TLV9351: 4-20mA Output Increases on increasing load resistor

What is VCC1?

What are the voltages at pins 1, 3, and 4?

Hi Harish, 

Per Clemens's comments, we need to know the Vcc1 supply rail, which it has to be >> 20mA*235ohm = 4.7Vdc. 

Harish Kumar Singh said:

The circuit functions correctly when the output current is calibrated (Zero: 4.00 mA, Span: 20.00 mA) by adjusting the PWM duty cycle at 1 kHz, using a digital multimeter (DMM) connected directly to the output.

In addition, the PWM is specified at 1kHz, and the circuit is converting the PWM signal to an average voltage. Your LPF filters' cutoff frequency is a bit high. The cutoff frequency needs to be <100Hz or prefer <50Hz in order to get rid of ripple noise - to have a stable average voltage with minimum ripple voltage. 

If  you have other questions, please let us know. 

Best,

Raymond

Hi Raymond,

VCC1 is a stable 24V DC supply, internally generated by the on-board SMPS circuit.

We made the following component changes:

• R38, R39 updated to 3.3k

• C10, C11 updated to 2.2µF

• R24, R37, R40, R41 updated to 1MΩ (all resistors are 1% tolerance; capacitors are 10%, X7R type)

After these modifications, we observed a significant improvement. The difference in output current between no-load and a 235Ω load at the 20mA SPAN calibration point is now approximately 35µA (lower with load). Previously, the output current increased under load.

Could you please review and suggest improvements to enhance the performance of the existing circuit?

Thank You

Thank you for your prompt response.

I would definitely try this configuration and if all ok, will implement this in upcoming revision of PCB.

For now, I am stuck with the circuit been built on PCB as shared previously. To avoid scrapping 100 mounted PCBs, do you foresee or may please suggest any modifications to the existing to improve the performance of the circuit. 

thank you

Hi Harish, 

Please try to cut and jump the following connections, and let me know how it worked out. 

Best,

Raymond

Hi Raymond,

Thank you for your kind support.

I will test the suggested and will let you know.

thank you

Hi Raymond.

Unfortunately, the suggested change did not make any difference.

Thank you

Hi Harish, 

Harish Kumar Singh said:

Unfortunately, the suggested change did not make any difference.

The modification is for the stability purposes, not the V-to-I in accuracy fixes. If you trigger a pulse load at 150ohm sensing resistor section, you will see the V-to-I circuit will oscillate drastically without the compensation. Although you are working with mainly the DC application, the transient events can happen if there is a power surge or transient events at the V-to-I's output or input, even EMI interferences. Please check it out. 

Can you tell me what is the bottleneck currently? If you want to improve V-to-I accuracy, you may have to use lower Vos op amp. But first, you could increase your 2.2uF LPFs to 4.7uF or 10uF and this will improve your input voltage ripple significantly. 

Best,

Raymond 

Hi Raymond,

The issue is not related to the accuracy or linearity of the 4–20 mA output, but rather to the variation in output current between no-load and loaded conditions.

We made the following component changes:

• R38, R39 updated to 3.3 kΩ

• C10, C11 updated to 10 µF

• R24, R37, R40, R41 updated to 1 MΩ (resistors: 1% tolerance; capacitors: 10% X7R type)

In several PCBs, the difference between calibrated no-load and loaded (with ~235 Ω resistor) output at 20 mA is within ±30 µA. However, in approximately 15 out of 50 boards, the difference reaches up to ±200 µA at 20 mA.

This variation appears to be linear and consistent across the full 4–20 mA range, and also changes proportionally with different load resistor values. Furthermore, the direction of the shift varies—positive in some boards and negative in others.

Could this be caused by input offset voltage variations of the TLV9351 op-amps used in the circuit?

Raymond Zhang1 said:

If you trigger a pulse load at 150ohm sensing resistor section, you will see the V-to-I circuit will oscillate drastically without the compensation.

After the above changes, I am not seeing the oscillations on 150ohm sense resistor, confirmed the same with the scope.

Do you suggest that changing the op amp with lower Vos (e.g. OPA196 or OPA182) would resolve the issue?

Thank You

Harish

Hi Harish, 

I was a bit confused where your 235ohm load is (you should draw the load). You have a Howland current pump.

You have to use 100kohm resistor with 0.1% tolerance if you are concerned about the accuracy. Yes, OPA196 op amp will help, but most important is the  resistor matching. Please revise the your circuit as shown below exactly.  

Howland Ipump 06232025.TSC

https://www.ti.com/lit/an/sboa436/sboa436.pdf?ts=1750670386820&ref_url=https%253A%252F%252Fwww.ti.com%252Fsitesearch%252Fen-us%252Fdocs%252Funiversalsearch.tsp%253FlangPref%253Den-US%2526nr%253D378%2526searchTerm%253Dhowland

If you have other questions, please let me know. 

Best,

Raymond

Hi Raymond,

Thank you for your support.

Since I’m not an expert in analog circuit design, I have few questions:

1. In a Howland current pump circuit, is it necessary to use a fixed load resistance? If that’s the case, does it mean that even with precision 0.1% 100k resistors or a better op-amp like the OPA196, the output current would still vary depending on the connected load? 

2. Can we calculate the maximum current drift on with 0.1% 100k resistors for 100 ohm & 500 ohm load resistor ?

3. I’m exploring the possibility of using the XTR111 for the 4–20 mA output circuit in next PCB layout revision. I’d appreciate your thoughts on this.

regards,

Harish

Hi Raymond,

- I have ordered 100k 0.1% resistors and OPA196. I would update the result. I also believe it would resolve the issue with my existing Howland circuit.

- In upcoming design revision I would go for XTR111.

Thank you for support.

Best regards

Harish