TPS61163A: Inconsistencies on the measured current

To add more info,

I just soldered a 62K 0.1% resistor on two prototypes, which should configure the TPS61163A, according to the datasheet formula, to source 20.41 mA per channel. However, I measured 17.04 mA on one prototype, and 18.23 mA on the other.

Hi, Lander

patrick is in a sick leave today, he will give you a feedback once he came back to the office.

Ok. Thanks.

Hi Lander,

I will give you update when I back to office.

BR

Patrick

Hi Lander,

Have you captured any waveforms during the test?

BR

Patrick

Hi Patrick,

I didn't. I only took current measurements with various multimeters.

But I can perform those captures without a problem. Is there a specific one you would like to see?

Regards.

Hi Lander,

The voltage waveforms on SW, COMP, IFB1 and IFB2 will be helpful for debug.

If you have current probe, the inductor current and LED current will also be good.

BR

Patrick

Hi Patrick,

I have attached the requested oscilloscope captures. If you need any additional one, let me know.

(By order: SW, COMP, IFB1, IFB2, Current on IFB1 and current on L1)

In addition, tomorrow I will desolder the critical components of one of the prototypes to measure them with an LCR meter.

Regards.

Hi Lander,

I am out of office for a biz trip, will give you feedback later.

BR

Patrick

Hi Patrick,

I just have measured the critical components of the board:

• C1, C2 and C3 (UMK107AB7105KA-T, according to provider): 894.6 nF, 899.5 nF and 944.3 nF respectively @ 10 kHz.
• C4 (CGA3E3X7R1H334KT000N, according to provider): 329.4 nF @ 10 kHz.
• L1 (LQH32PN4R7NN0L, according to provider): 4.728 µH @ 100 kHz and 0.192 Ω.
• R1: 9.988 Ω

LEDs specs: 30mA 70mW 1.2V-1.6V 939nm-946nm 1206 Infrared (XINGLIGHT XL-3216IRC-940)

If you need any more data, let me know.

Regards.

Hi Lander,

Ok, allow me some time to give you feedback.

BR

Patrick

Hi Lander,

Could you please reduce the number of LEDs in each string? With 14 LEDs in one string, the total output voltage might be too high for this device so it cannot output expected current because the voltage on IFBx is too small.

BR

Patrick

Hi Patrick,

I just took measurements after removing various LEDs from the circuit:

With 13 LEDs per string, the full string voltage is 16.58 V, and the current is 18.3 mA.
With 12 LEDs per string, the full string voltage is 15.3 V, and the current is 20.0 mA.

Knowing this information, is there a way to improve the circuit to support all 14 LEDs per string (16.8 – 22.4 V) at 20 mA? Maybe using a higher inductor value? (I will try it, as on the development board associated with the IC).

I find this behavior to be strange, as in the datasheet there are various charts showing the circuit operating at over 20 V and 20 mA per string (see figures 17 to 20 and 2 to 5). Additionally, I couldn't find any data regarding the circuit's maximum voltage per string.

Regards.

Hi Patrick,

I tested the circuit with different inductors and these are the results:

- 10 μH inductor: Both channels work at 9 mA, regardless of the value of R1.

- 6.8 µH inductor: The maximum current per channel is 12 mA. Lower values can be obtained by adjusting R1.

- 4.7 µH inductor: The maximum current per channel is 17.5 mA. Lower values can be obtained by adjusting R1.

- 2.35 µH inductor (two 4.7 µH inductors in parallel): Maximum current per channel is 24 mA. Lower values can be obtained by adjusting R1.

With a 2.35 µH inductor, the circuit seems to work with the parameters we need (28 LEDs @ 20 mA), but this value seems very strange to me, as there is no reference to it in the datasheet.

If you want me to test some other configuration, feel free to ask.

Regards.

Hi Lander,

Have you tried with a constant resistor on Iset instead of the adjustable resistor?

BR

Patrick

Hi Patrick,

Yes, the original oscilloscope captures and measurements are taken with a standard 62 kΩ, 0.1% resistor. The results are also consistent with the values obtained with the variable resistor.

In any case, after trimming the potentiometer, I always power cycled the board just in case, so the results should be identical to using a fixed resistor.

Regards.

Hi Lander,

OK, let me check with our EVM as well, could you please also share your layout? You can send me through private message if you have confidential concern.

BR

Patrick

Hi Patrick,

I have just sent you a private message with the layout.

Regards.

Hi Lander,

Ok, I will review it and give you feedback.

BR

Patrick

Hi Patrick,

While you are reviewing the data, is it possible to get the evaluation board (TPS61163AEVM-564) design or Gerber files?

Regards.

Hi Lander,

I need to check internally whether we could provide those files, but you could refer to our EVM user's guide for a rough layout view. I am setting up the bench test, will give you feedback later.

BR

Patrick

Hi Patrick,

We have ordered a pre-production batch of the board (with the tested 2.2 µH inductor) from two suppliers, so we can evaluate the part sourcing to discard the possibility of a fake part.

Have you achieved any results yet?

Regards.

Hi Lander,

Sorry for the late, I can get the expected output current on EVM. Could you please measure the real voltage of ISET pin (A1) when IC is operating?

BR

Patrick

Hi Patrick,

The measured voltage drop on R3 (same net as Iset) is 1.2248 V (with the 2.2 µH inductor).

I also took a picture of the IC:

Do you see anything strange?

Regards.

Hi Lander,

Was the 1.2248V on R3 measured with every condition you mentioned before (different R3 value)? Want to know whether it will vary with different working conditions.

Have you tried a larger output capacitance? After viewing your layout, I found the board size is kind of critical so the components parameters may need adjustments accordingly.

