some questions about high precesion constant current source
eehinjor
current source.My goal is 1uA whose error is blow 0.01%. In my circuit
I used a op-amp,p-mosfet and 5V voltage reference. But the result is so
bad,the error is almost 1%. Could somebody give me some hints? Thanks
very much. eehinjor
bill....@ieee.org
mounted fairly close to it?
MOSFETs have a nasty habit of oscillating at a few hundred MHz - fast
enough that a cheap oscilliscope won't see it - and this can invalidate
a DC analysis of the circuit. In particular, the inputs of the op amp
will tend to follow the peak of the AC envelope of the signal they are
looking at at, rather than the average, and this can look like an
out-of-spec input offset voltage.
A "gate stopper" somewhere between ten and a few hundred ohms usually
kills this sort of oscillation. At Cambridge Instruments, we had enough
trouble with oscillating MOSFETs that the 10R gate resistor was
compulsory.
It wasn't always necessary, and sometimes 10R wasn't enough, but at
least we always had space on the printed circuit board to fit a gate
stopper resistor if we needed one.
-------------
Bill Sloman, Nijmegen
cbarn...@aol.com
What exactly do you mean by error of 1%? I assume you have a variable
resistor somwhere in your circuit to trim the output current.
eehinjor
In my circuit,there is a variable resistor,but it is serial with
4990K,the goal is to achieve 5000K resistor.
I don't understant Bill by the 10R resistor.the voltage is 5V,so the
10R is so far.
bill....@ieee.org
- it just discourages the MOSFET from using the inductances of the
connecting wires and the stray capacitances around the circuit to make
a high frequency oscillator.
If the MOSFET is busy turning itself on an off every few nanoseconds.
your circuit won't behave the way you'd expect from an analysis based
on there being nice stable voltages and currents all around the
circuit.
Incidentally, what are you using to see the 0.99uA and 1.001uA numbers?
Can you borrow a 5-digit voltmeter, and see if it isn't going from -
say - 0.994uA to 1.0006uA? And how fast is moving up and down between
these limits?
-------------
Bill Sloman, Nijmegen
Walter Harley
news:1125320855....@z14g2000cwz.googlegroups.com...
What are you using for your 5V voltage reference? (Are you sure that it is
stable? .01% of 5V is 0.5mV; for comparison, an LM7805 is specced at
thermal drift of 0.8mV per degree C.)
Are you seeing thermal effects in your sense resistor? (Easy for air
currents to change the temperature of the resistor by .01%)
How well do you trust your meter? (Take a look at its specs - even on a
good meter, you shouldn't trust the least significant digit very much.)
There's good information on designing precision devices in Art of
Electronics. Getting .01% accuracy is not a matter of getting one thing
right, it's a matter of getting very few things wrong.
Joerg
From 1.000uA to 1.001uA it is only 0.1%, not 1%. Could the toggle
between 1.000uA and 0.99uA explained by the fact that you meter is not
able to display 0.999uA? (It should)
Anyway, a typical 5V reference isn't as precise as you desire. Go to
Analog Devices and see what they offer. Be prepared to pay a hefty price
since references in the sub 0.1% category are not cheap. Then keep your
circuit simple. The less parts the better. If you have to use an opamp
check its offset and drift specs since these will play a major role here.
Regards, Joerg
eehinjor
The value of the current is sway from abou 0.99uA to 1.01.Such as
0.995uA,my voltmeter is 8.5-digit.
The frequency is low,about 50Hz.of cource this data is read only from
the voltmeter.
eehinjor
the voltage reference is AD586,its accuracy is 0.04%,and temp. shift is
only 2ppm/c.
I am not familiar with pmosfet,maybe my pmosfet is not right,its type
is VP0610l from Vishay.
I have a 8.5-digit voltmeter,so I am sure the value is right.
Whan can I do next?
John Popelish
That is not so low a frequency at this current. If the current source
has a resistance high enough to keep the current from varying more
than 0.01% of 1 uA (0.1 nA) as the load voltage swings over just a 1
volt range, this implies a resistance of more than 10 giga ohms. A
capacitance of only a half pico farad passes current peaks of more
than .1 nA to a 1 volt peak to peak 50 hertz sinusoidal swing (200
volts per second peak slope). Good luck finding a P-channel fet with
less than 1/2 pF drain capacitance and a wiring method that has less
stray capacitance than what is left of that 1/2 pF.
