Timer and Multirole for PCA20006

RE: Timer and Multirole for PCA20006

Tobias Hofbaur — Thu, 25 Aug 2016 11:32:34 GMT

That is correct. Red dots show total number of infected vs. time. and blue are newly infected. I also added the graph showing who infected who.

RE: Timer and Multirole for PCA20006

tesc — Thu, 25 Aug 2016 11:17:33 GMT

This is so cool. Thank you for sharing the graph.

I suppose it is time on the x axis and count on the y axis, and that it shows number of infected individuals (red circles) and new incidents since the last data point (blue x markings).

It is great to hear that it was a success at the summer school, despite of relatively short battery life.

RE: Timer and Multirole for PCA20006

Tobias Hofbaur — Thu, 25 Aug 2016 10:38:03 GMT

Final conclusion with a lot of help from nordic support is that the implementation used too much power. I adapted the advertising and scanning cycles an achieved a running time of ~ 6 days.

The project itself was a great success at the summer school. I attached on figure showing the evolution of our disease.

The next picture show the succession of infections. We had some superspreaders who were well connected.

Thank you very much for your support.

RE: Timer and Multirole for PCA20006

Tobias Hofbaur — Tue, 26 Jul 2016 20:23:34 GMT

So, I did some more testing. It turned out, that it was simply the battery that was depleted. Using a new one everything worked fine. Only problem now is, that even the new one seems to be depleted after approx. 8h of testing. I deduce this, as the LEDs are now much weaker than at the start, barely visible and I get resets again.

I have updated my latest code above but I cannot test the power consumption on the DK. If you have any idea, what is eating my capacity i would be glad. I will try to find something in this thread in the meantime.

RE: Timer and Multirole for PCA20006

Tobias Hofbaur — Thu, 21 Jul 2016 13:30:20 GMT

Thank you. I will see what I can get on my device.

RE: Timer and Multirole for PCA20006

tesc — Thu, 21 Jul 2016 12:14:46 GMT

Sent you the file by email. With your .hex file it does count upwards, but the blinking seems to be much slower...

RE: Timer and Multirole for PCA20006

tesc — Thu, 21 Jul 2016 12:00:59 GMT

OK this is strange. I used the .hex file from your project and I think it behaves differently. I'll find a way to send you my .hex so that you can see if that one works as intended. If I can find a way, seems you can not add files in comments...

RE: Timer and Multirole for PCA20006

tesc — Thu, 21 Jul 2016 11:48:43 GMT

Do you have more beacons so that you can rule out faulty hardware on the one you are using for development? While it seems unlikely it would be good to just rule out the possibility.

RE: Timer and Multirole for PCA20006

Tobias Hofbaur — Thu, 21 Jul 2016 11:43:16 GMT

Damn... That would have been the easy solution for a mistake on my side. Is it possible that my hardware is fault? Not very likely that it only concerns scanning, though.

RE: Timer and Multirole for PCA20006

tesc — Thu, 21 Jul 2016 11:41:25 GMT

I simply downloaded the eclipse project folder you provided in the question (edit 2) and put it in /central_and_peripheral/experimental/ ...

RE: Timer and Multirole for PCA20006

Tobias Hofbaur — Thu, 21 Jul 2016 11:34:35 GMT

Thanks for your time. I was trying to figure out which error i ran into, but none of my breakpoint in app_error or app_error_weak triggered.

Could you zip and attached your project? Probably we differ in one of the includes.

I also went back to my early test versions for scanning and realized that the blinking I get there is not from received advertisments, but rather from the reset as well.

RE: Timer and Multirole for PCA20006

tesc — Thu, 21 Jul 2016 11:20:41 GMT

Hi,

Sorry for the delays. I have too much to do these days, and so it is limited how much I can look into this. Using the earlier version of your project I was not able to reproduce any errors, so I think it would be limited value in trying to reproduce it here.

AIRCR.SYSRESETREQ means you have probably ended up in the error handler (app_error_fault_handler(), via one of the error handlers in components\libraries\util\app_error.c.) See An introduction to error handling in nRF5 projects.

Sorry again for the limited follow-up, but at least I hope this will point you in the right direction.

Regards, Terje

RE: Timer and Multirole for PCA20006

Tobias Hofbaur — Wed, 20 Jul 2016 06:47:09 GMT

Ok. Misread this post, and read the register before enabling the SD. With enabled SD and cleared register in last run, I get as reset reason 100 = Reset from AIRCR.SYSRESETREQ detected

I have not figured out yet, what this means.

So apparently my device is getting a softreset. I just cannot figure out, where it originates.

It only appears if I call scan_init
The reset frequency seems linked to the scan interval
It is not linked to something in either on_ble_event or main_timer_event_handler (commented everything out)

Can you reproduce the error.

RE: Timer and Multirole for PCA20006

Tobias Hofbaur — Mon, 18 Jul 2016 16:30:48 GMT

I did not clear the register. I did it a little bit to fast in my lunch break. I will post the correct result on Wednesday.

RE: Timer and Multirole for PCA20006

tesc — Mon, 18 Jul 2016 11:45:10 GMT

Yes. After having read the register you should make sure to run sd_power_reset_reason_clr() to clear the old bits (since the register is retained). You must provide a mask (for what bits to clear) and using the value just read should work just as well as 0xFFFFFFFF.

