Trigger on variation of an analog signal

[pozz]

Hi all,

I have an analog signal (0-5v) that is normally stable or slowly
changing. I want to design a simple circuit (analog and digital)
that triggers a line when a strong variation occurs in a period
of time.
For example, if in 10 minutes the signal change from 3V to 1V,
than a signal must be raised from 0 to 1 in order to wake-up a
microcontroller. After doing its work, micro untriggers the digital
signal and goes in sleep mode again.

It's important to trigger a digital line on a strong variation on
an analog signal. I can do that with micro and internal ADC, but I
want it in sleep mode.

 

[John Devereux]

Hi all,

I have an analog signal (0-5v) that is normally stable or slowly
changing. I want to design a simple circuit (analog and digital)
that triggers a line when a strong variation occurs in a period
of time.
For example, if in 10 minutes the signal change from 3V to 1V,
than a signal must be raised from 0 to 1 in order to wake-up a
microcontroller. After doing its work, micro untriggers the digital
signal and goes in sleep mode again.

It's important to trigger a digital line on a strong variation on
an analog signal. I can do that with micro and internal ADC, but I
want it in sleep mode.

Do you mean you want to detect when the rate of change of voltage with
time (Volts per second), gets above a certain limit?

The obvious "analog" solution is a differentiator, but this might be
awkward for such long time periods. You would need quite high values
of capacitance and/or resistance, with low current leakages.

Can you not instead wake the micro every few seconds/minutes to check the
signal? This might only take 1ms every second, say, so your average
power would only be 1/1000 of the peak value.

 

[cirip cipcirip]

Hi,

There are micros that have interrupt on change on some pins. If 3V to 1V
could be thought as a change, that might be an option. Otherwise, use a
comparator and feed the micro with the comparator output.

If the transition is fast enough use a simple differentiator (RC) to
generate a pulse on change.

Just a few ideas,
Cirip

 

[Tim Wescott]

Hi all,

I have an analog signal (0-5v) that is normally stable or slowly
changing. I want to design a simple circuit (analog and digital)
that triggers a line when a strong variation occurs in a period
of time.
For example, if in 10 minutes the signal change from 3V to 1V,
than a signal must be raised from 0 to 1 in order to wake-up a
microcontroller. After doing its work, micro untriggers the digital
signal and goes in sleep mode again.

It's important to trigger a digital line on a strong variation on
an analog signal. I can do that with micro and internal ADC, but I
want it in sleep mode.

You really need to define what you mean.

Do you mean if the signal varies by a certain amount in a certain time,
or do you mean if the signal just moves by a certain amount in _any_
amount of time? The first requires a differentiator that needs to
remember things for minutes at a time; this may be easiest to implement
with your microprosessor. For the second you could set a a window with
a pair of DACs and use a pair of comparators to monitor the signal.

 

[Keith Williams]

You really need to define what you mean.

Do you mean if the signal varies by a certain amount in a certain time,
or do you mean if the signal just moves by a certain amount in _any_
amount of time? The first requires a differentiator that needs to
remember things for minutes at a time; this may be easiest to implement
with your microprosessor. For the second you could set a a window with
a pair of DACs and use a pair of comparators to monitor the signal.

....or combine the two. Have the DAC(s) set a window, when the input
goes out of that window the processor wakes up and figures out the time
interval, resets the window, and goes back to sleep if the rate of
change is below the desired threshold.

 

[Joerg]

Hi Keith,

...or combine the two. Have the DAC(s) set a window, when the input
goes out of that window the processor wakes up and figures out the time
interval, resets the window, and goes back to sleep if the rate of
change is below the desired threshold.

Yes, that might be nicely done with the MSP430. Even if it sleeps with
the realtime clock running it won't consume more than a few microamps.

Regards, Joerg

 

[Ken Smith]

The obvious "analog" solution is a differentiator, but this might be
awkward for such long time periods. You would need quite high values
of capacitance and/or resistance, with low current leakages.

I disagree with this part.

A differentiator is only valid for times that are much longer than the
time constant. It isn't hard at all to get 470K resistors and 1uF
capacitors and low bias current op-amps.

Chances are you are really more interested in something more like a band
pass filter. I assume a short spike or small step on the input isn't what
you are looking for.

