Just do not forget the capacitive loads.
At 10 kHz the impedance of 1 pF is ~0.0159 G ohm.
So try a (hf-) mosfet with low input capacitance.
4K is probably to low for a mosfet, but some cooling should lower
You can't usefully use a 1 MHz bandwidth with a femtoamp of current, at
least not if it has full shot noise, which I imagine it does. If your
current consists of N electrons per second, counting statistics predict
that your SNR will drop to 0 dB in a measurement bandwidth of N/2 Hz. A
femtoamp is only 6200 electrons/s, so assuming you want at least a 20 dB
SNR (because below that you haven't got a measurement really), your
bandwidth is going to be limited to 31 Hz.
Sorry about that.
You have two problems with basic physics:
To get a 3dB corner frequency of 1MHz you need to keep your total
capacitance down to about 160 atto farads, assuming I got my prefixes
right. That implies a capacitor that's about 0.02mm on a side, with air
You _may_ be able to get there from here with some custom IC operating
at 4K, but I wouldn't know.
I would suggest that you need some other, more reliable, way of
amplifying small currents. Hopefully someone will jump in with
suggestions. All I can think of is that if you can get your voltages up
high enough you may be able to do something with ionization, like a
One fA implies that you're flowing about 62400 electrons per second.
Even at 10kHz the shot noise is going to be enormous, and at a 1MHz
bandwidth you'll be seeing the electrons as individual events, not as
anything resembling a continuous current.
Wescott Design Services
Posting from Google? See http://cfaj.freeshell.org/google/
"Applied Control Theory for Embedded Systems" came out in April.
See details at http://www.wescottdesign.com/actfes/actfes.html
Tricky. 1 fA integrated over 500 ns (1/2 cycle at 1 MHz) is 0.003
electrons. You're going to need some serious signal averaging.
And a charge amp with very low input current, probably a cold jfet.
What's the physics?
Damn. 6240, per Phil Hobbs. Apparently I was visiting a universe where
6.24*10^3 = 62400, but I'm back now.
Heisenberg's uncertainty rules, at least in principle.
Bill Sloman, Nijmegen
>Now i see the problems - and am not that happy when i started this
>I was planning to use GaAs MESFETs with the 1G resistor at gate bias.
Mesfets are nasty, leaky, noisy, and have very strange changes of gain
at low (KHz) frequencies, some sort of trapping-state thing. Jfets are
often used cold, and work at liquid helium temps. The ultimate charge
amp uses pure capacitive feedback to eliminate resistor noise, but has
to be reset now and then to cancel offset drift buildup.
>However, now it seems it won't work as i thought.
>Here i am trying to detect a motion of a single ion in a penning trap -
>people use tuned circuits to pick up this ~50 fA image current. But i
>want to build something for broadband detection (upto 1 MHz).
That may not be possible... there's just not enough signal. Signal
averaging (lock-in technique, done in hardware or software) would
work, but that's just equivalent to reducing the bandwidth. At least
it's not fixed-tuned.
How do you get the signal out of the ion... tiny antennas? Optical?
Non-trivial, for sure.
My reading of the OPs comments doesn't rule out a narrow bandwidth around
some carrier in the range he suggested. If this is the case, this won't
be his biggest problem.
Figuring out how to amplify a signal that small accurately will be trouble
kens...@rahul.net forging knowledge
I will have you know, that *he* was NOT the principal of my school!
What kind of source ?
As others have noted, you're going to have a heck of a time with noise
if your signal is just 6200 electrons per second. Plus seeing a 1MHz
signal is going to require a whole lot of auto-correlation and
Do you have a known frequency and bandwidth to look for? I suspect
seeing that weak a signal is going to require like many many seconds of
observation and a very narrow bandwidth, like 1Hz or so.
I'd suggest forgetting about measuring the voltage, and instead measure
the current directly with a transimpedance circuit. You're getting so
few electrons it's a shame wasting them heating up a 1Gohm resistor. :)
Yeah, by watching oil drops in a microscope and changing the voltage
across two capacitor plates to levitate them. Doing that even 6200
times per second is a good trick, besides the fact that M. was wasting
almost all the oil drops while concentrating on one at a time.