I am a little confused that you got all those reasons, were they properly cleared between reruns?

RE: Timer and Multirole for PCA20006

Tobias Hofbaur — Mon, 18 Jul 2016 07:52:59 GMT

Thank you for your help.

If I read the register correctly (by inserting sd_power_reset_reason_get(&reset_reason); in main before sd is initialized) I get the following reasons after starting a debug session: A,B,C and once E

RE: Timer and Multirole for PCA20006

tesc — Mon, 18 Jul 2016 07:47:55 GMT

This sounds strange indeed, if anything I would expect the newer revision to work better than the old one. I will try to get hold of a new beacon and reproduce the behaviour there.

RE: Timer and Multirole for PCA20006

Tobias Hofbaur — Mon, 18 Jul 2016 07:43:17 GMT

I have CEAAE0 , so it seems I am running IC rev 3 vs. 2 on your side.

My power came from battery, but I will check the register (Unfortunately I have a business trip, so I can onyl check it tomorrow evening)

RE: Timer and Multirole for PCA20006

tesc — Mon, 18 Jul 2016 07:37:03 GMT

We may have different revisions of the nRF51822 chip on the beacon kit. The markings on mine is CEAAD0, what is yours? Knowing this would let me check the erratas for clues.

Also, I have been powering the beacon from a DK. The power consumption should increase slightly with temperature, may it be that you experience brownouts? You can check the RESETREAS register (See the Reference Manual, table 53 RESETREAS) to see if the chip has been reset from other sources than regular power-off or brownout.

RE: Timer and Multirole for PCA20006

Tobias Hofbaur — Sun, 17 Jul 2016 17:20:10 GMT

Each time with a flash of the LEDs.I could verify that this LED flashing is in fact the LED initialization I do in main. So it seems, that with scanning the beacon reboots at irregular intervals.
If I manually reset it from this state the reset frequency initially seems to correlate with the scanning interval (checked by increasing it). However later on the frequency increases s.t. it is practically always flashing.

Do you know what could be the reason for this? The only thing I found is that I am using Beacon rev. 1.2. with S130 which apparently was a problem one year ago. It would be great if we could solve this strange bug by the end of the week, as that is my deadline for ordering the hardware for the summer school. Thanks in advance Tobias

RE: Timer and Multirole for PCA20006

Tobias Hofbaur — Sun, 17 Jul 2016 17:14:25 GMT

Hello Terje. You are right, I still have a lot of safeguards to implement. I implemented a couple over the weekend and fixed some minor bugs in the functionality (updated atteachment). Now on the DK the code is behaving exactly as it should.

On the beacon however something is wrong. I oberserve the following:

If I do not call scan_start() everything seems to work alright. The advertisment is correct and updates to show the right time counter for a long time. No LEDs are blinking (as intended).
If I switch on scanning by calling scan_start I oberserve different behaviour. If I do a cold start, i.e. the beacon was without power for several minutes (~30), at the start everything looks fine, i.e. same behaviour as before. After some time (varying, observed up to 350 s) the LEDs flash and the counter is reset. The counter increases again and is reset at shorter and shorter intervals.

RE: Timer and Multirole for PCA20006

tesc — Fri, 15 Jul 2016 13:39:07 GMT

One last comment before the weekend: You are not checking that infect_source[sender_id >> 3] on line 195 is within the bounds of the infect_source array. Any chance the sender_id is 120 or higher (or read from the wrong position)? Although you specify IDs to be in the range 0-119 it would still be wise to do a check before writing the value, as that would let you know an error there right away instead of getting mysterious behaviour.

RE: Timer and Multirole for PCA20006

tesc — Fri, 15 Jul 2016 11:27:31 GMT

There is no point in changing the scan parameters, as passive scanning is pure RX (and as such can be done by an infinite number of devices simultaneously). Reducing the scan time would only decrease the chances of catching advertising packets (although it would of course save battery on the scanning device to scan half as much.)

The current s130 can only have one link as peripheral, but you can have multiple links as central. This has to do with timing constraints; the central decides connection parameters so you can schedule multiple connections in central role, but the peripheral has to accept whatever schedule the central provides meaning multiple connections in peripheral role would be a scheduling nightmare with a lot of scheduling collisions.

RE: Timer and Multirole for PCA20006

Tobias Hofbaur — Fri, 15 Jul 2016 10:36:12 GMT

Thank you for testing the code. The red LED/LED3 (indicating the status) should be on 1 sec, off 1 sec, etc. The DK works well for me, but the Beacon seemed to blink faster after some time. The blue LED/LED2 should signal any incoming ad from a tag with status I. The counter should show the seconds since power up (actually the number of times my main loop was called.) I planned to increase the adv_interval to 1000ms, the scan window to 100ms and the scan interval to 200ms to lower the collision rate. In connected mode, only 8 partners would work? But I did not plan to use that anyway. (Maybe in an upgrade to record the network, but then only one connection to a central)

RE: Timer and Multirole for PCA20006

tesc — Fri, 15 Jul 2016 09:26:06 GMT

The reason connected mode would prove difficult is the limit on simultaneous connections as peripheral and central, which would not allow for all-to-all communication.