Minimal design thoughts

[tbowmo]

@NotYetRated said:

Woah, wait.... 2 years of power?? Okay, if I could even get, say 18 months, I would go with battery powered over wall worts in a majority of my sensors.

I was assuming battery power may last 3-6 months or so, given the motion detection.

With motion detection you probably are further down, also one problem is that the PIR sensors I have seen is only 3.3V, which means you need a step up converter to power it from battery. This also increases the powerdrain on the batteries.

[NotYetRated]

Ahh okay. Ill stick with building my stuff out for wall power where possible, and probably test a few of your boards when available.

[axillent]

@tbowmo Cool staff!

Just one suggestion - to add so called FTDI programming interface like pro mini is using
It allows to use an external USB<->UART converter to work ith the board as usual (using programming through bootloader and Serial debugger)
It is simple, see PROGRAMMER connector for example here http://wiki.devicter.ru/images/f/f7/SNv3sch.png

[tbowmo]

@axillent
Hmm have been thinking about it.. Just for the versatility of the unit for other people. I have plans to use good old atmel programmers (I have a bunch of them, both jtagice3, and the older serial/usb programmers), so it's not that big an issue for me to skip serial for programming

And at the moment I can't find any room for it the serial (other than rxd/txd in upper right corner) It's almost become an obsession for me, to keep it at the small formfactor (Same size as a standard nrf24 module). Just to prove that it is possible to get it that small

btw. squeezed a couple of testpads in for A0, A1 and D3 last night.

[tbowmo]

An update.. I got room for the ftdi header, as well as a 4pin header where I got power, and 3 digital IO (D3, D6 and D7, if I my memory serves me). Just need to find location for the last couple of tracks

Only detail though, is that I have gone with 0402 size smd components..

/ Thomas

[hek]

[tbowmo]

And the latest schematic, and layout (After I scratched the old routing, and started all over, again, for the 5th time I think :))

Actually I enjoy this fiddling with placing components, and routing tracks.. Only problem is that the wife thinks it takes too much time, and I should do other more usefull stuff, than playing with my computer, making noncense drawings

Schematics

full layout

top layer

bottom layer

Think I am ready to press the GO button for first spin in china.. This is gonna be exciting

[hek]

Haha!.. I have the same problem with the wife over here. The modules and half build sensors I spread all over the house is most irritating.

[tbowmo]

And off they go to China.. Have ordered the first batch of prototypes.. (10 pcb's, panelized with 2 sensor boards per panel)

Now there is only the waiting time...

[NotYetRated]

@tbowmo Amaons 2-day shipping has killed me. When I started with MySensors projects, and the China waiting time of sometimes 2-3 weeks.... Difficult haha.

[tbowmo]

Here it's more like 3-6 weeks waiting time, for stuff to come back from China.. So maybe it arrives in time for Christmas.. And then I can't get components sourced from mouser at that time (waiting until I have seen the quality of the boards, before I order components)

[pgo]

@hek, @tbowmo and rest: ho ho guys, first think WAF, ie Wife Acceptance Factor

This variable has to be VERY high. Like: Darling, this whole fidling of mine would optimize our home and save on energy bills big time, so You can spend more on shopping craze. etc

Probably when I finish with testing, I start selling it to friends and we will have problem spending all of the money which suddenly comes our way. (Try here not to laugh )

[tbowmo]

The reason why I am building this, is that I always have been currious about collecting data at regular intervals, back in the days (8-9 years ago) I had a setup with 5 DS18B20's on long wires arround the house. One was in my furnace measuring the temperature of the water, one was outside, etc. Everything connected to my old 486 based linux machine with 20Mb ram, collecting data from all the sensors on a regular interval, and presenting it with digitemp (a php package for displaying temperatures).

Just want to make something like that again. And I think my wife can see the ideas in it. Will just use it for environmental measurements throughout the house. Maybe some of them coupled to a SMS gateway, letting the teenagers know when to open up a window for 5 minutes, in order to get fresh air in, and the humidity level down. (And another SMS to remind them to shut the windows again).

And then the fun in collecting data, analyzing it in graphs etc...

[axillent]

@tbowmo I also have some old programmers including jtagice3 and can well understand you)
and sure, your very small design is best to keep

The only problem I have to use jtagice3 (which I like a lot to do hardware debugging) is that Arduino is not supported by hardware debugging

Just an idea - if you have all pins needed for FTDI programming on the corner later you can have a special connector between your board and FTDI programmer. This case you can keep design without extra space needed

[tbowmo]

@axillent said:

The only problem I have to use jtagice3 (which I like a lot to do hardware debugging) is that Arduino is not supported by hardware debugging

Wouldn't using jtagice3 for debugging also collide with the radio on a mysensor module?

