Maker Pro
Maker Pro

for op-amp wizards: subtractor circuit

I'm looking at the schematic, Figure 3 at
http://www.analog.com/UploadedFiles/Application_Notes/58837442106604AN244.pdf

I want to use this for high-side current sensing. I start running into
problems when either op-amp outputs have to fall to near 0. I am using
an op-amp that supposedly works down to 0, but it's not getting close
enough. I'm working with a single polarity supply, and a 0.1 ohm sense
resistor. I am using 10K for R1-R4, and 2.5K for Rg....such that the
output is 10 times the input difference (1 volt per amp).

Now if I could shift the output up 1 volt, such that zero current is 1
volt, 1 amp is 2 volts, etc, that might solve my problem. But I'm
trying to avoid negative voltages. I realize if I tie the grounded end
of R4 to a voltage it will do exactly what I want, but the output of
one op-amp has to drop below 0 for low current measurements.
Any ideas?
 
J

John Popelish

Jan 1, 1970
0
I'm looking at the schematic, Figure 3 at
http://www.analog.com/UploadedFiles/Application_Notes/58837442106604AN244.pdf

I want to use this for high-side current sensing. I start running into
problems when either op-amp outputs have to fall to near 0. I am using
an op-amp that supposedly works down to 0, but it's not getting close
enough. I'm working with a single polarity supply, and a 0.1 ohm sense
resistor. I am using 10K for R1-R4, and 2.5K for Rg....such that the
output is 10 times the input difference (1 volt per amp).

Now if I could shift the output up 1 volt, such that zero current is 1
volt, 1 amp is 2 volts, etc, that might solve my problem. But I'm
trying to avoid negative voltages. I realize if I tie the grounded end
of R4 to a voltage it will do exactly what I want, but the output of
one op-amp has to drop below 0 for low current measurements.
Any ideas?
You need an opamp that has an input common mode range that includes
the positive rail (if the current sense resistor in in that rail) and
an output range that goes down to ground. Are you sure your problem
isn't the input common mode range? What opamp and supply voltage are
you dealing with?

Of course, since current shunt voltage measurements do not usually
require a high input impedance subtractor, you might do better with
the single opamp version shown in figure 2.
 
It's the LM324, at least for now. I've used the 324 in a previous
power supply design, and it works great. I may go to a better op-amp
later but that's beside the point.....

As for the positive rail thing, I'm not too worried about that now.
The op-amp will be powered with much more voltage than the sense
resistor will see.

Here's the thing. The 324 output gets very close to ground, especially
with a pull down resistor on the output. But not close enough. I have
a 47ohm pull down and it drops down to 24mV. I'm going for precision
here. In a previous, completely different current sense design, I
dealt with this problem by bringing up the output voltage floor to
2.5V. No op-amp output had to drop below 2.5V. Then I measured from
the op-amp output to the 2.5V reference. Don't laugh; despite using a
lowly 324 and dealing with millivolt levels, that current meter is
*very* accurate.

I may tinker with that other circuit you pointed out since it's simpler
and possibly easier to analyze and tweak for what I want.
 
It's the LM324, at least for now. I've used the 324 in a previous
power supply design, and it works great. I may go to a better op-amp
later but that's beside the point.....

As for the positive rail thing, I'm not too worried about that now.
The op-amp will be powered with much more voltage than the sense
resistor will see.

Here's the thing. The 324 output gets very close to ground, especially
with a pull down resistor on the output. But not close enough. I have
a 47ohm pull down and it drops down to 24mV. I'm going for precision
here. In a previous, completely different current sense design, I
dealt with this problem by bringing up the output voltage floor to
2.5V. No op-amp output had to drop below 2.5V. Then I measured from
the op-amp output to the 2.5V reference. Don't laugh; despite using a
lowly 324 and dealing with millivolt levels, that current meter is
*very* accurate.

I may tinker with that other circuit you pointed out since it's simpler
and possibly easier to analyze and tweak for what I want.

Have a look at the National Semiconductor LM10 - it includes a 200mV
reference output, which might replace your 2.5V reference in a useful
way.

The op amp itself is old, but interesting - Bob Widlar was years ahead
of everybody else.
Farnell still carries the LM10CN and the (slightly cheaper) LM10CLN.

http://cache.national.com/ds/LM/LM10.pdf
 
F

Fred Bloggs

Jan 1, 1970
0
The op amp itself is old, but interesting - Bob Widlar was years ahead
of everybody else.

