Maker Pro
Maker Pro

Auto Offset

M

Mark

Jan 1, 1970
0
I'm using a LEM current transducer to measure pulsed currents. The
current sensor produces a current output which I feed to a resistor
producing a voltage that represents the current flow. The voltage feeds
an opamp for offset correction and buffering. The sensor has good
accuracy (+/- 0.5%), but suffers from an initial offset
(+/- 0.15mA) and worse, a thermal induced offset (+/- 0.35 mA over temp).

Since I know when the current pulses are coming, I want to to force the
offset to zero before taking a measurement. I've seen circuits to do
this using the following techniques:

1. An A/D followed by a D/A. With zero current input, the offset is
driven to zero, and the offset correction is latched in the D/A until
the next "calibration" cycle.

2. A/D and a digital pot controlled by a processor, works similiar as
above.

What are the merits of either method. I prefer not to have a processor
involved, but do have access to FPGA resources.

Is ther a simple way to do this?

Mark
 
J

Jim Thompson

Jan 1, 1970
0
I'm using a LEM current transducer to measure pulsed currents. The
current sensor produces a current output which I feed to a resistor
producing a voltage that represents the current flow. The voltage feeds
an opamp for offset correction and buffering. The sensor has good
accuracy (+/- 0.5%), but suffers from an initial offset
(+/- 0.15mA) and worse, a thermal induced offset (+/- 0.35 mA over temp).

Since I know when the current pulses are coming, I want to to force the
offset to zero before taking a measurement. I've seen circuits to do
this using the following techniques:

1. An A/D followed by a D/A. With zero current input, the offset is
driven to zero, and the offset correction is latched in the D/A until
the next "calibration" cycle.

2. A/D and a digital pot controlled by a processor, works similiar as
above.

What are the merits of either method. I prefer not to have a processor
involved, but do have access to FPGA resources.

Is ther a simple way to do this?

Mark

What sort of pulses... amplitude, width, rate?

A simple DC restorer might work under some circumstances.

...Jim Thompson
 
J

John Larkin

Jan 1, 1970
0
I'm using a LEM current transducer to measure pulsed currents. The
current sensor produces a current output which I feed to a resistor
producing a voltage that represents the current flow. The voltage feeds
an opamp for offset correction and buffering. The sensor has good
accuracy (+/- 0.5%), but suffers from an initial offset
(+/- 0.15mA) and worse, a thermal induced offset (+/- 0.35 mA over temp).

Since I know when the current pulses are coming, I want to to force the
offset to zero before taking a measurement. I've seen circuits to do
this using the following techniques:

1. An A/D followed by a D/A. With zero current input, the offset is
driven to zero, and the offset correction is latched in the D/A until
the next "calibration" cycle.

2. A/D and a digital pot controlled by a processor, works similiar as
above.

What are the merits of either method. I prefer not to have a processor
involved, but do have access to FPGA resources.

Is ther a simple way to do this?

Mark


in--------c-------+----------out
|
switch
|
gnd


Just keep the switch closed when you know there's no pulse. A fet or
an CMOS analog switch thingie will work.

John
 
G

Genome

Jan 1, 1970
0
John Larkin said:
in--------c-------+----------out
|
switch
|
gnd


Just keep the switch closed when you know there's no pulse. A fet or
an CMOS analog switch thingie will work.

John

Serious? You might like to read that one again..... Oh, I'm sure you knew
anyway.

Has DM learnt about transient thermal impedance yet?

BLURP
 
L

Luhan Monat

Jan 1, 1970
0
Mark said:
I'm using a LEM current transducer to measure pulsed currents. The
current sensor produces a current output which I feed to a resistor
producing a voltage that represents the current flow. The voltage feeds
an opamp for offset correction and buffering. The sensor has good
accuracy (+/- 0.5%), but suffers from an initial offset
(+/- 0.15mA) and worse, a thermal induced offset (+/- 0.35 mA over temp).

