monster that
Destroys the Earth' and Latin for 'terrorizes pedestrians.'
Almost as funny as calling a huge 4wd pickup a 'Tundra'
John
John Woodgate
Sat, 4 Jan 2003:
>Another blast from the past. Robert seemed to often
> uncover bits of obscure but useful information.
It was/is his redeeming feature, and a skill that was/is probably a
parallel consequence of his other proclivities.
Winfield Hill
>
> John Larkin wrote...
>>
>> I did a post a while back: a 2N7000 with a 9-volt battery and an
>> LED in its drain, gate hanging open. This can be set on or off
>> and stay in that state for many days. You can also perch the
>> gate voltage just at the turnon point and get the LED to stay
>> on dimly, with no perceptable brightness change over many hours.
>> ... I figured that the gate leakage was in the 100 electron/second
>> range, maybe much less... attoamps.
>
> At low currents, such as 1 to 5mA, the transconductance of a
> 2n7000 approaches that of a transistor, where an 18mV change
> makes a 2x current difference. If we assume your FET stayed
> within 2x current over 5 hours, that's dV/dt < 1uV/s. Given
> a 40pF gate capacitance, we get I = C dV/dt = 4E-17A, or 40aA,
> which is 250 e-/sec.
OK, I've taken a few measurements and can use them to improve
the calculations. I measured a 2n7000 in the vicinity of 0.3
to 2mA, and found its transconductance to be about 30 to 40%
that of a transistor, with a 2x current change corresponding
to a 75mV change in gate voltage. Second, both Ciss and Crss
must be charged/discharged, which is more like 60pF total. So
a 2x change corresponds to a charge of q = C V = 4.5pC and if
it occurs over a 5 hour period, that corresponds to an average
gate current of 2.5 x 10^-16 or 0.25fA, which is 1560 e-/s.
> You don't mention the FET's gate-drain voltage, but assuming
> 2V, we get R > 5E16 or 50000000G, or 50M Gig-ohms, another
> interesting value. The high-quality plastics used by most
> semiconductor manufacturers must be impressive, especially
> considering that the gate leakage is likely in the gate oxide
> rather than the plastic.
Using the above values, revise this downward by about 6x to
8P (8 peta-ohms = 8E15), still assuming a 2V drop.
Also, I have just repeated John's experiment, a mini two-hour
version, largely confirming the above. But with a few rather
interesting results. :>)
Thanks,
- Win
Winfield Hill
>
> Perhaps Phil lives in an environment with an abnormally high number
Ocean spray at the end of the dock during a storm comes to mind :>)
Thanks,
- Win
Winfield Hill
> Hi, Win
>
> Try this some time, it's fun. I have a 9-volt battery wired to a
> resistor, an LED, and a pair of alligator clip leads, that I keep on
> my desk for checking continuity and LED polarities and stuff like
> that. So I just clipped it to the 2N7000 source and drain. With the
> LED dim, Vd was likely around 6 volts.
>
> You can set the gate voltage by touching the battery terminals with
> two fingers, and touching the gate briefly with a finger from the
> other hand... a human potentiometer. You can also ...
>
> Simple things like this amuse me disproportionately.
>
> But you're not allowed to play with stuff like this until you finish
> the Third Edition.
Your warning was too late, John, I already did some exeperimenting!
Thanks,
- Win
John Woodgate
THIStechPLEASEnology.etc> wrote (in <t0ee1vcfbm7e9it25kdrni2ta9r2r6o53h@
4ax.com>) about 'Low leakage parts', on Sat, 4 Jan 2003:
>Then there's the Nissan Xterra SUV. Xterra is Greek for 'Monster that
>Destroys the Earth' and Latin for 'terrorizes pedestrians.'
You only say that because you haven't got one. (;-)
>
>Almost as funny as calling a huge 4wd pickup a 'Tundra'
Sounds quite cool to me. (;-)
January Sales. Two one-liners in one post.
John Woodgate
>down longer, it will reach a stable partly-on equilibrium, which is
>'lamp dim'. That would fool a lot of people into thinking there's a
>PIC cpu inside!
>
>Oops... once you reach the 'dim' equilibrium state, are you stuck
>there? Yes!
Take an electret mic capsule to pieces, retrieve the 470 Mohm (or so)
resistor and put it across the 10 nF cap. With the NO switch open, the
lamp will go out after a while.
Phil Allison
"Tom Bruhns" <k7...@aol.com> wrote in message
news:3200347.03010...@posting.google.com...
> Perhaps Phil lives in an environment with an abnormally high number of
> ions floating around in the atmosphere, but the roughly 0.1fA of
> leakage I'm seeing in 0.1uF polyprop caps at 10V suggests about 10^17
> ohms effective shunt resistance,
** The insulation resistance for some ( LCR brand) polypropylene caps
( if that is what you used) is quoted in the Farnell cat as 1 x 10^11 ohms.
For LCR polystyrene it is given as 5 x 10^11 ohms.
Your suggested figure is some one million times better.
My doubts are reasonable.
..................... Phil
Boris Mohar
wrote:
>
>"Spehro Pefhany" <sp...@interlog.com> wrote in message
>news:eklc1v815rqjb2mfl...@4ax.com...
>> On Sat, 4 Jan 2003 14:21:33 +1100, the renowned "Phil Allison"
>> <bi...@bigpond.com> wrote:
>>
>> >
>> >"Spehro Pefhany" <sp...@interlog.com> wrote in message
>> >news:g3kc1v8nfssf4733m...@4ax.com...
>> >> On Sat, 4 Jan 2003 11:02:34 +1100, the renowned "Phil Allison"
>> >> <bi...@bigpond.com> wrote:
>> >> >
>> >> > I suspect there was a DC leakage path HOLDING the 0.01 uF cap
>> >voltage> >up that was the real story here.
>> >>
>> >> He did say he removed power for the 21 days.
>> >
>> >
>> > ** Does not prove there was *no* leakage path in his design.
>
>>
>> No, but no visible voltage source is a darn good start.
