LM51561-Q1: Flyback converter with hiccup

Hi Fred,

thank you for using the E2E forum.

There is no other trigger source for the Hiccup.

There might be an issue with the softstart. If there are over current triggered during softstart the LM5156 (without Hiccup) might be able to reach the output voltage with a kind of lump mode while the device with Hiccup continuously restarts.

Where did you check the CS pin voltage on the LM51561 or the LM5156?

May you can just give it a try and increase the softstart time (Cap).

Best regards,

 Stefan

I tried and no hiccup with a crazy 22u cap, but as soon as the softstart is over the cap is discharged and the process starts again.

Current limit resistor is 0.011 Ohm, current in the transformer is less the 3A.

The CS pin is less than 5mm from the current sensing resistor. PGND is connected to the ground side of the same resistor.

Hi Fred,

Thanks for the overview.
As Stefan already mentioned, there is no other trigger source for Hiccup besides OCP.

Do you use an RC filter at the CS pin? This can avoid falsely triggering the OCP due to noise.
You can also consider to reduce the sense resistance and check if the hiccup still occurs.

Best regards,
Niklas

I had a filter on CS 100R/100pf. When i increased the cap to 1n the MOSFET blew.

Without load, the output voltage comes up to the nominal voltage. But as soon as I dial in a load current of 100mA, the SS pin goes low, the unit restarts. Css is 10uf. The primary inductance is 60uH, fsw = 180kHz

Hi Fred,

Thanks for the update.
Would you be willing to share your LM5156 schematic and provide the input voltage range/ output voltage?

Then we can offer to review the design and check for possible failure causes.

Thanks and best regards,
Niklas

How to attach a .pdf?

Hi Fred,

You can either drag and drop the file into your reply textbox, or you can click below your message on "insert" and then on "Image/Video/File" and put in the local file location.

Best regards,
Niklas

Powersupply_75W_V1_6.pdf

Hi Fred,

Thanks for supplying the schematic.
When using LM5156 in isolated flyback topology, the FB pin is connected to ground. This essentially disables the internal softstart feature.
If the device goes into hiccup right from startup, it might be related to the higher inrush current due to charging of the output capacitors.

It is possible to implement an artificial softstart on the secondary side.
This will limit the duty cycle and avoid overcurrents during the ramp-up phase.
Please refer to this app note for additional details:
 https://www.ti.com/lit/an/snvaa44/snvaa44.pdf

Best regards,
Niklas

Hi Niklas,

without any load the system comes up, when I put 100mA load on the supply, it shuts down and restarts continuously.

Fred

Hi Niklas,

the blue trace is the transformer current (1A/div). After the initial high current, it stops for 3ms and then I works normal until the Css time expires and it starts again. Photo at 100mA load.

Maybe an extremely short Css time with a secondary softstart?

Hi Fred,

Thanks for sending additional info.

If this hiccup behavior does not occur without load is another indicator for the OCP being triggered during startup.
The LM51561 device has both cycle by cycle current limiting (OCP) and hiccup mode.
If 64 cycles of OCP are detected, the device goes into hiccup mode. This means it will shutdown for a fixed amount of switching cycles and Css will discharge in the meantime.

Changing Css will not have an effect on this behavior.
You will be able to see the OCP being triggered when measuring the CS pin signal. It should go above 100mV, which is the threshold for OCP.

The secondary softstart should solve this problem.

Best regards,
Niklas

Hi Niklas

I tried to no avail. The secondary softstart has no influence on the first few cycles, since it needs some voltage at the secondary.

The problem is, the LM51561 starts with about 45% duty cycle witch causes the trigger of the OCP.

Hi Fred,

Thanks for the feedback.
The mentioned app note mentions an additional way of limiting the duty cycle especially during the first cycles.
I am referring to the section promoting a BJT between SS and COMP pin.
This circuit will put an additional limit to the inrush current to avoid OCP during startup.

Would it be possible to implement this BJT circuit and check for improvements?

Thanks and best regards,
Niklas

will do

Hi Niklas,

this change seems to fix the high startup current. But the supply continuously shuts down at 100mA load (see screenshot above at 1A/div, hor 40ms/div, Css 100nF). No load works fine. Secondary softstart is implemented.

Hi Fred,

Thanks for the feedback.
So if I understand correctly, there are no high inrush currents during startup, but the device still goes into hiccup mode after startup?
First I would try to check if the 100mA is generally to high for the application.
As the device starts up normal with no load, can you run it in steady-state without load and then slowly increase the load to see when OCP will be triggered?
I am still missing the input voltage range and transformer inductance&winding ratio, so I cannot make a general review of the design yet.
You can also check if sense resistance is okay with our quickstart calculator tool:
https://www.ti.com/tool/download/SNVC240

Best regards,
Niklas

The input voltage range is 100V to 143V DC. The primary inductance is 60uH, winding ratio 4:1, 24V output, Fsw 180 kHz.

Keep in mind, that the LM5156 (without hiccup) delivers 24V 3.2A with the same transformer and current sense resistors, 0.011R => ~9A current limit.

The max. peak current should be about 6A at 75W load. As you can see from the screenshot, the peak current is below 3A at 100mA load at restart and  1A afterwards. I will run a test which load the supply can deliver without hiccup.

The adding of the BJT reduced the startup current to  <3A peak from 8A before

Hi Fred,

Thanks for the transformer information and the update.

I agree that this transformer is fit for the application and there should no current limit violations.
(LM5156 being able to support the full load in the same application confirms this)

It is still possible that the OCP is falsely triggered through noise at the CS pin.
I already see an RC filter with R31/C31 at the pin, so some degree of noise should already be filtered.

Still when checking the load capabilities of the the application, could you take a measurement of the Switch Note?
Here we will see if there is any subharmonic oscillation or noise.

Thanks and best regards,
Niklas

Hi Niklas,

Shortening PGND ad AGND directly at the chip, enabled 2.7A output current at 24V. At 2.8A the output voltage decreases to 10V.

4606.Powersupply_75W_V1_6.pdf

Hi Fred,

So the initial layout had no connection of power ground and analog ground?
This can indeed lead to shifts of the two ground levels, so sensing signals can become inaccurate, including current sensing or switching frequency setting.
It is recommended to have ANGD and PGND connected through one trace or via to avoid shifts.

Based on the layout snippet, it looks like there are only ground traces and no planes or polygons.
Large ground planes can improve thermal behavior and reduce noise.

Did shortening PGND and AGND solve the open issue, or is there still abnormal behavior on the board?

Best regards,
Niklas