Winfield Hill
>
> I am not familiar with pmosfet, maybe my pmosfet is not right,
Add the gate resistor suggested by Bill, and the inverting
input- to-output feedback capacitor suggest by me and others
many times here on s.e.d. (and in our book). It's imperative
that you prevent even a low-level high-frequency oscillation.
--
Thanks,
- Win
norman...@comcast.net
> the output is shifting from 0.99uA to 1.001uA.
Would by any chance mean 0.99uA to 1.01uA ?
Ban
> Thanks all.The circuit is shown as below.
> Q1:VP0610L;
> Q2:2N4402;
> RL:the load(resistor);
> IC1:OPA602;
> R1:the resistor(5000K ohm),it is composed of 4990Kohm fixed
> resistor(0.1%,15ppm/c) and a 20Kohm variable resistor;
> +12V,-12V:power;
> +7V:it is produce by INA105 and AD586(+5V voltage reference);
> +5.5V:to avoid the voltage of RL over this value;
>
> In fact,my goal is 1mA/100uA/10uA/1uA when R1 is 5K/50K/500K/5000K.
>
> By the way,the VP0610L has been obseleted from last year,would you
> like recommand one?
>
>> From simulation result by pspice,the circuit is right,but the real
> result is so bad.I don't know how to modify it.
>
> +12V
> |
> .-.
> | |
> R1| |
> '-'
> |
> |-----------------------o
> | +12V |
> | + |
> | IC1 | |
> | |\| ||-+
> o-------------|-\ ||-> Q1
> | >---||-+
> +7V--------------|+/ |
> |/| |
> | |
> - |
> -12V |-------------------o
> | |
> |< .-.
> +5.5V----| Q2 | |RL
> |\ | |
> | '-'
> | |
> o------------------ o
> |
> ===
> GND
What you do wrong is to put the reference between gnd and +7V, you have to
choose +12V as the reference terminal and have -5V from there, so you can
use the reference directly on the opamp.
--
ciao Ban
Bordighera, Italy
eehinjor
On the R1,if there is some noise on +12v,it will be reduced.
I don't understand what ciao Ban mean by -5V.
thanks.
Ban
You made it right, but there is a disadvantage with the reference hanging on
the +12V rail and common mode on the opamp inputs.
I propose you this circuit. You can fine tune R3 for infinite impedance and
the left R2 for exact value.
small caps(47p) across the feedback resistors will help stability.
___
.-------|___|---------------------------.
Uref ___ | ___ R3 ___ ___ ___ |
o-|___|--o--|___|--o--|___|--o--|___|--o--|___|--o--o
R2 | R2 | R2 | R2 | R1 |
o | |\ | | |\ | .-.
| '--|-\ | '--|-\ | | |
| | >---' | >---' Load| |
| .--|+/ .--|+/ '-'
| | |/ | |/ |
=== === R3=R2-R1 === Iload= Uref/R1 ===
GND GND GND GND
(created by AACircuit v1.28 beta 10/06/04 www.tech-chat.de)
I also don't like your transistor on the O/P to clamp overvoltage. It adds
(variable) capacitance and a temperature dependent cutoff current. Use a
comparator or something.
Tony Williams
eehinjor <eehi...@163.com> wrote:
> +12V
> |
> .-.
> | |
> R1| |
> '-'
> |
> |-----------------------o
> | +12V |
> | + |
> | IC1 | |
> | |\| ||-+
> o-------------|-\ ||-> Q1
> | >---||-+
> +7V--------------|+/ |
> |/| |
> | |
> - |
> -12V |-------------------o
> | |
> |< .-.
> +5.5V----| Q2 | |RL
> |\ | |
> | '-'
> | |
> o------------------ o
> |
> ===
> GND
As well as the R+C already suggested by others I'd
add a pulldown resistor on the output of the opamp,
say 15k to -12v This biasses the o/p transistors
away from the low gain (and sometimes oscillatory)
crossover region.
--
Tony Williams.
eehinjor
I don't know whether you have tested this circuit.In my opnion,so many
resistors and op-amps,it is almost impossible to reach the goal.
You used two op-amps to improve impdance,but the shift will be high.
If the accuracy is low,that is not problem.
During this several months,I found the result is very far from theory.
Robert Baer
It would help if you posted the circuit in question on abse.
Keep in mind that Vgs of a FET changes over temperature at a given
current..