>
>
> ** That is a massive assumption. There may be a switch, PCB tracks or
>simply plastic coated wires with leakage current of the order of 1 exp -13
>amps.
>
> Hey we *are* talking 1 exp 14 ohms here !!!!! Air is that leaky.
It is not.
Boris Mohar
Phil Allison
"Boris Mohar" <bor...@sympatico.ca> wrote in message
news:5nve1vo2ip2to6no4...@4ax.com...
** Try damp air.
........ Phil
Winfield Hill
>
> Tom Bruhns wrote ...
>
>> the roughly 0.1fA of leakage I'm seeing in 0.1uF polyprop caps
>> at 10V suggests about 10^17 ohms effective shunt resistance,
>
> ** The insulation resistance for some ( LCR brand)
> Farnell cat as 1 x 10^11 ohms. For LCR polystyrene it is given
> as 5 x 10^11 ohms.
>
> Your suggested figure is some one million times better.
>
> My doubts are reasonable.
Phil, your doubt are NOT reasonable, quite the contrary. That's
because you Are doubting engineers who have far better knowledge,
skill, experience and equipment than yourself.
In this thread you have been and are systematically doubting Tom's
measurements and insights, while putting up weak irrelevant points
to defend your doubts. Tom has measured what he says he's measured.
For you to postulate conditions he didn't experience is wrong and
irrelevant. [I have further opinions, but I'll remain silent.]
Furthermore some of the rest of us have made similar measurements,
are aware of "official" laboratory measurements on the properties
of air, etc., and have therefore repeatedly offered our opinion on
the reliability of Tom's reports, which knowledge and opinion you
have rejected.
As for your Farnell data, those are WORST case specs, thankfully
rarely seen in practice. A production engineer ignores them at
his peril, but you cannot use them to argue against laboratory
component measurements made by skilled engineers who know better
than you what they are doing.
For these reasons I say, "Drop it: Your doubts are NOT reasonable."
Thanks,
- Win
John Larkin
wrote:
>
> Also, I have just repeated John's experiment, a mini two-hour
> version, largely confirming the above. But with a few rather
> interesting results. :>)
>
> Thanks,
> - Win
Win,
it is *not* sporting to reference 'interesting results' without
telling us what they were. Cruel, in fact.
Hey, remember the old duck-and-cover days of the Cold War, when the
Civil Defense people had little pen-shaped radiation dosimeters? They
had a little lens on one end, and you'd hold it up to the light and
see a little quartz fiber in front of a scale. The scale went from
zero to maybe 1000 rads ('you're dead, Jack!') The fiber was charged
up to bend it to the zero mark, and radiation would discharge it. I
seem to recall long charge storage periods, many days. Anybody have
details?
Oh, here's one:
http://www.rpdinc.com/pdf/N1.pdf
claiming a typical leakage of 0.25% per day.
John
Phil Allison
> Phil wrote...
> >
> > Tom Bruhns wrote ...
> >
> >> the roughly 0.1fA of leakage I'm seeing in 0.1uF polyprop caps
> >> at 10V suggests about 10^17 ohms effective shunt resistance,
> >
> > ** The insulation resistance for some ( LCR brand)
> > polypropylene caps (if that is what you used) is quoted in the
> > Farnell cat as 1 x 10^11 ohms. For LCR polystyrene it is given
> > as 5 x 10^11 ohms.
> >
> > Your suggested figure is some one million times better.
> >
> > My doubts are reasonable.
>
> Phil, your doubt are NOT reasonable, quite the contrary.
** The figures I quote above *proves my point* - you should read my
post as a self defence and not an attack on Mr Bruhns claims.
Also read entire posts /thread so you can maybe see the purpose and
intent.
That's> because you Are doubting engineers who have far better
knowledge,
> skill, experience and equipment than yourself.
** That is just insane. You are so full of yourself, Win.
> In this thread you have been and are systematically doubting Tom's
> measurements and insights, while putting up weak irrelevant points
> to defend your doubts.
** Absolute bullshit. Tom (whom I do not know at all) put up ONE post
with NO details of his claim. Only today has he come back with any.
>> Tom has measured what he says he's measured.
** How do you know ? Were you there?
All claims that seem odd to others get challenged here - god knows
enough of mine have been and often very abusively. Every one of mine is
true - but if the "experts" here have no personal experience of something
then it can look very unlikely to some of them.
> For you to postulate conditions he didn't experience is wrong and
> irrelevant.
** Absolute rubbish - what you are saying here is WRONG and IRRLEVANT to
anything I actually posted. You are making a pompous, idiotic and offensive
speech, Win. That is very bad usenet practice - you should stick to
responding to someone's actual words and not indulge in wild imaginings of
their intentions.
[I have further opinions, but I'll remain silent.]
** I have some beauts about pricks like you too.
> Furthermore some of the rest of us have made similar measurements,
> are aware of "official" laboratory measurements on the properties
> of air, etc., and have therefore repeatedly offered our opinion on
> the reliability of Tom's reports, which knowledge and opinion you
> have rejected.
** This is all garbage - what are posts are you referring to ? Your
statements make no sense - it is all non-specific. You have enough
intelligence to realise how wrong it is to do this. No-one can answer a
non-specific complaint since that is merely abuse.
> As for your Farnell data, those are WORST case specs, thankfully
> rarely seen in practice.
** How do you know that ?? Look like typicals to me.
A production engineer ignores them at
> his peril, but you cannot use them to argue against laboratory
> component measurements made by skilled engineers who know better
> than you what they are doing.
** I did not do any such thing, Win. I posted them in self defence -
to show my reaction to the OPs claim was based on some evidence.
As I posted, PCB leakage phenomena is familiar territory to me from
my experience with designing condenser mic pre-amps and various repair work
on such mics.
> For these reasons I say, "Drop it: Your doubts are NOT reasonable."
** Yes they are entirely so. "Reasonable" and "correct" are not
synonyms you realise.
If you have some inside knowlede about the OPs expertise, fine - but I
do not and there was nothing in the OP to say how the amazing cct was built.