It will be hard to make an exact value of the RC time constant because you
will end up with 10% parts but that may not be a problem.

 

[John Devereux]

I disagree with this part.

A differentiator is only valid for times that are much longer than the
time constant.

Could you explain this a bit? I don't understand what you mean.

For example, a simple opamp differentiator would be:


V --||-+-[R]-------.
C | |
+----|- |
| >--+--- Vout
+--|+
|
---

Virtual ground at the opamp -ve input. So input current is CdV/dT. So
Vout=RC.dV/dT. If we take the OPs example, dV = 2V, dT = 10 minutes =
600 seconds. Assume we want a 5V output (roughly), so we have

5 = RC*2/600 or RC = 5*600/2 = 1500 seconds.

Hmm, perhaps I see what you are saying, in that after switch-on the
output will take much longer than this to settle. Perhaps an analog
switch accross the R, to allow the capacitor to be quickly charged to
the initial value?

It isn't hard at all to get 470K resistors and 1uF
capacitors and low bias current op-amps.

A differentiator followed by an amplifier perhaps?


V --||-+-[470k]----.
1u | | .-[R2]-.
+----|- | | |
| >--+--[R1]-+-|- |
+--|+ V1 | >-+- Vout
| +--|+
--- |
---

One could then have a comparator on Vout, to interrupt the
microcontroller. V1 = RC.dV/dT = 1u*470k*2/600 = 0.466mV. This is
effectively still a "high impedance" circuit however, since say a 1G
resistor accross the 1u leaks 5V/1G=5nA, which turns into 2.35mV,
dwarfing the required signal.

I guess everything is OK if you use low bias current opamps, as you
say, and make sure everything is kept clean and dry!

 

[Ken Smith]

A differentiator is only valid for times that are much longer than the
time constant.

Could you explain this a bit? I don't understand what you mean.

For example, a simple opamp differentiator would be:


V --||-+-[R]-------.
C | |
+----|- |
| >--+--- Vout
+--|+
|

I real life you will need a resistor in series with the C.


Try a simpler circuit:

Vin ---!!----- Vout
!
\
/
\
/
GND


If T << RC, this is a differ.

All we need to do it add enough gain to make a big enough signal to work
with and we can go into the comparitors.

Just getting a trigger on the rate of change doesn't require that big of
values.

[...]

Perhaps an analog
switch accross the R, to allow the capacitor to be quickly charged to
the initial value?

Yes this is a good idea.

A differentiator followed by an amplifier perhaps?


V --||-+-[470k]----.
1u | | .-[R2]-.
+----|- | | |
| >--+--[R1]-+-|- |
+--|+ V1 | >-+- Vout
| +--|+
--- |
---

You may want resistors in the non-inverting inputs. The bias current of
an op-amp is usually more than the offset current.

One could then have a comparator on Vout, to interrupt the
microcontroller. V1 = RC.dV/dT = 1u*470k*2/600 = 0.466mV. This is
effectively still a "high impedance" circuit however, since say a 1G
resistor accross the 1u leaks 5V/1G=5nA, which turns into 2.35mV,
dwarfing the required signal.

Spend more on your capacitor if a 1G leakage is the best you've got.

I guess everything is OK if you use low bias current opamps, as you
say, and make sure everything is kept clean and dry!

Yes everything must be clean and dry and in a closed box.

 

[John Fields]

Hi all,

I have an analog signal (0-5v) that is normally stable or slowly
changing. I want to design a simple circuit (analog and digital)
that triggers a line when a strong variation occurs in a period
of time.
For example, if in 10 minutes the signal change from 3V to 1V,
than a signal must be raised from 0 to 1 in order to wake-up a
microcontroller. After doing its work, micro untriggers the digital
signal and goes in sleep mode again.

It's important to trigger a digital line on a strong variation on
an analog signal. I can do that with micro and internal ADC, but I
want it in sleep mode.

---
If you connect your changing signal, VIN, to a resitive voltage
divider,:

VIN>---+
|
[R1]
|
+--->VOUT1
|
[R2]
|
GND>---+--->GND

Then the amplitude changes of VOUT1 will follow the amplitude changes
of VIN and will be in phase with them.


