MPM3810 cannot regulate under no load

DC-DC Power Converters
1

Dear MPS,

Part: MPM3810GQB-Z

Condition: Load = Open
Cout = 10uF x2 X5R 16V+100nF X7R
Cin = 10uF x1 X5R 16v

When no load the output is ~= vin which is very wrong!!!
when load increase from 1mA to 5mA vout start drop to targeting 3.3V

Both feedback resistor configurations are tested:

  1. 80k, 17k7 as datasheet suggested.
  2. 180K, 40k2 MPS web designer.

So this is very puzzling why the output did not regulated until 5mA.

I hate to say but MPS COT is really really bad over different designs either module or driver IC.

Thanks,
Brian

2

Hi @briansune ,

Some things to try out:

  1. What’s your input voltage, and can you share your schematic and layout?
  2. We can also try to investigate the switch node by probing one of the NC pins (pin 3 or 4). Please also probe the FB pin, EN pin, VIN pin, OUT pin, and PG pin. Please share the waveforms for those pins for no-load condition and the load where you see output = 3.3V.
  3. Can you add a 1kOhm resistor and 330ohm resistor from Vout to GND, and then check what the output voltage is for both cases?

Let me know if you have questions.

Thanks,
Rubas

3

@Rubas.FAE

Some things to try out:

  1. What’s your input voltage, and can you share your schematic and layout?

    image
    image519×503 11.1 KB

  2. We can also try to investigate the switch node by probing one of the NC pins (pin 3 or 4). Please also probe the FB pin, EN pin, VIN pin, OUT pin, and PG pin. Please share the waveforms for those pins for no-load condition and the load where you see output = 3.3V.

    Pending on experiments and testing.

  3. Can you add a 1kOhm resistor and 330ohm resistor from Vout to GND, and then check what the output voltage is for both cases?

    For 330R to 430R we need to place it next to 100nF in order to introduce a dummy loading to the COT buck otherwise it will not regulate and always result in a very high voltage output i.e. 4.7 to VIN.

    For 1k the 3.3/1k = 3.3mA which is a regulation loading. The observations are when no load to light loading aka >5mA the Vout will drop from Vin level to 3.3V targeted regulation level.

Thanks,

Brian

4

@Rubas.FAE

Update:

Previous observations are found on 2 MPM3810 chips, which using the same DUT on 1.8V shows no issues.

However when voltage set above some threshold the unregulated issue happens.
Due to board development, the DUT with that 2 buck chips are under use.

A blank PCB with different MPM3810 chip is assemble and tested.

The unregulated behavior is gone on this chip. I will forward the full test report to you privately.
I plan to test several more to confirm this behavior is cased by internal inductor degradation or PCB copper trace etching sanity issues.

@Rubas.FAE

Additional updates:

After fetching back the previous board (w/ DCDC load unregulate under light loading).
Removing the MPM3810 chip under the same blank PCB layout board (no other components).
And replace the tested light load w/o issue MPM3810 chip to the board that observed light load issue.

Result:
Same issue, 3.3V feedback resistor figuration and Vout w/o any loading result in ~4.xxV.
Only adding 430R will remove this over voltage issue.

  1. Do this caused by noise from Vin?
  2. Same supply node aka Vin from the light load issue chip also attached 2 MPM3810 chip and with 1.8V and 1V configuration which shows no issue.
  3. Any noise spurious could be leak from VIN back to the supply node that could introduce MPM3810 misbehavior?

Thanks,

Brian

5

@Rubas.FAE

Sorry for the noise.
The same layout with same BOM is concurrently placed and the other board design shows no issue at all.

So the issue turns out is the first batch PCB production copper/via sanity is having issue.
PCB cross dissecting is pending on PCB factory and we also test another blank board with only 3.3V configuration and also shows no issue.

Thanks,

Brian

6

@Rubas.FAE

Sorry Rubas,

Issues are still found on the following conditions:

When FPGA is unprogrammed, where VCCIO is connected to 3.3V MPM3810.
Measuring result: 3.22V stable.

When FPGA is programmed, where IOs in VCCIO with 3.3V bank is not used.
Measuring result: 3.5V to 3.6 over-voltage but stable voltage.

The light load issues are still persisted until 680R is place or stronger loading.

Thanks,

Brian

7

Hi @briansune

Thanks for the detailed updates. Will look into it and follow up.

Thanks,
Rubas

8

@Rubas.FAE

Background:

There are total of [3] MPM3810 connected to 5V and output voltage set to [1V2 | 1V0], 1V8, and 3V3.

We had tried using programmable load w/ tran. mode to trigger any possible switch current on 1V2 or 1V8 MPM3810 regulator to see any ambiguous behavior on 3V3 output rail.

Founding shows when 1V2 / 1V8 transient loading is applied the output switching of 3V3 changed from tringle to square which is a bit questionable.

While the noise really do affect between MPM3810 modules.

Again do this COT module is subjected to any harmonic spurious between modules?

Again MPS COT is really really bad on experience every time we use COT based product from MPS uncontrollable problem happens.

Thanks,

Brian