Can you try *real hard* Win to SEE a situation on a NG from *other*
than your OWN ivory tower point of view ??
.................... Phil
Boris Mohar
wrote:
Surface contamination is easily activated by damp air. Memories of one
humid summer when we had a batch of 1fA ionization chamber electormeters
misbehave. Humid air by itself is not conductive. Ionized air dry or
humid is a different story.
Regards,
Boris Mohar
Viatrack Printed Circuit Designs
Mike Monett
[...]
> Surface contamination is easily activated by damp air. Memories of one
> humid summer when we had a batch of 1fA ionization chamber electormeters
> misbehave. Humid air by itself is not conductive. Ionized air dry or
> humid is a different story.
>
> Regards,
>
> Boris Mohar
> Viatrack Printed Circuit Designs
Boris, can you tell how you cleaned the chambers to achieve low enough
leakage to measure 1fA?
Mike
Winfield Hill
>
> it is *not* sporting to reference 'interesting results' without
> telling us what they were. Cruel, in fact.
Sorry, see my new thread.
> Hey, remember the old duck-and-cover days of the Cold War, when the
> Civil Defense people had little pen-shaped radiation dosimeters? They
> had a little lens on one end, and you'd hold it up to the light and
> see a little quartz fiber in front of a scale. The scale went from
> zero to maybe 1000 rads ('you're dead, Jack!') The fiber was charged
> up to bend it to the zero mark, and radiation would discharge it. I
> seem to recall long charge storage periods, many days. Anybody have
> details?
>
> Oh, here's one: http://www.rpdinc.com/pdf/N1.pdf
> claiming a typical leakage of 0.25% per day.
Indeed, I remember those from my Nuclear Physics labs in college.
Another blast from the past.
Thanks,
- Win
Boris Mohar
There was only one part to clean was large glass feed through that
supported the collection electrode. The chambers were polished nickel and
were cleaned with deionized water and alcohol. We actually had very
little problems with leakage once the electormeter was encapsulated in
wax. The only insulating part exposed to humidity was the outer surface
of the glass feed through. The feed through support ring was grounded.
Do you know of any good subs for 3N155A?
Mike Monett
> On Sat, 04 Jan 2003 22:52:35 -0800, Mike Monett
> <z06hg...@sneakemail.com> wrote:
[...]
> >Boris, can you tell how you cleaned the chambers to achieve low enough
> >leakage to measure 1fA?
> >
> >Mike
>
> There was only one part to clean was large glass feed through that
> supported the collection electrode. The chambers were polished nickel and
> were cleaned with deionized water and alcohol.
Anything special about the alcohol? Some have mentioned that plain
store-bought isoprop is unsuitable, but I've cleaned keyboard switches
with it and they measured over 1 gigohm, which is the top range on the
HP3456A.
>We actually had very
> little problems with leakage once the electormeter was encapsulated in
> wax.
You mentioned this before. Where do you get it, and how do you ensure it
is not contaminated? Any there special precautions in applying it, like
disposable rubber gloves?
>The only insulating part exposed to humidity was the outer surface
> of the glass feed through. The feed through support ring was grounded.
> Do you know of any good subs for 3N155A?
That was my next question - what input device do you use. I can't even
find a datasheet on it, but perhaps the 2N7000 that Win and John are
playing with might be suitable.
> Regards,
>
> Boris Mohar
> Viatrack Printed Circuit Designs
Thanks for the good info!
Mike
Boris Mohar
<z06hg...@sneakemail.com> wrote:
>Boris Mohar wrote:
>
>> On Sat, 04 Jan 2003 22:52:35 -0800, Mike Monett
>> <z06hg...@sneakemail.com> wrote:
>
>[...]
>
>> >Boris, can you tell how you cleaned the chambers to achieve low enough
>> >leakage to measure 1fA?
>> >
>> >Mike
>>
>> There was only one part to clean was large glass feed through that
>> supported the collection electrode. The chambers were polished nickel and
>> were cleaned with deionized water and alcohol.
>
>Anything special about the alcohol? Some have mentioned that plain
>store-bought isoprop is unsuitable, but I've cleaned keyboard switches
>with it and they measured over 1 gigohm, which is the top range on the
>HP3456A.
Use as close as you can get to100% isopropyl. Rubbing alcohol may have
oils added to it.
>>We actually had very
>> little problems with leakage once the electormeter was encapsulated in
>> wax.
>
>You mentioned this before. Where do you get it, and how do you ensure it
>is not contaminated? Any there special precautions in applying it, like
>disposable rubber gloves?
Dilmont Wax
Product Dilwax#35 tel 514 272-5714 It has been a while. I do not know
if the number or the company are still valid.
I do not think that minor contamination is a problem as wax encapsulates
the contaminants. I used to solder right through it when doing tests.
You might want to try some soft cross country ski wax. Paraffin is to
stiff.
>>The only insulating part exposed to humidity was the outer surface
>> of the glass feed through. The feed through support ring was grounded.
>> Do you know of any good subs for 3N155A?
>
>That was my next question - what input device do you use. I can't even
>find a datasheet on it, but perhaps the 2N7000 that Win and John are
>playing with might be suitable.
Low leakage, low threshold matched pair fets are the holy grail for
this kind of stuff. 3N155A is obsolete and the reason that you cannot find
it is that Motorola or whatever thy call themselves these days is that
they have a policy of obliterating information on any component that they
obsolete. I have not even been able to fine their obsolete list.
The data is in their Small Signa Transistors, Fets and Diodes book.
It is a P channel enhancement mode FET with Rgs of 1x10E16 ohms and
Vgs(th) of about -2.5V
Mike Monett
[... good info]
> Regards,
>
> Boris Mohar
> Viatrack Printed Circuit Designs
Thanks for the excellent info. If you find a sub for the 3N155A please
let us know.
Mike
Boris Mohar
3N163 from Siliconix. The specify the IGSS leakage of <-1pA at VGS of
-40V. They were also used in same instrument that measured 1fA and up.
I would still like to find something better.
