Custom board not seen by IOS device but it is seen by NRF-Connect Desktop

[rogerarchibald]

Hi There,

I've been developing a product using the NRF52DK.  I had some custom PCB's made after I had my circuit designed, and I programmed them with the NRF52DK.  Now my boards and the NRF52832 on the DK are running the exact same firmware.

With the NRF52DK, I can connect to my service on NRF Connect for IOS and Desktop (with an NRF51 dongle). 

With my boards, I can't see them with IOS (using NRF Connect or LightBlue) but I can see them using NRF Connect Desktop.  

When I scan with NRF Connect Desktop, my boards' RSSI indicates a fairly strong signal (~ -55dBm) and I can connect and view/control all the characteristics fine.  

I'm going to continue poking around to try to understand what's different but I'm quite stumped as to why IOS will only see the NRF52832 on the DK, but NRF Connect Desktop will see either the DK or the custom board.  I've tried an iPhone and an iPad and got the same result.

Any suggestions would be greatly appreciated!

Thanks,

Roger

[AmbystomaLabs]

Make sure you don't broadcast with an Apple manufacturer ID (ie, 0x004c).  iOS segregates all 0x004c broadcasts to the CoreLocation API's, only non 0x004c show up in the CoreBluetooth API's. 

Or, if you are using CoreLocation in iOS then you can only broadcast with an Apple ID. 

[rogerarchibald]

Thank you for the suggestion, but I am fairly sure it's not an addressing issue since I can connect just fine using the same firmware on the PCA10040 DK.   

[Edvin]

sorry. I didn't see that you responded to the other comment.

Ok. Then this sounds like there could be some issue with the antenna or the pcb. We can check the PCB layout if you like. I am not the one doing this around here, but if you can upload the entire schematics, someone can take a look at it. If you don't want to upload it in public, you can create a private ticket.

 

Best regards,

Edvin

[rogerarchibald]

I'm happy to share files if any advice can be given from them.  If I eventually get this working it will all be posted as Open Source anyway.  Attached are Eagle files, I can upload Gerbers if that's preferable.

FYI, I am trying to make a BLE version of this.  The current version is built on a pair of 8-bit AVR's with NRF24L01+ transceivers.  It works well enough and a couple of local animal rescues have had success using it to trap animals for TNR (Trap Neuter Release) but the interface is clunky and a BLE version would be significantly less expensive due to only having one 'smart' box and not requiring separate transceivers.  I also hope that an app would be more user-friendly for your typical 'cat lady' then a remote with small buttons.OpenTrap_BLE1.schOpenTrap_BLE1.brd

Thanks again for the support and I welcome any feedback.  

Cheers,

Roger

[AmbystomaLabs]

All files received and work fine.  Plus I got the chance to try out the latest Eagle. It has turned into a pretty nice layout package.

I will go through and review the entire thing since you are doing a respin.

It will take about a day to complete. There should be a lot of useful feedback for you.

[rogerarchibald]

Any feedback is appreciated.  A few items on there are copy-pasted from my first board when I was using a 7.2V Nimh battery and gate drivers to drive the ultrasonic transmitter and the boost converter.  I intend to use a 3.7V lipo here and some of the design decisions I'd made for power saving with the big battery aren't going to yield much results here.  Just wanted to explain the weird power-up circuit.  I think on my next spin of this I'm just going to use an on-off switch rather than the push-button, and I'll probably leave the battery level divider on all the time rather than switch it through the pair of FETS.

Cheers,

Roger

[AmbystomaLabs]

As an amusing coincidence, my company makes Cat Tailer (www.cattailer.com). It is a +9dBm, bluetooth tracking device for pet cats. During a stint in the UK, we embraced the RSPCA's approach of, if you are going to have a pet cat, let it live a natural life by experiencing the outdoors where it is reasonably safe to do so. Cat Tailer serves to enhance this by giving you a way to find your cat to get them is at night, or as a way to keep track of them and know they are safe. A scared or injured cat won't call out for you when you look for them. Cat Tailer give you a way to always find your pet.

We are also big advocates of controlling feral populations and your live trap and spay/neuter and release program with the ASPCA sounds great!

