Pine A64 Plus Fails to Boot

[TinkerBear]

I sorta had my doubts that it could be the power supply, so I sat down to do some rigourous testing - three different power supplies to try, recording fluctuations on the power input (DC in and 5V lines) via Euler header and a DSO (Siglent SDS1052DL).

The good news is that there's not much to see on the power line. Some fluctuations, but no major departures that I can see.

The bad news is... No crash when running off my bench power supply. Definite repeatable crashes on both a "PQI" brand 3.4A USB charger and a cheap chinese 5V3A micro-USB charger. For giggles, I tried a genuine Apple iPad 2.1A charger - these things are generally regarded as very well made, with little ripple and high reliability... and it crashes. Different iPad charger... still crashes.

It crashes (mostly) just after "[ 13.159862] cfg80211: Calling CRDA to update world regulatory domain". No idea why, I've had my WiFi/BT module disconnected for all this testing.

The average current pulled by the board is only around 0.3A. Is there a decoupling capacitor missing somewhere?

Let's try bodging a 10µF low ESR tantalum capacitor on the 5V power rail on the Euler connector... no crash with bench power supply, still crashes with PQI wall wart. Maybe the 3.3v rail would be better... nope.

Maybe I'm missing the obvious solution - bin this POS and buy an RPi 3.

Wait - ONE MORE THING TO TRY - move that 10µF cap to the DC IN pin of the Euler connector... And I'm booted and running off the PQI supply. But not the 5V3A chinese supply. Let's try that again. Yep, works on 2nd try. Hmmm, maybe bigger capacitor.

[pfeerick]

I sorta had my doubts that it could be the power supply, so I sat down to do some rigourous testing - three different power supplies to try, recording fluctuations on the power input (DC in and 5V lines) via Euler header and a DSO (Siglent SDS1052DL).

....

Wait - ONE MORE THING TO TRY - move that 10µF cap to the DC IN pin of the Euler connector...  And I'm booted and running off the PQI supply.  But not the 5V3A chinese supply.  Let's try that again.  Yep, works on 2nd try.  Hmmm, maybe bigger capacitor.

Yeah, me also, since I have a 8000mah lipo strapped to the underneath of the pine64, so it ain't gonna gonna get any ripple in the supply that way. It all seems rather strange... as this particular unit (which does have a WiFi module installed, but no ethernet) had been pretty stable prior to now, and then started having issues on reboots. However, I tell a lie (unless I had a different SD card inserted at the time - which is possible... I often swap that one between vanilla ubuntu and Armbian ubuntu depending on what it is being used for... it's a mobile test rig) it is currently on "ARMBIAN 5.25 stable Ubuntu 16.04.2 LTS 3.10.104-pine64", so if it was this image that was having issues, it is not 105 kernel related. However, I'm pretty sure it exhibiting reboot roulette (i.e. have to keep rebooting it until it decides to work) until I plugged the HDMI in, but that could be something going on with the WiFi driver from what longsleep was saying.

If I didn't have a battery connected I would probably also try 1nf, 100µF (and a 1F supercap :-P) on the euler pins just to make sure things are smoothed out, but in my case it has a battery connected... so already has a bit one Or the PMIC just isn't doing it's job.

[TinkerBear]

I sorta had my doubts that it could be the power supply, so I sat down to do some rigourous testing - three different power supplies to try, recording fluctuations on the power input (DC in and 5V lines) via Euler header and a DSO (Siglent SDS1052DL).

....

Wait - ONE MORE THING TO TRY - move that 10µF cap to the DC IN pin of the Euler connector...  And I'm booted and running off the PQI supply.  But not the 5V3A chinese supply.  Let's try that again.  Yep, works on 2nd try.  Hmmm, maybe bigger capacitor.

Yeah, me also, since I have a 8000mah lipo strapped to the underneath of the pine64, so it ain't gonna gonna get any ripple in the supply that way. It all seems rather strange... as this particular unit (which does have a WiFi module installed, but no ethernet) had been pretty stable prior to now, and then started having issues on reboots. However, I tell a lie (unless I had a different SD card inserted at the time - which is possible... I often swap that one between vanilla ubuntu and Armbian ubuntu depending on what it is being used for... it's a mobile test rig) it is currently on "ARMBIAN 5.25 stable Ubuntu 16.04.2 LTS 3.10.104-pine64", so if it was this image that was having issues, it is not 105 kernel related. However, I'm pretty sure it  exhibiting reboot roulette (i.e. have to keep rebooting it until it decides to work) until I plugged the HDMI in, but that could be something going on with the WiFi driver from what longsleep was saying.