Mike Monett
Do you have to select devices to get low enough leakage?
Have you thought about the LMC662: 2fA typ, 1.3uV/C tempco? This would
raise the minimum current that can be measured, but the drift with
temperature is attractive for long-term meaurements.
How important is the low-fA region anyway? Are there many instances where
being able to measure it directly really matters?
Mike
John Woodgate
<z06hg...@sneakemail.com> wrote (in <3E184B...@sneakemail.com>)
about 'Low leakage parts', on Sun, 5 Jan 2003:
>Anything special about the alcohol? Some have mentioned that plain
>store-bought isoprop is unsuitable, but I've cleaned keyboard switches
>with it and they measured over 1 gigohm, which is the top range on the
>HP3456A.
Measure it after a day or so in a humid atmosphere. Some ionic compounds
dissolve in alcohols and of course they dry out to the anhydrous form,
which is often hygroscopic. Calcium chloride is a particularly
significant example, because of its extreme affinity for water, but even
the ubiquitous sodium chloride is slightly soluble.
Of course, you may well have been lucky - up to now.
Mike Monett
>
> I read in sci.electronics.design that Mike Monett
> <z06hg...@sneakemail.com> wrote (in <3E184B...@sneakemail.com>)
> about 'Low leakage parts', on Sun, 5 Jan 2003:
>
> >Anything special about the alcohol? Some have mentioned that plain
> >store-bought isoprop is unsuitable, but I've cleaned keyboard switches
> >with it and they measured over 1 gigohm, which is the top range on the
> >HP3456A.
>
> Measure it after a day or so in a humid atmosphere. Some ionic compounds
> dissolve in alcohols and of course they dry out to the anhydrous form,
> which is often hygroscopic. Calcium chloride is a particularly
> significant example, because of its extreme affinity for water, but even
> the ubiquitous sodium chloride is slightly soluble.
>
> Of course, you may well have been lucky - up to now.
> Regards, John Woodgate
Well, the problem usually arises after a coffee spill. I have to take the
keyboard apart, unsolder the switch, then disassemble it. There are two
parallel plates separated by a spacer that looks like nylon. I was
surprised to find the initial resistance was quite low - perhaps 10k or
so.
Swishing it around in a glass with a small amount of isoprop seemed to
fix the problem, but keyboard switches are less demanding than some
applications:)
Mike
John Woodgate
about 'Low leakage parts', on Sun, 5 Jan 2003:
>surprised to find the initial resistance was quite low - perhaps 10k or
>so.
That would make me wonder what is in the coffee! 'Soluble solids of pure
coffee' should not make ionic solutions, I think. (;-)
Mike Monett
>
> I read in sci.electronics.design that Mike Monett
> <z06hg...@sneakemail.com> wrote (in <3E188A...@sneakemail.com>)
> about 'Low leakage parts', on Sun, 5 Jan 2003:
> >I was
> >surprised to find the initial resistance was quite low - perhaps 10k or
> >so.
>
> That would make me wonder what is in the coffee! 'Soluble solids of pure
> coffee' should not make ionic solutions, I think. (;-)
> Regards, John Woodgate
Only sugar and coffee made with plain tap water.
I just stuck the dvm probes in a fresh cup. It measures about 4k. A fresh
glass of cold water is about 16k. A glass of plain hot water is about 5k.
In each case, the resistance starts to drop rapidly as electrolysis
releases metal ions into the solution. I will finish my cup of coffee
knowing it contains ions of copper, lead, tin, iron, and who knows what
else. I doubt there will be any harm, but just in case, it was nice
knowing you, John:)
Mike
John Woodgate
about 'Low leakage parts', on Mon, 6 Jan 2003:
>I just stuck the dvm probes in a fresh cup. It measures about 4k.
That's not unreasonable for the hot liquid.
>A fresh
>glass of cold water is about 16k. A glass of plain hot water is about 5k.
It's the 10 k you measured on the keyboard, presumably due to a dry or
almost dry film residue, that seems odd to me.
Tom Bruhns
...
> Wow. A brief push toggles the lamp full on or full off; if you hold it
> down longer, it will reach a stable partly-on equilibrium, which is
> 'lamp dim'. That would fool a lot of people into thinking there's a
> PIC cpu inside!
>
> Oops... once you reach the 'dim' equilibrium state, are you stuck
> there? Yes!
Yep, you analyzed it correctly, except that if you put it in the "dim"
state by holding down the button, it only needs to drift off by a very
tiny amount to get into a state where short taps on the switch,
separated by say 10 second waits, rapidly get it into the full-on,
full-off state. Even with a rock-solid gate-source voltage, tiny
changes in temperature would get you into that mode. That might even
add to the illusion of a processor at work, since you get a series of
gradually increasing changes in the light intensity with the taps.
(Be really sure the switch's insulation is good!)
Cheers,
Tom
>
> John
>
> Oh, when I was a kid I had a long-wire antenna on the roof. I could
> hook it to my VTVM (with the 10M input resistor lifted) and watch the
> induced voltage as clouds passed overhead.
Probably do that these days quite nicely with one of the National
ultra-low-bias op amps! Good idea to have a spark-gap from antenna to
ground and a high value resistor (capable of not arcing over, also)
from antenna to op amp.
Tom Bruhns
> Tom's been working with, and thinking about, this low-leakage stuff
> for a *long* time. Here's an interesting thread he started back in
> July of 1999:
>
> <http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&oe=UTF-8&threadm=37A5DB4E.1DF4F7B4%40lsid.hp.com&rnum=4&prev=/groups%3Fas_q%3Dcapacitor%2520leakage%26safe%3Dimages%26ie%3DUTF-8%26oe%3DUTF-8%26as_uauthors%3DTom%2520Bruhns%26lr%3D%26hl%3Den>
>
It may be time for another measurement on those caps. They are still
salted away (well, not with much salt!) in an unused drawer of my
filing cabinet. The decay of the voltage on the polyprops has been so
gradual that I don't yet have enough info to know if the mechanism
looks like a constant resistance or a constant current, though I think
it's at least safe to say that we're past the point of worrying about
dielectric absorption and dC/dt effects. :-) The polyester (Mylar)
caps have discharged enough that I can say the curve looks more like
the constant-R exponential than like a constant current.
