Since I experienced a few sudden shutdowns the last days under high load or when plugging in USB peripherals or when I let an older connected disk spin-up, I exchanged my PSU/cable and tried to measure what happens:
I connected USB keyboard+mouse, a SSD (running iozone test), a HDMI display and put some load on the SoC: My powermeter showed ~8W useage (please remember: the crappy micro USB connector is rated for 1.8A max --> 9W) and according to my multimeter voltage dropped already from 5.1V down to 4.7V.
C’mon: What is micro USB rated for? 1.8A maximum. Have you ever had a look at the tiny contacts?
Did you measure whether voltage drops occur or not? That’s the question! With micro USB they occured and using the very same PSU (with USB type A to 5.1/2mm barrel cable) and the soldered barrel plug they don’t. Problem solved!
Do you know why everyone in the industry who’s not mad in the brain doesn’t use this shitty connector? Do you realise that responsible vendors don’t use this crap even for charging?
Do you know when voltage drops will occur when you use a crappy DC-IN? Only when consumtion is high!
Did you already measure (of course not!) what happens when you cool the A83T with a fan and let it operate under high load? Consumption of the CPU cores alone will easily exceed 8W (and your connector is rated 1.8A/9W max. so there’s “not that much room” for the additional 5-6W connected USB peripherals, a SATA disk and the display engine driving a connected display will need).
Choosing micro USB for a power hungry SBC is braindead. And the current available from the PSU is nearly irrelevant because you run into undervoltage/undercurrent situations way before the current limit is reached. Again: micro USB is rated for 1.8A max and might be able to provide 2A under best conditions (heating up the tiny little contacts!)
The PSU I use is rated 5V/2.1 and I used it over the years to power/measure all sorts of things. When you combine it with a cable with barrel plug it’s always fine. When you combine it with micro USB it simply sucks when consumption increases:
This is what happened with a boring old Banana Pi M1 (not able to consume more than 600mA/3W under full load when overclocked). The very same PSU and cable used with the M3 and its A83T isn’t sufficient enough. And this is just general knowledge, no need to discuss.
I would suspect it’s still the voltage. That’s important since everyone complaining about power problems is asked “how much is your PSU rated?”, then they answer “2A” and will be advised to go with 5V/3A (USELESS since the shitty connector isn’t able to exceed 2A and only rated for 1.8A).
Sometimes these ‘stronger’ PSU seem to work. But most of the times this is due to a better/different cable between PSU and board and not the more theoretically available current
This is already general knowledge for boards that can’t exceed 2A in consumption (like Banana Pi M1). The M3 when running under full load with good heat dissipation, a few USB peripherals and a connected disk might easily consume 3A. Using micro USB here is just moronic.
Why did BPI-M3 use micro USB otg for power? It seems though causing problem because I get SBC turned off time to time. I am not sure what really causes it because I am not measuring anything, but I suspect voltage drop or low current due to too many devices or chips connected.
What can be solution for this? I hope SINOVOIP comes up with something soon because this is not very good. Voltage or current shortage can’t really be the answer logically because other devices (cell phones or other smart devices) probably use more DC then BPI-M3. What do you all think?
If you know someone who can solder, it is easy to fix. Parts you get at FARNELL, CONRAD & such.
You can try limiting the CPU frequency (i.e. max 1GHz), make it slow. It will then take less power.
Do not attach USB devices or only via external-powered-USBhub.
No SATA Harddisk, only via external-powered-USBhub.
And that’s the reason customers want to buy it. At least that was one of the reasons I thought about ‘upgrading’ my M1+ to M3. If I understand the thermal issues right it’s not possible to exceed 1.2 GHz without a heatsink? And according to this thread where heatsink and fan are used it seems not possible to exceed 1.8Ghz anyway: Set CPU clock speed to 2GHz
There’s something wrong with the product specs when the advertised CPU performance can not be used or even the clockspeed has to be limited to half of the maximum to use the board without power troubles!
Well, unless there are different versions, I guess we are all affected by the same problem with BPI-M3. As you suggested right below, if it can be resolved by dropping its clock speed, it should be reviewed. 8 cores of 1GHz will still be faster than 4 cores of 1GHz.
Stability must be achieved with BPI-M3 soon. Otherwise other SBC’s will catch up very soon. I don’t think competitors are just watching.
I have been using a few power adapters came with my Samsung phones. They are all 5V 2.0A otg adapters. Then I also had a couple of power adapters which were given by a phone store. These are not made by Samsung nor anyone well known. However, I found out that its specification saying 5.0V 2.5A OUTPUT!!! Wow, I was just checking to see if they provide 2.0A, but now I have two microUSB power adapters providing 2.5A!
Immediately, I tried the adapter on my BPI-M3. It seemed to work fine. Nothing blew up. I just wanted to make sure nothing becomes too hot. Actually, USB-SATA bridge chip was somewhat hot after I connected SATA HDD.
With 2.5A power, the board was pretty stable even with SATA, HDMI out, and USB devices.
I don’t know if we can say that our power issue with M3 is at rest now if we use 2.5A adapters yet. At least it is working OK for now. I am planning to run for awhile and find out unless SINOVOIP comes up with something better.
I am trying to backtrack to purchase many of these 2.5A adapters for now.
Have you already measured how much the board consumes when you use it with these other power adapters? If not, you can’t draw any conclusions.
Have you already made a full load test? If not, you can’t draw any conclusions.
It’s a well known fact that micro USB can’t exceed 2A so the cables/adapters you use now just suck less than the chargers you used before.
Due to the dvfs settings Vcore will be increased to 1000mV or even 1080mV when reaching 1800 MHz. And then CPU activity alone is responsible for a whopping 8W consumption. And while this might work in a test environment with just a serial console attached it can’t work in a normal environment with a connected display, a few USB peripherals and a connected disk needing another 4-5W exceeding 12W easily. Micro USB is rated 5V/1.8A max: 9W or maybe 10W under best conditions (most likely due to crappy USB cables the board’s PMU will diagnose undervoltage way earlier and shuts down).
In case you improve heat dissipation be prepared for shut-offs unless you solder a sane DC-IN solution and use 5V/3A at least (won’t help with micro USB due to the tiny contacts)
(quoted from the sunxi wiki). So please optimise heat dissipation, run a full load test and provide consumption numbers using a power meter.
Dracon, why do you waste your time with @seities - if a person in this area (SBC) does not know
how to solder a wire to a hole in the board or recognize limitation of a plug, it is just not worth to talk about.
Unless this person is a hell of a coder
Same about @Beaver