I would love to provide assistance on your respin.  When it is complete, please pass it by me either directly via email or the devzone before you have it built.

Below are my comments on your design:

1. You said you were losing the boost converter, but notes on it anyway: On the boost converter, it appears you are doing PWM to modulate the ultrasonic transducer you mentioned.  I  assume you would be using the onboard pwm on the nRF to facilitate this.   Not sure why you need the level shifter, Vgs should be low for your fet. If not then choose a better one.   Diodes Inc. always has nice products in useful packages,especially for heat dissipation. Vg threshold should be less than 0.5volt and the device should be full on by 3volts. ie, nRF_VDD

2. On U2, linear regulators just throw away power.  Always use a switcher(ideally there will be one switcher for the whole device, see comment 4).  Again, look for a synchronous buck  converter. Many operate at 1-3MHz so the inductors are small. TI, ST, and Diodes Inc.are all good companies.  The topology is super easy, and they will all come in a small QFN package  with ground slug. Also I no idea what the point of Q8 is. I would suggest you could remove a lot of U2, resistors and Q8 and replace everything with a proper LiPo controller. And,  resistors just suck power so a good design should have 10k voltage dividers. ie, r38, r37

3. There is no point in running the ultrasonic transducer off of vbat(ie, vin). vbat will vary from 4.2 to 2.7 for a single LiPo.  Correspondingly the output from the transducer will vary  and give you unreliable performance.  Just run it off of the switcher you choose from comment 4 which will make Vdd_nRF.

4. If you use a LiPo as you mentioned, look at some of the TI solutions.  You can get charge management, over voltage protection, under voltage protection, over current protection,  different charge voltages to optimize the life and a synchronous buck converter that will go to 100% duty all one tiny QFN.

5. I thought I saw a tantalum on your design.  Never use tantalums.  They have static losses that are substantial and will shorten your battery life.  Also, they fail easily. They can  dump current so fast that an accidental short while you are debugging causes them to blow apart from the thermal stresses.  Chip caps are safe and fast, plus you can get small 0805  devices at 47uF cheaply. Always use a mix of cap SRF's though.  This is required due to the noise from all the electronics.  I've gone on about it elsewhere on the devzone if you want to  read up on it.  Generally these things will have a mix of 47uF, 2.2uF or 4.7uF, and 0.1uF's on VDD.

6. Unless you have a real reason not to, then you should use the buck converter on the nRF, then you can have 3-3.6vdc gpio, better power management software and maybe lose some of the  level shifters. If you are really trying to push your LiPo then you can run the nRF at 2.7 while still using the internal buck converter for the 2.7 to 1.8 solution.

7. Generally you don't need all the 1k gate resistors.  They show them on data sheets simply because you can fry the gate with a large inrush of current. Most cmos devices can't deliver  that much current so generally the need is not there.  But you should always go through the specs to make sure.  I bet they take up a lot of room though.

8. D7 seems to be a massive LED. The via pattern will not be liked by a fabhouse. The vias are too close together and this tends to weaken the board. Also, they don't seem to serve much  purpose since they didn't connect to the ground pours. It looks like you copied it from the datasheet. Since they obviously didn't complain when you got the board made, just make sure to  connect the vias to ground so it floods on all 4 layers.  Then it should be fine. It's not pretty though.

9. You will save a lot of space on your board by using a current standard for the pad sizes.  Looks like your models are following an old standard. You should update to a more modern  take on this.  Also, if you are hand soldering the parts, I advise against it. Really easy to cause thermal cracks on chip parts and they are nearly impossible to see even with a  microscope. Also, most home built ovens are a bad idea they tend to overheat the parts. For hand built prototypes or one-off designs, I normally buy a cheap metal stencil from OSH  Stencils for the paste, then once the parts are placed preheat the board on a rather cool hotplate to 100c then hit it with hot air reflow until done about 60 seconds. Been doing that  for years and never had a problem. Plus it is easy to stick to the recommended reflow profile.

10. So, all this brings us to why doesn't it transmit well? It'd be great if you had a spectrum analyzer, but I'm guessing you don't. There are really cheap USB WiFi card based ones if  you want to consider it.