I'll post an update when I have a chance to make a careful
measurement. (It's possible the 'speriment will end when we move to a
different facility in about six months.)
Cheers,
Tom
Tom Bruhns
started my capacitor experiment, I biased myself with a look at a
graph in a reasonably new edition of Sams' "Reference Data for
Engineers." It shows insulation resistance (ohm-farads) for various
dielectrics commonly used in film caps. Polyprop was the best, but
the graph shows it with only an 8-day time constant at 20C. So I
expected I'd see a rather rapid decay over a couple weeks. Polyester
is shown about an order of magnitude lower resistance, so that should
have taken only a couple days to drop to a tenth the initial charge.
I was not prepared for the >10^4 factor relating what I actually
measure and the values in the graph. Also, the manufacturers' data
for the caps I'm testing matches the graph's values pretty well.
Again, I believe the attention to process details and to improved
materials has had some big benefits outside the reasons I suspect
they've been done in the first place. I doubt if very many people
were asking for caps with 100-year time constants, but I'm sure a lot
were asking for more reliable caps, and I'll bet a lot of the
improvements are the result of a constant search for better
reliability.
I wouldn't _rely_ on such performance in the products I design and put
into production, but that doesn't keep me from playing with circuits
and measurements that let me know how good things actually are. With
my hobby circuits, I don't have to please anyone but myself. :-)
Cheers,
Tom
"Phil Allison" <bi...@bigpond.com> wrote in message news:<MxKR9.16806$jM5....@newsfeeds.bigpond.com>...
Spehro Pefhany
wrote:
>I wouldn't _rely_ on such performance in the products I design and put
>into production, but that doesn't keep me from playing with circuits
>and measurements that let me know how good things actually are. With
>my hobby circuits, I don't have to please anyone but myself. :-)
One big difference I see between your results and the graphs in the
capacitor manfacturers' data sheets is that theirs are measured at
100V or 200V and yours are at relatively low voltages.
Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
sp...@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
Mike Monett
[...]
> Again, I believe the attention to process details and to improved
> materials has had some big benefits outside the reasons I suspect
> they've been done in the first place. I doubt if very many people
> were asking for caps with 100-year time constants, but I'm sure a lot
> were asking for more reliable caps, and I'll bet a lot of the
> improvements are the result of a constant search for better
> reliability.
>
> I wouldn't _rely_ on such performance in the products I design and put
> into production, but that doesn't keep me from playing with circuits
> and measurements that let me know how good things actually are. With
> my hobby circuits, I don't have to please anyone but myself. :-)
>
> Cheers,
> Tom
Tom,
As you point out, process improvements and improved reliability probably
account for the low leakage. If so, then leakage may be a good way to
select vendors for critical applications and to monitor their quality.
You probably couldn't reject a batch of caps if the leakage dropped to
the spec values, but it would alert you that something was degrading
their process.
The only problem is it takes so darned long to tell if there's a
problem:)
Mike
Phil Allison
> Indeed, Phil, I think your doubts are quite reasonable!
** I am releaved you did not think I was calling you a liar !!
When I
> started my capacitor experiment, I biased myself with a look at a
> graph in a reasonably new edition of Sams' "Reference Data for
> Engineers." It shows insulation resistance (ohm-farads) for various
> dielectrics commonly used in film caps. Polyprop was the best, but
> the graph shows it with only an 8-day time constant at 20C.
** Isn't polystyrene even better ?
So I
> expected I'd see a rather rapid decay over a couple weeks. Polyester
> is shown about an order of magnitude lower resistance, so that should
> have taken only a couple days to drop to a tenth the initial charge.
> I was not prepared for the >10^4 factor relating what I actually
> measure and the values in the graph. Also, the manufacturers' data
> for the caps I'm testing matches the graph's values pretty well.
** High voltage electros often hang on to their charge for days rather
than minutes as the data would suggest. I think they do it conscioulsy to
get revenge on techs.
> Again, I believe the attention to process details and to improved
> materials has had some big benefits outside the reasons I suspect
> they've been done in the first place. I doubt if very many people
> were asking for caps with 100-year time constants, but I'm sure a lot
> were asking for more reliable caps, and I'll bet a lot of the
> improvements are the result of a constant search for better
> reliability.
** Improvements in films have made plastic caps much smaller than a
decade ago - I suspect this is a bigger sales point than the number of
gigohms.
> I wouldn't _rely_ on such performance in the products I design and put
> into production, but that doesn't keep me from playing with circuits
> and measurements that let me know how good things actually are. With
> my hobby circuits, I don't have to please anyone but myself. :-)
* I *did* attempt to make pre-amps for condenser mics once ( 25 years
ago) and could not source the essential 1000 Mohm resistors in small
quantities so I resorted to drawing the component, zig zag and all, on the
PCB with a fountain pen filled with resistive ink.
I got my 1G resistors OK but the PCB leakage was in the same order
order of magnitude if the weather was even slighly damp.
Any clear coating I applied over the area to solve this ate the
resistor. Very frustrating.
.................. Phil
Tom Bruhns
...
> Have you thought about the LMC662: 2fA typ, 1.3uV/C tempco? This would
> raise the minimum current that can be measured, but the drift with
> temperature is attractive for long-term meaurements.
An interesting experiment is to plot the input bias current versus
common mode voltage. You may find that it passes through zero at some
voltage. It did with the old CA3130 op amps. If that happens, and
it's stable enough, you can get to exceedingly low input bias current.
It may even be possible to bootstrap the power supplies to keep it
low over a moderate range of input voltages.
What was the best one could do with "electrometer" vacuum tubes? What
are the current best practices for low input bias current amplifiers?
I assume there are ways to get well below 1fA, but with what sort of
bandwidth? (How few electrons can we reliably sense electrically?)
