NVMe power issue?

[tsys]

Hi all,

I've just experienced what appears to be a power issue with the VCC3V3_SSD and probably also the VCC3V3_SYS rail. When my battery is below 40 % communication with the NVMe SSD fails on bootup. This manifests itself in either kernel warnings on bootup and a kernel panic as soon as i enter lspci or a kernel panic immediately on bootup.

My current theory is that during probing of partitions on the NVMe SSD the SSD enters a high power mode and U3 can't keep VCC3V3_SYS at 3.3 V.
The SSD I'm using is a Corsair MP510. Its supported power states are

Code:

St Op     Max   Active     Idle   RL RT WL WT  Ent_Lat  Ex_Lat
0 +     6.80W       -        -    0  0  0  0        0       0
1 +     5.74W       -        -    1  1  1  1        0       0
2 +     5.21W       -        -    2  2  2  2        0       0
3 -   0.0490W       -        -    3  3  3  3     2000    2000
4 -   0.0018W       -        -    4  4  4  4    25000   25000

I'm in general somewhat puzzled about how the Pinebook Pro is supposed to stay operational when the battery is low. The discharge cutoff voltage of the battery is specified as 2.75 V. However U3 is just a step-down regulator. Thus it won't be able to maintain VCC3V3_SYS at 3.3 V once the battery discharges below 3.3 V.

Has anybody else experienced anything similar yet?

[hmuller]

I'm in general somewhat puzzled about how the Pinebook Pro is supposed to stay operational when the battery is low. The discharge cutoff voltage of the battery is specified as 2.75 V. However U3 is just a step-down regulator. Thus it won't be able to maintain VCC3V3_SYS at 3.3 V once the battery discharges below 3.3 V.

That's a great observation. I suspect the 2.75 V is specified to keep it in U3's operational range. With that info, it's prudent to monitor battery voltage while on battery as system behavior may be abnormal below 3.3 V.

I am waiting on the replacement M.2 PCIe NVMe adapter before I install one. I will circle back when I have and share observations. I have my eye on the MyDigitalSSD SBXe, as it is cheap and has an "active power" rating  of 3 W with an operating voltage of 3.3 V +/- 5%.

[tsys]

Hey all,

I've just designed a small daughterboard with a step-up/step-down regulator that might fix the issue.


  pcb_revA_top.png (Size: 69.49 KB / Downloads: 687)

KiCad project + some rough ideas on assembly can be found at https://github.com/TobleMiner/pinebook-p...-regulator

PCBs from JLC will take some time to arrive though. So there will probably not be any results until in a few weeks.

[hmuller]

I've just designed a small daughterboard with a step-up/step-down regulator that might fix the issue.

Nice work Tobias. It looks easy except for the part where you join the through holes on the adapter board and the daughter board. How do you ensure heat transfer through via pairs to get a solid connection?

[tsys]

I've just designed a small daughterboard with a step-up/step-down regulator that might fix the issue.

Nice work Tobias.

Thanks 

It looks easy except for the part where you join the through holes on the adapter board and the daughter board. How do you ensure heat transfer through via pairs to get a solid connection?

Usually that is not really a problem. Once solder has flown into the holes thermal conductivity is pretty good and the capillary effect will just suck the solder through to the mating surface of the two boards. Worst case I'll have to put a short piece of wire through the holes before soldering them.

[hmuller]

@tsys ,

I opened up my PBP today to take a closer look at the audio amp chips U19 and U20 (xa9108). Because I don't have a depth guage I wanted  to look at the chip datasheet. I could not find the datasheet for these chips online, but I found one similar (ant8108) and confirmed their physical sizes are approximately the same. The datasheet of the similar component shows a max height of 1.75 mm.

I haven't ordered from JLC but their standard PCB thickness is 1.6 mm according to their website. So a potential issue is that the adapter board may not provide enough clearance to get the daughter board cleanly (parallel) over the audio amp chips.

[Arwen]

Would it be possible to make a replacement NVMe board with the power chip, (and storage capacitor), on it?

Maybe a tiny heat sink for the power chip too.

If it also solved the installation problem from the original NVMe boards, I'd pay more for a clean solution like this.

[tsys]

I haven't ordered from JLC but their standard PCB thickness is 1.6 mm according to their website. So a potential issue is that the adapter board may not provide enough clearance to get the daughter board cleanly (parallel) over the audio amp chips.

.15 mm is probably not going to be a big deal. Should have really ordered 1.8 mm PCBs though. Guess we'll see in few weeks ¯\_(ツ)_/¯

Would it be possible to make a replacement NVMe board with the power chip, (and storage capacitor), on it?

Maybe a tiny heat sink for the power chip too.

If it also solved the installation problem from the original NVMe boards, I'd pay more for a clean solution like this.

Sure. Might try that next. Would be a lot easier to retrofit than replacing the main 3v3sys regulator.

[hmuller]

.15 mm is probably not going to be a big deal. Should have really ordered 1.8 mm PCBs though. Guess we'll see in few weeks ¯\_(ツ)_/¯

=) Worst case, you will need to use the potential adapter/daughter board mating solution you identified earlier (short wire segments in the mating vias).

I think I may be making an order of your design after looking it over one more time. If so, I will buy you a cup of coffee!  I haven't attempted soldering castellations or board joining through vias, so practice is a good idea.

[hmuller]

@tsys ,

I performed a physical clearance test of the PCB and it did not pass on my unit.

I took card stock and built it up (including scotch tape) to a 3.0 mm height, creating a test dummy. The rectangular dimensions of the test dummy were approximately the same width and length of the PCB (14.6 mm x 7.8 mm). I placed the dummy over the site it would be installed, over the audio amplifiers. Then the top surface was coated lightly with black pastel. A strip of white Post It note (with adhesive) 0.1 mm thick was stuck to the bottom magnesium case over the site the PCB would be over.

The combined height of the Post It note and the test dummy is 3.1 mm. This is the same height as the PCB (1.6 mm) combined with the height of the tallest component, L1, the inductor (1.5 mm). Note that the full height of the dummy extends to the full face of the rectangle, and not the surface area of the top of the inductor which is indeed smaller.

Unfortunately the test dummy created a full impression on the Post It note indicating there is little or no clearance for the regulator board in my particular unit. As there are bound to be variations between units, you may find this is not an issue with your unit.

But I may have a solution which may work but it will require a slightly different design. I will share the details later. I have not been in a rush, as I have to spend about $200 USD on equipment to do the actual PCB rework, aside from the board manufacture costs.