nrf51dk sniffer capturing advertisement of nrf52 blinky - Inconsistency and missing packets, especially on chan 37

[Ron]

Hi,

I am working with the nRF52 DK running the Blinky example, captured BLE packets using the nRF51 DK sniffer and analyzed the interval timings across the advertising channels (37, 38, 39). The results for both tests are attached as graphs (intervals on y axis, time since last advertisement on channel) -

I also ran the blinky modified to use 100ms advertising intervals- see here- 

There seems to be a lot of missing packets, especially on chan 37

This test was done after I experience some inconsistencies with my new nRF52 design scanning results, I wanted to make sure I am advertising correctly adn to make sure I am testing properly using nrf51dk sniffer to do so, so first I wanted to see how it looks like when both HW and FW from nordic is used -  so -

What is causing all those missing packets? nRF51DK sniffer capture limitations? the packets are actually not being transmitted? is this normal and expected? 

Is it expected for performances on 37 to be worse than 38, and for 38 to be worse than 39? why is that?

-------

Another thing- If I need to get a more professional sniffer solution - what (low-cost solution, lets say 1000$ price range) would you recommend?

I am attaching the logs, on which the graphs were based on,  they are were exported to json, see below

Thanks!

11-2f_blink_a01.json

11-2f_blink_a01_100ms.json

[Hung Bui]

Hi Eran, 

Could you let me know how you preform the test ? Did you set the scan channel to one single channel on each test ? 
Also could you try to test with a nRF52 as a scanner ? The radio on the nRF52 has a lot of improvement compare to the nRF51. 

Have you tried to test on different environment ? There could be a chance that channel 37 has some interference with other devices in your test area. 

[Ron]

Hi Hung

Thanks for the input, 

regarding my wireshark setup?, I just just filter with "btle.advertising_address contains 11:2f"

regarding trying with nrf52 , I will try it, but what would it mean if the results are better? our device interfaces many different smartphones, perhaps some may be just as as good as nrf51, so are my existing results with nrf51 normal? does the nrf51 misses packets because of lacking FW? or is it a HW/RF issue?

Thanks!

[Hung Bui]

Hi Eran, 
Have you configured the sniffer to sniff only on one advertising channel ? 
The way the sniffer works is that when you run it by default,  it will listen on only one channel, I don't remember the exact order. Can be that it listen to 37 first and then switch to 38 to catch the 2nd packet and then 39 for the third packet. But this doesn't explain what you observed. 

My suggestion is to set the sniffer to listen to only one channel at a time (Adv Hop) 

[Ron]

Dear Hung

Following with recommendations, I have ran another experiment - this time with nRF52833-DK and focusing only on Chan 37, it does look much better with 77% success rate, but is this good enough? Is that what we would expect? 

I had one unit which failed to transmit any ADV packets whatsoever for a time, this issue came and went after some reprograming sessions, I was not able to recreate it since than, and although it was a unique singular issue, it made me suspect that there is a significant variance between different units in terms of the BLE performance, and that it might be observable using the sniffer, and lead me to find certain HW and FW bugs, that is why I have undertaken this analysis, it might be, for example, the the clocks of some units drift, that would cause more failures I assume, so I want to have a better insight into my system's BLE performance for such cases, and so let me repeat one of my earlier questions - Should I get a more professional sniffer solution? what (low-cost solution, lets say 1000$ price range) would you recommend? 

Thanks and best regards

Ron

[Hung Bui]

Hi Ron, 

The issue with one particular part may come from other passive component on the board, especially the crystals either 32MHz crystal or the 32kHz crystal. So you need to take them into account as well. 
It's unavoidable that there will be variant difference from chip to chip but they should all pass our intensive test procedures before reaching the market. 

Regarding your question, I'm not familiar with the price of our equipment but what I use here for more advanced sniffing is a Ellisys Bluetooth Tracker. It's a small but handy sniffer, I'm not so sure about the price. 
Other than Ellisys, we also use Front line (Teledyne) sniffers. But again, I don't know about their price. 

The nRF5 sniffer is still my main everyday tool. 

[Ron]

Hi Hung

Thanks for your answer, from issues stemming from crystal inaccuracies, do you think I need more advanced sniffin like Ellisys Bluetooth tracker or Front line sniffers? or is the nRF5 enough?

I am trying to see if my sniffing test setup is sufficient and than to start comparing between different oscillators, different FWs etc.. but if that 70% error rate is stemming from sniffer issues and not actual device/crystal/FW so its probably not good enough, what do you think?

Best regards

Ron

[Hung Bui]

Hi Ron, 
Could you 

Have you tried to test in a RF chamber ? We need to rule out the interference in the environment. Could you try a quick test inside a microwave oven ? 

Have you tried to test with a nRF52 as the sniffer , just for comparison  ?
When you scan on channel 38 and 39 (scan on single channel) do you have better result ? 

30% packet drop is quite high in my opinion. Could you use a phone to scan and compare ? The nRF Connect app has an option to draw the same graph on each advertising packet it receives.