Which component to detect any changes on analog circuit?

[Eki27]

Hello,

Could you please help me to find the reference of a product able to do the folowing job:

Supervise any voltage change (positive or negative) of at least 10mV on an analog circuit (which can take a variable value from 0V to 1,2V) and produce an output pulse (ideally a 0V for +-300 mS) for each detected changes?

I will use the output of the component to wake another µP which will convert the analog value to numeric...
The input of the component will be connected in parallel with the input of the another µP, the component should therefore been high impedance input.

Thank you in advance for your suggestions and have a nice evening!

Eki

 

[alec_t]

It would help to know over what periods do the 10mV change and 0-1,2V variation occur?
What do you mean by +-300mS?

 

[Eki27]

The circuit is normally stable, if the voltage change (rarely, 1 time per minute maximum) I would like to active the output for 300mS (output pulse time).

All these values are not really important because all treatment are made by another µC, I just need a kind of supervisor (if the voltage change, I just need to be informed...)

Thanks,
Eki

 

[alec_t]

Let me get this right. Are you saying the voltage is normally set (i.e. is intended to be fixed) somewhere between 0 and 1.2V (the stable voltage) but you want to detect any deviation from that setting by ~10mV in a 1 min or so period?
A rapid change of ~10mV could be easily detected by analogue means, but in view of the long time period in which voltage drift could occur I think using a microcontroller with an accurate high-resolution (10-bit at least) ADC to monitor the voltage over time would be preferable to an analogue solution.

 

[Eki27]

The goal is to supervise a large kind of sensor devices, these devices uses dual resistors in order to get 4-states information.

I will use a stabilized 1,2VDC as source, crossing a voltage divider resistor where R1 is the resistance value inside the sensor. The sensor can have any kind of resistor values, even no resistors (dry contact).
I could define the resistors value to use in sensors and it's done, but I would like to keep the freedom of using any kind.

Don't care about Adjusting the voltage tolerance and threshold, that will be made at higer level by the "master IC ADC" (14bits resolution), and can be changed for each input by software...

I dont want to detect a value with the component that I'm searching for, I just want to give an information of any change of minimum 10mV from the sensor circuit, in order to wake my ADC and sample the value...

-> the voltage change will be always fast, because the information come from a switch

I think my needs is so simple, that you look too far

Now I rephrase my question; do you now an IC able to sense (for example at 4Hz) an analog voltage (in a range of 0 to 1,2V), record the previous value and compare with the next in order to pulse (200mS, automatic reset) an output in case of voltage variation (10 mV minimum)?

Thank you
Eki

 

[Gorgon]

Ok, so what you want is a tracking comparator, to adjust to the different levels.

You say that the changes comes fast. How fast?

What is your budget for power here?
What is your budget for cost?
How many inputs is used in each construction?

Just to mention it, if you put 4Hz into this circuit you'll get 8 triggers, a bit crowded for 200ms pulses? If this pulse is reset for processing MCU you will not get any processing time.

 

[Eki27]

Indeed, an IC comparator with an output pulse auto-reset.

Only the first trigger will be consider by the other µP (it is independant working), it will wake an do it's job until it finish, if other triggers comes while the job is in progress, it will be ignored, thats right for my needs.

The power supply is already provided by the other µP.
Only one input is enough, 4 inputs is the best.
The budget dont care actually (will be considered if there is any alternatives)

Do you have an IC part number in mind?

Thank you,
Eki

 

[crutschow]

Because of the long times involved, an analog approach is difficult (due to leakage drift), so a microprocessor approach would be easiest. A small 8-bit processor with a built-in A/D converter would work.

Upon a trigger input the processor measures the voltage with the A/D and stores the value. Then you continue to take readings and compare them with the stored value. When the difference is more that 10mV, the processor sends out the pulse. An 8-bit A/D would give better than 5mV resolution if you amplified the 1.2V with an op amp to the full-range value of the A/D input (typically 5V).

 

[Gorgon]

Indeed, an IC comparator with an output pulse auto-reset.

Only the first trigger will be consider by the other µP (it is independant working), it will wake an do it's job until it finish, if other triggers comes while the job is in progress, it will be ignored, thats right for my needs.

The power supply is already provided by the other µP.
Only one input is enough, 4 inputs is the best.

Ok, what I asked for is how much power you can use for this.

If power is not a question, I would use the MCU you already got for this. You can send it to sleep and wake it up to read the inputs at a certain interval, check the value and do the rest of the processing if any of the the read values is 10mV changed. If the problem is that you use a premade solution, I would go for the microcontroller solution as sketched before.

If you on the other hand, need to make a solid state variant, I would make a 'combined' version. You can use a comparator to trigger the change, and use a DAC(or more) triggered from the existing MCU to set the test voltage(s). This removes the problem with the timing constant in the pure analougue version.

But, I think this is the last resort.

 

[Eki27]

Sorry Gorgon, I misunderstand the question about the power, indeed it's important to save power (cause of battery powered solution) and it's the reason why I can't let the MCU always online (otherwise my problem will be solved).

The constant timing is not enough, it already does it for beacons reasons but it needs to react immediately if a "big" change occurs.

All of that is exactely what I'm looking for, an IC "all in the box" consumming a few µA dedicated to do the wake job of the MCU.

Even I searched over thousand technical sheet of product without success these last days, I still can't believe a suitable taylored device doesn't already exist on the market for this kind of purpose!?

Crutshow thinks that's not workable with analog and I dont anderstand everything about the problem of leakage drift in this case. However, remember it's not a problem if the IC wakes the MCU sometimes for "nothing".

Thank you,
Eki

 

[Gorgon]

Then we're back to the question of how fast the changes are. Could you expand on what kind of signal you are measuring?

The reason a pure analogue solution is problematic is that to detect a deltaV of 10mV you need to have a reference point to the last value measured. This should be the last sample in the MCU. If you only reference against a slow changing analogue value, you may glide over the triggerpoint not knowing.
If you can set a reference for each input that is passed for each change (more or less a 2 point input), and only adjust this point for each input, you may use a simple comparator. If not you need a dynamic feedback from the MCU like I said in my last post.

You could use a sample and hold circuit, but the hold part need to be very good and with little leakage.

 

[alec_t]

As I said in post #4, and as Crutschow repeats in post #8, digital is the preferred way to go. You are asking a lot of an analogue circuit, as the reference 1.2V may be subject to noise or drift and has to be filtered and tracked. That implies some sort of integrator/smoothing circuit with a time constant of several minutes, hence large-value electrolytic capacitors and consequently leakage currents to consider. Lengthy time periods are child's play for an MCU, of course.

 

[crutschow]

It's difficult to store an analog voltage for a minute or so without the voltage drifting more than 10mV. That is why I and others suggested the microprocessor solution.

To minimize power you could use a low power microprocessor (PIC among others build them). You put it into the sleep mode between conversions (where some only take nA of current), and periodically (4 times/sec) wake it with an input strobe to take a conversion.

 

[Eki27]

Ok, I exclude the analog solution, thank you for explanations.

Programming PIC is not really in my field (no equipment, no experience).

I thought that some specialized microcontroller with ADC+stuff for this kind of purpose could be ordered from electronics suppliers... with the possibility of choosing the sampling rate, voltage range, power supply type, offset range... just by specifying the right product code into a device family (I'm probably a dreamer )

Conclusion, I will use simple comparators with fixed threshold, resistors in sensors will always be of the same value.

Thanks to all for your advices,
Eki