Current source design (tricky?)

[[email protected]]

Hi,

I'm trying to design a power current source. It has to have some
features that make it a bit of a challange to design (for me at
least...)
it has to do the following:

- supply about 2-3 amps (no need for programmability, it can be preset
to a value when constructed).
- have a low minimum voltage drop, of 1 Volt or less (0.5V would be
better). That is, it should work even if Vcc=12V and the load needs 11V
to keep the preset current. of course, it should also work with a
higher Vcc.
- keep a reasonably constant current for both changes in the load, and
(the bigger problem) for a changing Vcc.
- the load is returned to ground, so the source has to be on the "high"
side.
- NO need for good precision, +/-5% is more than ok. the conditions are
DC so bandwidth isn't an issue as well.


it seems that each design I find or make meets only some of the
criteria above.
for example, an LM317 or similar regulator in a current loop would have
been fine, except the large minimum voltage drop (definitely not less
than 2 or even 3 volts).

am I requiring too much? help anyone?

thanks,
Guy.

 

[Larry Brasfield]

Hi,

I'm trying to design a power current source. It has to have some
features that make it a bit of a challange to design (for me at
least...)
it has to do the following:

- supply about 2-3 amps (no need for programmability, it can be preset
to a value when constructed).
- have a low minimum voltage drop, of 1 Volt or less (0.5V would be
better). That is, it should work even if Vcc=12V and the load needs 11V
to keep the preset current. of course, it should also work with a
higher Vcc.

An important parameter is: How much higher?
This will set the dissipation required in the pass element.

- keep a reasonably constant current for both changes in the load, and
(the bigger problem) for a changing Vcc.

Obviously, "reasonable constant" is too vague to
permit a design. How fast does the load change?
How much current variation can be permitted as
that change occurs?

- the load is returned to ground, so the source has to be on the "high"
side.
- NO need for good precision, +/-5% is more than ok. the conditions are
DC so bandwidth isn't an issue as well.

Changing load and Vcc means "not DC".

it seems that each design I find or make meets only some of the
criteria above.
for example, an LM317 or similar regulator in a current loop would have
been fine, except the large minimum voltage drop (definitely not less
than 2 or even 3 volts).

am I requiring too much? help anyone?

You will find that a power PMOSFET with a current
sense resistor in the source or drain, together with an
op-amp circuit to control the gate in response to
sensed current deviation from a setpoint, will likely
do the job easily. A design will probably show up
here shortly after you firm up your requirements.

 

[John Popelish]

Hi,

I'm trying to design a power current source. It has to have some
features that make it a bit of a challange to design (for me at
least...)
it has to do the following:

- supply about 2-3 amps (no need for programmability, it can be preset
to a value when constructed).
- have a low minimum voltage drop, of 1 Volt or less (0.5V would be
better). That is, it should work even if Vcc=12V and the load needs 11V
to keep the preset current. of course, it should also work with a
higher Vcc.
- keep a reasonably constant current for both changes in the load, and
(the bigger problem) for a changing Vcc.
- the load is returned to ground, so the source has to be on the "high"
side.
- NO need for good precision, +/-5% is more than ok. the conditions are
DC so bandwidth isn't an issue as well.

it seems that each design I find or make meets only some of the
criteria above.
for example, an LM317 or similar regulator in a current loop would have
been fine, except the large minimum voltage drop (definitely not less
than 2 or even 3 volts).

am I requiring too much? help anyone?

thanks,
Guy.

I think I would use a big P-channel mosfet with a fraction of an ohm
source resistor, with a rail to rail input opamp driving the gate.
Add a voltage reference chip tied to the positive rail and you have
something for the opamp to compare the source resistor drop against.

 

[Fred Bloggs]

A design will probably show up
here shortly after you firm up your requirements.

Well it sure as hell won't be posted by you now will it pussy,
pretentious, pseudo-intellectual, delusional born-again xtian trash.

 

[Larry Brasfield]

Well it sure as hell won't be posted by you now will it pussy,
pretentious, pseudo-intellectual, delusional born-again xtian trash.

Fred, you need not waste any more bandwidth with
such drivel. I will gladly assume your low opinion of
me and my posts without further notice. And if you
should change your mind, I would just as soon not
learn about that unless you have undergone a total
personality transformation.