No question about it, recently finished downloading all of his published
papers, letters, and briefs. But I think the modern process improvements
and high fidelity physics simulations would have deprived him of the
work he liked best.
 
T

Tim Wescott

Jan 1, 1970
0
I'm looking at the schematic, Figure 3 at
http://www.analog.com/UploadedFiles/Application_Notes/58837442106604AN244.pdf

I want to use this for high-side current sensing. I start running into
problems when either op-amp outputs have to fall to near 0. I am using
an op-amp that supposedly works down to 0, but it's not getting close
enough. I'm working with a single polarity supply, and a 0.1 ohm sense
resistor. I am using 10K for R1-R4, and 2.5K for Rg....such that the
output is 10 times the input difference (1 volt per amp).

Now if I could shift the output up 1 volt, such that zero current is 1
volt, 1 amp is 2 volts, etc, that might solve my problem. But I'm
trying to avoid negative voltages. I realize if I tie the grounded end
of R4 to a voltage it will do exactly what I want, but the output of
one op-amp has to drop below 0 for low current measurements.
Any ideas?
feedback
o---------------------.
|
|
|\ |
-|-\ | out
| >--->|---o-o--------o
-|+/ |
|/ .-.
| |
| |
'-'
|
|
===
GND
(created by AACircuit v1.28.6 beta 04/19/05 www.tech-chat.de)

Granted, there will still be some leakage from the diode, but what do
you want for free?

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Posting from Google? See http://cfaj.freeshell.org/google/
 
F

Fred Bloggs

Jan 1, 1970
0
I'm looking at the schematic, Figure 3 at
http://www.analog.com/UploadedFiles/Application_Notes/58837442106604AN244.pdf

I want to use this for high-side current sensing. I start running into
problems when either op-amp outputs have to fall to near 0. I am using
an op-amp that supposedly works down to 0, but it's not getting close
enough. I'm working with a single polarity supply, and a 0.1 ohm sense
resistor. I am using 10K for R1-R4, and 2.5K for Rg....such that the
output is 10 times the input difference (1 volt per amp).

Now if I could shift the output up 1 volt, such that zero current is 1
volt, 1 amp is 2 volts, etc, that might solve my problem. But I'm
trying to avoid negative voltages. I realize if I tie the grounded end
of R4 to a voltage it will do exactly what I want, but the output of
one op-amp has to drop below 0 for low current measurements.
Any ideas?

You're nearly there. Replace the GND at R4 in Figure 3 with your
reference voltage. This will shift the output upwards by the same
voltage. It also reduces the existing CM range headroom by the same amount.
 
I was thinking of something like this, but that pull down resistor
should ideally be zero to achieve zero output voltage....something that
isn't going to work.
Hmmm, but if I used larger value resistors in the rest of the circuit
this might work.

BTW, the circuit you suggested for the switching and linear power
supply (the diff amp with 2 resistors, BJT and diode) works
*beautifully*. The op amp circuits jumped all over the place. Simple
and stable. I love it.
 
K

Ken Smith

Jan 1, 1970
0
I'm looking at the schematic, Figure 3 at
http://www.analog.com/UploadedFiles/Application_Notes/58837442106604AN244.pdf [...]
Now if I could shift the output up 1 volt, such that zero current is 1
volt, 1 amp is 2 volts, etc, that might solve my problem. But I'm
trying to avoid negative voltages. I realize if I tie the grounded end
of R4 to a voltage it will do exactly what I want, but the output of
one op-amp has to drop below 0 for low current measurements.
Any ideas?

How about this for a different attack:

You can make Vcc/2 as your "signal ground". This will mean that the
op-amps will all be at the center of their swing for zero volts. R-R
opamps tend to have extra bias current when they are running near the
rail. Putting the "zero" at the center of the swing avoids this issue.
 
J

John B

Jan 1, 1970
0
[email protected] scrobe on the papyrus:
I'm looking at the schematic, Figure 3 at
http://www.analog.com/UploadedFiles/Application_Notes/58837442106604AN
244.pdf

I want to use this for high-side current sensing. I start running
into problems when either op-amp outputs have to fall to near 0. I
am using an op-amp that supposedly works down to 0, but it's not
getting close enough. I'm working with a single polarity supply, and
a 0.1 ohm sense resistor. I am using 10K for R1-R4, and 2.5K for
Rg....such that the output is 10 times the input difference (1 volt
per amp).

