Power supply and cable issues

tkaiser

destroyedlolo replied at Tue Oct 28, 2014 16:38
So with an external power supply, right ?
The Banana only provides 5v ... and I'm worrying for the D5 if your disk consume so much

As far as I understand the situation is as follows:

a) using a Micro USB connector to power something with 2A current looks like a design flaw

b) the SATA power connector on the Banana Pi is not managed by the AXP209 but instead directly connected to the pwr-in Micro-USB connector (utilizing a ferrite bead(FB3) acting as a resistor so voltage available to a connected disk/SSD drops again)

c) the "burning D5 problem" seems to be related to overall power consumption which must be then related to USB peripherals (either consuming too much or being a 'bad' USB hub backpowering the BananaPi and destroying it due to missing overcurrent protection). I don't know whether the SATA power connector is involved in this situation.

In my setup I want to use a large 3.5" disk. They need both 12V and 5V so a capable PSU providing both voltages might be a good idea and I won't use the SATA power connector on the Banana Pi (to power the disk -- maybe for something else if it's true that this connector is totally indepedent from the functionality of the AXP209 PMU and devices powered by the SATA power connector won't increase the load on the PMU?)

tkaiser

destroyedlolo replied at Tue Oct 28, 2014 05:20
Now if I want to dispatch several 1wire probe in my home, the problem you reported has to be tracked and potentially, I'll have to find an better way to provide power

I just started with temperature probes using the DHT22 sensor. I'm using Olimex' nice RPi-Uext adapter with a ribbon cable connected to the BananaPi (which works flawlessly as long as you have the different PIN layouts in mind: rpi_bpi_GPIO.pdf or here (german)




To read out temperature/humidity I used https://github.com/technion/lol_dht22. First impression: Everything's unreliable. DHT11/DHT22 seem to be very bitchy regarding timings. Clocking the Banana Pi above 816 MHz always resulted in the software not being able to read out the temperatures at all. So measuring ambient temperature inside an enclosure (my next project) won't be possibly with the Banana Pi already inside the enclosure so I need a second one for this purpose

Now that I upgraded my temp-daemon.sh script to also read out the DHT22 probes (and substitute them in case they couldn't be read) I just realized that the temperature values reported are wrong. First I fixed the DHT22 just above PMU and A20 SoC (Banana Pi operated vertically) to measure convection effect. While temperatures increase and decrease as expected the values reported are a few degrees to low:



At 10:39 I changed the position of the DHT22 to not measure rejected heat from AXP209/A20 but to measure the real ambient temperature around. But it's not that cold here. And the spike recorded at 11:32 can't be real too.

Conclusion: If one wants to do real probes there's much more to think about and verify than just graphing nice looking stuff without any real meaning...

seblucas

Hi,

I also installed a DHT22 two days ago and it seems to work fine. Some informations about my case :
* My banana pi is mostly idle so running at 600Mhz
* I used a 5.1kΩ as a pull-up
* I used 40cm cables to put the DHT22 as far as possible to any heat source
* I'm also using Bananian with lol_dht22

I check with a oregon scientific temperature sensor and it seems mostly good for the temperature (not so good for humidity), I also opened a window and I saw the temperature and humidity drop.

I'll receive a DS18B20 in a few day so I'll be able to compare.

tkaiser

Interesting

I made little efforts (read https://www.sparkfun.com/products/10167 and tried some things).

In my case it makes no difference whether I use 3.3V or 5V (currently I run two DHT22 side by side: one with 3.3V and the other on the 5V rail). But cabling is important. I used cheap jumper wires before and what I got were huge delays or complete fails to read the probes. I fixed it now with better cables/connectivity and the reliability improved a lot.

But the temperatures reported are approx. 3 degrees too low at normal room temperature level (no idea how the DHT22 behave when it's about reporting much higher temperatures -- I want to test them in an enclosure under load). When adjusted via a formula to increment by 3 then it looks good right now:

tkaiser

seblucas replied at Sun Nov 2, 2014 12:33
I used 40cm cables to put the DHT22 as far as possible to any heat source

BTW: Interactions with 'heat sources'. This here was just me sitting next to Banana Pi and the two DHT22 for half an hour after we came home from Dinner:



Even the SoC's and the PMU's internal thermal sensor were able to record the presence of a human being 0.5 meters away

seblucas

My Banana Pi / DHT22 are on the first floor (the DHT22 is hanged on the side of a bookshelf and is quite protected) and I never saw anything like that (we spend most of our time on the ground floor that's maybe why).

tkaiser

seblucas replied at Mon Nov 3, 2014 02:42
My Banana Pi / DHT22 are on the first floor (the DHT22 is hanged on the side of a bookshelf and is quite protected) and I never saw anything like that

But in case you're willing to try this experiment (standing 15 minutes nearby the sensor) it should be noticeable (not by reading sensor values but by using plotted graphs) since human beings are effective heaters emitting heat in any direction

Anyway this was more a 'note to myself'. I plan to buy a dozen of cheap DHT11 to record temperatures inside and above an enclosure to measure effectiveness of heat dissipation through convection. The first important step would be to calibrate all the DHT11 in the needed temperature range (I assume that i'm not done measuring all of them together in 10° steps and get a simple increment/decrement value for each of the probes but must instead develop more sophisticated formulas for every sensor -- fortunately RPi-Monitor supports this sort of stuff)

And the second important factor is to always consider ambient temperature as well and how this will affect reading out thermal probes from within chips or an enclosure.

seblucas

If you're only interested in temperature (and no humidity), you'd better buy some DS18B20 (1€ / piece in china). DHT11 is just too unreliable for my taste.

I'll try to test your theory using one of my daughters as a heat source (just for the fun).

tkaiser

I thought humidity could also be an issue inside enclosures due to possible condensation (I will need the sensors just a few days for the initial measurement stuff and wanted to reuse them afterwards in other enclosures). But thx for the DS18B20 recommendation. I thought they would use analog readings (which might be a problem over longer distances) but this is not the case. They're available for just 0.67 € on aliexpress with an accuracy not that promising: "±2°C(Max)" ;)

seblucas

The accuracy of DS18B20 is not really that bad ;) :

See : http://www.maximintegrated.com/e ... erface/DS18B20.html

±0.5°C Accuracy from -10°C to +85°C as far as I know they're pretty good.