How to put a constant v-drop in a signal?

[Michael]

Hey there - I've now run into this problem twice in the last week. I
want to be able to insert a constant voltage drop into a signal.

The first place where this would have been nice was that I had a
comparator watching the drain and the source of an N-FET. I wanted the
comparator to toggle the instant the drain went higher than the source
(and the other way around, naturally). Unfortunately, the comparator
that I was stuck using had fairly lousy input offset. It was OK for me
to error more towards one state of the comparator, but not towards the
other. So what I figured was that I'd drop one signal by as much as
the worst case input offset (1.5mV). But I know of no way to do that.
Instead, I just used a voltage divider that just barely dropped it by
a constant amount, and since I knew the range of voltages that I'd be
able to see I was able to make the minimum constant amount large
enough to always be larger than the input offset.

Anyways - that worked - but dropping by a constant voltage would have
better.

Now I have another application of a similar thing. I need to monitor
the voltage of a signal that ranges from about 12-18V. It will be
monitored by a singled ended ADC with a 3.0V max input. So I can just
divide the signal by 6. However, that voltage will always range from
2-3V, so I'm losing two thirds of my ADC range. What a waste. What'd
be awesome would be if I could put in a constant drop in that voltage
so that I could use the full range of my ADC - either a 12V drop or a
2V drop would work, clearly. There are shunt references with voltages
this large - so that'd definitely do it (but be fairly costly and
large, as it seems most shunts are fairly large parts). Zeners are
typically what I think about 5% or so, so that would be less than
ideal.

Am I missing some better way of doing this? I suspect a 10V or 2V drop
is a lot easier to achieve than a 1.5mV drop, but I've been wrong
before!

Thanks!

-Michael

 

[James Arthur]

Hey there - I've now run into this problem twice in the last week. I
want to be able to insert a constant voltage drop into a signal.

The first place where this would have been nice was that I had a
comparator watching the drain and the source of an N-FET. I wanted the
comparator to toggle the instant the drain went higher than the source
(and the other way around, naturally). Unfortunately, the comparator
that I was stuck using had fairly lousy input offset. It was OK for me
to error more towards one state of the comparator, but not towards the
other. So what I figured was that I'd drop one signal by as much as
the worst case input offset (1.5mV). But I know of no way to do that.
Instead, I just used a voltage divider that just barely dropped it by
a constant amount, and since I knew the range of voltages that I'd be
able to see I was able to make the minimum constant amount large
enough to always be larger than the input offset.

Anyways - that worked - but dropping by a constant voltage would have
better.

Now I have another application of a similar thing. I need to monitor
the voltage of a signal that ranges from about 12-18V. It will be
monitored by a singled ended ADC with a 3.0V max input. So I can just
divide the signal by 6. However, that voltage will always range from
2-3V, so I'm losing two thirds of my ADC range. What a waste. What'd
be awesome would be if I could put in a constant drop in that voltage
so that I could use the full range of my ADC - either a 12V drop or a
2V drop would work, clearly. There are shunt references with voltages
this large - so that'd definitely do it (but be fairly costly and
large, as it seems most shunts are fairly large parts). Zeners are
typically what I think about 5% or so, so that would be less than
ideal.

Am I missing some better way of doing this? I suspect a 10V or 2V drop
is a lot easier to achieve than a 1.5mV drop, but I've been wrong
before!

Thanks!

-Michael

You can put a dropping element in series with the signal.

Zeners around 6.2v are pretty stable, or you could use
an LMV431-type thing for an adjustable drop.

Another way is to insert a series resistor, then create
the desired offset with a constant current sink.

Flying capacitors are another way--accurate & stable.

HTH,
James Arthur

 

[Joerg]

You can put a dropping element in series with the signal.

Zeners around 6.2v are pretty stable, or you could use
an LMV431-type thing for an adjustable drop.

Another way is to insert a series resistor, then create
the desired offset with a constant current sink.

Flying capacitors are another way--accurate & stable.

Can't see the OP (please use something better than Google if that's the
domain you used).

Anyhow, the OP might not gain much if the ADC is already 12bits or
better. Zeners are horrible in precision and a LMV431 type thingie will
be around 1%. So it's usually better to waste 2/3rd of the ADC range.
More precise drops are possible but then the cost will be so painful
that upping the ADC by two bits should cost less.

 

[James Arthur]

Can't see the OP (please use something better than Google if that's the
domain you used).