Just an idea - if you have all pins needed for FTDI programming on the corner later you can have a special connector between your board and FTDI programmer. This case you can keep design without extra space needed

There is a full ftdi header on my board now. And I just got information from China that my boards are shipped... so just waiting for them to arrive now

[axillent]

@tbowmo said:

Wouldn't using jtagice3 for debugging also collide with the radio on a mysensor module?
jtagice3 is using only single pin of atmega328p (RESET pin) to do programming and debugging.

There is no reason to have conflict with radio.
The only problem with jtagice3 is that arduino environment is not supporting hardware debugging.
I'm wondering on having ability to compile Mysensors using regular Atmel Astudio to be able to use debugging.

There is a full ftdi header on my board now. And I just got information from China that my boards are shipped... so just waiting for them to arrive now

Cool! I also like 0603 footprints. But also is fine to use 0402. Sometimes you will probably need it.
If you need support I can prepare a bootloader for you with needed settings for boards.txt

[tbowmo]

@axillent said:

@tbowmo said:

Wouldn't using jtagice3 for debugging also collide with the radio on a mysensor module?

jtagice3 is using only single pin of atmega328p (RESET pin) to do programming and debugging.

Aaahh, yeah, I forgot about their "new" debugWire interface, so thought that a fullblown spi interface was nescessary for the ISP to program the device. Hmm.. With THAT info, I could have used a smaller pinheader, instead of the 6pin ISP header..

Could become handy for next revision of the board (if there'll be anny)

There is no reason to have conflict with radio.
The only problem with jtagice3 is that arduino environment is not supporting hardware debugging.
I'm wondering on having ability to compile Mysensors using regular Atmel Astudio to be able to use debugging.

There is a full ftdi header on my board now. And I just got information from China that my boards are shipped... so just waiting for them to arrive now

Cool! I also like 0603 footprints. But also is fine to use 0402. Sometimes you will probably need it.
If you need support I can prepare a bootloader for you with needed settings for boards.txt

I'm used to the small components.. And for a couple of handbuild sensor boards, I thought that I might as well cope with them, in order to keep the layout small If this is becomming a success, and others would like similair boards, the boards should be assembled with an pick & place machine, and then it doesn't matter that the components are 0402.

I'll keep it in mind with the bootloader

[axillent]

@tbowmo said:
With THAT info, I could have used a smaller pinheader, instead of the 6pin ISP header..

you cannot remove ISP header because debugWire is activated through ISP )))

@tbowmo said:

I'm used to the small components.. And for a couple of handbuild sensor boards, I thought that I might as well cope with them, in order to keep the layout small If this is becomming a success, and others would like similair boards, the boards should be assembled with an pick & place machine, and then it doesn't matter that the components are 0402.

Me and Henrik we have a list of partners for PCB production, If you fine to cooperate more closer with Mysensors we can help you on putting the board unto production. Just let Henrik to know if you have an interest on closer cooperation. I will help with production.

[tbowmo]

@axillent said:

Me and Henrik we have a list of partners for PCB production, If you fine to cooperate more closer with Mysensors we can help you on putting the board unto production. Just let Henrik to know if you have an interest on closer cooperation. I will help with production.

I know, I have been talking with Henrik about it. Waiting for the first prototypes, so the design can be validated.

[blacey]

@tbowmo Nice design exciting to see it come together so well. Good luck with the first board bring-up and I look forward to hearing/seeing the results as the unfold.

[tbowmo]

I have been looking at prices (Using mouser, as they appear to have local european storage, so I can avoid import taxes in Denmark).

for a temperature / humidity sensor pcb, fully mounted, the price comes up to arround 13$ per unit. The most expensive part is the Si7021, which accounts for 4.5$ per unit (even in larger quantities).

for BMP180 on board, it's another 4.5$ added.

The above prices are material only...

[bjornhallberg]

@tbowmo Too bad the Si7021 is still too new to show up on AliExpress. The BMP180 is like $1.5 ... ironically the complete soldered module is also $1.5 ... Chinese economy ...

Component sourcing is the killer of fun things.

Hmmm. Am I correct to assume then that Mouser has free shipping over €65, and that they add country taxes on top? When I check out I seem to get a 25% VAT charge for instance. Ugh. Still better than paying an extra €10+ to the Swedish Post or UPS or whoever.

[tbowmo]

@bjornhallberg

Hmm.. I don't trust aliexpress that much, for example is the BMP180 you get there a counterfeit?

Take the bmp180, I only need 1 of those, for the whole installation, so I don't mind paying 4-5$ more for a single unit. However, for the rest of the units, all needs the Si7021 (which is pin compatible with HTU21x device, if I remember correctly? However, the specs for the Si7021 is way better)

I don't know about the taxes from mouser, as I haven't actually bought anything there yet, only set up my project, and added the components needed, so I could get a price indicator on it. But their website says that they pays import duties etc..

[tbowmo]

And the first prototypes have just arrived

Next step is ordering components. so I can get the thing up and running..

It's going to be a fun christmas

[tbowmo]

on a side note.. this IS a prototype.. probably only 1 or 2 will be assembled, as version 2 is almost ready to hit the press..

