MySensors battey board revision 1.0

[Anticimex]

@axillent Looks good. But is not the LED symbol reversed in the schematics?
I am considering making a board as well for personal use (open HW of course), but using through hole components a bit more extensively for ease of customization. And probably a sockeded MCU so I don't have to include programming interfaces. In the future, I (and hopefully all of you) will push new FW to the board OTA so there will not be a need for holes in the case :)
But I might be interested to join in if your are ordering a batch of these as well. Will it be panelized and pick/placed as well?

[axillent]

@Anticimex thank you for your attention ) yes, led had a wrong orientation, it was fixed.
I have an DIY experience for making PCB and like you at the beginning willing to use through hole components mostly.
But through the real practice I came to a conclusion that SMD is easy to use, it is cheaper and need much less space.
I also used to make zero-hole PCB using SMD only. This type of PCB requires from me 20-30 minutes of time only for the full production circle. There is 0.5mm copper plates, I use office scissors to cut it in a second.

The programming interface (ISP) is not only for the bootloader programming. It can be used for bootloader upgrade or hardware debugging.
Even Nano have it.
The programming interface (FTDI) can be used also for usual Serial.print debugging.
Both interfaces also can be used for additional I/O.

Sure, there is no absolute ideal solution. We spend a few months discussing what is a good trade off for this board and you see the result.

What do you mean by "panelized and pick/placed as well"?

[Anticimex]

@axillent By panelization that I mean that you send the board to the fab with a BOM, and some reels, and have them solder the components to the board as well. Most if not all fabs will then require you to panelize the board if also picking&placing (that is, multiply the board on a bigger standard-size board, that is fed to the pick&place machine).
Example of a PCB panel
Having a panelized board could also reduce cost for a non pick&placed design, as it allows the manufacturer to make them with less effort (less cutting and less waste).

For larger batches this is a good way to save a lot of work, and if good decisions are made on layout, design and component selection, it is not overly expensive either.
I also like SMD, but for "my" board, I was more thinking of having a generic back-bone which can accept a variety of components, so the user can choose what to put on it and use if for various purposes.
Some "standard" interfaces need to exist of course (socket for MCU, power, RF), but the rest (decoupling, IO, etc) is just left "empty" and for the user to populate with the components they would like.

And don't get me wrong, ISP/FTDI is really good to have, especially on "in progress" projects. But for the gritty end product, I personally prefer to program and debug the MCU on a breadboard, and then put it in a socket in my optimized board/casing. I even consider using the Arduino "as is".
But it could be that I cannot get my casing to be as small as I want it, without soldering the "bare" MCU directly on the board, and then of course programming interfaces will be needed :)

[marceltrapman]

@axillent said:

There is 0.5mm copper plates, I use office scissors to cut it in a second.

Do you use this thickness for all boards?
And what thickness is the copper?

[bjornhallberg]

Well, needless to say it sounds interesting, when it's been built and tested I'd surely buy a couple of boards. Right now I'm not sure how many actual battery sensors I need and if the sensors I want to run are even viable for battery operation anyway.

I think that for me anyway, more practical things are an issue, like avoiding Swedish customs charges, and finding a GOOD open source automation software so that I can actually use my sensors and not just build them ;-)

If you could also manufacture a separate 3.3V boost circuit based on a modern chip that would be fantastic. I think that is turning out to be a hurdle for most of us that aren't comfortable with SMD/SMT and/or how to nice integrate it into DPI prototyping.

[axillent]

@marceltrapman

Do you use this thickness for all boards?
And what thickness is the copper?

No, it is only for hole-less fast-processed boards like this:

It was a weekend one day project to automate lighting of my DIY desk

[axillent]

@bjornhallberg I'm an experienced z-wave user and 1/3 devices I currently have managed by my smart-home are battery based

battery based is most easiest way to launch smart-home without big disaster to the home
I also expect an interest to use this battery as outdoor sensors, while z-wave do not have any alternative to this
with double power outdoor sensor can lasts from a single battery for year or even years depending on solar power.

for open source software I personally looking forward to have an integration between mysensors and openhub

A separate boost development was finished today and I'm expecting quotes from our partner.
Any preliminary reservations are welcome.
It will be board 15 x 14.8 mm with a switch between 3.3V (up to 100mA) or 5V output (up to 70mA)
it can be sourced by 1-3 cells alkaline or 1-4 cell ni-mh or 1 cell lion

Henrik are build already a boards documentations section on the site :)
We expect to have a good documentation to the boards

[Zeph]

@axillent said:

radio you can connect to lion using 3.3V LDO or 1-2 1n4148 connected simultaneously to drop voltage from 4.2 (fresh lion) bellow 3.6V (maximum allowed for radio)

It's probably moot, but the diodes may drop substantially less voltage when the current is very low. Of course, there may be no harm done with such low currents - raising the current brings the Vfwd up too. I'm just noting that the "almost constant Vfwd" only applies to larger currents (eg: 10ma not 10 uA).

