How to build a Voltage limiter for a millivolt signal?

anoopak

Joined Nov 24, 2023
37
Do you mean it is a 4 R DC-bridge with a 1k load across the differential output??
or something else?

View attachment 313558or View attachment 313559

or a single supply method

View attachment 313562

Here the output from bridge is based on small imbalance between two 180 Ohm resistors. 2K is fixed.
Power supply of bridge and power supply of AD620/ADS1115 is derived from a common 26V DC Power supply [i.e.They all share common GND]

Last edited:

anoopak

Joined Nov 24, 2023
37
hi anoo,
There is also this option.
E
View attachment 313527
I have a 5Volt, 0.5Watt Zener in my box. Can you tell whether the following series resistor value calculation is correct?
Step-2 Voltage to be dropped across resistor= 15V-5V= 10V
Step-3 Maximum current through Zener = 0.5W/5V= 0.1A
Step-4 Minimum Resistor value = 10V/0.1A= 100 Ohm
Step-5 Minimum Wattage of resistor= 10V X 0.1A = 1W
**I have a doubt regarding Voltage to be taken in Step-1. Since AD620 is working between +15 and -15V , whether i have to take 30V for calculation?

ericgibbs

Joined Jan 29, 2010
19,027
Hi anoo,
The AD620 cannot output a current of 100mA.
If you wanted to consider the worst case where the Vout is say+15V, means that the voltage across the output resistor would be ~15V-5Vz = +10V.
Assuming an output limit of 20mA the series resistor would ~500R

But the maximum Vout of the AD620 is (15V-1.2V), so that is (13.8V-5Vz)/20mA = 440R.
A 470R though 1K would be my choice.

E

Attachments

• 45.4 KB Views: 2
Last edited:

ericgibbs

Joined Jan 29, 2010
19,027
V1 is not a voltage source but some unknown values of a Wheatstone bridge, but may work if they are low enough R values. We need to see the bridge datasheet link. Also STP cables ought to be used to reduce stray line noise, preferably with PE gnd connection to shield.
Hi tony,
You obviously have not read my posts correctly, they are related to my suggestion of fitted a resistor/zener clamp on the Output of the AD620, no reference in my posts to modifying the AD620 inputs.

As I said, my simulation shows the effect of the Zener clamp, nothing else.

The attached simulation image shows three options in driving the clamp, they give the same result.

E

anoopak

Joined Nov 24, 2023
37
Hi anoo,
The AD620 cannot output a current of 100mA.
If you wanted to consider the worst case where the Vout is say+15V, means that the voltage across the output resistor would be ~15V-5Vz = +10V.
Assuming an output limit of 20mA the series resistor would ~500R

But the maximum Vout of the AD620 is (15V-1.2V), so that is (13.8V-5Vz)/20mA = 440R.
A 470R though 1K would be my choice.

E

View attachment 313601
Thanks E.

tonyStewart

Joined May 8, 2012
132
Hi Eric,
Apologies for my reading and misinterpretation. Brain fart. Thanks for your patience. I had no major issues with Zener clamp. But now as I think about it...

That will prevent accidental overvoltage, although undervoltage of -0.7 est. of Vf of Zener would exceed limit of ADSxxxx from SCR effect. I recall, this is why I recommended current limiting with 2kohm R's which would drive the fast little ESD protection diodes.

Hi anoo,

The peak signal must also stay within Vcm range to 0V of ADS. Although it is a unipolar gravity force +ve input, vibrations x 100 gain "could" undershoot -300 mV, if not mechanically damped. (which may result in CMOS latchup)

GND – 0.3
VDD + 0.3 V"

1. Therefore I think current limiting input to < 5mA or even 1mA with 10k is safer, given that input current is in nanoamps. (e-9) (see below plot)

Notice how they avoid absolute input levels of 0V which is your nominal DC level. So I cannot predict exact input current even though it is small.