Next version will have more pins available on pinheaders along each side of the board, and power input to the board is placed a more central place (electrical wise).

So this prototype will be verification of schematics, and check that radio etc. is wired correctly.

[bjornhallberg]

Great progress!

If you still use dirtypcbs (what did you think of their work btw?) and have the ability to pretty much create any sort of routing and tabs, you could look into the upcoming Gerber Panelizer:
http://blog.thisisnotrocketscience.nl/projects/pcb-design-tools/
Might be more cost effective to order 10x10 pcbs and panelize the hell out of that space. You could fit a ton of different layouts. Tried the tool myself and it saves a ton of time. It can load and rotate gerbers any way you want.

Only snag is they haven't released the tool yet. You have to request it from from their twitter account. And it is still beta.

[tbowmo]

@bjornhallberg said:

Great progress!

If you still use dirtypcbs (what did you think of their work btw?) and have the ability to pretty much create any sort of routing and tabs, you could look into the upcoming Gerber Panelizer:
http://blog.thisisnotrocketscience.nl/projects/pcb-design-tools/
Might be more cost effective to order 10x10 pcbs and panelize the hell out of that space. You could fit a ton of different layouts. Tried the tool myself and it saves a ton of time. It can load and rotate gerbers any way you want.

Only snag is they haven't released the tool yet. You have to request it from from their twitter account. And it is still beta.

Seems like dirty pcb is doing a great job, compared to the price, only have experience with prinline.dk from previous jobs, but they are expensive like hell (just looked at their prices earlier today, together with Hek, They want 314$ (plus taxes) for the same job, that I pay 14$ for at dirtypcb, only advantage is that they have a turn arround time of 8 days, instead of 4-5 weeks from dirtypcbs.

It like a nice tool, Have to look at it Only problem is, that I have to generate gerbers myself then, instead of just uploading the eagle file, which dirtypcb turns into gerber files for me. (Remember a project 10 years ago, where it was a lot of work in creating gerbers in eagle).

[bjornhallberg]

@tbowmo said:

Seems like dirty pcb is doing a great job, compared to the price, only have experience with prinline.dk from previous jobs, but they are expensive like hell (just looked at their prices earlier today, together with Hek, They want 314$ (plus taxes) for the same job, that I pay 14$ for at dirtypcb, only advantage is that they have a turn arround time of 8 days, instead of 4-5 weeks from dirtypcbs.

Ouch, that seems a bit steep. Never heard of any cheap board houses outside of China. OSHPark may be the exception? I thinks they have fabrication in the US right? I heard they were planning to set up a factory in Europe ... don't know if anything ever happened.

It like a nice tool, Have to look at it Only problem is, that I have to generate gerbers myself then, instead of just uploading the eagle file, which dirtypcb turns into gerber files for me. (Remember a project 10 years ago, where it was a lot of work in creating gerbers in eagle).

I don't know how CAM exports have changed over the years but compared to how backwards everything else works in Eagle, exporting Gerbers is easy. DirtyPCBs have a CAM-file listed on their about page. I sent them Gerbers of Meanpenguin's PCB designs and that seemed to have worked out. Knock on wood. Hopefully I'll have the boards back before xmas. The reason I didn't send the Eagle file was their new warning on the front page regarding Eagle v7+ files. Some unspecified problem with their Eagle 6.x installation. Only iffy thing I've run into so far (on some exports) is the drill file showing the holes all over the place because of the wrong precision settings. Now I'm sure the board house would have fixed this for me (their Gerber tools are bound to a great deal more advanced than mine and can probably detect Excellon formats automatically), but choosing EXCELLON_24 as output solved that problem. It's enough to send them impossible routing and endless stop mask errors

Oh and btw, that Gerber Panelizer also helps to circumvent the 5x5cm limit on freeware Eagle since you can take your Gerbers and join them any way you please as a greater whole. Just to point out the obvious.

[tbowmo]

@bjornhallberg said:

I don't know how CAM exports have changed over the years but compared to how backwards everything else works in Eagle, exporting Gerbers is easy. DirtyPCBs have a CAM-file listed on their about page. I sent them Gerbers of Meanpenguin's PCB designs and that seemed to have worked out. Knock on wood. Hopefully I'll have the boards back before xmas. The reason I didn't send the Eagle file was their new warning on the front page regarding Eagle v7+ files. Some unspecified problem with their Eagle 6.x installation. Only iffy thing I've run into so far (on some exports) is the drill file showing the holes all over the place because of the wrong precision settings. Now I'm sure the board house would have fixed this for me (their Gerber tools are bound to a great deal more advanced than mine and can probably detect Excellon formats automatically), but choosing EXCELLON_24 as output solved that problem. It's enough to send them impossible routing and endless stop mask errors

Hhm.. don't tell my boards that dirtypcbs have problems with eagle v7 files Haven't seen that warning before now.. So cross my fingers, that everything is ok.

