Radiocrafts Rf Transmitter Failures.., A RC1180 AMR RF info and advice....

Have any of you guy's worked with the Radiocrafts wireless M-Bus modules..

Been busy trying to get some serious comms out of this thing, because the meter will be running on a 2.1Ah battery, and these little RF modules suck 30mA when in listen mode and 35mA when transmitting. so we trying to limit on time.



So Test #1, How fast does this transmit. We scope it out, and we get 20ms for the UART (~10mA) and 20ms for the RF transmit.

Test #2, how often can it transmit. Again scope it on RF TX and the RF RX side...
We pushed these suckers upto a RF packet every 50ms, problem was on the RX side we lost every second packet..

However after Test #2, the RF Sucker that we were transmitting with, started to behave erratic, and shortly there after, DIED. Pulled out a spare to carry on..

Dont worry about tests 3 & 4, as these were range(which we still have to work on) and time tests.

So yesterday we proceed with Test #5, full 2 way Comms..this time it was with data packets every ~250ms (2TX and 2RX per second) and then the second RF module (the RX one from test #2) starts with a weird problem..

Send a UART command, and it will respond..
Send a RF TX packet, and it will transmit it (LED and RF confirms)
Set the Config or Reset pins, UART will respond accordingly..
However, on a RF RX, Squat comes out the UART.. (LED confirms receive)...

Guess I just killed a second one of the modules...

Your Thoughts on this, did i simply push these way to hard ?? or is there some other issue i need to look at.

BTW, just finished replacing the RF Module and am back in testing mode...

any good rf design would be rated for "brick-on-key" mode .
this means that the rf transmit section would be perfectly happy running 100% output for as long as the power was applied .

it may be that the design is marginal or just "one of those deals" .
it is hard to say without knowing more information .
was the antenna for the device correct ? this has got to be one of the most common problems .

does the transmitter have adjustable power output ?
it may just be that the thing was designed for a certain duty cycle depending on output level .
(a piss-poor design in my book) .

Agreed. Sounds like an antenna or matching issue.

169MHz is really low for PCB based antennas and most RF devices in that range have a helical springy-thingy, not a stripline.

Distance to interfering objects, like metal or people, must be farther away, too... half wavelength at least, or about 89cm.

Shielding... does that thing have to be in a case? RF feed back can cause all sorts of grief.

Cheers!

Lol not these little buggers... The RF only starts once the UART is complete, and the data-packet signature and length is confirmed. if there's a problem with the data on the UART, is does not get sent via RF, so at best you have the RF transmit at 50% duty cycle...

the Antenna track is 50ohm matching and the antenna is a Rainsun 1/4 wave length surface mount, with sufficient ground plane at 90deg (top right corner of the board pictured previously)

Yeah it has 5 levels ..

1: -20 dBm
2: -10 dBm
3: 0 dBm
4: 5 dBm
5: 10 dBm

but absolutely no mention on Duty cycles for any of them..(Hell i've got 32 PDF doc's on these modules, and have read just about every page on them...)

We running at full balls 10 dBm, because in the field we'll need all the power it can give..

So Far i'm agreeing with ya...

what more ... I'll tell you everything i know...

Actually we running the 868 MHz version, and the antenna is a 1/4 Wave length, although just recently we did receive some "helical springy-thingy" 1/4 wave length surface mount antenna, but we have not got round to testing them yet..

Distance to interference, well its laying on the workbench, under the scope,hooked to a SMPS, next to my laptop (with 802.11g wireless network link to the net.),and a few other 'Std workbench' items... However in the field the interfering items are going to be plentiful.. (Celular, Cars, Water Pipes, Electrical cabling). so the few on the bench are nothing compared to what will happen in best case scenario's..

The board will be mounted inside a plastic case (X 2), with minimal metal in the connection side..

good engineering practices never go out of style - eliminating the possible problems is first order .

provide rf shielding for the unit ... leaving the antenna section exposed .
use a feed through (rf) connector for the antenna (center mounted on a sheet of aluminum .
use feed-through decoupling caps/connectors/ferrite beads on the power lines and data/control lines (as applicable) .

it is possible to use 52 ohm coax to go between the terminus of the rf network and the rf connector/antenna .
use the good stuff .

stay at low power until the bugs are worked out .
if you have (access to) a spectrum analyzer with a return loss bridge you can check the matching of the network and antenna .
(a cheaper route would be a bird watt meter - i do not know if bird makes a slug in that range and low power levels) .

once you have done all you can do to make sure that your design is "acceptable" ...
increase power output and test .
if the rf section blows up then you have something intelligent to tell the engineers/designers of the board .

if you can't make it work decently on the bench , it aint gonna work worth a s*** in the field .