> So for 1000 electrons, theoretically that could be done in 10mSec or a
> rough bandwidth of 100Hz.
> Following that to 1MHz, one would need (crudely now) a flow of 10,000
> But do not use a high value resistor, as displacement currents will
> kill what it would "see"; if you insist, then look into measuring the
> displacement current or the voltage it generates on a known capacitor.
> Say, use a huge 1.00 pF capacitor....
> Or, try to be more nasty and get them electrons to "jump" thru a
> chamber with a view port, at one electron per oil drop (or other
> insulating liquid) - and *count* those or determine when they pass by
> (132 this time interval, 45 next one, etc...).
If the current starts out in a wire, i.e. it comes from displacement
current due to ion motion, the OP going to have to use bandwidth
narrowing of some sort. Signal averaging is generally much better than
using a narrow bandwidth near DC, because you can get out of the 1/f
noise and the drift pretty well. Tuned circuits are not a stupid idea
at all, because you can set the fields in the trap up to get a
particular cyclotron resonance frequency--i.e. you can tune the signal
to the filter. (Superhets do the same thing.)
If the signal starts out as free electrons in a vacuum, then the
situation is much better--a Channeltron or other electron multiplier is
the way to go. It's trivially easy to put 140 dB of gain on those
pulses, at which point you can probably even detect them with a neon bulb.
Thanks for the description. Interesting!
I understand we have 1 nV or less at 20 pF. I believe this is possible to
detect provided data acquisition time is longer than life time of the ion.
It is not a problem with Heisenberg, IMHO.
A capacitance of 20 pF is a lot. The signal would increase if C could be
reduced. If this is not possible maybe one could raise the impedance by
means of an inductor to form a tuned circuit at the amplifier input.
I hope someone can give advice on the best choice of 4 Kelvin charge
amplifier. John Larkin suggested a cold jfet. I guess that is a good
suggestion. Some semiconductors do not work at 4K as minority carriers are
not generated thermally as they are at room temp.
> detect provided data acquisition time is longer than life time of the ion.
Sorry, I mean
"I believe this is possible to detect provided the NEEDED data acquisition
time is NOT longer than life time of the ion."
I saw Werner Heisenberg once. The closer he came, the more uncertain I was
whether it was really him.
> Sorry, I mean
> "I believe this is possible to detect provided the NEEDED data acquisition
> time is NOT longer than life time of the ion.":
This can be assured by detecting for along time, approx. 1 sec - the
digitizer we have goes at 1 MHz sampling rate - this makes 1 million
data points to store, not bad at all.
Also, i am looking use the transimpedance amplifier with a resistor in
feedback, described by Bob Pease in
This might help me to put a high value resistor in feedback without
compromising the bandwidth.
>I have to do broadband detection -can't go with tuned circuits.
>John do you know of any commercial cold JFETs which won't have
>freeze-out at 4 kelvin? One which i can buy?
It's my unofficial understanding that regular silicon jfets work cold.
I'm sure google has refs, and the Review of Scientific Instruments is
rife with stuff like that.
What do you estimate as the spindown time constant of a trapped ion? I
guess you'd want minimal resistive losses in the amp to reduce damping
of the orbit.
Really low-level charge amps use pure capacitive feedback; see Knoll's
book on radiation detectors. Or how about a jfet amp with no feedback
at all, just the floating gate on your pickup electrode?
> Really low-level charge amps use pure capacitive feedback; see Knoll's
> book on radiation detectors. Or how about a jfet amp with no feedback
> at all, just the floating gate on your pickup electrode?
How would the FET will be biased in this case? In my system there are 2
detection electrodes for which i am planning some sort of differential
I'm not well versed in this stuff, so please, guys, be gentle in your
flames if I'm completely out of order.
After following all the good advice on your analog signal conditioning,
it's my impression from this thread that you'll still be pretty far down
in the noise.
If that's not correct, you could simply acquire broadband data at, say,
5Mhz for several seconds and run an FFT on it. This would give you at
one shot all the ion species present in the experiment.
