Precise constant current source for NTC

AI Thread Summary
The discussion focuses on designing a precise constant current source for an NTC thermistor, with resistance ranging from 1kΩ to 13kΩ. Users are experiencing issues with current precision, noting a 1µA to 2µA tolerance that persists despite circuit modifications, including switching from a BJT to a MOSFET and using a voltage reference instead of a zener diode. Suggestions emphasize the importance of increasing the supply voltage to ensure the BJT operates in the active region and recommend using high-gain transistors or dedicated constant current ICs for better accuracy. Users are encouraged to build the circuit physically, as simulation results may not reflect real-world performance. The conversation concludes with a recommendation to choose MOSFETs with lower capacitance for improved circuit stability.
wasabilumen
Messages
9
Reaction score
0
Hello, everyone,

I have to design a constant current source for NTC, to have a more precise measurement.
The NTC resistance used range is 1kΩ~13kΩ. I have tried several circuits, but the problem is the the current is not precise enough even in simulation.

1.png

2.png

1uA difference

All the suggestions will be wellcome!
 
Engineering news on Phys.org
Please provide the data sheet of the zener diode RD2.0S

What do you mean by not accurate? Do you mean the current read by the meter change as you change the resistance R1?

The circuit you have is not a bad design, the voltage of the zener is questionable. It might drift or change with the current drawn. It is particular bad when you use 330Ω R4. That is very low and might draw too much current if the Vz is too low. If you want precision, you need a voltage reference.

If it is the current change due to adjusting R1, the only improvement is to use a MOSFET instead of BJT, but you only have 5V and that can cause problem.
 
hello, yungman, thanks for answer.

1. datasheet of zener diode is attached.

2.yes,by not accurate. I mean the current read by the meter change as I change the R1.

3.Then should I change a zener diode or change a R1?
I calculate the R1, IDiode = (Usource - Uzd) / R1
a.R must be small enough that the current through D keeps D in reverse breakdown.
5mA>(5v-2v)/R1 so R1<600Ω
b.R must be large enough that the current through D does not destroy the device
Id*Ud<200mw so R1>30Ω
so ,I choose 330Ω,which is between 30Ω~600Ω

4. It is the current due to adjusting R1, so I can only increase 5v? then change BJT to MOSFET? Drive gate of Mosfet need 3v, so 8v will be enough?

5.In simulation ,I tried to check the voltage an current drawn of zener during changing the R1, none of them change. Will this situation be different when it comes to the real world?

6.In simulation, I try to provide a 8v, and use a mosfet instead of BJT, the situation is not getting better. And I also try to replace the zener diode with a voltage reference, the situation still remain the same.
 
Here is the datasheet
 

Attachments

wasabilumen said:
hello, yungman, thanks for answer.

1. datasheet of zener diode is attached.

2.yes,by not accurate. I mean the current read by the meter change as I change the R1.

3.Then should I change a zener diode or change a R1?
I calculate the R1, IDiode = (Usource - Uzd) / R1
a.R must be small enough that the current through D keeps D in reverse breakdown.
5mA>(5v-2v)/R1 so R1<600Ω
b.R must be large enough that the current through D does not destroy the device
Id*Ud<200mw so R1>30Ω
so ,I choose 330Ω,which is between 30Ω~600Ω

4. It is the current due to adjusting R1, so I can only increase 5v? then change BJT to MOSFET? Drive gate of Mosfet need 3v, so 8v will be enough?

5.In simulation ,I tried to check the voltage an current drawn of zener during changing the R1, none of them change. Will this situation be different when it comes to the real world?

6.In simulation, I try to provide a 8v, and use a mosfet instead of BJT, the situation is not getting better. And I also try to replace the zener diode with a voltage reference, the situation still remain the same.

1)The zener part of the circuit is fine, but if you are looking for precision, use a voltage reference as zener voltage change with temperature.

2) Your Vcc is too low, when R1 is 50% of 13K=6.5K, you have 6.5X0.2=3.25V drop across R1. VCE of the BJT is 5-3.25=1.75V. That is way too low for the BJT to work as constant current source. Raise the +V to +8 or +10V, you should see big improvement. Look at the data sheet of the BJT, you need to make sure you have the VCE above the knee voltage at the current you are running and the VCE stay at the region where the collector curve is in the flat part.
 
Last edited:
yungman said:
1)The zener part of the circuit is fine, but if you are looking for precision, use a voltage reference as zener voltage change with temperature.

2) Your Vcc is too low, when R1 is 50% of 13K=6.5K, you have 6.5X0.2=3.25V drop across R1. VCE of the BJT is 5-3.25=1.75V. That is way too low for the BJT to work as constant current source. Raise the +V to +8 or +10V, you should see big improvement. Look at the data sheet of the BJT, you need to make sure you have the VCE above the knee voltage at the current you are running and the VCE stay at the region where the collector curve is in the flat part.


Hello, yungman,
really really thanks for you replying.

1. I changed zener diode to a voltage reference.
2. For the BJT work as constant current source means BJT work in active region, Ic value was controlled by Ib, right?
In simulation,I use a BJT 2n2219, change +v to 10v, and adjust the Ie to 0,076mA, now when R1 100%13k, voltage across BJT is 10v-0,076*(R1+R3)=8v, this is enough for BJT work in active region.

But still the same situation, during I change R1, constant current source has a 1uA or 2uA tolerance. I have also change BJT to a mosfet, situation did not get better. Is this the limitation of the circuit itself or there is something that I have not understanding it correctly?