The Radio Module is RF shielded, as supplied.

Hmm total track length of the RF line is ~2cm. from Rad to antenna. Connectors and cables for the antenna is out.. we doing all surface mount. (unit need to be 100% water proof)

Done..

lol... Already had the RadioCraft engineers approve the design and antenna placement (well they proposed it)

Yeah .... thats the problem.... I'm trying to get near 100% comms on the bench (~5 % packet loss at the mo)

And just to let you know... Another flipping Rad blew the RF side... Uart works fine but 0 RF ......

So i decided to remove the RF shielding on the module and found a TI's CC1110-F32 inside with a few Control items (tiny little buggers)..

Managed to order a few samples from TI, so i can play with it directly..

This will not get me beat..

hmmm ... pages 36-40 in the pdf ...
these show the typical values for the sundry rf band/frequencies for matching the differential output to a 50 ohm load .
i would make sure that these values are in place and all (the design) is correct .
then i would sub in a 50-52 ohm metal film/non-inductive resistor for the antenna (and the short connecting line you mentioned) at the matching network output .
this will eliminate/prove the possibility of an erroneous antenna/line design .
(i really cannot see that company selling stuff that blows up constantly is going to stay in business)

elsewhere in the pdf ...
depending on the baud/data rate and power output level set current consumption can go up to 33mA .

to my notion , there is a helluva mismatch *somewhere* and the reflected power is hammering the final .
or ... the output level is being run too high for the amount of mismatch .

the point is to eliminate/prove the design of the rf matching network as a problem .

i mentioned running the rf into an external antenna as a means of testing/proof .

if the chaps at the company *proposed* the design and you are having grief ... have them whip up the same thing on their end and see what happens .
hey , i had to stand behind what i engineered ... there aint no bones in ice cream ...
well , at least normally , that is .

Eeeeee! The ancient Forward Error Correction used on Shortwave bands does about that

Eeeeee! The ancient Forward Error Correction used on Shortwave bands does about that

Lol ... and i'd be happy to accept 1% loss due to channel noise....

BTW, lost two more rad's in the last week. My supply of rads is now finished, and we may have to order more to continue dev. (fitted the last one this morning)

Funny thing is sometimes when the rads shut down the UART RX output oscillates,( well actually outputs a empty frame with start/ stop bits only) and this ends up locking up the Proc as it sits constantly receiving and discarding the extra bytes.. (interrupt driven RX)

It feels like a one step forward, 10 steps back on this one...

Another something to note.. I inherited this project, a few months ago, from a techy who appears not to know how to handle CPU cycles and interrupts properly, and decided to take up a position elsewhere, and 'Walked out' on this project after 9 months of dev..

Been fixing indefinite loops galore in the flipping code, and initially got comms up from a ave of 1 in 10 packets...(bad code timing) and now suddenly it looks like these buggers cant handle it... ..

Been emailing RadioCrafts support team directly, however so for the only reply's are, --"Fowarded to xxxxxxx - Please respond regarding the issues below"

no replies with a notification of "forwarded to ..." is a means of passing the buck .
suspicions of this are elevated by the lack of even a perfunctory/cursory reply stating :
"hi , we have received word from customer support about the problem you are having , rest assured that we are looking into this" .

with as many failures as you have had this pretty much eliminates the occasional bad unit due to tolerance run-out .
it boils down to a problem with the design on your end (albeit inherited) and/or a design problem on the manufacturers end .

if indeed you are being stonewalled by the manufacturer you do not have many options when it comes to "moving forward" .
eliminating design problems (from either side) is the most prudent course of action at this time ...
as you have inherited the project it is up to you to show to management where the problems/fault(s) are at and who is responsible (to the best of your abilities considering the situation at hand) .

just as you have to answer to management ... management has to answer to someone else .
being able to "tell them something intelligent" shows that you are perusing the matter in attempts to resolve the issue(s) .
keeping a paper trail in a folder will help out with this ...
there is a difference between talking with someone and saying "yeah , i tried this , and this and this ..." versus slapping a folder down in front of them with the proofs in black and white .