A 100 ohm resistor in series with the gate of a FET is to help
prevent HF oscillations.
eehinjor
In the circuit shown before,the accuracy is infulenced by
+12V,+5V,resistor and MOSFET.If a pulldown resistor is added between
the output of op-amp and -12v,the error of -12V will be included.
Are you sure?
best regards.
Ban
I have had good results with this circuit for low currents and it gives you
the possibility not only to trimm the absolut value, but also for
infinite impedance, which was what your circuit is suffering. The resistors
are only of 2 values and they are commercially available in the required
precision, why should a second opamp decrease precision?
You have to study abit more before you can make these kind of remarks. sorry
to say.
eehinjor
I will try again and find the result.
best regards.
Tony Williams
eehinjor <eehi...@163.com> wrote:
> In the circuit shown before,the accuracy is infulenced by
> +12V,+5V,resistor and MOSFET.If a pulldown resistor is added
> between the output of op-amp and -12v,the error of -12V will be
> included. Are you sure?
A 1mA pulldown (or pullup) simply takes the output
transistors of the opamp out of Class-AB into Class-A.
Done for the reasons already mentioned. No additional
error mechanism is created.
--
Tony Williams.
eehinjor
Your last circuit is suitable for 1mA or more higher constant current
source.
Normally,the voltage reference is +5V or +2.5V,if the output is 1uA,R1
should be 5000K.
Because of R3=R2-R1,the result is R2 is very big.
After my simulation,the voltage of load should not more than 60% of
Uref.
Tony,thank you.
Because the goal is so precesion,something little will be important.I
will try your way later.
Fred Bloggs
At 1uA and with the VP06101 @ VDS=7V(ish), he is running the FET
subthreshold, maybe nS transconductance, attempting to regulate
complicated IDS leakage which is almost certainly 100's nA, low
composite loop gain even with diFET at 100dB Aol, and regulation against
power supply variation will not be very good.
Winfield Hill
>
> As well as the R+C already suggested by others I'd
> add a pulldown resistor on the output of the opamp,
> say 15k to -12v This biasses the o/p transistors
> away from the low gain (and sometimes oscillatory)
> crossover region.
Editing your drawing, eehinjor, showing our suggestions:
. +12V
. |
. R1 5k to 5M
. |
. +-------------------------,
. | 470pF |
. | ,---||---, |
. | | | |
. '---+-|-\ | 220 ||--' S
. | >-+-+--/\/\--||->
. +7V -----|+/ | ||--, VP0610L
. 15k | BS250
. | | VP0106
. -12V +-----------o
. | |
. |< .-.
. +5.5V----| Q2 | |RL
. |\ | |
. | '-'
. | |
. o-----------o
. |
. ===
. GND
--
Thanks,
- Win
eehinjor
For me this task is a chanllege.I have done some experiment during past
months.
My goal is 1mA/100uA/10uA/1uA,the accuracy is 0.01%,the result is that
the output can be switched between them.Then some mux-chips will bring
error.
I have tried three ways.One,single op-amp & resistor.Second,two
op-amps,one of them as the feedback.Third,the op-amp & PMOSFET.Before
experiments,I simulated them by pspice.
The first two ways can acheive 1mA/100uA easily,but they can not
realize 10uA/1uA.
So I have to change the way to the third.But until now,I can not
realize 1mA by the third way.Maybe because I am not familar with
PMOSFET.
I will buy some chips tommorrow to do some experiments on your advice.I
am sure I have to study more from all of you.
best regards.
Tony Williams
Fred Bloggs <nos...@nospam.com> wrote:
> At 1uA and with the VP06101 @ VDS=7V(ish), he is running the FET
> subthreshold, maybe nS transconductance, attempting to regulate
> complicated IDS leakage which is almost certainly 100's nA, low
> composite loop gain even with diFET at 100dB Aol, and regulation
> against power supply variation will not be very good.
Good thought. Perhaps drop the 12-0-12 and run it off a single
24v rail. All voltages are within most opamp's CMV and Vout
ratings.
--
Tony Williams.
Winfield Hill
And, eehinjor, a small surface-mount p-channel MOSFET, BSS84.
--
Thanks,
- Win
Ken Smith
Tony Williams <to...@ledelec.demon.co.uk> wrote:
[...]
> As well as the R+C already suggested by others I'd
> add a pulldown resistor on the output of the opamp,
> say 15k to -12v This biasses the o/p transistors
> away from the low gain (and sometimes oscillatory)
> crossover region.