Since I know when the current pulses are coming, I want to to force the
offset to zero before taking a measurement. I've seen circuits to do
this using the following techniques:

1. An A/D followed by a D/A. With zero current input, the offset is
driven to zero, and the offset correction is latched in the D/A until
the next "calibration" cycle.

2. A/D and a digital pot controlled by a processor, works similiar as
above.

What are the merits of either method. I prefer not to have a processor
involved, but do have access to FPGA resources.

Is ther a simple way to do this?

Mark

Mark,

Since you are incorporating a processor here, make 2 a/d readings: one
just before and one during the pulse. Then subtract the first from the
second.
 
J

John Fields

Jan 1, 1970
0
Mark,

Since you are incorporating a processor here, make 2 a/d readings: one
just before and one during the pulse. Then subtract the first from the
second.
 
S

scada

Jan 1, 1970
0
Mark said:
I'm using a LEM current transducer to measure pulsed currents. The
current sensor produces a current output which I feed to a resistor
producing a voltage that represents the current flow. The voltage feeds
an opamp for offset correction and buffering. The sensor has good
accuracy (+/- 0.5%), but suffers from an initial offset
(+/- 0.15mA) and worse, a thermal induced offset (+/- 0.35 mA over temp).

Since I know when the current pulses are coming, I want to to force the
offset to zero before taking a measurement. I've seen circuits to do
this using the following techniques:

1. An A/D followed by a D/A. With zero current input, the offset is
driven to zero, and the offset correction is latched in the D/A until
the next "calibration" cycle.

2. A/D and a digital pot controlled by a processor, works similiar as
above.

What are the merits of either method. I prefer not to have a processor
involved, but do have access to FPGA resources.

Is ther a simple way to do this?

Mark

Usually the Hall Current Transducers are matched with a signal conditioner.
The conditioner handles excitation current, and offset balancing. A visit to
LEM's web site might be helpful.
 
J

John Larkin

Jan 1, 1970
0
Serious? You might like to read that one again..... Oh, I'm sure you knew
anyway.

Er, eh, wot? Did I miss something?

switch closed when no pulse

switch open when we have a pulse

looks OK to me. OP sez he knows when it will hit.
Has DM learnt about transient thermal impedance yet?

BLURP

Geez, don't confuse the poor slob; he's having enough trouble with the
steady-state stuff.

John
 
J

John Popelish

Jan 1, 1970
0
Mark said:
I'm using a LEM current transducer to measure pulsed currents. The
current sensor produces a current output which I feed to a resistor
producing a voltage that represents the current flow. The voltage feeds
an opamp for offset correction and buffering. The sensor has good
accuracy (+/- 0.5%), but suffers from an initial offset
(+/- 0.15mA) and worse, a thermal induced offset (+/- 0.35 mA over temp).

Since I know when the current pulses are coming, I want to to force the
offset to zero before taking a measurement. I've seen circuits to do
this using the following techniques:

1. An A/D followed by a D/A. With zero current input, the offset is
driven to zero, and the offset correction is latched in the D/A until
the next "calibration" cycle.

2. A/D and a digital pot controlled by a processor, works similiar as
above.

What are the merits of either method. I prefer not to have a processor
involved, but do have access to FPGA resources.

Is ther a simple way to do this?

Mark

There is a single chip solution for this sort of problem.

http://cache.national.com/ds/LF/LF198.pdf
 
T

Tony Williams

Jan 1, 1970
0
I'm using a LEM current transducer to measure pulsed currents.
The current sensor produces a current output which I feed to a
resistor producing a voltage that represents the current flow.
The voltage feeds an opamp for offset correction and buffering.
The sensor has good accuracy (+/- 0.5%), but suffers from an
initial offset (+/- 0.15mA) and worse, a thermal induced offset
(+/- 0.35 mA over temp).
Since I know when the current pulses are coming, I want to to
force the offset to zero before taking a measurement. I've seen
circuits to do this using the following techniques:

The old way of doing this would be a zero-crossing
comparator on the output volts, setting the direction
of an up/down counter, which drives a DAC. With the
counter clock running, the system auto-zeroes, with a
hunt of +/- 1-lsb around zero. The size of 1-lsb is
determined by the hysteresis of the comparator and
the resolution of the DAC. Stop the counter clock to
Hold the auto-zeroed compensating offset.