Anyhow, the OP might not gain much if the ADC is already 12bits or
better. Zeners are horrible in precision and a LMV431 type thingie will
be around 1%. So it's usually better to waste 2/3rd of the ADC range.
More precise drops are possible but then the cost will be so painful
that upping the ADC by two bits should cost less.

It all depends on the missing details.

Current sink + precision resistor could be quite accurate; flying
cap stuff the same, and cheap.

If all he needs it monotonicity, anything works.

Cheers,
James Arthur

 

[[email protected]]

Hey there - I've now run into this problem twice in the last week. I
want to be able to insert a constant voltage drop into a signal.

The first place where this would have been nice was that I had a
comparator watching the drain and the source of an N-FET. I wanted the
comparator to toggle the instant the drain went higher than the source
(and the other way around, naturally). Unfortunately, the comparator
that I was stuck using had fairly lousy input offset. It was OK for me
to error more towards one state of the comparator, but not towards the
other. So what I figured was that I'd drop one signal by as much as
the worst case input offset (1.5mV). But I know of no way to do that.
Instead, I just used a voltage divider that just barely dropped it by
a constant amount, and since I knew the range of voltages that I'd be
able to see I was able to make the minimum constant amount large
enough to always be larger than the input offset.

Anyways - that worked - but dropping by a constant voltage would have
better.

Now I have another application of a similar thing. I need to monitor
the voltage of a signal that ranges from about 12-18V. It will be
monitored by a singled ended ADC with a 3.0V max input. So I can just
divide the signal by 6. However, that voltage will always range from
2-3V, so I'm losing two thirds of my ADC range. What a waste. What'd
be awesome would be if I could put in a constant drop in that voltage
so that I could use the full range of my ADC - either a 12V drop or a
2V drop would work, clearly. There are shunt references with voltages
this large - so that'd definitely do it (but be fairly costly and
large, as it seems most shunts are fairly large parts). Zeners are
typically what I think about 5% or so, so that would be less than
ideal.

Am I missing some better way of doing this? I suspect a 10V or 2V drop
is a lot easier to achieve than a 1.5mV drop, but I've been wrong
before!

Search for an adder/subtractor circuit such as

http://homepages.which.net/~paul.hills/Circuits/Adder/Adder.html

Use a stable voltage reference as the source of the offset you want
and scale it up or down as needed.

 

[whit3rd]

Hey there - I've now run into this problem twice in the last week. I
want to be able to insert a constant voltage drop into a signal.

... signal that ranges from about 12-18V. It will be
monitored by a singled ended ADC with a 3.0V max input.

So, use an op amp and make an amplifier that has gain of
0.5 and DC offset of -12V. If you use the reference voltage
from the ADC for the offset, it'll be as stable as your
converter, and two calibration points will be the only
setup requirement.

I've also seen (in very low-noise circuits) a mercury battery
and a potentiometer in series with the signal wire.

 

[James Arthur]

So, use an op amp and make an amplifier that has gain of
0.5 and DC offset of -12V. If you use the reference voltage
from the ADC for the offset, it'll be as stable as your
converter, and two calibration points will be the only
setup requirement.

I did exactly that way back when in a product that needed
custom offset and gain tailoring per channel.

I second the op-amp adder approaches, but they're kind of a pain:
then you have gain and offsets to adjust, and common-mode errors
to consider.

I'd thought the OP just wanted a simple offset, but I see
he needs a scale factor too. Missed that.

I've also seen (in very low-noise circuits) a mercury battery
and a potentiometer in series with the signal wire.

Neat idea.

Cheers,
James Arthur

 

[Joerg]

I did exactly that way back when in a product that needed
custom offset and gain tailoring per channel.

I second the op-amp adder approaches, but they're kind of a pain:
then you have gain and offsets to adjust, and common-mode errors
to consider.

I'd thought the OP just wanted a simple offset, but I see
he needs a scale factor too. Missed that.



Neat idea.

Those are frowned upon by the EPA and others. I had to modify my old
Minolta camera to 1.55V since mercury cells have become unobtanium.

 

[[email protected]]

I did exactly that way back when in a product that needed
custom offset and gain tailoring per channel.

I second the op-amp adder approaches, but they're kind of a pain:
then you have gain and offsets to adjust, and common-mode errors
to consider.

I'd thought the OP just wanted a simple offset, but I see
he needs a scale factor too. Missed that.


Neat idea.

The Weston standard cell is a better voltage reference if you want to
go that way, but the better semiconductor references are more stable
(though the tolerance on the output voltage tends to be worse) and
they are a lot more compact.

http://en.wikipedia.org/wiki/Weston_cell