If I didn't have a battery connected I would probably also try 1nf, 100µF (and a 1F supercap :-P) on the euler pins just to make sure things are smoothed out, but in my case it has a battery connected... so already has a bit one Or the PMIC just isn't doing it's job.

My somewhat vague knowledge of electrical engineering says that to fill a really fast dropout, you need something that can flow a lot of current in a hurry - and that typically means a ceramic capacitor.  Very low ESR and cheap... and not available in larger capacities.  Electrolytics can store a lot more energy, but can't give it up as fast.  (And supercaps are even slower, batteries slower yet.)  So I'm not sure if I need a bigger cap, or a ceramic.  And I have other things to work on today, so maybe tomorrow.

[MarkHaysHarris777]

I sorta had my doubts that it could be the power supply, so I sat down to do some rigourous testing - three different power supplies to try, recording fluctuations on the power input (DC in and 5V lines) via Euler header and a DSO (Siglent SDS1052DL).

....

Wait - ONE MORE THING TO TRY - move that 10µF cap to the DC IN pin of the Euler connector...  And I'm booted and running off the PQI supply.  But not the 5V3A chinese supply.  Let's try that again.  Yep, works on 2nd try.  Hmmm, maybe bigger capacitor.

Yeah, me also, since I have a 8000mah lipo strapped to the underneath of the pine64, so it ain't gonna gonna get any ripple in the supply that way. It all seems rather strange... as this particular unit (which does have a WiFi module installed, but no ethernet) had been pretty stable prior to now, and then started having issues on reboots. However, I tell a lie (unless I had a different SD card inserted at the time - which is possible... I often swap that one between vanilla ubuntu and Armbian ubuntu depending on what it is being used for... it's a mobile test rig) it is currently on "ARMBIAN 5.25 stable Ubuntu 16.04.2 LTS 3.10.104-pine64", so if it was this image that was having issues, it is not 105 kernel related. However, I'm pretty sure it  exhibiting reboot roulette (i.e. have to keep rebooting it until it decides to work) until I plugged the HDMI in, but that could be something going on with the WiFi driver from what longsleep was saying.

If I didn't have a battery connected I would probably also try 1nf, 100µF (and a 1F supercap :-P) on the euler pins just to make sure things are smoothed out, but in my case it has a battery connected... so already has a bit one Or the PMIC just isn't doing it's job.

My somewhat vague knowledge of electrical engineering says that to fill a really fast dropout, you need something that can flow a lot of current in a hurry - and that typically means a ceramic capacitor.  Very low ESR and cheap... and not available in larger capacities.  Electrolytics can store a lot more energy, but can't give it up as fast.  (And supercaps are even slower, batteries slower yet.)  So I'm not sure if I need a bigger cap, or a ceramic.  And I have other things to work on today, so maybe tomorrow.

TinkerBear,  this is also for Pete, the problem is that the wall wart PSU(s) are not clean especially under load ( like the spikes which appear during bootup ).   ( yes, the DCIN could have been more robust with decoupling caps and inductors ).

I power all three of my pine boards from 5v 2.5A PSU(s)  (official RPi power supply)  using the euler bus and using a PI (passive) low_pass filter on the input.

Take a look at the DC Power In  section of this forum (stickes) for several articles that I wrote regarding the PI filter (with pics) and the overall differences between the official RPi power supply and the cheaper one provided for the Pine board  lite weight 5v 2A.  I pasted the links below.

Both supplies will work IF you provide the PI filter on the DCIN and IF you use the euler bus. 

A quick note about euler;   the microusb connector is defective on the Pine boards.  It will not correctly handle 2.5+ amps  (such as occurs during bootup) and the pine board will crash. (this is documented on the wiki).

Please read this entire link:

Please also read this link:

[TinkerBear]

Hey Mark, thanks for your help the other night! You got me unstuck when I was pulling my hair out - much appreciated.

I'm not sure blaming the power supplies is the right answer here. I'm measuring roughly 0.3A drawn - it's not taxing any of these power supplies, nor the connector. A quick google says brand-name Micro USB connectors are rated for 1.8A for VCC and GND, and even the crap chinese ones must be good for half that.

I don't think ripple is a problem either - or it'd be causing crashes much more often than just at boot. My board has run 24 hours without a crash (until I shut it down), once it made it through booting.

So, my working theory is that at some point in the boot, a bunch of hardware is switched on at once, causing a short/sharp sag in the voltage. (Why it doesn't cause a problem with an HDMI monitor, I don't know, timing issue?) That's the sort of thing I've seen fixed with a decoupling capacitor, storing a wee bit of energy locally, rather than in the power supply filter caps at the far end of that - probably thin - wire.