Depending on the op-amp:
I'd pull from the op-amp to the +12V not ground. You don't want noise
from the power supply to be able to get near the VGS voltage of the power
MOSFET.
--
--
kens...@rahul.net forging knowledge
Winfield Hill
Nah. First, there's nothing wrong with running MOSFETs in the
subthreshold region - they work fine (it's the typical spice
subthreshold MOSFET model that fails to work properly). Second,
the transconductance/current ratio actually improves for MOSFETs
in the subthreshold region, approaching BJTs in many cases, so
that argument is wrong. Third, typical leakage for these small
MOSFETs is in the low pA region, not 100s of nA, for Vds drain
voltages below 80% of Vdss, which = 30 to 40V for these parts.
Fourth, any Ids leakage current is part of the current-source
output, measured by the servo, so contributes NO error anyway,
unless the leakage exceeds the desired current. So 100nA would
be fine in a 1uA current source.
All this said, I'd prefer to use a BJT over a MOSFET in this
application, because its high gate capacitance slows circuit
response at low currents. E.g., a VP0610L, BSS110 or BS250
has Ciss = 25pF typ, so f_T = Id / (2pi n Vt Ciss) = 50kHz
at 1uA (assuming n = 4), which I'm guessing is perhaps 15 to
20x less than a small Darlington transistor like the MPSA64
or MPSA14 (50MHz at 500uA according to the datasheet curve).
This means an active BJT current source can be compensated
for faster operation than a MOSFET one. If that matters...
--
Thanks,
- Win
Ken Smith
Winfield Hill <Winfiel...@newsguy.com> wrote:
[...]
> All this said, I'd prefer to use a BJT over a MOSFET in this
> application, because its high gate capacitance slows circuit
> response at low currents.
How about switching to a device like the 3N163?
Ciss is only about 2.5pF on them. The Cds is in about the same range.
This would make for less frequency roll off problems. If you really want
to go fast, you can AC boostrap the substrate on them.
Tony Williams
Ken Smith <kens...@green.rahul.net> wrote:
> I'd pull from the op-amp to the +12V not ground. You don't want
> noise from the power supply to be able to get near the VGS
> voltage of the power MOSFET.
OK.... wouldn't argue with that.
--
Tony Williams.
eehinjor
I think there is some difference between two ways.
In fact,we must differ -12v from ground in this circuit.
Today I found a strange thing,Vs=6.97V,Vg=6.35(the mosfet is
VP0610l)
Fred Bloggs
Winfield Hill wrote:
>
> Nah. First, there's nothing wrong with running MOSFETs in the
> subthreshold region - they work fine (it's the typical spice
> subthreshold MOSFET model that fails to work properly). Second,
> the transconductance/current ratio actually improves for MOSFETs
> in the subthreshold region, approaching BJTs in many cases, so
> that argument is wrong. Third, typical leakage for these small
> MOSFETs is in the low pA region, not 100s of nA, for Vds drain
> voltages below 80% of Vdss, which = 30 to 40V for these parts.
> Fourth, any Ids leakage current is part of the current-source
> output, measured by the servo, so contributes NO error anyway,
> unless the leakage exceeds the desired current. So 100nA would
> be fine in a 1uA current source.
I don't believe that for a minute- this MOSFET is leaky as a sieve. They
specify Vgs(th) at Id=1mA, which is much higher than most, the gate body
leakage Igss is bounded by 10nA, and the Idss at Vds=25V is bounded by
0.5uA. One aspect of this you have not considered is that these numbers
are on the order of 100x to 1000x the specified error band of 100pA at
1uA source current. These leakage currents will all have components of
some unknown proportion that exhibit standard deviations not governed by
the usual shot- and Johnson- noise distributions, so that it is
unreasonable for you to discard a complete unknown as being 40 to 60 dB
down from the mean when you have so little information. If the regulated
current is in fact exhibiting the flicker described, then it has to be
related to this ratio of allowed variation to average leakage
magnitudes- you cannot expect reliable performance with arbitrarily
small errors.
Winfield Hill
>
> Today I found a strange thing, Vs=6.97V, Vg=6.35
> (the mosfet is VP0610l)
Why do you consider that to be strange? Was that
for your 1uA or 10uA range? If you look at figure
3.14 in AoE, you'll see that MOSFET gate voltages
are pretty low for < 1mA sub-threshold operation. An
observed Vgs of 0.6V would be low, but not amazing.
--
Thanks,
- Win