If you are indeed buried in noise, you could run correlations on the
same acquired data with offsets equal to the period of each expected ion
species, which should dig any ion signals out of a lot of noise.
>> It's my unofficial understanding that regular silicon jfets work cold.
>> I'm sure google has refs, and the Review of Scientific Instruments is
>> rife with stuff like that.
>I have found some Ge JFETs from NASA, will try to get some. For the
>time being i am thinking to use the GaAs FETs i have and measure their
>leekage current. I know it has to decrease when i cool them to 4
As I mentioned, GaAs fets are nasty dudes. The defect densities and
trapping states are awful. I'd excpect the Ge's to be leakier than
>> What do you estimate as the spindown time constant of a trapped ion? I
>> guess you'd want minimal resistive losses in the amp to reduce damping
>> of the orbit.
>I have to check the numbers for that, if i recall correct the lmiting
>factor which causes the ion orbit to decay - is collisions with other
>gas molecules not due to the resistive decay of the preamp. The
>detection duration is typically in the order of couple of seconds,
Cool, pretty high Q once you get up to a MHz. Even so, a resistor will
add Johnson noise.
>> Really low-level charge amps use pure capacitive feedback; see Knoll's
>> book on radiation detectors. Or how about a jfet amp with no feedback
>> at all, just the floating gate on your pickup electrode?
>How would the FET will be biased in this case? In my system there are 2
>detection electrodes for which i am planning some sort of differential
Just use it common-source with floating gate. Gate voltage will be
close to zero, so drain current will be close to Idss, gain will be
high, and you'll boil a little bit of helium. I did an unbiased jfet
infrasonic microphone preamp once, using a GR ceramic mic element, for
listening to static tests of the Saturn V main engines, with the mics
scattered around southern Mississippi. They accused my amp design of
intermittent motorboating but eventually figured out they were picking
up the subsonic mating calls of alligators.
Who is that specialty fet company... Interfet? I think they may have
some cryo-rated jfets.
I am thinking he would be doing dam well to squeek out 100KHz
>>I have found some Ge JFETs from NASA, will try to get some. For the
>>time being i am thinking to use the GaAs FETs i have and measure their
>>leekage current. I know it has to decrease when i cool them to 4
>As I mentioned, GaAs fets are nasty dudes. The defect densities and
>trapping states are awful. I'd excpect the Ge's to be leakier than
However, GaAs works down to 4K. Si doesn't. Si JFETs perform beautifully
at 77K though.
>Just use it common-source with floating gate. Gate voltage will be
>close to zero, so drain current will be close to Idss, gain will be
>high, and you'll boil a little bit of helium.
Make that nitrogen.
>Who is that specialty fet company... Interfet? I think they may have
>some cryo-rated jfets.
Yeah, but not for LHe temperatures. I once did some literature research on
the subject, and the upshot is: Depletion-mode Si JFETs don't work at 4K
due to carrier freeze-out. Enhancement-mode Si MOSFETs work if you're
lucky. GaAs works in general.
In my drawer I have a dual JFET that is rated at 4K. It is actually
mounted on a thermally insulating stud inside a TO78 metal can package
together with a small heating resistor. It was bought 15 years ago for
$500 from a company that doesn't exist any more. Talk about "Not
recommended for new designs" ;-)
Thanks for the update. The application would benefit from minimum
capacitance, so plumbing the signal out to 77K would be an issue.
Maybe a driven guard back from a jfet amp at LN temperature would
One approach would be to attach the JFET with high thermal resistance
leads, such as fine brass wire--the thermal resistance is very high at
low temperature, so you might be able to run a normal JFET at 77K in a
4K environment without adding a significantly increased heat load. You
can set the dissipation remotely by changing VDD.
The thermal conductivity integral of stainless steel at 77K is about
200W/m, so if you could stand 50 mW dissipation, you could get 77K with
three 1-mm stainless steel wires, 1 cm long--with lots of room to reduce
these numbers if 50 mW is too much. You can't go too low, of course,
since radiation will eventually dominate.
Of course, you'd either have to apply the heat externally, or make sure
the JFET was turned on before you transfer the helium.