I remember the trouble with excellon format, and getting drill files alligned, but got some files made for printline back then, and they are still making the pcb, so must have done something right

Oh and btw, that Gerber Panelizer also helps to circumvent the 5x5cm limit on freeware Eagle since you can take your Gerbers and join them any way you please as a greater whole. Just to point out the obvious.

Think that the free eagle is 10x8cm board space, isn't it?

[bjornhallberg]

@tbowmo Yes you're right, the Eagle board size is 10x8! Don't know where I got the 5x5 from. Nevertheless, since DirtyPCBs can do pretty mindblowing routing at no extra cost and have a pretty good price on 10x10 also ... you could probably save a great deal by doing what the Not Rocket Science guys did and create a heavily panelized board that covers a handful of different PCBs.

[RJ_Make]

It will be interesting to see if that radio causes any interference for the MCU. Looks like it will be right over the MCU.?

[tbowmo]

@ServiceXp said:

It will be interesting to see if that radio causes any interference for the MCU. Looks like it will be right over the MCU.?

the antenna is not above the mcu, and the nrf module is using the entire backside as gnd. So should be shielded.. Also, I am kind of more worried about mcu generating noise for the radio..

Specially for the first prototype, where I was a bit daft at placing powersupply pins on the wrong side of the pcb, as far away as possible, from the radio. And letting it run under the CPU to pick up noise from that.. Version 2 has this fixed, where I have the power input pins moved closer to the radio, and using wider tracks for VCC arround the board. (Should have done more reviews on the first design before sending it off to dirtypcbs)

[RJ_Make]

@tbowmo Ah.. I didn't even think about the MCU causing interference for the radio. @Bandra's design suffers from this close proximity interference, which causes, I think, the MCU to lockup. I had to move the radio module from over top the Arduino for stability.

[tbowmo]

@ServiceXp

On one of my first mockups, I removed the connector on the radio and used some double sided tape to stick it to the back of an arduino micro, and then soldering wires between the boards. Works like a charm on battery. Antenna is just hanging off the edge of the arduino on the end opposite the serial port pins.

Using it as a range tester for mysensors, logging into a raspberry pi with my mobile using ssh, and then running a serial port terminal emulator on the raspberry to connect to the serial GW. Then I can walk arround with the battery powered node, and check if the GW receives the data packages on my phone

Haven't looked at @Bandra's design, but how close is the nrf antenna to the mcu?

[RJ_Make]

@tbowmo
This close.

[tbowmo]

This is my node, with radio close to the mcu, this works really good. So I would expect that it also works for the small module that I have made.

[tbowmo]

just ordered components for the first 10 boards. Just about 190$. My wife looked funny at me, as she saw the amount (she was sitting next to me when i ordered).

Well birthay present for myself

[bjornhallberg]

@tbowmo That was a tad more expensive than you expected I guess? Did Mouser add taxes etc at checkout?

[tbowmo]

@bjornhallberg

Yeah, they added Danish VAT to the order..

Just double checked, it's actually only 154$ (925dkk).. Remembered wrong with the exchange rate when I converted earlier, and just took some numbers from the top of my head. The VAT is around 31$.

But still it's a lot of money, specially this time of year..

[Ivan Z]

If use one-sided mounting and on other side set CR2450 holder. Will get a small deviceю

[tbowmo]

And now the components arrived :d

Just miss one capacitor, which is out of stock at Mouser. Got a notice that they will get it in stok again in March 2015! So I'll have to see if I can find an alternative (It was decoupling for the radio).

Hopefully I get some time off to do the first board friday evening.

[tbowmo]

So, it's time for board mounting..

One advice, DOUBLE CHECK the component orders, and TRIPLE check them.. After I got the shipment from mouser, I discovered, that I accidentally had ordered some capacitors in 0101 housing, instead of 0204.. that's DAMN small components.. And I can't use it.. Well thats about 10$ down the drain.. (It was all the decoupling capacitors for the board).

Also found out that I forgot to order a ~100R resistor for the status LED's.. What a bummer!

So I'll have to see if I can source the components quickly from some friends in the next couple of days

Anyways, pic's from the build:

first depanelling in the "garage"

(And just for the fun of it a picture from the garage.. one of my other projects)

And then to the electronics workshop, mounted the atmega

And with NRF module attached...

[bjornhallberg]

@tbowmo Sorry to hear about your sourcing troubles. Wish there was an easier and cheaper way to source things.

Smart depaneling, I have a saw like that standing around as well. I have used it to cut some aluminium profiles before but never considered it for cutting pcbs. Still, I will continue to try and look into tabs and "mouse bites" to make depaneling a bit easier.

Is that an old Ferguson that you're renovating? Sorry, but I immediately thought of Frank Erichsen

So, no problems hand soldering then? Can you complete the circuit for testing or do the missing components get in the way?

[tbowmo]

@bjornhallberg

Sorry to hear about your sourcing troubles. Wish there was an easier and cheaper way to source things.