 

[Fred Bloggs]

Fred, you need not waste any more bandwidth with
such drivel. I will gladly assume your low opinion of
me and my posts without further notice. And if you
should change your mind, I would just as soon not
learn about that unless you have undergone a total
personality transformation.

Hey- post your circuit, windbag. We're calling your bluff, cowgirl.

 

[Fred Bloggs]

Fred, you need not waste any more bandwidth with
such drivel. I will gladly assume your low opinion of
me and my posts without further notice. And if you
should change your mind, I would just as soon not
learn about that unless you have undergone a total
personality transformation.

Anybody know which cartoon character the punk is imitating? A Fox
network Jeeves or something?

 

[[email protected]]

Hi,

thank you for taking the time to look through my requirements.

as for the calrifications:

I'd say the maximum difference between Vcc and the load will be 3 to 4
volts. the minimum should be as low as possible, say 0.5 Volt. on 2
amps, that gives 8 watts of dissipation in the pass element. is it
reasonble for to-220 and only a small heatsink?

as for the "reasonably constant current", I need about +/- 5%, as I
specified later in the post.

you were right, since the load and Vcc are changing, the conditions
should be more accurately called "nearly DC" I suppose. I approximate
the load would only change at a max rate of 1-2% per second (because of
heating). the voltage would also drop or rise slowly. the exception is
that if this would be connected to a 12V source from a car, and the car
starts, the voltage would rise from 12V to 13.8 or so pretty quickly
(do you have an idea how fast, or how bad would transient spikes
durning ignition would be?)

the idea of an opamp controlling a P mosfet sounds interesting. I
suppose that for 0.5V minimum drop, I need a mosfet with Rds(on) of
0.25 Ohm. is that a reasonble requirement? (for a single digit price
tag, that is). would the opamp work well if the drop on the sense
resistor be only 100mV or so?

if you can give me a recommendation for a single supply rail to rail
opamp, as well as a mosfet the will do, I'll be most grateful. I don't
do much analog, so some pointers would save me lots of time.

thanks again for the responses!
Guy.

 

[Winfield Hill]

John Popelish wrote...

I think I would use a big P-channel mosfet with a fraction of an ohm
source resistor, with a rail to rail input opamp driving the gate.
Add a voltage reference chip tied to the positive rail and you have
something for the opamp to compare the source resistor drop against.

There are a few subtleties in such a circuit, such as isolating the
opamp from the high gate capacitance of the FET, and using a Kelvin
connection for the low-value sense resistor, including the return
for the reference divider chain. If the "12V" is a battery in an
operating vehicle, provision should be made so "load-dump" spikes
won't damage the opamp. Inductive loads may be applied, so we'll
add a protective diode. Fairly high power may be dissipated in the
MOSFET, so we'll add a heat-sink thermal cutout. And it's useful
to switch the power at low current levels, so we'll add that too.

Although the concept is simple, a fully-fleshed-out circuit begins
to show some signs of complexity. viz,

.. 2.5A current source r1.0
.. by Winfield Hill 10Mar'05
.. IRF4905
.. 12V battery 5A fuse Rs p-channel ---> 2.5A
.. (O)====o---o===o== 0.1 ==o====== S D =====+=========(O)
.. | Kelvin | _|__|_ |
.. ,---+--------+---+------+ wired 22k ---- | Vr R2
.. | | | | | | | heat | Io = -- -------
.. | | LM385- | | ,----| ----+---' G sink | Rs R1 + R2
.. | \_|_ 1.2 | | | | _|_ C2 | : _|_
.. | /_\ R2 | | _|_ --- 0.01 | : /_\ Io R1 Rs
.. | | 1.23V | | '--|- \ | R3 | : | R2 = ----------
..\_|_ | | | | >--+----- 680 --' : | Vr - Io Rs
.. /_\ +-- R1 --+---| ---|+__/ : |
.. | | | | LMC6482, LM8261, : | for Io = 2.5A,
.. | | _|_ | OP284, etc, : | R1 = 2.0k
.. | 3.3k --- 0.1 | RRIO type thermal | R2 = 510, or
.. | | | | power cutout | a pot may be
.. '---+------------+------+-- 220 ---o / o----o / o---+ used for R2
.. 12V zener |
.. (O)========================================+=========(O)
.. RETURN

Some experimenting may be necessary with the values of C2 and R3,
depending on the output-stage properties of the opamp choice.

