RPi3 powering and overclocking

[pajper]

Hi all!
The first of all, I want to say "Hello" to all of you, because Ihave been using RPi for two years, but it's my first post on this forum.
I have some questions about powering and OC of RPi3. I was using RPi2 previously, but I unfortunatelly damaged it, so I decicded that it's time to try some newer.

I'm using RPi as on-board computer in very extreme conditions with very big accelerates, like 50g, on big altitude.
It isn't simple to power this device in that condition, because the max weight of all devices (RPi and other) is 200-250g.
I created simple powering system, but it's max current is 1,2..1,5A.
To RPi I have connected camera and more devices using SPI, UART and I2C.
RPi2 was working properly on that current, but the RPi3 has bigger current requirements.
I haven't any periferals via USB connected to it, excluding one device with external powering.

I also don't use any displays, connecting to RPi by signals on GPIO pins.

So, my question is: did anyone try to use RPi with as small current?

My second question is overclocking. Did anyone try 1,7GHz? My problem isn't heating, I'm interested in voltage of it.

On RPi, I'm using modified Raspbian, the current of devices on GPIO (from 5V pins) is maximum 100mA.

Thanks for replies,
DP

[Rive]

My second question is overclocking. Did anyone try 1,7GHz? My problem isn't heating, I'm interested in voltage of it.

Stable 1.3 GHz is as high as I have gotten (and that is with a massive heatsink and a fan). Your problem will be heat (or you will throttle back to below 1.2 GHz).

Others have reported going higher, but I doubt they are very stable, or adequately cooled, and will probably crash on stress/cpuburn-a53 if they tried it.

As for voltage...I can't help you there. I use 5.25v 2.4A power supply. I tried lower but the overclocked arm with the required over_voltage struggled. I am thinking the higher the arm overclock, the more juice it needs.

see: viewtopic.php?p=952371#p952371

[pajper]

Thank you for help.
I got 1,5GHz on RPi2, but, yes, it wasn't stable.

I used liquit oxygen as cooling, because it was also used in the project, so it wasn't a problem to use it, when it was also on-board. Also, I used powering by power supply, not by on-board batteries, so I had 5A of maximum current.
So, as I said, my problem is voltage.
I got 4,98V 1,5A, so it isn't too big, but I'm thinking about using small LiPo batteries, actually I'm using one dual LiPol 800mAh, but I'm using voltage converter with output 1,5A.
But I'll try to find some better or, may be, buy second RPi, 0 will be the best option if I find it. And, then, use another powering from another LiPol.
Thanks for help.

[MaxK1]

Do you need to overclock? Do you _really_ need to overclock or just want to? If you _need_ an overclocked Pi3 then a Zero isn't going to cut it for you. If you do overclock, you'll need to test, test and test again to verify that it is stable. I have a couple of Pi3's that passed memtester and stress running for hours (Sometimes days) - I thought I was good to go. When I fired it up with it's intended application it ran fine for about a week and then crashed when I was backing it up (tar -czf ...) quite repeatable. Backed off a bit (1350 down to 1320) and it worked fine. Then increased to 1340 which is where it is today. So, memtester and stress are good for finding ballpark overclocking figures, but can't be relied on to give the final, definite answer. If you don't have the time to fine tune/test every pi3 you are going to use, don't overclock. If you can't afford to wait the 30 or so seconds it takes to reboot in mid-flight, don't overclock.

It sounds like you have the cooling part licked and you understand the power issues. Just make sure you have mechanicly sound (ie can withstand 50 g's ) connections without spraying the the board with LOX or loose power connections/sensor wiring... It would really suck if you launched and had no control, sensor data or images to show for it!

[pajper]

Hi!
Thank you very much for all suggestions.
First, I tried RPi3 with all devices with current 1,3A and it's working fine without overclocking obviously.

I'm creating a rocket, we want to fly, in the first stage of the project, on the pulap of 15-20km. I want to use RPi as on-board computer.
After first test on pulap of 500 metres, so very low, with accelerates 30-35Gs it was working properly to the time... Unfortunatelly, the soldered cables was damaged and the RPi2 was damaged. It is working, but USBs aren't working properly and I didn't want to use computer which can be more damaged.
The problem is that the RPi must register data from about 25 devices and drive the rocket, by setting the position of engines, shot parachutes and second stage of the rocket and drive parachutes to land near by launchpath.
In 900MHz the RPi2 was working little unstable, so I overclocked it, on the ground, to 1,5GHz, but It flew with 1,2GHz.
At moment, tomorrow we'll try again on bigger pulap of 1km with only one stage, without overclocking.
If it works, I'll start testing, as you proposed.
I hope it'll work properly on 1,5A current.
If it is succeeded, we'll shot the rocket in May on 1,5km, and, then, in June, on 3-4km with supersonic speed.

Thank you, again, for the propositions. Also, if you have any ideas about powering, overclocking or, mainly, about the project, please, write.

Greetings,
DP

[Rive]

have a couple of Pi3's that passed memtester and stress running for hours (Sometimes days) - I thought I was good to go

No. Stress is weak (but is 4 times more power hungry than sysbench, for comparison) .

You should test your overclock with cpuburn-a53 (which is significantly more power hungry than Stress, and runs NEON which is about as "Stressful" as you can get....BUT only if you can keep the SoC from throttling back). Chances are very good your unstable (and/or improperly cooled) pi3 would have throttled back/crashed within minutes (thus indicating you have an issue with being unstable, and/or being improperly cooled).

A stable 1.3 GHz overclock on a pi3 that is properly cooled, can run cpuburn-a53 in the low 60's without throttling. A 1.2 GHz stock pi3 that is properly cooled should be able to run cpuburn-a53 at slightly lower temps (approx 60C).

[Rive]

For those who want to overclock and/or properly test their overclocks, I created this thread:
viewtopic.php?p=952371#p952371