As a businessman, I'd drop that brand like a hot rock, write off the investment and look elsewhere....

Hmm... who doesn't try to pass the back when you have no idea what's going on ..

OMG, I'd change a Sit Load on this design, however with about a year into dev and the final client crying for these (plus another project coming onto the bench) there's just not enough time to reverse right now.. some things are already set in stone because of other items. Casing Molds are been manufactured, with things like the board shape/size ,push button & LCD position and Antenna location now not possible to be relocated. Also because of this I cant even move something to the opposite side of the board because it'll change the overall thickness and require mold adjustment. But i will push on and see if i cant make some adjustments to move some items around and clean up some of the signal paths...

I love paper trail's .. Save my ass plenty with signed and dated pages...

One of my thoughts too...

paper trails (better not make it the possessive form in your report) are a life saver .
you may just wind up talking with the firm's (the correct use of a possessive form) legal department when they want to sue for damages .
but , this is getting ahead of the issue of facts .

it looks as if you are "up against the wall" :
no time to scrap the design and start over .
the customer wants delivery of the product yesterday .
you have to make the current design play with a minimum of changes and ensuing time delays .

the last part may be easier than you think .
still , it depends on which part is at fault ; the antenna side (including the interconnecting trace) or the "final side" (including the matching network) .

has/does the matching network design followed the parameters/circuit as outlined in the data sheet ?
double check the values and circuit used in the network to ensure that they are correct .
now boils down to isolating the matching network from the antenna/interconnecting trace side and substituting a dummy load smack dab on the output of the network ...
a simple chip resistor that matches the characteristic impedance of the design will suffice .
start out at low power and low baud rate ... gradually ramp these items up while monitoring the current (etc) .
if at any point things do not look right (excessive current draw for the given voltage supplied , output , baud rate (etc) , then stop and figure out what is going on .

once the rf side is confirmed as being functional/correct this would leave the antenna and interconnecting line/trace as the culprits .
as i see it , i believe the matching network would have a better chance of being correct than the antenna .
however , proof is needed .

when the new parts arrive and you get a unit repaired then you can start down the path .

personally speaking ...
your predecessor may have been experiencing the same problems that you are .
did he leave behind any notes/schematics/correspondence after he highballed the gate ?

also ... do not be afraid to generate progress/state of development reports on a daily basis and ship these to others .
make it known (aka : document) the "mess" you were handed and the steps you are taking to correct it .
i was lucky in this department at times ... i could take pictures of screwed-up stuff ... some of it was really "sore thumb" .

Well some further info..

Just blew the last of our radio stock. The Last two rad's i actually used wire bridging rather than surface mount them (Burning off tracks trying to replace each time)


Not a very clear pic, but you get the idea..

So First rad like this lasted about 2 hours.. and blew out the RF to RX link..

N.P. Install the last radio in stock (end of samples now).. and for days the Radio worked enough for us to properly debug the Protocol, and trace and fix some of the crash's, funny loops, and even incorrect Bit manip on the uC's registers.

Yesterday had the System (with Rad comms) running for ~30 min before it would crash, but recover (Much better than before). Well this morning i think i got the last of the problems and also added DB strength (RSSI) display to the code. got to a 0.5% packet loss, (5 needed retransmit out of 1000 sent)

Hmm .... strange thing now with the rads ~ 1 meter apart and getting only -45db (was getting that at ~15M before, PC Side RSSI display has been in from the beginning, just didn't check it for the last few days).

Gee wonder whats wrong, lets look under the hood...
Kill power and unplug the ICSP port....
Pick up tweezers and probe the Bridging wires....
AHHH RF line did not contact properly, Lets solder it back in place..

Power up and start the rads..

Woot -25db from across the room...

and 10 min later... Blew out the Rad... Same thing... RF to RX link broken...

So.. the design of the Aerial is definitely the issue (Proved with out specifically trying too). and the Board is going back to the 'Pros' for recalibration...

BTW: found out that is was not the Radiocrafts engineers that approved the antenna placement, It was the Antenna supplier/manufacture that proposed the placement, etc..