Now if I could shift the output up 1 volt, such that zero current is 1
volt, 1 amp is 2 volts, etc, that might solve my problem. But I'm
trying to avoid negative voltages. I realize if I tie the grounded
end of R4 to a voltage it will do exactly what I want, but the output
of one op-amp has to drop below 0 for low current measurements.
Any ideas?

Have you considered the AMP04? Have a look at:

http://www.analog.com/en/prod/0,,759_782_AMP04,00.html
 
J

John B

Jan 1, 1970
0
[email protected] scrobe on the papyrus:
I'm looking at the schematic, Figure 3 at
http://www.analog.com/UploadedFiles/Application_Notes/58837442106604AN
244.pdf

I want to use this for high-side current sensing. I start running
into problems when either op-amp outputs have to fall to near 0. I
am using an op-amp that supposedly works down to 0, but it's not
getting close enough. I'm working with a single polarity supply, and
a 0.1 ohm sense resistor. I am using 10K for R1-R4, and 2.5K for
Rg....such that the output is 10 times the input difference (1 volt
per amp).

Now if I could shift the output up 1 volt, such that zero current is 1
volt, 1 amp is 2 volts, etc, that might solve my problem. But I'm
trying to avoid negative voltages. I realize if I tie the grounded
end of R4 to a voltage it will do exactly what I want, but the output
of one op-amp has to drop below 0 for low current measurements.
Any ideas?

You should also look at this app note:

http://www.maxim-ic.com/appnotes.cfm/appnote_number/746
 
T

Terry Given

Jan 1, 1970
0
Fred said:
No question about it, recently finished downloading all of his published
papers, letters, and briefs. But I think the modern process improvements
and high fidelity physics simulations would have deprived him of the
work he liked best.

is that all nicely lumped together somehwer? care to provide a link?

Cheers
Terry
 
T

Tony Williams

Jan 1, 1970
0
Here's the thing. The 324 output gets very close to ground,
especially with a pull down resistor on the output. But not
close enough. I have a 47ohm pull down and it drops down to
24mV.

The LM324 and 47 ohm are being asked to sink the
current through that final 10K feedback resistor,
whose top end is at the voltage on the shunt.
That 24mV suggests that the supply being measured
is around the 6V mark.

With a low impedance source carrying Amps you can
afford to pull the odd 1mA off it, so you don't need
to struggle with diff amp.

0R1
Vs(6V?) ------+/\/\/+----------------->Load+
| |
[100R] +-------------------------+
| |
+------[1k]----+---+---------+ |
| | | | |
| 100p=== /_\D _ | |
+--| | | / -|--+ |
ptype||--[100R]--+---+--<324| |
+--| \_+|-----+
| +----------+
| | _ |
1mA/A \|/ +--|- \ |
| |324>---+----->Vout+
1V/A-->+--------|+_/ |
| |
[1k] [1k]
| |
0v------------+----------------+----->Vout-
|
+---------------------->Load-

The top opamp simply develops a current of 1mA/A
into the 1k down on 0V. Generating a constant
current is the piece of elastic that takes away
the CMV. The lower opamp is just a buffer which
no longer has the offset due to having to sink
current. The ptype MOSFET could be a pnp darlington.
 
The LM324 and 47 ohm are being asked to sink the
current through that final 10K feedback resistor,
whose top end is at the voltage on the shunt.
That 24mV suggests that the supply being measured
is around the 6V mark.

With a low impedance source carrying Amps you can
afford to pull the odd 1mA off it, so you don't need
to struggle with diff amp.

0R1
Vs(6V?) ------+/\/\/+----------------->Load+
| |
[100R] +-------------------------+
| |
+------[1k]----+---+---------+ |
| | | | |
| 100p=== /_\D _ | |
+--| | | / -|--+ |
ptype||--[100R]--+---+--<324| |
+--| \_+|-----+
| +----------+
| | _ |
1mA/A \|/ +--|- \ |
| |324>---+----->Vout+
1V/A-->+--------|+_/ |
| |
[1k] [1k]
| |
0v------------+----------------+----->Vout-
|
+---------------------->Load-

The top opamp simply develops a current of 1mA/A
into the 1k down on 0V. Generating a constant
current is the piece of elastic that takes away
the CMV. The lower opamp is just a buffer which
no longer has the offset due to having to sink
current. The ptype MOSFET could be a pnp darlington.
This is intriguing and I'm going to try it.
 
Top