People use fine manganin wire for this, too. And somebody makes fine
How about a silicon diode as a temperature sensor, and a resistor
heater, with the control circuit at 77K or room temp? Adds 2 more
wires. Below about 20K, a silicon diode's voltage drop gets huge, and
gives tons of signal vs temperature around 4K, typically biased at 10
Or maybe just monitor the jfet current as the temperature indication,
and drive the heater resistor as needed. Wow, that's a new jfet bias
scheme! Who is it that claims there are no new circuits?
I guess there's still the fundamental problem... is there enough
signal? If the thing is to be untuned, I guess it'll need an FFT to
find the orbital frequency. How long does it take to do a 2M point FFT
I'd expect they do. If you show up with a bag full of money wanting
special testing of an existing part, they are happy to do it for you.
So how you'll be able to get a bandwith of 10Khz to 1Mhz with such a
National Institute for Isotopic and Molecular Technology,
I like Phil's idea, to run the JFETs at around 70 K in 4K environment.
i have Phosphor bronze wire which i w'ld use.
The Thermal conductivity of this wire is merely 1.6 W/m.K around 4
> The thermal conductivity integral of stainless steel at 77K is about
> 200W/m, so if you could stand 50 mW dissipation, you could get 77K with
> three 1-mm stainless steel wires, 1 cm long--with lots of room to reduce
> these numbers if 50 mW is too much. You can't go too low, of course,
> since radiation will eventually dominate.
> Of course, you'd either have to apply the heat externally, or make sure
> the JFET was turned on before you transfer the helium.
I have to heat it externally as the amplifier will be inside the
cryostat well before we start to detect. For this as John said we can
use a resisitive heater and monitor the JFET current.
Doing 2M point FFT is not a big deal which yes is done to find the
frequency of the ion. Its typically done on the fly, you won't even
notice on a good PC.
Make sure you're using the thermal conductivity integral, and not just
the value at 4K times the temperature drop--you'll significantly
underestimate the thermal loading otherwise. Flip to the second plot in
that Lakeshore app note and make sure you understand the distinction.
Yeow, is that an understatement. All semiconductors not specially designed
to operate below liquid Nitrogen temperatures (about 77K) suffer from
"freeze-out" and do not work at all.
Look at devices called "charge amplifiers" to integrate very small charges
Thanks for the reminder, Coulombs number is something like 6.245x10^18;
below picoamperes you are approaching counting basic charge units.
> Yeow, is that an understatement. All semiconductors not specially
> designed to operate below liquid Nitrogen temperatures (about 77K)
> suffer from "freeze-out" and do not work at all.
> Look at devices called "charge amplifiers" to integrate very small
> charges <10uColumbs.
Very interesting, joseph. Here is some more information that might be
...the lower temperature limit is typically determined by the
ionization energy of the dopants. Dopants usually require some
energy to ionize and produce carriers in the semiconductor. This
energy is usually thermal, and if the temperature is too low, the
dopants will not be sufficiently ionized and there will be
insufficient carriers. The result is a condition called
"freeze-out." For example, Si (dopant ionization energy ~0.05 eV)
freezes out at about 40 K and Ge (ionization energy ~0.01 eV) at
about 20 K. Thus, for example, Ge devices in general operate to
lower temperatures than Si devices.
The various effects described above can be illustrated in a graph
such as the one below (the shape of the curves should not be taken
literally, only as an indication of trends). Ordinarily, the usable
temperature range corresponds roughly to the flat region of each
curve. As can be seen, increasing the doping concentration can
extend both the low and high temperature limits; however, the heavy
doping may not be suitable for a particular device.
On the low-temperature end, there are additional effects that allow
devices to operate below their "freeze-out" temperature. First, if
the semiconductor is doped to a certain concentration, it can attain
degeneracy, a condition in which the dopants require no energy for
ionization. For example, this happens in n-GaAs at a fairly low
doping concentration (~1016 cm-3) that is common in standard
devices. Thus, standard GaAs MESFETs can operate down to the lowest
temperatures, essentially to absolute zero. For Si, degenerate
doping requires a much higher a concentration (~1019 cm-3). On the
other hand, there are effects that prevent operation even before the
device is cooled to the "freeze-out" temperature. For example,
standard Si bipolar transistors cease operating well above the Si
"freeze-out" temperature, as described later.
