LM27762: LM27762 startup issue

Andy,

Please provide layout figures of LM27762 and surrounding components. Typically start-up issues of negative output have been related to layout issues. (parasitic inductance in GND pin wiring).

 Regards,

Juha

Hello Juha,

super fast response, thank you. You might be right, the layout could be the problem. Attached are pictures of all PCB layers (it is just a 2 layer board) and a photo of the empty PCB. 

Thank you for your support and best regards

Andy

Picture 1: Both layers

Picture 2: Front layer

Picture 3: Back layer

Picture 4: Photo

Andy,

Thanks for the layout figures. It looks like path from GND pin to ground plane could be shorter. This is from our EVM:

Notice how GND pin is connected through via to GND plane. Similarly your board could be modified by adding via next to GND pin:

Modification could be done to existing board if you have small enough drill bit.

This is the waveform in C1+ pin after failed start-up due to not so good layout:

And this is the waveforms when charge pump is operating normally:

Maybe you can compare these scope shots to your situation.

Regards,

Juha

Hello Juha,

thank you for the hint. I did not manage to get it working yet but I keep trying.

Best regards

Andy

Andy,

If you can share oscilloscope shot of start-up (VIN,CPOUT,C1+,VOUT-,EN), I can take a look at it.

Regards,
Juha

Hello Juha,

I have made some measurements. After power up, my CPU waits 100msec before EN goes high, therefore VIN (+5.3V) sould be stable at the moment the LM27762 gets enabled.

Picture 1 shows: 

CH1: EN

CH2: VIN

CH3: C1+

CH4: VOUT-

Picture 2 shows:

CH1: EN

CH2: CPOUT

CH3: C1+

CH4: VOUT-

When measuring I sometimes manage to get the LM27762 working, in particular when it happens that I accidently short circuit the input voltage (VIN). In those cases the circuit seems to start up properly but unfortunately this is not repeatable and definetely not a solution for a production product.

Thank you for any hints and best regards

Andy

Hi Juha

I connected the enable pins through a 1k ohm resistor to VIN an now the LM27762 starts up. It seems that controling the enable pins from the microcontroller causes the issue.

Best regards

Andy

Andy,

Sorry for late reply, I was out of office last week.

Looking at your scope shots, it seems similar to the malfunctioning case I sent picture of. I am little worried that you only masked the real issue by pulling enable pins up to VIN via resistors. I think it fixed the problem because device is now enabled earlier, VIN has not yet risen to its target voltage. Startup happens with smaller VIN => smaller in-rush currents. We have seen this exact behavior when debugging start-up issues caused by excessive parasitic inductance in GND path.

Did you manage to modify the PCB layout in a way I suggested earlier?

Regards,

Juha

Hi Juha,
I agree that my solution does not solve the root cause but at least it allows me to test and evaluate the Overall design.
The PCB modification is not that easy to do so we need to assemble another prototype next week. I plan to modify the PCB before mounting the components - this should be easier to do. The WSON package does not make rework on the PCB easy. May I suggest to consider someting like MSOP package for this device?
As an alternative to the LM27762 I consider using DC/DC buck converter for generating my +/-5V. Any suggestions for a low noise device?
Regards
Andy

Hi Andy,

I understand - if the fix makes it possible to do further testing, it is good enough. About the PCB modification, my idea was to make additional via from GND pin to GND layer by drilling hole through the PCB and soldering wire in it. It is not very sophisticated method, but it should reduce the inductance of path to GND plane. I admit that it is challenging as the space for the drill hole is limited, but with very small diameter drill size it should be possible. If this would be done next to footprint, you really would not have to touch LM27762 itself.

As for buck alternative, you can take a look at TPS65132. However it does not have negative side LDO if I remember correctly.

Regards,

Juha