Here is what I see wrong on the nRF stuff:
A. Your grounding is just really bad. I see a lot of shared grounds You even seem to stress them in the schematic. As an aside, it is always better to never have a 4 point junction since  it is ambiguous whether you have 4 things connected or just a dot over a couple of connected things. Also, if you break out the parts and put a ground symbol on all the caps you tend to  think about how it will have to get laid out to still have good grounding. So, always avoid shared grounds. A via is about 0.7nH and a lot of impedance can show up at high enough  frequencies.  Where possible always put a via in for every ground.  For this reason I always use big 15 or 20 mil stitching vias in the floods then a bunch of 8-10mil vias for local  grounding and ground slugs (eg, nRF). You can't turn the board into swiss cheese but you need grounds that won't talk to each other.

B. You will never be able to match that antenna to the nRF with 2 components.  A quarter wave monopole has an ideal impedance of 36.5+j21.25. The two components on there are just the  ones Nordic specs to get the nRF to 50ohms, after that you need to match the antenna to 50ohms.  Even Nordic uses about 5 elements to get a 1/4 wave to match to the nRF. You could try to  do a complex impedance match between the two, but without a VNA you won't succeed. You might consider a more forgiving antenna like an F-meander hybrid. TI wrote a nice whitepaper on  one.  It has even been referenced in the devzone.

C. You could save some space by using an SMT header for the SWD. I think the standard jtag/segger ribbon cable is 50mil pitch and mates well with an SMT header. You can get the correct  orientation of the pins from the segger site. I did similar on a programming jig for one of my companies products.

D. Ah, power planes! I don't like power planes.  Over 20 years, I have never had a problem on boards with routed power and one common ground, yet I have been called in many times to  debug boards with power planes and split ground planes. The power plane results in a lot of unexpected coupling between power and signal routing. It is just one big capacitor sitting  underneath all that stuff on the bottom.

E. Your microstrips on the nRF output aren't characterized.  Not really important since you have to match it anyway. But matching is more deterministic when the strips are 50 ohms. I  have a standard 0.062" 4 layer stackup I use with FR4 that is very reliable.  Also too, you could easily do this as a 2 layer design and save money.  This could be useful since it is an  open project for the community.

F. Also, your RF grounding is pretty bad. But that's all part of (A).

[rogerarchibald]

Thanks for taking the time to write such a detailed response.  

I'm going to ditch the level shifters on the ultrasonic transmitter on the next spin and drive it off of VCC via some buffers (It's a highly capacitive piezo transducer and I'm weary to drive it direct from the NRF GPIO).  I'll have to increase the gain on my ultrasonic receiver due to the lower drive voltage, but at least it will be a constant voltage and not varying with Vbatt as you say.

On my previous board I had a much higher battery voltage and I used gate drivers in the places where there are currently level shifters: the gate drivers' enable line was tied to VCC, so when my LDO was off they drew virtually no current...On that note, Q8 serves to disconnect the resistor divider for battery voltage to further reduce current consumption, and on the previous board, when VCC is disabled and Q8 was off, I was down to a few hundred nA consumption. Next time around I'm not going to be so miserly about a few hundred uA here or there and just have a power switch that turns the whole thing on.

I didn't mean to give the impression that the boost is going away: the boost is a separate circuit from the ultrasonic and it actually develops the voltage required to fire the solenoid when the user actuates the trap.  Q6 is a pretty beefy part with a decent amount of gate capacitance and I was concerned about driving it directly from the NRF, but again the level shifter probably isn't the best device to use here...That drive was so much easier with a >6V Vbatt and a proper gate driver!

I've used quite a few buck converters in hobby designs before, but usually constant current devices for hi-power LED's.  I understand that LDO's are somewhat wasteful, but in a device with a power requirement of ~100mW I'm generally not too concerned about it....In my day-job of working with downhole oilfield electronics where power budgets can get tight, switchers are a necessity, so sometimes I like to play with different toys in my hobby electronics.  I'll look into a buck for the next spin.  I'm a fan of Diodes Inc. switchers: they seem to work as well as TI parts but they're substantially cheaper for an equivalent topology.   