[Zeph]

@Anticimex said:

In the future, I (and hopefully all of you) will push new FW to the board OTA so there will not be a need for holes in the case :)

For that, it's handy to have some extra flash on board. (Ref Moteino and anarduino)

[hek]

@Zeph

Yep, extra flash for OTA would be nice! Great idea for the next revision.

[marceltrapman]

@hek @Zeph:

Yep, extra flash for OTA would be nice! Great idea for the next revision.

Would we need extra hardware for this?
If so what is your suggestion?

[Yveaux]

@marceltrapman said:

Would we need extra hardware for this?

The flash downloader + flash code could be stored as a bootloader, but that would drop Arduino bootloader compatibility and it will be a challenge to get a stripped MySensors implementation in the bootloader area.

Maybe we can add a small i2c or spi eeprom which is used for temporary storage of the new firmware, but then again, how to get the image data from the eeprom into the ATMega's flash?

[hek]

@ToSa is working on it.
https://github.com/ToSa27/Bootloader

Reports of successful OTA flash yesterday. But work still remain to make it safer. An extra eeprom would probably help safing things up a bit.

[Yveaux]

@hek that's looking very promising :+1: Thanks for the tip!
More toys to play with ;-)

[marceltrapman]

@hek said:

Reports of successful OTA flash yesterday.

Nice :)

But work still remain to make it safer. An extra eeprom would probably help safing things up a bit.

I have started my own board design so it would be very interesting for me to know where this is heading at. Any idea what would be preferred to use?

[Anticimex]

In my opinion, keeping Arduino bootloader compatibility (or rather, supporting the Arduino IDE flash protocol) is a must for a OTA bootloader. Even if it would mean some more flash need to be reserved for the bootloader.
But that of course makes it even more interesting to have some external memory support. I2C based EEPROMS have been around for quite some time :)

[Zeph]

I mentioned the Moteino, which has an optional 8 pin 4 Mbit (512KByte) SPI flash chip.

Felix has a modified Uno class bootloader, the Dual Optiboot https://github.com/LowPowerLab/DualOptiboot which takes up 1KByte (vs 512 Bytes for normal OptiBoot).

As I understand it, you send the new code OTA where it's written to the external SPI Flash memory. Then when booting, Dual Optiboot looks for a signature in the Flash memory, and if it found, burns the code into application Flash on the ATMega chip (and it obviously removes that signature). If not signature found in SPI flash, it will boot normally as a normal OptiBoot system.

The Moteino uses the RF12B or RF69 sub-GHz radios. Another cool design. I however like the 2.4GHz nRF24L01+ because it's cheap and very fast, useful in controlling Christmas lights, which take much more bandwidth. I see periodically reporting MySensor style data as another option using the same nRF24L01+ hardware, either instead of the light control function, or in addition to it.

[axillent]

OTA is a very promising thing
for the battery board selecting any external chip is also a question of power consumed

the idea with flash is good because it builds the ability of transactional while you able to safely recover from any situation.
but from other hand. you can safely recover even without flash if your second party will take care
look how upload from USB is organized. It do not have a transactional mechanism, it cannot recover by itself
if upload failed the only way to recover is to upload again
why we need OTA to be more safe than regular USB update?

but while you a thinking on this i found an issue in the board schematics. I have to add one diod and one 10k resistor to isolate D0 of the MCU from the parasite power coming from the external FTDI programmer. Otherwise the parasite power can destroy nordic chip

[axillent]

I have news

prototype of the board fully tested except radio
stepup is working fine, power switching is working
programming using external FTDI is working, schematics corrected to get rid from parasite power
al I/O is fine
battery and solar voltage measurement is working
temperature sensor is excellent
it was not simple to solder, the package is unbelievably small - 1.7 x 1.2 mm, 6 pins

next and last is radio, going to solder it today

[Zeph]

Interesting that you are choosing GROVE connectors. Are they easy to find (other than SEEEDStudio)?

This makes for an almost drop in alternative to the DevDuino's :http://www.seeedstudio.com/wiki/DevDuino_sensor_node