So this is on the shelf till we get more rad's...

hmmm ...
at this point (once you get replacement parts installed) it would be best to sub in a resistor for the antenna on a couple of the units and use them in close proximity to allow debugging and long run checks while fixing the problems with the rf side .
this will eliminate the dead time .

still , the proofs of which area is at fault (or both) have not been determined :

an broadcast engineer was on his way to a rural site , his boss had arranged for a rancher to give him a ride from town up to the mountain top site .
he met the old boy at a local coffee shop and from there they headed to the site .

on the way , the engineer noticed a flock of sheep and that they were all facing in the same direction , broadside to the road they were on .
a little further down the road he noticed the same thing with another flock .
the engineer asked the old boy why the sheep were all facing the same direction .
the old boy replied that there was a storm rolling in and that sheep point into a storm .
the engineer looked out at the horizon ... sure enough , a storm was building .

a little further down the road the engineer noticed some pinkish animals all pointing in the same direction .
he asked the old boy what they were .
the old boy replied that those were sheep and the pinkish color was due to the fact that they had been sheared very recently .
the engineer said : "why do they shear sheep on one side only ?" .

Well not one to be cornered by a supply problem, decided to do the Equiv of Dumpster dive and dug around in the stores a little. eventually found some of the old proto's with Rad's still on them. Woot what a luck, pulled out 3 rads from the proto's..

So stuck one back onto the board, but this time took the SM Antenna and related off, and decided to go back to basic's.. Dbl checked and found that for 868, you need 82mm antenna, so clipped a piece of 20 AWG winding wire and cut it to 84mm (2 mm for solder overlap) and put it directly onto the Rad's RF pad.

Ran if for a while, NP, then bent the Antenna to match the board profile (only approx profile match) and found that i barely lost any db..

Been running it Balls to the wall now for the past hour and so far, no problems... Some one screwed up on the ant design and it ain't me, and all it took was to put a simple piece of copper wire (cut to correct length) on the rad and then hit it hard (distance between Rads ~ 500mm) with RF data... Previous one popped at twice this in less than 10 min, others went at ~5 meters in about double the time (~500ms RF data packets)

Report back more later...

[EDIT] Because i forgot the pic...

Because your an idiot.

2 1/2 hours later, and still going strong...



Less than 1% packet loss

BTW RX count is higher as the meter initiates connection then pc passes commands.

Also the processor has not crashed once in all this time...

Finally after weeks, I truly feel like i've accomplished this task...

At time of pressing submit the counts are..

7630 RX, 6970 TX, 54 lost.

well , that is good news .
the original antenna was not designed correctly or it was made to work into a different impedance .

a few measurements of the factory made antenna should show what happened .

now , make a 5/8 wave for the frequency .

5/8 wave = 205.4 mm .. so that will be a bit difficult to wrap inside. Also, right now my thinking is Screw the Matched SM antenna, and lets go with the wire.. Quick, easy, Low cost, and currently effective.

But i will do a side by side on the 1/4 and 5/8, and to make them comparable, they both will be bent to fit the profile of the board..

Then i will do some RSSI stats.. report back later...

yep , a chunk of wire is pretty robust .
stick it down with a thin layer of good epoxy and go home .

it is hard to tell from the pictures you have posted ...
how much room to you have to play with ?
a good picture (just about anything beats a smellphone picture) with the pertinent dimensions imposed would be helpful .

i take it that you can remove the "approved design" antenna assembly ?

you might experiment with a "wavy antenna" ... it is similar to an "S" or double vertical "S" shape .
no hard bends .
i have seen this design used a few times over the years .

i would really like to have a look at that "factory approved" antenna section .

Okay so i don't really have a 'Better' Picture taker.. however I did notice that if i lay things on the scanner, i can take a Color Scan of them (at 600DPI), allowing almost 100% size reproduction prints

So here's the Pic (and Linky to large)


Included in pic is the Pad side of the Radio Module..

So for the RSSI stats... Hmmmmm....

the 1/4 wave copper has a better db than the 5/8 wave, then again it's indoors and not over large distance... there are two values.. How the antenna Transmitted, and how it received.

