nrf52832 ble problem

[Navdeep]

I have designed nrf52832 customized board with tace antenna antenna and every pin of the controller is connected to the male headers present on both side of the boards. I havent used lpkf , so I changed the sdkconfig.h file to use internal RC but it wasnt working. By mistake , the P0.07 pin gets connected to Vcc and the led start blinking i.e. the bluetooth is advertising. Now ble is working fine but everytime I have to connect this P0.07 with 3.3v or also with little variation on that pin (like when I Just touch that pin with my fingers). I want to kmow that P0.07 is just an I/o pin , so why this behaviour? DOes it mean there is problem in my layout design.

There is no problem in the code as I'm using the standard code of ble of nrf52 dk by changing the sdkconfig.h as we dont have lpkf

[Navdeep]

Ohk I will post you the schematic and layout soon. But a quick question,as you have said, the capacitor size matters!!.

 I have used various 0201 capacitors instead of 0402 and 0603 but with the same specs given in nrf datasheet. Does that honna make any difference?

[AmbystomaLabs]

I should have said “capacitance value”. For the logic portion of the chip, the physical size is not too important. But since the SRF ( series (or self) resonant frequency ) is strongly correlated with the value, it is best to use a combination of large and small value caps as they do in the reference design. 

On the RF the physical size of the cap influences the size of the pads and microstrips that should be used. This is more a method of controlling parasitic reactances on RF.

Our own designs make use of 0201 parts for much of the logic and all of the RF. Only bulk storage capacitance is made up of 0402 and 0805s. We utilize 0805s since our main product is a coin cell powered device. Thus we need a lot of local capacitance.

happy to review your design and provide feedback.

[AmbystomaLabs]

You can send your updated layout for review to my email if you wish. 

Yes the local capacitors are not doing anything because the length of board surface the current must flow is so long that there is significant resistive losses and parasitic inductance. Also since the ground currents for the entire nRF pass through that same narrow connection they talk to each other ( or couple). These are often called ground loops and are a large source of noise in a device.

[FormerMember]

(It is also possible to upload layout for review in private case here on devzone.)

Thanks @AmbystomaLabs for such good and detailed answers!

[Navdeep]

I have done 2 connections with ground to the nrf pad (49) by scratching off the solder mask one connection through the cap c7 and c8 via , but still the problem persists. I think this is the only issue. SO should I go for new layout ? I will mail you the new design for the review.

[AmbystomaLabs]

You should do 4 minimum on the ground plane side as I showed in the drawing.  It really should start working.  Let me know.  It shouldn't take you more than an hour to accomplish and test.  Time well spent.

Also, please post of photo of your re-work when you reply.

[Navdeep]

HI please find the attachments. I think because of small size, I am not able to connect wires in a proper way. May be, that's the issue

[AmbystomaLabs]

No, no, no, that won't work at all. The reason the ground problem is bad is due to the physical length of the ground connection. Ideally we are looking for about 1mm of wire to bridge your ground. What you did will not work.

See below for the correct way to do this. A trace on the 52840 DK was bypassed for to demonstrate this technique.

[AmbystomaLabs]

And finally here it is with the excess wire snipped off and an nRF52832 placed alongside for size comparison.

[AmbystomaLabs]

So out of curiosity, I'm guessing you didn't populate this board?

And, did the person who populated it with components do correct reflow work?

[Navdeep]

I think so because its one of the best companies for component placement here in canada and I'm working on the connections , will let you know when the board works