1fA*1sec at 1pF is only 1mV, but even 0.01fA for a long enough time
(an hour, or a day) on a 1pF cap becomes easy to measure. If you
connected that LMC662C as a follower, leaving the NI input floating,
it would eventually drift to the zero bias current point (if such
point is within the input common mode range) but it would take a while
to get there, at probably less than 1mV/second slew (2fA and probably
a bit more than 2pF input capacitance). ... Maybe some of the
lurkers are doing work in the very-low-current area and can offer some
insights. For me, it's just an area of casual interest.
Cheers,
Tom
John Woodgate
wrote (in <CkoS9.18167$jM5....@newsfeeds.bigpond.com>) about 'Low
leakage parts', on Tue, 7 Jan 2003:
>I *did* attempt to make pre-amps for condenser mics once ( 25 years
>ago) and could not source the essential 1000 Mohm resistors in small
>quantities so I resorted to drawing the component, zig zag and all, on the
>PCB with a fountain pen filled with resistive ink.
>
> I got my 1G resistors OK but the PCB leakage was in the same order
>order of magnitude if the weather was even slighly damp.
>
> Any clear coating I applied over the area to solve this ate the
>resistor. Very frustrating.
Did you try a soft pencil? Very soft pencils were used to make 1 to 2
Mohm grid leaks for valves/tubes in the 1920s. The solvent in the
lacquer wouldn't affect the graphite.
Were you able to measure the resistors? I still have a 'Twenty Million
Megohmmeter' from the 1960s. It has a huge mirror-scale meter, far too
impressive to throw away, but it doesn't get enough use to keep it
working. Lots of EF37/6J7G.
Phil Allison
"John Woodgate" <j...@jmwa.demon.contraspam.yuk> wrote in message
news:OrTmMPAB...@jmwa.demon.co.uk...
Phil wrote :
>
> >I *did* attempt to make pre-amps for condenser mics once ( 25 years
> >ago) and could not source the essential 1000 Mohm resistors in small
> >quantities so I resorted to drawing the component, zig zag and all, on
the
> >PCB with a fountain pen filled with resistive ink.
> >
> > I got my 1G resistors OK but the PCB leakage was in the same order
> >order of magnitude if the weather was even slighly damp.
> >
> > Any clear coating I applied over the area to solve this ate the
> >resistor. Very frustrating.
>
> Did you try a soft pencil? Very soft pencils were used to make 1 to 2
> Mohm grid leaks for valves/tubes in the 1920s. The solvent in the
> lacquer wouldn't affect the graphite.
** The drawing ink I used ( Penguin ...) was suggested for coating plastic
to make one's own ESLs.
>
> Were you able to measure the resistors?
** Using a 30 volt supply and a DVM's 10 Mohm input Z as nanoamp meter -
yes.
> I still have a 'Twenty Million Megohmmeter' from the 1960s. It has a huge
mirror-scale meter, far too> impressive to throw away, but it doesn't get
enough use to keep it
> working. Lots of EF37/6J7G.
** I used to own a Sanwa 501 - 2 uA FSD ( 500 kohm/volt) taut band
multimeter.
It fell off the bench - very sad.
................. Phil
Boris Mohar
<z06hg...@sneakemail.com> wrote:
>Boris Mohar wrote:
>>
>> On Sun, 05 Jan 2003 12:28:28 -0500, Mike Monett
>> <z06hg...@sneakemail.com> wrote:
>>
>> >Boris Mohar wrote:
>> >
>> >[... good info]
>> >
>> >> Regards,
>> >>
>> >> Boris Mohar
>> >> Viatrack Printed Circuit Designs
>> >Thanks for the excellent info. If you find a sub for the 3N155A please
>> >let us know.
>> >
>> >Mike
>>
>> 3N163 from Siliconix. The specify the IGSS leakage of <-1pA at VGS of
>> -40V. They were also used in same instrument that measured 1fA and up.
>>
>> I would still like to find something better.
>>
>> Regards,
>>
>> Boris Mohar
>> Viatrack Printed Circuit Designs
>
>Do you have to select devices to get low enough leakage?
Not only select but select for matched pairs
>
>Have you thought about the LMC662: 2fA typ, 1.3uV/C tempco? This would
>raise the minimum current that can be measured, but the drift with
>temperature is attractive for long-term meaurements.
>
>How important is the low-fA region anyway? Are there many instances where
>being able to measure it directly really matters?
This was for measuring Tritium in air. It was important that once the
instrument was zeroed that it stayed zeroed. Speaking of zeros, to my
dismay I discovered that some high value resistors have a non linear
voltage coefficient. At very low voltage drops the value of one Tohm
resistor start to go through the roof. So just as you zero your
electrometer your feedback resistor value increases making it even harder
to zero.
fred bartoli
Boris Mohar <bor...@sympatico.ca> a écrit dans le message :
19ll1vs5a3p3saum6...@4ax.com...
> On Sun, 05 Jan 2003 13:58:38 -0500, Mike Monett
> <z06hg...@sneakemail.com> wrote:
>
> >Boris Mohar wrote:
> >>
> >> On Sun, 05 Jan 2003 12:28:28 -0500, Mike Monett
> >> <z06hg...@sneakemail.com> wrote:
> >>
> >> >Boris Mohar wrote:
> >> >
<SNIP> still [... good info]
> Speaking of zeros, to my
> dismay I discovered that some high value resistors have a non linear
> voltage coefficient. At very low voltage drops the value of one Tohm
> resistor start to go through the roof. So just as you zero your
> electrometer your feedback resistor value increases making it even harder
> to zero.
>
Interesting.
What kind of physical phenomenon could be at work here.
Can you tell us more / points to some URL ?
Thanks,
Fred.
Roy McCammon
> Phil, your doubt are NOT reasonable, quite the contrary.
Win, I think that his doubts are reasonable. I would
have those same doubts except that I hold you and Tom
to have very high credibility.
Of course, my opinions were formed circa 25 years ago;
things must have gotten better.
I do think his allegations of net abuse are unfounded,
but the very short (even one word) responses did come
across as a little bit antagonistic.