1/4 W Straight
@ 1.0m : -18 to -20 and -22 to -23
@ 3.5m : -29 to -33 and -34 to -37

1/4 W profile bent
@ 1.0m : -19 to -21 and -22 to -23
@ 3.5m : -24 to -26 and -28 to -30

5/8 W Straight
@ 1.0m : -30 to -32 and -34 to -35
@ 3.5m : -35 to -38 and -41 to -43

5/8 W profile bent
@ 1.0m : -30 to -32 and -34 to -35
@ 3.5m : -35 to -37 and -39 to -41

Each one is the min and max values over 100 RX & TX packets.

Also during all tests it did not lose one packet...

When range testing i did notice that if the RSSI drops below ~ -70db it drops packets like mad..

i have dl'ed the full file ... i'll do a little cleanup on it .

at first glance and unless there is some stuff on the other side of the main board and the rf board ...
something is not adding up .

that antenna looks like an off the shelf pre-fab type .
"rainsun" ... do you have a part number for it ?
it seems that is no where near resonant for the tx/rx frequencies used .

is there a ground plane on the flip side of the board that covers that triangular area ?
from the photo it looks like a small area is covered .
this might explain the shift in resonance of the antenna .
or , it is a PPD .

you may want to experiment with pruning the 1/4 wave ... start out long and nip a little off in order to sneak up on the resonant point .

the impedance of the "line" at/from the rf module may not be what you would expect .
in theory the "line" should have no reactance and act like a piece of coax (for practical purposes and intent) that matches the characteristic impedance set by the network at the business end of the rf module .
this may not be the case and as such and there will be a "sweet spot" that the antenna length will compensate for .
as a guess , i think that C17/18 might be/have been used to "tune out" the inductive reactance of the line ... it does not look quite right for a "stripline" (as it were) BPF/LPF/HPF .

Opposite side has a few caps, crystals, etc that would stand a little prouder than the LCD and hence are placed on solder side. Nothing of significant importance that side..

Part No: AN1603-868 - Chip antenna..

PPD ???
Its a 4 layer Board, and layer 2 is full ground plane (Matching shape of layer 1) Layer 3 has a few linking tracks and ground plane filled to match shape, as well as solder side (Layer 4) matches the shape. That triangle area is Clear of any copper, short of a small pad to hold the far end of the antenna. The ghosting you might see is a sticker on the other side that says "DEAD" !!
which could explain why IC3 has been removed...

Posible, the few Proto's of this board that i found 'dumpster digging' have only a 0 ohm link in place of L2, with C17/18 empty... The Schematic has bold 0R and DNF for these parts.

As for clipping back on the 1/4 W till resonance.. could i do this while running comms, and checking the RSSI ?? At what range do you recon i should perform this..

BTW. this will be on hold for a week as i'm also going on a business trip to Windhoek on Monday. Spending a week in 'Meetings' with the City Council about their prepaid water meters...

Talk about been thrown in the deep end ..

"i'm also going on a business trip to Windhoek on Monday."
is that anything like "ren hoek" ?


antenna pruning ...
while it is best to trim antennas for best radiation (this is not to be confused with impedance matching as sometimes the two do not coincide) in "free space" (no junk and reflective surfaces nearby) , the next best solutions include :
use lower power
a decent amount of separation that precludes severe interference of reflected rf
physical placement that avoids affecting the signal by moving objects (like your body in the path) and "near field" effects .
the receiver should not "overloaded" ... as the transmitted signal comes up , increase the distance of the receiver . it is best to work with a low signal strength on the receive end but keep it out of the mud and grass .

one can increase the bw of the antenna by using small diameter copper or brass tubing (think hobby store) as opposed to a skinny piece of wire .
this makes the antenna less "finicky" and will broaden out the response/resonance curve and may increase the tx/rx efficiency (google "radiation resistance") .
once the "about right" length is determined this will allow for variances in cutting/mounting/soldering .
(and i'll bet that a piece of brass tubing is cheaper than one of those pre-fab units)


i am looking over the specs of the antenna ...
it shows 2:1 SWR (their site lists this as VSWR - a misnomer)
the BW is 10MHz
the impedance is 50 ohm

the antenna is "cut" for 868MHz ...
i would take it that this is the center frequency of the band being used in the unit design ?
the 2:1 ratio is a bit high ... this is where being able to prune an antenna comes into play .
if your actual tx frequency is towards one end or the other of the bw of the antenna the efficiency will suffer .

i have been busy the last couple of days ...
i'll get a chance over the weekend to do some head scratching on the area allowed for the antenna .