15) How do temperature capabilities differ between the two main
types of devices: field-effect transistors and bipolar transistors?
Field-effect transistors (FETs): Characteristics of FETs generally
improve with cooling, such as transconductance, leakages, and white
(high-frequency) noise (although Si JFETs degrade below about 100
K); low-frequency noise is less predictable.
The low-temperature limit of field-effect devices depends on the
particular type and material: Si JFETs, are limited by their
freeze-out temperature (about 40 K), but their performance actually
degrades at a higher temperature. Ge JFETs have a similar behavior,
although the relevant temperatures are lower, and under proper
biasing can operate to the lowest cryogenic temperatures. Properly
designed n-channel GaAs JFETs can also operate to the lowest
temperatures, although they are uncommon.
Si MOSFETs, enhancement type, can also operate to the lowest
temperatures because the carriers needed for conduction in the
channel can be ionized by an electric field from the gate. Si
MOSFETs and CMOS circuits are often used at deep cryogenic
temperatures, below the freeze-out of Si.
Various types of heterostructure FETs (HEMTs or MODFETs), usually
based on III-V semiconductors, do not require thermal energy to
ionize the dopants. As a result, they can also be used over the
entire cryogenic temperature range down to the lowest temperatures.
Bipolar transistors: Ordinary Si bipolars (Si BJTs) suffer a rapid
decline in gain with cooling and are unusable below about 100 K.
This in not a result of "freeze-out" but of low emitter-base
injection efficiency. This effect can be avoided by adjusting the
band gaps through "bandgap engineering" as in heterojunction bipolar
transistors (HBTs), such as those based on SiGe. HBTs have
demonstrated operation down to very low cryogenic temperatures and
show increased performance on cooling. On the other hand,
conventional homojunction Ge and GaAs bipolar transistors have also
been reported to operate to very low cryogenic temperatures.
Antiviral, Antibacterial Silver Solution:
SPICE Analysis of Crystal Oscillators:
Noise-Rejecting Wideband Sampler:
>> I am thinking he would be doing dam well to squeek out 100KHz
> Thanks for the reminder, Coulombs number is something like 6.245x10^18;
> below picoamperes you are approaching counting basic charge units.
The Radio Frequency Single Electron Transistor (RF-SET) can go past 100MHz.
Sorry, all my old links have gone bad and I don't have time to google for
more right now.
> Very interesting, joseph. Here is some more information that
> might be useful:
Interesting post Mike, well worth the read. Thanks for
taking the effort to type it in.
> Interesting post Mike, well worth the read. Thanks for taking the
> effort to type it in.
> Tony Williams.
Thanks for the nice comment - actually, the credit goes to Randall
Kirschman for gathering and typing the info. All I did was cut and
paste. But people hate to go to a bare link, so I like to post the
relevant info and give a link in case anyone wants to dig some more.
But I will take credit for the nice square justification, however.
That's from my very own personal DOS editor that I wrote and still
Anyone who still uses a DOS editor has run into a problem
transferring information between DOS and Windows. If you edit a file
in DOS, and have another program running in Windows with the same
file loaded, the Windows program may not know the file was updated.
So when you edit the Windows version of the file, you overwrite the
DOS version of the file and lose your information.
This problem plagued me for years, until I found EditPad. The author
must be the smartest Windows programmer I have ever seen. He checks
to see if the file was edited by another program, and automatically
loads the most recent version so you don't lose the info you just
entered. Very slick - and it completely solves the problem. The free
version is available at:
> joseph2k <quiett...@yahoo.com> wrote:
> I am thinking he would be doing dam well to squeek out 100KHz
>> Thanks for the reminder, Coulombs number is something like
>> 6.245x10^18; below picoamperes you are approaching counting basic
>> charge units.
> The Radio Frequency Single Electron Transistor (RF-SET) can go
> past 100MHz.