I am trying to activate my local nerd network, to see if anyone could get the components within the next couple of days (we are 10-18 nerds having regular meetups, some of them are in the electronics business). One of the nerds have some resistors and caps in 0402, but not exactly the values I want (1uF cap, and 760R resistor if I remember correct)

Smart depaneling, I have a saw like that standing around as well. I have used it to cut some aluminium profiles before but never considered it for cutting pcbs. Still, I will continue to try and look into tabs and "mouse bites" to make depaneling a bit easier.

for the next batch of boards, I might go for easier depanelizing options, with tabs and so on (I got the gerber panelizing tool from http://blog.thisisnotrocketscience.nl/projects/pcb-design-tools/)

So, no problems hand soldering then? Can you complete the circuit for testing or do the missing components get in the way?

I managed the soldering so far.. I must admit that I had forgotten how small 0402 components are But soldering, and double checking everything with a "monocle" to magnify things, it's doable..

Also I have tried to power up the atmel, it seems to be ok, at least I can get the correct device checksum when I use my jtagice3 on it with avrdude. It would just be more fun to have all the correct components mounted, for testing and verification of the board, before I push the next board revision off to China

Is that an old Ferguson that you're renovating? Sorry, but I immediately thought of Frank Erichsen wink

Yeah it's an old fergie TEA-20 from '54, that I'm trying to renovate Not everything should be computers / electronics (Btw. is Frank, or "bonderøven" as he's called in Denmark, also known in Sweden?)

[bjornhallberg]

@tbowmo said:

Yeah it's an old fergie TEA-20 from '54, that I'm trying to renovate Not everything should be computers / electronics (Btw. is Frank, or "bonderøven" as he's called in Denmark, also known in Sweden?)

Yep, bonderøven has been on TV over here for a couple of years now. I think I've seen most episodes. I just love that guy's enthusiasm ... and patience.

[tbowmo]

I succeeded in getting a couple of components more (Only missing a 2.2uF 0603 capacitor, that was out of stock at mouser.)

I have vacation, so in theory I could go to the "man cave" and mount the last couple of components. Only problem is, that I have 2 "tasmanian devils" aged 3 and 5 years running wild at home today, So no possibility to have time for my self

[tbowmo]

A quick christmas status update, finaly got arround to power the first board up, just with basic code (blink an LED to check that it will work).

Used a couple of hours trying to get my LED to blink. all other pins would toggle, but not my LED, until I finaly digged into the datasheet. I had used A6 for the LED, which turns out to be an input pin only, facepalm. So the board already have the first "green wire" patch

Next up is to get serial up and running. before i can start testing temperature / humidity sensor, and radio interface.

[RJ_Make]

@tbowmo Thanks for the update!!

[tbowmo]

And today, radio connection verified, and Si7021 communication is up and running..

Only problem though, is that the Si7021 reports a temperature of 45-50 degrees celcius, where I know it should be closer to 22 degrees..

Also current consumption is rather high at the moment, arround 40mA, It feels like the SI7021 is getting a little bit warmer, than the rest of the components, which (to me) indicates that there is something wrong here.. I accidentially supplied the board with 5V for at test of the internal RC oscilator stability, which is way above the maximum supply voltage of 3.6V. So that might have killed it.

[RJ_Make]

@tbowmo Did you get to do anymore testing?

[tbowmo]

@ServiceXp

In the "lab" at the moment, testing the first prototype build.. I got the humidity / temperature "bug" sorted out. Replaced the Si7021 chip on my board, and the new one responded as expected, at least with correct temperature. (humidity is a bit off, I think.. I don't have another humidity sensor at hand here in the lab, so can't compare it..)

I also got the radio up. I haven't done any distance measurements yet, but I can reach the gw in the main building (6-7m away throug a couple of brick walls). Added a 2.2uF cap to the radio. It's still running off the bench power supply. I don't see any problems at the atmel side of things when radio is transmitting.

I just made a test, supplying the setup with 1.9V, and the GW still received the temperature updates (somehow my humidity readings don't get through to pidome on the GW, so have to check my setsup again).

I'm waiting for a peer review of the pcb layout, before I push the button for the next batch of proto pcb's. btw. on the next prototype build, I have removed the bmp180, but exposed the SDA/SCL pins on a connector, so it can be attached externally..

And here is a small preview of the upcomming board prototype.. I had to make the board 1.5mm wider, to make it fit a multiple of 0.1mil between the pinheaders on each side of the pcb. So it will be more versatile to other people as well.. (if any one should want to build them)

[tbowmo]

small update again

Did a small rangetest, and it seems that the range is equal to my other sensor (based on arduino mini and batteries). Also moved the node to the main builidng, besides the weatherstation that I got for christmas (which have hygrometer buildin). It seems that the humidity level on both my sensor node, and the weatherstation is almost comparable.

[RJ_Make]

@tbowmo That is great news! Did you have any thoughts/plans on packaging these up for DIY kit sales? Did you have a chance to test the EEPROM?