Now, if Vout1 is connected to an integrator,:

VIN>--+
|
[R1]
|
+-----+----->VOUT1
| |
| [R3]
| |
[R2] +----->VOUT2
| |
| [C1]
| |
GND>--+-----+----->GND







then as VIN changes amplitude, the amplitude of VOUT2 will also
change, but more slowly than VOUT1. That is, it will always _lag_
VOUT1. As a consequence of that, and with two exceptions, while VIN
is rising the voltage at VOUT1 will always be higher than the voltage
on VOUT2, and when VIN is falling, the voltage at VOUT1 will always be
lower than the voltage on VOUT2. The exceptions are If VIN was rising
and then starts to fall or if it was falling and then starts to rise,
when it will take a little while for VOUT2 to catch up. So, if VIN
was rising and then starts to fall there will be a short time between
when VOUT1 starts to fall and when it actually gets lower than VOUT2
and, conversely, if if VIN was falling and then starts to rise, there
will be a short time between when VOUT1 starts to rise and when it
actually gets higher than VOUT2, but for this application I don't
think that matters.

Now, assuming that VIN is stable and that you want to start a timing
interval by setting VOUT1 equal to VOUT2 we can do that by momentarily
shorting them together like this:

VIN>--+
|
[R1]
|
+-----+---+-->VOUT1
| | |
| [R3] O
| | |
| O-->|S1
[R2] |
| +------>VOUT2
| |
| [C1]
| |
GND>--+-----+------>GND

Now, (neglecting C1's leakage current) as soon as C1 charges to VOUT1,
the switch can be opened, no current will flow through R3 and, as long
as VIN doesn't move, VOUT1 will stay equal to VOUT2.


Let's hook this stuff up to a window comparator:


VIN>--+
| VCC
[R1] |
| [R6]
+------------------|-\ |
| | >--+----->DOWN
[R2] +--|+/
| S1 | U1A
+----O-->| |
| | | VCC
+--[R5]--O------+ |
| | [R7]
[R3] +--|-\ |
| | | >--+----->UP
+---------------|--|+/
| | U1B
[R4] [C1]
| |
GND>--+---------------+-------------->GND

We've added R2 and R3, equal value low resistance resistors for the
purpose of setting U1A- slightly more positive than U1A+ and U1B+
slightly more negative than U1B- when S1 is closed. This will make a
little window which will assure that the output of both comparators
will be low when S1 is closed or when when the voltage on C1 falls
within the window.

Now, (with S1 open) if we ramp up VIN to a more positive voltage, the
voltage on C1 won't be able to follow instantaneously, so the voltage
on U1A- and U1B+ will go more positive than the voltage on U1A+ and
U1A-. Since the voltage on U1A- was already more positive than the
voltage on U1A+ its output will stay low. However, if the time rate
of change of the ramp voltage is great enough, U1B+ will go more
positive than U1B-, forcing the output of U1B high for as long as the
ramp can stay far enough ahead of the changing voltage on C1. The
speed at which C1 charges and discharges is set by the voltage
difference between the ramp voltage and the voltage on C1, the
resistance of R5, the capacitance of C1, the currents associated with
the leakage current of C1 and the bias currents of the comparators so,
to a first approximation, it's RC.

Just like the output of U1A will be low and the output of U1B will
high while the ramp woltage is rising, the output of U1A will go high
and the output of U1B will go low when the ramp voltage is falling.
Then, since the output of both comparators is low when VIN is stable,
these three outputs can be used to control an up-down counter used as
a timer. That is, when both outputs are low inhibit the counter, when
U1B goes high force the counter to count up, and when U1A goes high
force the counter to count down.

Then, by using another comparator to set the zero-time point,
(momentarily closing S1 or its non-mechanical equivalent) another to
detect the voltage limit, and a magnitude comparator to detect the
counter timeout, you can do whatever you need to before you wake up
the micro. For instance, you could easily detect the condition you
described by detecting that the voltage had gone past its limit in the
time allocated since the voltage limiter comparator's output would
have gone true before the magnitude comparator's output did.

Although the implementation of hysteresis hasn't been addressed,
although you must surely consider it if your input voltage will be
varying slowly, and to do it properly you'll probably have to change
the sense of one comparator and decode the two outputs to reflect that
sense change.