-Roy
--
Thank you for reading and or replying
If you are one in a million, there are 6000 people just like
you.
Local optimization almost never yields global optimization.
Opinions expressed here are my own and may not represent those of my employer.
Boris Mohar
wrote:
I could take a lame guess and propose that quantum physics could provide
an answer. No other info available at the moment.
--
Mike Monett
[...]
> This was for measuring Tritium in air. It was important that once the
> instrument was zeroed that it stayed zeroed. Speaking of zeros, to my
> dismay I discovered that some high value resistors have a non linear
> voltage coefficient. At very low voltage drops the value of one Tohm
> resistor start to go through the roof. So just as you zero your
> electrometer your feedback resistor value increases making it even harder
> to zero.
>
> Regards,
>
> Boris Mohar
> Viatrack Printed Circuit Designs
Amazing. How do you measure the value of a 1T resistor at low voltages?
Mike
Tom Bruhns
> ** Isn't polystyrene even better ?
(than Polypropylene)...Well, not according to the graphs I have. But
since the graphs are so understated, the reality may be something
different. I do have some 0.01uF polystyrenes at home, but they are a
big no-no for reliability and have been "outlawed" in our production
for many years. Teflon would be another one to have a look at, as
would fused quartz (though we probably won't find any 0.1uF fused
quartz caps kicking around). But my goal in my simple experiment was
to see what the caps we actually use looked like, not necessarily to
find the very best (lowest-leakage) possible cap.
...
> * I *did* attempt to make pre-amps for condenser mics once ( 25 years
> ago) and could not source the essential 1000 Mohm resistors in small
> quantities so I resorted to drawing the component, zig zag and all, on the
> PCB with a fountain pen filled with resistive ink.
>
> I got my 1G resistors OK but the PCB leakage was in the same order
> order of magnitude if the weather was even slighly damp.
>
> Any clear coating I applied over the area to solve this ate the
> resistor. Very frustrating.
I have a small catalog around here somewhere that lists high-value SMT
resistors; it was from at least 10 years ago (may have tossed it by
now!) Anyway, http://www.irctt.com/pdf_files/HR.pdf shows values in
0805 size up to 50 gigohms.
My experience with modern FR4 pc material, at least from the vendors
we use, is that it's very low conductance IF it's been properly
handled...run through a controlled no-clean process or perhaps
thoroughly cleaned after processing. When a very high resistance
between two nodes is required, it's possible to slot the board; you
may then have to guard around one or more of the separated nodes as
well. (It can be quite a challenge to make the guard track the node
you're guarding closely enough to do the job you want. With 1E10 ohms
resistance, 10uV gets you 1fA of leakage.) Some of the boards we use
have a little "diving board" area, a peninsula if you will, with a
slot routed on three sides. With light weight, small components, it's
possible to make an area like that which won't give too much trouble
in vibration. Of course, as Win notes, you generally don't need to
isolate very many nodes, and a single teflon standoff or feedthrough
may do the trick. They just don't fit well into 100% SMT processes.
Cheers,
Tom
Tom Bruhns
> On 6 Jan 2003 12:46:17 -0800, the renowned k7...@aol.com (Tom Bruhns)
> wrote:
>
> >I wouldn't _rely_ on such performance in the products I design and put
> >into production, but that doesn't keep me from playing with circuits
> >and measurements that let me know how good things actually are. With
> >my hobby circuits, I don't have to please anyone but myself. :-)
>
> One big difference I see between your results and the graphs in the
> capacitor manfacturers' data sheets is that theirs are measured at
> 100V or 200V and yours are at relatively low voltages.
Yes, you're absolutely right Spehro. That's something I've thought
about, but I've been too lazy to do anything about it. That is, my
HP3456A DVM has an input bias current around 1pA, and very high input
resistance, IF it's running on a range up to 11V full scale. So if I
connect it to the 0.1uF capacitor for 5 seconds to make a
(triggered-and-held) measurement, it will induce a change of voltage
that's small enough to be neglected, so long as I only measure the
caps every few months. I don't have a tool that will let me measure
100+V so nicely. I'm sure I could come up with something, but it's
just not quite a high enough priority. On the other hand, it's
generally at 10V and below that I use the caps anyway, so my testing
matches how I use them. Maybe someone else would like to try at high
voltage?? I'd welcome the results!
Cheers,
Tom
John Larkin
Tom,
Thank you for doing a very interesting long-term experiment, and
sharing your results with us. I never doubted your numbers... why
should I?
I suggest that if anybody who *does* doubt experimental data posted
here, they should repeat the experiment themselves and post *their*
numbers.
Oh, did you precharge the input of your DVM before connecting to the
caps? Some meters have a high-z input but need a lot of charge.
John
Spehro Pefhany
wrote:
>I have a small catalog around here somewhere that lists high-value SMT
>resistors; it was from at least 10 years ago (may have tossed it by
>now!) Anyway, http://www.irctt.com/pdf_files/HR.pdf shows values in
>0805 size up to 50 gigohms.
Ohmcraft makes 0603s up to 300G and 2512s up to 2T ohm. The latter
is rated at 2W. ;-)
Mike Monett
> Mike Monett <z06hg...@sneakemail.com> wrote in message
news:<3E1880...@sneakemail.com>...
> ...
>> Have you thought about the LMC662: 2fA typ, 1.3uV/C tempco? This
>> would raise the minimum current that can be measured, but the
>> drift with temperature is attractive for long-term meaurements.
> An interesting experiment is to plot the input bias current versus
> common mode voltage. You may find that it passes through zero at
> some voltage. It did with the old CA3130 op amps. If that happens,
> and it's stable enough, you can get to exceedingly low input bias
> current. It may even be possible to bootstrap the power supplies
> to keep it low over a moderate range of input voltages.
That would be a nice trick!
> What was the best one could do with "electrometer" vacuum tubes?
It must have been pretty good. I recall a class at MIT in 1967 where
Prof. Searle mentioned measuring the diode IV curve down to
attoamps. While everyone sat stunned at the concept, another
professor laughed and said this was less than one electron per
second. So they were in the right ballpark. I assume they used
vacuum tubes for the measurement.