Your PI filter undoubtedly works, but I'm not sure it's necessary. A single capacitor on a connector on the Euler bus is smaller and easier to add than circumventing the entire power input connector and filters. I think I'll try the biggest ceramic I can find first and see what happens. Might also see if I can get the o'scope to capture a voltage drop. But... tomorrow.

[MarkHaysHarris777]

Hey Mark, thanks for your help the other night!  You got me unstuck when I was pulling my hair out - much appreciated.

I'm not sure blaming the power supplies is the right answer here.  I'm measuring roughly 0.3A drawn - it's not taxing any of these power supplies, nor the connector.  A quick google says brand-name Micro USB connectors are rated for 1.8A for VCC and GND, and even the crap chinese ones must be good for half that.

I don't think ripple is a problem either - or it'd be causing crashes much more often than just at boot.  My board has run 24 hours without a crash (until I shut it down), once it made it through booting.

So, my working theory is that at some point in the boot, a bunch of hardware is switched on at once, causing a short/sharp sag in the voltage.  (Why it doesn't cause a problem with an HDMI monitor, I don't know, timing issue?)  That's the sort of thing I've seen fixed with a decoupling capacitor, storing a wee bit of energy locally, rather than in the power supply filter caps at the far end of that - probably thin - wire.

Your PI filter undoubtedly works, but I'm not sure it's necessary.  A single capacitor on a connector on the Euler bus is smaller and easier to add than circumventing the entire power input connector and filters.  I think I'll try the biggest ceramic I can find first and see what happens.  Might also see if I can get the o'scope to capture a voltage drop.  But... tomorrow.

hi again,  you're welcome !  ... my pleasure.

... on the connector,  its not a theory;  the engineers at the plant have confirmed that the microUSB connector does not work properly, and they have decided to replace the microUSB with a barrel connector;  the new sopine base boards are now shipping with the new barrel connectors; at some point the pine boards will get this same revision.

marcus

[dkryder]

yeah, but in their infinite wisdom they put a non-standard [3.5mm/1.35mm] barrel jack on the board which means to use it requires an adapter of some sort. or a rewiring of an existing jack/plug PS setup. or buying the pine PS which i guess has the proper plug. i happened to have a plug to fit and made an adapter patch cord type thing so i can use 2.5mm/2.1mm plugs. Mark, suggest to the engineers that they use 2.5mm/2.1mm jack on pine64

[TinkerBear]

... on the connector,  its not a theory;  the engineers at the plant have confirmed that the microUSB connector does not work properly, and they have decided to replace the microUSB with a barrel connector;  the new sopine base boards are now shipping with the new barrel connectors; at some point the pine boards will get this same revision.

marcus

It's a good idea - particularly if you want any shields/hats on there, battery charging, and display, and ....  1.8A is weensy.  I assume they used a micro USB because RPi did, and they wanted to slot in as a replacement or upgrade.  I don't think it was a great choice for the RPi either, if I'm honest.

Oh, the irony: Just today I got my two 5V3A wall warts with micro-USB connectors from China, AND learned the current-carrying capacity of a micro USB port is 1.8A.

I'm just sayin I don't think it's the source of my crash-on-headless-boot - because I'm only seeing 0.3A.  My meter doesn't update that fast, but there should be enough energy cached on the board to get through current spikes.

Though now that I think about it, maybe a sense resistor there and the oscilloscope across it would show me current spikes more easily than trying to find a voltage dropout.

[xalius]

I had no troubles with a proper micro-USB supply so far as long as the cables are not too thin, I use those AWG 20 phone charger USB cables and off the shelf 5V/3A wallwarts (simple SMPS, no chargers since those try to be smart about current limits...) that come with like 10 different adapters including micro-usb and the new 3.5/1.35mm barrel plug...

e.g. https://www.amazon.de/PortaPow-spezielle...B00RQ5AZL6

[MarkHaysHarris777]

yeah, but in their infinite wisdom they put a non-standard [3.5mm/1.35mm] barrel jack on the board which means to use it requires an adapter of some sort. or a rewiring of an existing jack/plug  PS setup. or buying the pine PS which i guess has the proper plug. i happened to have a plug to fit and made an adapter patch cord type thing so i can use 2.5mm/2.1mm plugs.  Mark, suggest to the engineers that they use 2.5mm/2.1mm jack on pine64

I agree with you completely.  I'm sure they went with cost of components not standards and convenience.

I was disappointed with the choice of barrel spec;  was really hoping the standard plug for the arduino would suffice for the pine board too.  We can still push this, since they have not begun manufacturing the new pine boards yet-- I believe.