Thanks for the Update!

[tbowmo]

@ServiceXp

Haven't tested the eeprom yet, as I don't have any use for it right now. And also need to verify the 32khz crystal oscillator (currently running on the internal rc oscillator). Come to think about it, I need to think up some scenarios where the crystal could come in handy.

Investigations are undergoing, to make something for the "public". Can't say anymore at the moment

[RJ_Make]

@tbowmo Gotcha, thanks again for all your hard work, for those of us who simply don't have the skill set to create a good pcb, I really appreciate your work.

I'm looking forward to your other MySensors pcb projects...

[tbowmo]

And sensor #2 is build.. This time the radio module is soldered directly to the sensor module, making it a realy small device now..

[tbowmo]

hmm.. Now I've had the two sensors on the table beside each other (sensor #1 and #2).

Temperature is off by 1.6 degrees celcius, and humidity 2% difference. Could be it's the different temperature, that makes the RH calculation differ a bit..

And for some build data.. It took me a good hour to populate all the parts. This one also have the additional eeprom mounted. but still need something for verifying that it works..

[RJ_Make]

@tbowmo what is the margin of error for the sensors?

[RJ_Make]

@tbowmo Oh and is that an hour using hand soldering?

[tbowmo]

@ServiceXp

according to the datasheet the precision on temperature is +/- 0.4 degrees celcius, and +/- 3% RH.

They did however release a new chip revision (Si7021-A20) just after I bought mine (which is A10), they had some problems where the chip didn't enter standby correctly. Maybe that is the problem (also battery drain has been quicker than what I expected on the new node)

and yes, it's about an hour handsoldering things, and also programming the sensor, which also caused some troubles. Had to reprogram a lot of times, because of checksum errors. This could be caused by the fact that I have both eeprom, and nrf24l chip hanging on the SPI bus on the atmega, this is the same bus that is used by my ISP programmer (jtag ice3).

Think I have to get a proper bootloader into the device, so I can use the serialport to download instead.

[tbowmo]

and now the external SPI flash is verified as working, just with a simple program, reading out the manufacturer ID, and writing / reading data to/from the device.

Next step I wanted to get a bootloader working, so I could use standard arduino to download firmware, and also (at some point in time) utilize the flash for OTA updates. However, I can't seem to get bootloader support working..

I have tried to get DualOptiboot from lowpowerlabs working now, with several rebuilds etc., it seems like it is running fine (LED is blinking a couple of times, when it boots) but arduino wont sync to the bootloader, so either it runs on a wrong baudrate, or something else is wrong in the setup. But so far I haven't suceeded in figuring out what it is..

(Anyone else that has some knowledge on bootloaders, that could have an idea to what to look at?)

/ Thomas

[tbowmo]

I am almost ready for a second prototype spin of the PCB, but before doing that I could use some input from the community,

If this board was going into "mass production", is there any features that I should consider adding? Any missing parts? Anything that you think that I have forgotten?

The schematics for revision 2 is a couple of posts back in the thread.

I have considered adding a ATSHA204 for future security purposes, but can't seem to find a suitable spot for it (maybe I'm just too tired to see things clearly at the moment :)).

[boozz]

@tbowmo
Does this link help you? Some explanation on bootloaders and practical stuff to help you burn botloaders on the ao Atmega326...

BR
Boozz

[tbowmo]

@boozz said:

@tbowmo
Does this link help you? Some explanation on bootloaders and practical stuff to help you burn botloaders on the ao Atmega326...

BR
Boozz

I can program a standard arduino bootloader, which works. I just want to use the one from lowpowerlabs, dual optiboot, so i can utilise the external Flash chip. But i Can't get it to work at all.

[boozz]

@tbowmo
Ok, I misinterpreted your question. Can't help you on that one, sorry

[RJ_Make]

@tbowmo said:

I am almost ready for a second prototype spin of the PCB, but before doing that I could use some input from the community,

If this board was going into "mass production", is there any features that I should consider adding? Any missing parts? Anything that you think that I have forgotten?

The schematics for revision 2 is a couple of posts back in the thread.

I have considered adding a ATSHA204 for future security purposes, but can't seem to find a suitable spot for it (maybe I'm just too tired to see things clearly at the moment :)).

I would work on trying to get the ATSHA204 included, even if it means the dims get a little larger. The one thing that I really don't like about MySensors is the lack of any wireless security.. Would love to see this implemented.

[rdevine]

@tbowmo Maybe i missed it but do you have an image with the si7021 attached to the radio and pcb?

[tbowmo]

@rdevine said:

@tbowmo Maybe i missed it but do you have an image with the si7021 attached to the radio and pcb?

Look in post #93 the small white dot on the pcb is the si7021

[tbowmo]

@ServiceXp said:

I would work on trying to get the ATSHA204 included, even if it means the dims get a little larger. The one thing that I really don't like about MySensors is the lack of any wireless security.. Would love to see this implemented.