2. To avoid potential vibration undershoot, you could use the REF pin to raise the output and also Vin- of ADC to some value above 0V such as 0.5V which limits input ringing protection to (500 + 300 mV)/100 gain= -8mV which you can verify if that is a practical physical limit for your scale.

Rigid platforms for scales tend to have a high Q > 5 even with elastomers with step or pulse input excitation. There are correlation tables between % overshoot (or under) and Q resonance factor. (if you are interested). I recall verifying even "Lord mounts" have a Q of 5.
3. How you choose to prevent overshoot and undershoot depends on your desired design specs which includes environment and voltage or current limiting.
4. I also noticed your strain gauges of 180 ohms will each dissipate 1 Watt at 24 or 26V which causes a large temperature rise. This is why I suggested a current source or at least a lower voltage source, which also reduces gain which is why I added a gain of 10 in ADC. Any ripple on this source will also add noise. Although I do not know what the supplier recommends. A constant current (CC) source is called the Howland CC using an Op Amp with a Vref.
I have read there is no clear advantage of CV vs CC for driving a bridge, yet 2 watts total seems excessive.

FWIW, traditionally if your noise is low by design, analog voltage of 3.3 = AVdd is adequate and also suited to ADC's which use the same Vref.
Then using a common mode of AVdd/2 still allows a full input differential swing.
Cheers,
Tony

Last edited:

anoopak

Joined Nov 24, 2023
37
Hi Eric,
Apologies for my reading and misinterpretation. Brain fart. Thanks for your patience. I had no major issues with Zener clamp. But now as I think about it...

That will prevent accidental overvoltage, although undervoltage of -0.7 est. of Vf of Zener would exceed limit of ADSxxxx from SCR effect. I recall, this is why I recommended current limiting with 2kohm R's which would drive the fast little ESD protection diodes.

Hi anoo,

The peak signal must also stay within Vcm range to 0V of ADS. Although it is a unipolar gravity force +ve input, vibrations x 100 gain "could" undershoot -300 mV, if not mechanically damped. (which may result in CMOS latchup)

GND – 0.3
VDD + 0.3 V"

1. Therefore I think current limiting input to < 5mA or even 1mA with 10k is safer, given that input current is in nanoamps. (e-9) (see below plot)

Notice how they avoid absolute input levels of 0V which is your nominal DC level. So I cannot predict exact input current even though it is small.

2. To avoid potential vibration undershoot, you could use the REF pin to raise the output and also Vin- of ADC to some value above 0V such as 0.5V which limits input ringing protection to (500 + 300 mV)/100 gain= -8mV which you can verify if that is a practical physical limit for your scale.
View attachment 313617

Rigid platforms for scales tend to have a high Q > 5 even with elastomers with step or pulse input excitation. There are correlation tables between % overshoot (or under) and Q resonance factor. (if you are interested). I recall verifying even "Lord mounts" have a Q of 5.
3. How you choose to prevent overshoot and undershoot depends on your desired design specs which includes environment and voltage or current limiting.
4. I also noticed your strain gauges of 180 ohms will each dissipate 1 Watt at 24 or 26V which causes a large temperature rise. This is why I suggested a current source or at least a lower voltage source, which also reduces gain which is why I added a gain of 10 in ADC. Any ripple on this source will also add noise. Although I do not know what the supplier recommends. A constant current (CC) source is called the Howland CC using an Op Amp with a Vref.
I have read there is no clear advantage of CV vs CC for driving a bridge, yet 2 watts total seems excessive.
View attachment 313620

FWIW, traditionally if your noise is low by design, analog voltage of 3.3 = AVdd is adequate and also suited to ADC's which use the same Vref.
Then using a common mode of AVdd/2 still allows a full input differential swing.
Cheers,
Tony
This circuit is for calorimetry and its necessary to pass 50-60mA current through 180 Ohm resistors for heating purpose. So the supply voltage of bridge cannot be reduced. [Giving different supply for heating and bridge excitation can complicate things].
I am wondering whether a 10k resistor along with Zener option will have advantage of both Tony's and E's suggestions.