Problems with constant current source composed of op amp + MOS tube

S3S4S5S6

Problems with constant current source composed of op amp + MOS tube [Copy link]

 

PWM is generated by the microcontroller, with a frequency of 20K. After RC filtering, it becomes DC, that is, the voltage at point A is DC voltage.

The problem during debugging is that I don't give PWM, that is, the PWM port of the microcontroller is directly pulled low, the output is zero, but the LED is still on at this time.

Then I used an oscilloscope to measure the voltage at point B, which was 2V. As I gave PWM and controlled the duty cycle, the voltage at point B slowly rose from 2V, and the LED gradually became brighter, but the LED could not be turned off.

I would like to ask you why the voltage at point B is not 0V but 2V when the input is zero.

maychang

"Then I used an oscilloscope to measure the voltage at point B, which was 2V. As I applied PWM and controlled the duty cycle, the voltage at point B slowly rose from 2V, and the LED gradually became brighter."

Why not measure the voltage at the source of Q1 (upper end of R31)?

Fred_1977

It may be caused by capacitor C28, you can remove this capacitor and try it first.

S3S4S5S6

maychang posted on 2022-6-17 14:38 "Then I used an oscilloscope to measure the voltage at point B, which was 2V. As I gave PWM and controlled the duty cycle, the voltage at point B slowly rose from 2V, and the LED also gradually became brighter...

The voltage test shows 35mV

chunyang

The op amp has input offset...

S3S4S5S6

maychang posted on 2022-6-17 14:38 "Then I used an oscilloscope to measure the voltage at point B, which was 2V. As I gave PWM and controlled the duty cycle, the voltage at point B slowly rose from 2V, and the LED also gradually became brighter...

I suspected that the pull-down of the microcontroller was unreliable, so I directly soldered the PWM port to GND. I found that the 2V voltage at the G pole of the MOS tube slowly decreased over time. After about 5 minutes, it dropped to 700mV, and it was still decreasing over time.

S3S4S5S6

chunyang posted on 2022-6-17 15:13 The op amp has input offset...

Input offset, is it such a big offset?

chunyang

S3S4S5S6 posted on 2022-6-17 15:26 Input imbalance, is it so big?

Low-end models generally have an input offset voltage of several mV, which can be obtained by directly grounding the non-inverting input and measuring the inverting input. However, accurate measurement requires the internal resistance of the meter to be high enough.

chunyang

S3S4S5S6 Published on 2022-6-17 15:24 I suspect that the pull-down of the microcontroller is unreliable. I directly soldered the PWM port to GND and found that the 2V voltage of the G pole of the MOS tube slowly decreased over time, about 5 minutes...

The MCU's IO cannot actually output 0V, as long as it conforms to the logic level of 0. When testing, the op amp's non-inverting input should be directly grounded, otherwise the residual charge of C31 will have an impact over a period of time.

S3S4S5S6

Fred_1977 posted on 2022-6-17 14:54 It may be caused by capacitor C28. You can remove this capacitor and try it first.

You can try it

maychang

S3S4S5S6 published on 2022-6-17 15:24 I suspect that the pull-down of the microcontroller is unreliable, so I directly soldered the PWM port to GND and found that the 2V voltage of the G pole of the MOS tube slowly decreased over time, about 5 minutes...

"I directly soldered the PWM port to GND and found that the 2V voltage at the G pole of the MOS tube slowly decreased over time. After about 5 minutes, it dropped to 700mV and continued to decrease over time."

I suspect that you used a capacitor with large dielectric absorption, such as an electrolytic capacitor, for C31 in the first post. Due to the dielectric absorption of this capacitor, the potential of the op amp's in-phase input terminal can only drop slowly to zero. The time constants of C28 and R27 are also over 10ms. These two items make the current in Q1 change slowly.

S3S4S5S6

maychang posted on 2022-6-17 16:02 S3S4S5S6 posted on 2022-6-17 15:24 Suspecting the pull-down of the microcontroller, it is unreliable. I directly soldered the PWM port to GND and found that the 2V of the G pole of the MOS tube...

C31 uses a 0603 packaged ceramic capacitor, and there may be problems with the selection of the values of C28 and R27.

S3S4S5S6

maychang posted on 2022-6-17 16:02 S3S4S5S6 posted on 2022-6-17 15:24 Suspecting the pull-down of the microcontroller, it is unreliable. I directly soldered the PWM port to GND and found that the 2V of the G pole of the MOS tube...

The entire circuit structure is based on TI's documentation.

zmldty

Learn about ,,,,,,,,,,,,,,,,,,,,,,,,,

Gen_X

At least a 120K resistor should be connected to the ground at point A to achieve basic input current balance. At the same time, the offset voltage also needs zero point adjustment.

S3S4S5S6

Gen_X posted on 2022-6-17 17:42 At least a 120K resistor should be connected to the ground at point A to balance the input current. At the same time, the offset voltage also needs zero adjustment.

How does this work? Can you explain it in more detail?

Fred_1977

The reference circuit is 100PF, so the capacitor of C28 is used incorrectly. There is a big difference between 100PF and 0.1UF.

平漂流

Excuse me, what is the function of the 100pF capacitor in this reference manual?

平漂流

maychang posted on 2022-6-17 14:38 "Then I used an oscilloscope to measure the voltage at point B, which was 2V. As I gave PWM and controlled the duty cycle, the voltage at point B slowly rose on the basis of 2V, and the LED also gradually became brighter...

Teacher, does this op amp require rail-to-rail op amp?

maychang

Pingpiaoliu published on 2023-7-11 09:53 Teacher, does this op amp require a rail-to-rail op amp?

[Does this op amp have to be a rail-to-rail op amp?]

The power supply of the op amp is supplied separately. The on-state voltage of the MOS tube is usually 2V, which is higher than the lowest voltage that an ordinary op amp can output, so this op amp does not require rail-to-rail.