An out-of-regulation indicator, and a voltage-compliance-limit would
be nice, etc., but I'd better stop here...

 

[John Larkin]

Well it sure as hell won't be posted by you now will it pussy,
pretentious, pseudo-intellectual, delusional born-again xtian trash.

What's an xtian?

John

 

[Jonathan Kirwan]

What's an xtian?

Someone who uses _xmas_ to honor superstitious beliefs.

Jon

 

[John Woodgate]

I read in sci.electronics.design that Jonathan Kirwan

Someone who uses _xmas_ to honor superstitious beliefs.

I thought it was someone who used to work for Texas Instruments.

 

[John Larkin]

Someone who uses _xmas_ to honor superstitious beliefs.

Jon

Oh, that explains the bumper sticker, "Keep the X in Xmas."

John

 

[Larry Brasfield]

Hi, Hi.
thank you for taking the time to look through my requirements.

You are welcome.

as for the calrifications:

I'd say the maximum difference between Vcc and the load will be 3 to 4
volts. the minimum should be as low as possible, say 0.5 Volt. on 2
amps, that gives 8 watts of dissipation in the pass element. is it
reasonble for to-220 and only a small heatsink?

That helps, but you are still looking at 18 W dissipation under
"normal" circumstances. (I rely on the 3 Amp output you
mentioned in your earlier post and battery charging voltage
of 16V.) During a load dump transient, that could briefly go
to 75W. (The transient is brief enough that you need not size
the heatsink for it, but it may mean that you want to keep the
junction cooler to provide some headroom for the thermal
transient.) Whether that can be managed with a small heatsink
is questionable. I would rely on the heatsink vendor's datasheet.

as for the "reasonably constant current", I need about +/- 5%, as I
specified later in the post.

So, the response to a voltage transient must stay within
that bound. This makes the design a little more interesting
and likely makes it a good idea to low pass filter the
incoming supply for both the FET and op-amp. You will
likely want to put a resistor in series with the integrator
feedback capacitor to improve the transient response
of the current feedback loop.

you were right, since the load and Vcc are changing, the conditions
should be more accurately called "nearly DC" I suppose. I approximate
the load would only change at a max rate of 1-2% per second (because of
heating). the voltage would also drop or rise slowly. the exception is
that if this would be connected to a 12V source from a car, and the car
starts, the voltage would rise from 12V to 13.8 or so pretty quickly
(do you have an idea how fast, or how bad would transient spikes
durning ignition would be?)

I was not trying to quibble about the meaning of "DC".
Ignition transients are pretty easy to filter out before
they hit the active part of your circuit. The load dump
transients change much faster than you appear to
suspect. You will need to research them.

the idea of an opamp controlling a P mosfet sounds interesting. I
suppose that for 0.5V minimum drop, I need a mosfet with Rds(on) of
0.25 Ohm. is that a reasonble requirement? (for a single digit price
tag, that is). would the opamp work well if the drop on the sense
resistor be only 100mV or so?

Yes. Maybe.

if you can give me a recommendation for a single supply rail to rail
opamp, as well as a mosfet the will do, I'll be most grateful. I don't
do much analog, so some pointers would save me lots of time.

I will not claim these are the most cost effective parts,
but they will do the job:
http://ec.irf.com/v6/en/US/adirect/ir?cmd=catProductDetailFrame&productID=IRF9Z24
http://www.national.com/pf/LM/LM8261.html
This op-amp, suggested by Mr. Hill, is well suited for this application
due to its positive rail input range and capacitive load tolerance.

Just a few tips for your design:

Be sure to zener protect the op-amp supply.
Transients on an automobile "12V" rail can be
surprisingly large due to quickly changing loads,
especially removed loads, and the way alternators
are regulated. (This effect is often called "load
dump".) I would use a 24 or 27 Volt zener diode,
grounded at one end.

Unless you are willing to see an output current
spike sometimes, the MOSFET output stage
should get similar protection. I would use a
power rectifier from the low end of the current
sense resistor to the zener mentioned above.
You should satisfy yourself that the zener's
power rating is sufficient to absorb a load
dump transient. (I would have to research
this to know what that might be.)