For one second, I thought you where danish :), as the word "dims" usually refers to a small gadget like device, that you don't have a name for at the moment (or even components, screws, bolts, anything small could be a "dims")

[Tibus]

@tbowmo said:

ATSHA204

This module is verry interesting but how it work? What you have to do in code to make it work, crypting messages,...?

[Dirk_H]

@tbowmo I don't see a Load Capacitor on the crystal. Especially if you need a precise clock you should use some. Have a look at https://www.adafruit.com/blog/2012/01/24/choosing-the-right-crystal-and-caps-for-your-design/ for example about Load Capacitors.

[tbowmo]

@Dirk_H said:

@tbowmo I don't see a Load Capacitor on the crystal. Especially if you need a precise clock you should use some. Have a look at https://www.adafruit.com/blog/2012/01/24/choosing-the-right-crystal-and-caps-for-your-design/ for example about Load Capacitors.

I know that normally the crystal require load capacitors, but for the low freq. oscilator (32Khz) its not necessary (as far as I have read), that's why I omitted them from the design. I'll try and dig out the datasheets when I'm at home and double check things.

[tbowmo]

@Tibus said:

@tbowmo said:

ATSHA204

This module is verry interesting but how it work? What you have to do in code to make it work, crypting messages,...?

There are other threads about the ATSHA204 around in the forum. It's basically just a method for signing messages between units, not encryption. I haven't made any code for it myself, just thinking about implementing it for future plans (and for others to pick up on)

[RJ_Make]

@tbowmo said:

@Tibus said:

@tbowmo said:

ATSHA204

This module is verry interesting but how it work? What you have to do in code to make it work, crypting messages,...?

There are other threads about the ATSHA204 around in the forum. It's basically just a method for signing messages between units, not encryption. I haven't made any code for it myself, just thinking about implementing it for future plans (and for others to pick up on)

Ah but from what I've read it can indeed encrypt the data stream. Did I completely get it wrong?

[tbowmo]

@ServiceXp said:

@tbowmo said:

@Tibus said:

@tbowmo said:

ATSHA204

This module is verry interesting but how it work? What you have to do in code to make it work, crypting messages,...?

There are other threads about the ATSHA204 around in the forum. It's basically just a method for signing messages between units, not encryption. I haven't made any code for it myself, just thinking about implementing it for future plans (and for others to pick up on)

Ah but from what I've read it can indeed encrypt the data stream. Did I completely get it wrong?

It's a crypto authentication device, datasheet is at
http://www.atmel.com/devices/ATSHA204A.aspx

For sensor data it should be enough, it help to prevent replay attacks, so an intruder can't switch of the lights for example, because the attacker don't have the presharred keys available

[daulagari]

It's a crypto authentication device

As discussed in the security thread I think authentication alone can be useful but encryption is providing both security and a reasonable form of authentication so why not go for an encryption IC instead of the ATSHA204A.

Atmel has a nice IC doing AES:

http://www.atmel.com/devices/ATAES132.aspx

A bit larger (8 pins SOIC), but reasonable cheap € 0,782 (@10)

It is also part of CryptoCape, so there should be some software support for it.

[tbowmo]

@daulagari

That chip is still only for authentication, only it's with aes-128 message auth instead of sha 256

Also another thing i noted, is that the supported supply range is from 2.5 to 5v, whereas the atsha204 could operate down to 2.0v.

[daulagari]

That chip is still only for authentication, only it's with aes-128 message auth instead of sha 256

If I have a look at the datasheet I see:

Opcode Name        Description
0x06   Encrypt     Encrypts 16 or 32 bytes of plaintext data provided by the Host.
0x07   Decrypt     Decrypts 16 or 32 bytes of data provided by the Host after verifying the integrity MAC.

So it does encrypt/decrypt.

The CryptoCape board has both ATSHA204A and ATAES132 where the ATAES132 can "Encrypt up to a 32 byte packet at a time with AES-128-CCM"

Not sure if it is handy to have the integrity MAC combined as that does require 16 additional bytes, but that is the same story for SHA-256.

[tbowmo]

Hmm I was tricked by the "crypto-authentication" note on atmel site, so thought it only did auth.

Still, lowest vcc is 2.5v. If they could make a device that could operate down to 1.8v instead instead, it would be more interesting.

[RJ_Make]

@tbowmo Yeah the 2.5v would be a problem, unless we can build an extremely efficient boost...

[tbowmo]

@ServiceXp

Yes, but I am trying to avoid booster circuits at all costs in this design. Also the main function is to collect temperature and humidity values from around the house. So for my part encryption is not a must have for this device.

btw I'm almost there with rerouting rev 2 to make room for the atsha204 (in sot23 housing)

[tbowmo]

Just reading through the datasheet of ATAES132, and realise that I already have support for it on the board. It uses a standard I2C / SPI eeprom footprint, and that is already on the board. So if anyone would like to use it, they only need to replace the flash.