BR

Patrick

Hi Patrick,

At this point, I ran out of available boards to perform the tests, as the previous tests were (unintentionally) destructive.

It will have to wait 1-2 weeks until the ordered assembled batches arrive. I will perform those tests you suggested once they arrive.

With regard to the board size you mentioned, to add more info, the circuit barely heats up: around 10 °C above the ambient temperature. Is there any other factor that would be affected by the board dimensions?

Regards.

Hi Lander,

Because the components placement will be limited by board size, like the SW copper is a little larger and some traces have to go through inner layers by vias to connect external components, which will introduce some parasitic inductance. So I think a little larger output capacitance might help.

BR

Patrick

Hi Patrick,

Ok, thanks for your support, I will test the circuit with different output capacitance values and then give you an update when we receive the ordered batch.

Regards.

Hi Lander,

Noted, I will wait for the update.

BR

Patrick

Hi Patrick,

I finally have some news to share with you. I have received 2 batches from 2 different assemblers: one from Europe and another from China (this last one on the same site we bought the components for the prototypes). The BOM of both boards contained the same component values that worked for me on the prototypes (i.e., a 2.2 µH inductor on L1 [this is the only value outside the datasheet parameters]). The results are as follows:

• On the batch from China: approximately the 60 % of the boards worked flawlessly (at around 18.5 mA per string with very little drift of about 0.3-0.5 mA when it warms up to the working temperature). Modifying the Iset resistor of these working boards they can operate at a constant 20 mA per string without a problem (and up to around 25 mA).

The remaining 40 % work at insufficient current (around 12 mA). Every board falls into one of these two categories and appears stable on it, there are no intermediate values (it works at 18.5 mA or 12 mA per string).

• On the batch from Europe (from a high-quality assembler and, presumably, components sourced from authorized distributors), all boards work at about 12 mA. If I change the inductor to the standard 4.7 µH, some of them function properly (although the current drifts up to 2.5 mA, which seems excessive to me).

I must say that both manufacturers received the same manufacturing files and BOM. Although the European one made me change minor things for manufacturability reasons (moving a couple of pads and tracks a few micrometers).

Additionally, I have tested on one of the faulty (or less LED current sourcing capable boards) soldering 10 µF/50 V X7R MLCC capacitors on C1, C2 and C3. The result of this is that it appears to have no effect on the string's current (I haven't taken any oscilloscope waveform).

Just in case you can see anything strange, these are two microscope captures of the driver. The first one corresponds to a correctly working board and the second one to one working at 12 mA per channel.

If you are interested, I can send you one or two boards of each batch for you to analyze.

Thank you very much.

Hi Lander,

Actually I don't expect such high "failure" rate of our IC, so this is still more likely related to application design.

And yes, you can send your board to my site for test. But firstly, do you have the EVM board in your site? I think having a test with EVM to check the IC performance first will help us locate the root cause of such issue.

BR

Patrick

Hi Patrick,

I understand that if this behavior is related to a design flaw of the PCB, all boards of the same batch should behave similarly, isn't it?

I have measured the values of all components on both boards (working and not working), and I cannot see any significant difference in the measured values. This would limit the problem to the board tolerances (which seems unlikely since all boards fall into two well-defined string current values, without measurements in the middle), or the IC.

I have not yet ruled out the possibility of counterfeit parts, especially since those provided by a reputable assembler did not work with the 2.2 µH inductor value that is clearly outside the recommended specifications in the datasheet.

We don't have the evaluation board. However, it would not be helpful at this point as we cannot use it to test the suspected ICs due to the inability to rework this type of footprint.

If you would like us to send the boards, please send me a DM with your shipping address.

If you have exhausted all options, it may be necessary for us to redesign the board using a different driver.

Best regards.

Hi Lander,

Before you ship the board, could you please capture the waveform of +VLED? Because the output capacitor is not so close the the GND pin, so want to see if there is any unstable.

BR

Patrick 

Hi Patrick,

Sorry for the delay, we have been very busy the last few weeks.

Below are the oscilloscope captures:

1. +VLED on D29 pin of a non-working board:

2. +VLED on D1 pin of a non-working board:

3. +VLED on D29 pin of a non-working board, with an additional 1 µF capacitor added on D29 pin:

4. +VLED on D1 pin of a non-working board, with an additional 1 µF capacitor added on D29 pin:

5. +VLED on D29 pin of a working board (1):

6. +VLED on D29 pin of a working board (2):

7. +VLED on D1 pin of a working board:

Regards.

Hi Lander,

Thanks for the update, so have you measured the IC case (top) temperature during operation?

BR

Patrick

Hi Patrick,

Yes, I did. The maximum temperature we measured in the worst case (on a climate chamber at 40°C and enclosed in its plastic assembly) is 62°C.

Also, I just took two thermal images of both working and non-working boards:

Non-working board:

Working board:

(all taken at 22 °C ambient temperature).

Regards.

Hi Lander,

I think the last thing need you to check before sending the boards to me is that whether the input power supply is good enough? Have you tried a DC source with more output capability instead?

BR

Patrick

Hi Patrick,

We tested the boards in all cases with a high-end laboratory power supply. I just measured the ripple on both the working and non-working circuits just to check, and it's about 30 mV (ripple frequency of about 1 MHz), which seems reasonable to me.

We decided to replace the TPS61163A with two TPS61165 (one for each string). But if you are still interested in finding out what the problem is with this board, we can still send you the boards and you can ask for more information.

Best regards.

Hi Lander,

Noted, I will let you know if I need additional information. Do you still need to find the root cause of TPS61163A?

BR

Patrick

Hi Patrick,

At this point, it would be good to know where the problem is, but as we will not produce boards with this chip, it is not strictly necessary for us to know. However, I can still send you the boards in case you want to further investigate this problem.

Thanks for your time.