To avoid the need for a bipolar DAC put the comparator
ref input on a small positive offset, which is taken out
with an identical negative offset downstream.

For your app it looks like the DAC needs the equivalent
of at least 2mA full scale (+/- 1mA), with a resolution
of (say) better than 0.1mA. An 8 bit counter+DAC with
5mA full scale (+/- 2.5mA) would have a resolution of
0.02mA..... probably more than enough.
 
M

Mark

Jan 1, 1970
0
Tony Williams said:
The old way of doing this would be a zero-crossing
comparator on the output volts, setting the direction
of an up/down counter, which drives a DAC. With the
counter clock running, the system auto-zeroes, with a
hunt of +/- 1-lsb around zero. The size of 1-lsb is
determined by the hysteresis of the comparator and
the resolution of the DAC. Stop the counter clock to
Hold the auto-zeroed compensating offset.

To avoid the need for a bipolar DAC put the comparator
ref input on a small positive offset, which is taken out
with an identical negative offset downstream.

For your app it looks like the DAC needs the equivalent
of at least 2mA full scale (+/- 1mA), with a resolution
of (say) better than 0.1mA. An 8 bit counter+DAC with
5mA full scale (+/- 2.5mA) would have a resolution of
0.02mA..... probably more than enough.

Yes, this was one of my thoughts also. Thanks,

Mark
 
M

Mark

Jan 1, 1970
0
Luhan Monat said:
Mark,

Since you are incorporating a processor here, make 2 a/d readings: one
just before and one during the pulse. Then subtract the first from the
second.

Another foos suggestion! Thanks,

Mark
 
M

Mark

Jan 1, 1970
0
John Larkin said:
in--------c-------+----------out
|
switch
|
gnd


Just keep the switch closed when you know there's no pulse. A fet or
an CMOS analog switch thingie will work.

John

I don't see how this helps, The offset is inherent in the sensor, so
when I open the switch I still get measurement + offset.

Mark
 
F

Fred Bloggs

Jan 1, 1970
0
Mark said:
I don't see how this helps, The offset is inherent in the sensor, so
when I open the switch I still get measurement + offset.

Mark

You're right- this Larkin is off in left field most of the time- in lala
land of egomaniacal euphoria over delusion of being God's gift to
engineering- when he's not acting the shameless NG toady that is...
 
F

Fred Bloggs

Jan 1, 1970
0
Fred said:
You're right- this Larkin is off in left field most of the time- in lala
land of egomaniacal euphoria over delusion of being God's gift to
engineering- when he's not acting the shameless NG toady that is...

Then again if that "c" means capacitor-he is actually trying to
communicate an age-old technique of charging the cap to a Vos-from
left-to-right so your digitizer or whatever then reads
Vpulse+Vos-Vos=Vpulse during the measurement, so you would put your
burden resistor to the left of "c" in the diagram, drop it from "in" to
GND, to convert current to voltage, and select "c" to pass Vpulse with
negligible droop.
 
J

John Larkin

Jan 1, 1970
0
Then again if that "c" means capacitor-he is actually trying to
communicate an age-old technique of charging the cap to a Vos-from
left-to-right so your digitizer or whatever then reads
Vpulse+Vos-Vos=Vpulse during the measurement, so you would put your
burden resistor to the left of "c" in the diagram, drop it from "in" to
GND, to convert current to voltage, and select "c" to pass Vpulse with
negligible droop.


Welcome to la-la land. Can I get you a smoothie?

John
 
Top