Anyway, I have finished re-routing things.. and the ATSHA204 is now incorporated in the design, so challenge/response authentication could be implemented.

Just need to do the last couple of reviews, before I press the order button for a new set of pcb's

[RJ_Make]

Is it not possible to just encrypt the entire packet with an AES library on both ends?

[Anticimex]

@tbowmo Does your design support both I2C and single-write ATSHA204?

[tbowmo]

@Anticimex

It's single wire atsha204, it's connected to A3. I couldn't fit in the 8 pin variants of the atsha204, so that ruled out the full I2C bus version. It even took me a couple of hours of re-routing to make enough room for the sot23 housing of the ATSHA204.

Anyway, schematic / pcb layouts are as follows

[Anticimex]

@tbowmo Excellent.

[RJ_Make]

@tbowmo Once again,... Well Done!

[tbowmo]

And for the fun of it.. Just made a github repository of the design

https://github.com/tbowmo/MySensorMicro

Just in case anyone would like to have a closer look in eagle.

[RJ_Make]

I wonder what kind of "hit" on battery life authentication/encryption will have?

[Anticimex]

I do not think it is a big issue. Authentication is normally only needed on nodes that have actuators. And this implies that they always have to listen for incoming data and are therefore inherently non-battery friendly The ATSHA has a very low power consumption, so the added cost for message signing procedures is probably negligible compared to the cost of running the radio continuously.

[tbowmo]

Hmm thinking about it, authentication could be valuable on sensors as well.

If you use the sensor readings to control another actuator, then an attacker could send in his own bogus sensor values in order to trigger system events. He does need to know the specifics about your setup though, so the question is if it's affordable to the mischief to do anything like that

[Anticimex]

Yep. This is true, and something I eventually have to add support for.

[tbowmo]

@tbowmo said:

@Dirk_H said:

@tbowmo I don't see a Load Capacitor on the crystal. Especially if you need a precise clock you should use some. Have a look at https://www.adafruit.com/blog/2012/01/24/choosing-the-right-crystal-and-caps-for-your-design/ for example about Load Capacitors.

I know that normally the crystal require load capacitors, but for the low freq. oscilator (32Khz) its not necessary (as far as I have read), that's why I omitted them from the design. I'll try and dig out the datasheets when I'm at home and double check things.

Got around to check up on datasheet for atmel328p, when using lowfrequency oscilator, it has internal load capacitors. If you look at page 33 in the datasheet. So it's not necessary for external capacitors

[tbowmo]

very soon the next prototype pcb revision will be ordered from China. Gonna be exciting. Almost like Christmas, where the kids just wanna open their gifts

[tbowmo]

And pcb's are ordered.. Again, using dirtypcbs.com, for the order..

I have used a panelizer tool from http://blog.thisisnotrocketscience.nl/ to panelize the boards this time, so it should be easier to depanelize it when they get here..

[bjornhallberg]

@tbowmo Looks great!

Haven't actually ordered anything produced by the panelizer yet, but the results look really amazing:

I was planning to finally get around to some boost regulators. But I need to design mine from scratch (these are just scavenged from other places) so they follow the same design guide lines and can be plugged in seamlessly to the main pcb.

Would be great if there were some sort of common design to follow for designing shields and such.

[tbowmo]

As you all probably have noticed, there is an mysensor logo on the boards, that I just ordered..

I have been cooperating with @hek for a while, and this is going to be an "official Mysensor pcb"

We have been in discussions with a vendor in china, for doing SMT assembly as well.. We are currently waiting for the pcb's to arrive and get them populated, to verify that everything is as it should be, before we start up production in China.

For a batch of 100 units, the price is about 13$ per unit, plus shipping/handling fee, this is without any profits to the project

One question though, I made room for an ATSHA204 chip on the board, that could be used for authentication purposes, we want to know if this should be mounted as default on the board, as it will add arround 1$ to the unit price.

Also how many would be interested in ordering one (or 10, 20, 50?) of them? (just so that we get an indication if we should make a batch of 100 units, or 500 units)

[RJ_Make]

@tbowmo

That's 2 questions...

1. Yes, It's worth it to me (Hopefully there will be MySensors software support down the road)
2. I would probably start off with 3 to 5.

Will there be a "kit" price for those who want to populate and reflow themselves?

[Dwalt]

@tbowmo

I would also be interested in at least 5 boards (initially) if they were populated, not ready for smt soldering yet.

[tbowmo]

@ServiceXp said:

@tbowmo

That's 2 questions...

I edited it a couple of times while talking with @hek last evening, so it slipped. Hope you can forgive me

1. Yes, It's worth it to me (Hopefully there will be MySensors software support down the road)
2. I would probably start off with 3 to 5.

Will there be a "kit" price for those who want to populate and reflow themselves?

No kit version, the pcb will be released under oshw so you could order it yourself, and source the components. However my unit price for the initial units will hit 11-12$, so it's only 1 or 2$ extra to get it smt assembled in China.

The more units we can order the cheaper it will become.