Thank you.

Regards,
Jining
 
wasabilumen said:
Hello, yungman,
really really thanks for you replying.

1. I changed zener diode to a voltage reference.
2. For the BJT work as constant current source means BJT work in active region, Ic value was controlled by Ib, right?
In simulation,I use a BJT 2n2219, change +v to 10v, and adjust the Ie to 0,076mA, now when R1 100%13k, voltage across BJT is 10v-0,076*(R1+R3)=8v, this is enough for BJT work in active region.

But still the same situation, during I change R1, constant current source has a 1uA or 2uA tolerance. I have also change BJT to a mosfet, situation did not get better. Is this the limitation of the circuit itself or there is something that I have not understanding it correctly?

Thank you.

Regards,
Jining
The current is controlled by the current through R3 which is the voltage drop across R3.

2N2219 is the wrong transistor to use, the β is quite low. try this one:

http://www.onsemi.com/pub_link/Collateral/MPSA18-D.PDF

This one has minimum β=400 at 10uA.

MOSFET should work. I question your simulation program. MOSFET don't draw gate current, so all current has to pass through R3. So if the voltage across R3 don't change, you should get constant current.

What is your current range you need the circuit to work? If it is possible, increase R3 so you get more voltage across it. Repeat the experiment, instead of measuring the voltage of the reference, measure the voltage across R3 and change R1 like you did, see whether there is a voltage change that indicate current change.

Another thing, to guaranty stability, put a 10K resistor from the emitter/source of the transistor to the negative output of the op-amp. Then put a 100pF cap from the output of the op-amp to the negative input. This create a high frequency bypass so the feedback loop don't have to include the transistor. This is particular important if you use the MOSFET where the input capacitance is high.

MOSFET is the best, I designed a high current supply of 40A using this design with MOSFET and is stable to 0.02%. What you show is like 2%, that is not correct. Build the circuit and really measure it. You can't trust the simulation program. I just don't buy the result.

There are constant current IC, check them out. But your circuit is particular good for higher frequency application. There are a lot of designs using op-amp to simulate constant current source, but there is down fall using a voltage amp to act like a current source.
 
Last edited:
Hello, yungman,

Thanks for you detailed explanation. I will follow your suggestions, try to build the circuit.
Having a good weekend!

Regards,
Jining
 
Why use an NTC to measure a temperature?

NTC are not linear, not reproducible, nothing.
One working method is to measure the Vbe of a bipolar transistor with B and C tied. It gives you -2.1mV/K, only calibrate the zero.
The better method is to buy the chip that measures the temperature. Some give you 1µA/K already calibrated to better than 1K.

If really you want to create a constant current, there are chips that do it, already calibrated.
Or use one LM317 and one resistor to create a constant current. The model in TO92 could work at 200µA, but check it.
 
Last edited:
  • #10
Enthalpy said:
Why use an NTC to measure a temperature?

NTC are not linear, not reproducible, nothing.
One working method is to measure the Vbe of a bipolar transistor with B and C tied. It gives you -2.1mV/K, only calibrate the zero.
The better method is to buy the chip that measures the temperature. Some give you 1µA/K already calibrated to better than 1K.

If really you want to create a constant current, there are chips that do it, already calibrated.
Or use one LM317 and one resistor to create a constant current. The model in TO92 could work at 200µA, but check it.

Thanks for your advise,Enthalpy, any suggestion for a constant current ship? I have checked some chips, but all of them is not precise enough for my applycation.
 
Last edited:
  • #11
  • #12
Enthalpy said:
By googling "temperature to current" I got for instance the AD590
http://www.analog.com/static/imported-files/data_sheets/AD590.pdf
which gives 1µA/K between -55°C and +150°C. Initial accuracy is 5K, 2.5K or 0.5K.
Or the ancestor LM334
http://www.linear.com/product/LM334S

By googling "current source IC" for instance
http://www.ti.com/product/ref200
http://electronics.stackexchange.com/questions/21961/alternative-to-ref200-current-source
LT3092 (minimum 500µA)

Thanks a lot.
I have searched for "constant current source IC", seems like I searched the wrong word. Ref200 seems really good for me.
 
  • #13
I don't think there is anything particular wrong with the original circuit using a MOSFET. I don't know why the simulation program don't show good result. When using a MOSFET, the current coming out from the Drain has to be from the Source as there is no gate current. The source current is closed loop feedback controlled by the op-amp. It should be as accurate as the components used.

As I said, I designed a precision big magnet controller using the same design except I had many FETs parallel for 40A.

The original circuit is quite fast if need to be. It jump from step to step and settle very fast.
 
  • #14
yungman said:
I don't think there is anything particular wrong with the original circuit using a MOSFET. I don't know why the simulation program don't show good result. When using a MOSFET, the current coming out from the Drain has to be from the Source as there is no gate current. The source current is closed loop feedback controlled by the op-amp. It should be as accurate as the components used.

As I said, I designed a precision big magnet controller using the same design except I had many FETs parallel for 40A.

The original circuit is quite fast if need to be. It jump from step to step and settle very fast.

Hello, yungman,
May be it is somethingwrong with the simulation program.Book the elements need sometime, but I will try to build the circuit and also try to use a current-chip, to see which one performs better. Any suggestion which mosfet to use?

Regards,
Jining
 
  • #16
Back
Top