> Sorry, all my old links have gone bad and I don't have time to
> google for more right now.
OK, got a minute and searched google for:
rf-set transistor bandwidth
Lots of info if you can get down to -459F. Here's an old one:
NEW HAVEN, Conn.- Scientists at Yale University have developed the
world's most sensitive electrometer, a transistor so sensitive it
can count individual electrons as they pass through a circuit. The
detector could be useful not only in developing and testing
miniaturized electronic devices but also as a highly sensitive light
detector in powerful new microscopes and telescopes.
Made from aluminum, the device is about 1,000 times faster than the
best electrometer on record and 1 million times faster than other
single electron transistors, according to a report by Yale applied
physicist Daniel E. Prober in the May 22 issue of the journal
Science. Working with him on the device were Yale postdoctoral
associate Robert J. Schoelkopf; former graduate student Peter
Wahlgren, now in Gteberg, Sweden; and graduate students Alexay A.
Kozhevnikov and Per Delsing.
"Single electron transistors have been around for about a dozen
years, but our laboratory has developed a new type called a Radio
Frequency Single Electron Transistor (RF-SET) that can measure
charges as small as 15-millionths of an electron. It detects an
extremely large bandwidth," said Prober, an expert in
high-temperature superconductivity as well as electron conduction in
metal films, wires and semiconductors.
The goal of many scientists for the last 10 years has been to
develop more precise frequency measurements and to devise current
voltage standards, said Schoelkopf, who began working on the RF-SET
design while a graduate student at California Institute of
Technology. Without that, researchers cannot study and perfect
extremely miniaturized electronic devices and computer chips at the
level where quantum mechanical effects become important.
Currently, the RF-SET works only at temperatures near absolute zero
Kelvin, or about -459 degrees Fahrenheit, thus requiring a large
refrigerator. The Yale scientists are exploring ways to make the
detector work more effectively at higher temperatures.
On the plus side is the device's high operational speed.
Conventional single electron transistor electrometers have been
limited by slow speeds, typically below frequencies of 1 kilohertz
(1,000 cycles per second), Schoelkopf said. The RF-SET can operate
even at frequencies exceeding 100 megahertz (100 million cycles per
second), where the noise due to background charge motion is
completely negligible. In their report, the Yale researchers
describe how improved versions of this device could even approach
the quantum limit, yielding the best electron detectors possible.
Because the device effectively monitors a wide range of photons -
including X-rays, ultraviolet radiation, light, infrared radiation,
and microwaves - the RF-SET design is "the best by many criteria,
very exciting," Prober said. Among the many potential applications
are far-infrared detectors, being considered by the National
Aeronautic and Space Agency (NASA) for use in astronomy, and
high-resolution electron microscopes that can amplify light for the
study of molecular structure in medicine.
Schoelkopf et al. improved the sensitivity to 6.3ue/sqrt[Hz] in
1998, so it has to be much better by now:
The Radio-Frequency Single-Electron Transistor (RF-SET): A Fast and
Schoelkopf et al.
Science 22 May 1998: 1238-1242
Radio-Frequency Single-Electron Transistor as Readout Device for
Qubits: Charge Sensitivity and Backaction
A. Aassime, G. Johansson, G. Wendin, R. J. Schoelkopf, and P.
Received 21 November 2000
We study the radio-frequency single-electron transistor (rf-SET) as
a readout device for charge qubits. We measure the charge
sensitivity of an rf-SET to be 6.3ue/sqrt[Hz] and evaluate the
backaction of the rf-SET on a single Cooper-pair box. This allows us
to compare the needed measurement time with the mixing time of the
qubit imposed by the measurement. We find that the mixing time can
be substantially longer than the measurement time, which would allow
readout of the state of the qubit in a single shot measurement.
Hm. It seems to me that he can acquire data at any rate he can achieve
after (or even before) amplification. Whether there's any signal to be
detected is the question you're addressing; correlating a lot of (maybe
unnecessary) data will reveal whether there's any there to be detected, no?
I think you should start playing with one of these:
If you're an optimistic guy, then up to 1KHz bandwith maybe it's
(do you have a map?)