Your op-amp circuit will be operating with
inputs at the positive (filtered, protected)
rail, and its output is referenced to that rail
by the feedback condition, so be sure not
to ground reference those parts of the
circuit that influence the controlled current.

thanks again for the responses!

You're still welcome.

 

[Spehro Pefhany]

Oh, that explains the bumper sticker, "Keep the X in Xmas."

John

Our Roman X represents an upper-case chi (the first letter in the word
pronounced "Christos" (Christ), as written in Greek).

So it's not the same as the x in xfrmr. But we can keep the x in xfmr,
xducer, xmtr and so on. It could result in some obscure bumper
stickers.


Best regards,
Spehro Pefhany

 

[Jamie]

Someone who uses _xmas_ to honor superstitious beliefs.

Jon

no santa clause? damn!

 

[Rich Grise]

Our Roman X represents an upper-case chi (the first letter in the word
pronounced "Christos" (Christ), as written in Greek).

So it's not the same as the x in xfrmr. But we can keep the x in xfmr,
xducer, xmtr and so on. It could result in some obscure bumper
stickers.

"Let's keep the X in Xvestite!"

;-P

 

[Rich Grise]

You are welcome.


That helps, but you are still looking at 18 W dissipation under
"normal" circumstances. (I rely on the 3 Amp output you
mentioned in your earlier post and battery charging voltage
of 16V.) During a load dump transient, that could briefly go
to 75W. (The transient is brief enough that you need not size
the heatsink for it, but it may mean that you want to keep the
junction cooler to provide some headroom for the thermal
transient.) Whether that can be managed with a small heatsink
is questionable. I would rely on the heatsink vendor's datasheet.


So, the response to a voltage transient must stay within
that bound. This makes the design a little more interesting
and likely makes it a good idea to low pass filter the
incoming supply for both the FET and op-amp. You will
likely want to put a resistor in series with the integrator
feedback capacitor to improve the transient response
of the current feedback loop.


I was not trying to quibble about the meaning of "DC".
Ignition transients are pretty easy to filter out before
they hit the active part of your circuit. The load dump
transients change much faster than you appear to
suspect. You will need to research them.


Yes. Maybe.


I will not claim these are the most cost effective parts,
but they will do the job:
http://ec.irf.com/v6/en/US/adirect/ir?cmd=catProductDetailFrame&productID=IRF9Z24
http://www.national.com/pf/LM/LM8261.html
This op-amp, suggested by Mr. Hill, is well suited for this application
due to its positive rail input range and capacitive load tolerance.

Just a few tips for your design:

Be sure to zener protect the op-amp supply.
Transients on an automobile "12V" rail can be
surprisingly large due to quickly changing loads,
especially removed loads, and the way alternators
are regulated. (This effect is often called "load
dump".) I would use a 24 or 27 Volt zener diode,
grounded at one end.

I'd use a TransZorb - they're like a Zener, but optimized for transient
suppression. They can take a much higher instantaneous glitch.
http://www.vishay.com/diodes/protection-tvs-esd/trans-zorb/

Good Luck!
Rich

 

[Michael A. Terrell]

"Let's keep the X in Xvestite!"

;-P

Are you trying to tell us something?

 

[Ken Smith]

[...]

A few comments:

Make the parallel combination of R1//R2 have the same resistance as the
inverting pin of U1 sees. This drops the bias current out of
consideration.

Wire the (+) supply of the OP-AMP and the 12V zener with their own wire to
the Rs current pin. Only R2 and the LM385 should go to the voltage leg of
Rs. This keeps those DC currents out of the measurement circuit.

The 22K to the inverting input of the OP-AMP and the 0.01, C2 feedback on
the op-amp look a bit too big to me as does R3. R3 and the gate
capacitance of the IRF4905 make a time constant. You want the 22K and the
C2 to make one that is bigger than that by perhaps a factor of 10 but not
a factor of 100. I did not look up the MOSFET.

Thermally isolate the non-power parts away from the heat of the power
section if your mechanical situation allows.

A constant current diode like the 1N5313 can be used in place of the 220
Ohm resistor. If not, I suggest making the 220 a 10W resistor.

If you add a Schottky across the 12V zener, the circuit will be fairly
well protected against reversed supply leads, assuming the 0.1, the MOSFET
and the output diode can all take the current while the 5A fuse blows.