> What are the current best practices for low input bias current
> amplifiers?
The Keithley model 6430 claims 10aA resolution on the 1pA range
using a 5 second RC filter. The voltage burden is < 1mV.
http://www.keithley.com/kei_assets/downloads/5544.PDF
Their graph shows the noise at 200aA p-p.
http://www.keithley.com/kei_assets/downloads/386.PDF
> I assume there are ways to get well below 1fA, but with what sort
> of bandwidth? (How few electrons can we reliably sense
> electrically?) 1fA*1sec at 1pF is only 1mV, but even 0.01fA for a
> long enough time (an hour, or a day) on a 1pF cap becomes easy to
> measure.
An exciting technology is the single electron transistor. NIST
claims a sensitivity of 10-5 e in a 1 Hz bandwidth:
http://emtech.boulder.nist.gov/div814/pubs/downloads/set/KellerFermi.pdf
Using an RF-SET, Schoelkop claims a bandwidth of 100MHz:
http://www.eng.yale.edu/rslab/papers\RFSETScience.pdf
C. P. Heij shows an inverter operating at 25 mK, and why room
temperature logic devices will be difficult to make. He shows a
curve of gain vs temperature up to 200mK:
http://qt.tn.tudelft.nl/~hadley/publications/inverter/inverter.html
These are cryogenic devices, but many companies are working on room
temperature versions. Matsumoto shows clear Coulomb staircase curves
at 300K:
http://luciano.stanford.edu/~shimbo/set.html
It will be a while until we can run down to Radio Shack and buy one,
but they seem quite promising. Just think about the RFI/EMI
problems!
> If you connected that LMC662C as a follower, leaving the NI input
> floating, it would eventually drift to the zero bias current point
> (if such point is within the input common mode range) but it would
> take a while to get there, at probably less than 1mV/second slew
> (2fA and probably a bit more than 2pF input capacitance). ...
> Maybe some of the lurkers are doing work in the very-low-current
> area and can offer some insights. For me, it's just an area of
> casual interest.
> Cheers, Tom
That is a great idea! I hope someone tries it and posts the results.
Mike
Phil Allison
> I have a small catalog around here somewhere that lists high-value SMT
> resistors; it was from at least 10 years ago (may have tossed it by
> now!) Anyway, http://www.irctt.com/pdf_files/HR.pdf shows values in
> 0805 size up to 50 gigohms.
** Gohm resistors are easy to get now - just not in 1978 when I had to
bug to make mic pre-amps. BTW I have a couple of old 1 G ohm resistors
fitted inside sealed glass tubes - seems the way to go to me.
Commercial fet pre-amps for condenser mics all seem to use tiny
carbon comp 1 G ohm resistors and the PCBs are normally coated in the
sensitive area to keep out moisture - essential for a hand held mic.
................. Phil
John Woodgate
wrote (in <19ll1vs5a3p3saum6...@4ax.com>) about 'Low
leakage parts', on Tue, 7 Jan 2003:
>At very low voltage drops the value of one Tohm
>resistor start to go through the roof.
1 Tohm is already on the roof; it must climb up the lighting conductor
to get higher.(;-)
I suppose the effect is caused by an interface layer developing where
the lead wires contact the resistive element.
John Woodgate
leakage parts', on Tue, 7 Jan 2003:
>multimeter.
>
> It fell off the bench - very sad.
That's down in the Cambridge Instruments mirror galvanometer range.
Those are repairable (by experts) but I don't know about the Sanwa, of
course.
It's astonishing to see (or it was; they may be all dead now) one of
these guys re-shape the hairspring of a Model 8 AVO by pulling it
between a thumb and finger-nail.
They used to hand-calibrate the Model 7s individually with a mapping pen
and Indian ink.
Boris Mohar
By using a better electrometer, Keithley 614 as a voltage source and a
current meter. It can measure 1x10-14A. Better than I ever could.
--
John Woodgate
<sp...@interlog.com> wrote (in <ov5m1v8dmolnpfq09vfq5oma8a6o469mre@4ax.c
om>) about 'Low leakage parts', on Tue, 7 Jan 2003:
>Ohmcraft makes 0603s up to 300G and 2512s up to 2T ohm. The latter
>is rated at 2W. ;-)
How do they test them? (;-)
Tom Bruhns
> Tom,
>
> Thank you for doing a very interesting long-term experiment, and
> sharing your results with us. I never doubted your numbers... why
> should I?
>
> I suggest that if anybody who *does* doubt experimental data posted
> here, they should repeat the experiment themselves and post *their*
> numbers.
>
> Oh, did you precharge the input of your DVM before connecting to the
> caps? Some meters have a high-z input but need a lot of charge.
:-) I'm happy that anyone out there is noticing it and finding it
interesting, as I do.
And yes, as a matter of fact, I have been pre-charging the DVM input.
Based on the input bias current I measured and the dV/dt I see after I
precharge, I'd say there's not very much capacitance there, but it was
easy enough to do and I figured it was worthwhile. I just use
(approx.) the value predicted by extrapolating the measured
performance up till the present.
I actually have hoped that someone else would independently make
similar measurements. It would be nice to know if other brands and
types (e.g. metalized versus foil) of caps show similar performance,
or are significantly different. If they are different, it would be
fun to try to understand why.
Cheers,
Tom
Mike Monett
>
> On Tue, 07 Jan 2003 12:11:56 -0500, Mike Monett
> <z06hg...@sneakemail.com> wrote:
[...]
> >Amazing. How do you measure the value of a 1T resistor at low voltages?
> >
> >Mike
>
> By using a better electrometer, Keithley 614 as a voltage source and a
> current meter. It can measure 1x10-14A. Better than I ever could.
> Regards,
>
> Boris Mohar
> Viatrack Printed Circuit Designs
So the resistance must start increasing at some voltage greater than
1e12 * 1e-14 = 0.01V?
Thanks Boris. Your information is really helpful.
Mike