D7 is a power LED as you state.  I found that package on Luxeon's website.  I was weary of the concentration of vias but I've had that part made at two different board houses so far (I use it on a few other projects as well) and nobody's complained.

Thank you again for your input!  I've shared the link for cattail with my neighbour, whose cat has set off my OpenTrap alarm several times in the course of my back yard testing!  I'm making some progress with porting my firmware over to SES and SDK14.2, and once that's all solid and I understand all the new changes I'm going to resin this thing.

Cheers,

Roger

[AmbystomaLabs]

Yes the SN74 makes a good buffer.  Prob not necessary to use its inverting cousin in the design, you can easily use two gpio out of phase of each other to make your bi-phase arrangement.

Though I understand the prudence in using the SN74, a MuRata spec on a 40kHz transducer shows it as 2,500pF. Assuming 5mA gpio you can easily switch this thing rail to rail at 200kHz. For more margin, you could lash two gpio onto each phase for 10mA total sink/source capability. GPIO state changes for all gpio on the same clock edge, so they will always stay in phase as long as you write the register correctly. But as you stated, more assurance with a buffer.

On the boost then, assuming you have something like a 1amp 12vdc solenoid on the door latch, then I would still use a packaged converter.  Just use the chip enable on the converter to turn it on and off. Then you can lose Q1, D1, Q6 and the sn74 is just replaced with the boost chip.  There are many out there, but here is an example of one from Diodes that give 12vdc at up to 5amps from a 2.5-5.5vdc source. Assuming 1amp out and 3vdc in that means switch current of around 4amps. You will need to buy a good inductor.  But there are plenty of smt devices that will handle that kind of current. Something with a saturation current of around 8amps or so should do it. Boost chip example: www.diodes.com/.../PAM2421_22_23.pdf

I agree with you LDO's are easy, but on something with a LiPo you are going to want a complete solution with charge, protection, buck switcher.  Then you end up with a user friendly device that you plug into a wall wart once in a while to charge it up.

You didn't specify what frequency the ultrasonic is running at, buy you may want higher than 40kHz.  The web  says cats can hear to 64kHz. I know our cats get freaked out by high frequency noises. Though it is possible a feral cat might wonder what the noisy cage is all about and be inquisitive.

[AmbystomaLabs]

I made a mistake.  You will still need your turn off fet for the solenoid due to the boost topology.

Sorry, I missed that. 

Though, I would still use a packaged device instead of the way you have it in the schematic.

[rogerarchibald]

The solenoid current is pretty substantial (~2A @12V) and the prepackaged boost drivers aren't an option here.  The 5A rating is for the internal FET, and if I'm looking at say 3.5V Vbatt boosting up to 12V @ 2A, the internal FET is going to get fried.  Those solutions are great for a regulated supply at a moderate load, but my discreet boost is sacrificing simplicity, size and regulation for pure beefiness.  I'm planning to run this off of a 26650 battery which will supply more than enough current, provided I have a decent efficiency (and make sure that the power LED is turned off as part of the solenoid fire function!).  One thing that my current rev doesn't have, but is needed, is a 'solenoid sense' circuit so I don't fire up an unregulated boost designed to provide that kind of power when there's no load.  I have a simple circuit in mind that I think will work nicely. 

I've used little battery managers (Microchip 73831) with smaller lipos.   I'm not sure if I'll go that route here...I haven't looked into compatibility with the big batteries and had just assumed on using an external charger, but an on-board charger with a micro USB connector might be a good option here.  That is what I used for the battery on the remote of the first revision of OpenTrap.

Again, I'm sure you're a busy man and I appreciate you taking the time to provide feedback on my design.  If by any slim chance you're planning to be at the SF Maker Faire this weekend, I would like to offer to buy you a frosty beverage!

Cheers,

Roger

[AmbystomaLabs]

Sorry, my company HQ is out of Florida and currently I'm working out of Massachusetts.  I get out to SF once in a while, sometimes for Arm Tech Con or ESC Silicon Valley.

If I can work it into my schedule, I will spin a design for your live trap.  The ultrasonic stuff looks interesting and I'm thinking about doing a PCB artwork based output match for the RF and antenna that Nordic might like.

I'll pass along the design info if it happens.