When 1-wire goes wrong

[MN]

I have a big 1 wire network which works fine 90% of the time

But when it goes wrong – what a pain it is to try to figure which part is at fault.  I basically have to take out section and rewire the loop.

The software just goes from 30 devices to zero and not a clue
And stay at zero for an hour then suddenly go back up to 30

I think I might rewire it using a network hub and zone off runs – so basically unplug a run and plug in a ‘blank’ to return the loop

Anyone do anything similar or are all your loops hardwired?

MN

[ericw]

It looks as though you have one device which is intermitently shorting the one wire bus. Not that this helps you find which one though.
Breaking up the bus seems to be only way to proceed unless one device is giving suspicious readings. (for a DS18s20 85 degrees)

[dhaslam]

It is useful to have joiners that can be easily detached so that you can split up the circuit easily.

http://cpcireland.farnell.com/jsp/search/productdetail.jsp?SKU=CN12061

[hiccup]

Hi

In my experience loops are a problem (don't want to start a religious war though). The 1 wire bus should be either a single bus with appropriate termination at the end, or a limited star where the legs are not too long.

Do you have access to an oscilloscope? Quite easy to see what's wrong with one.

The Dallas 1-wire guide is worth a read, I'll dig out a link if you need it.

HTH

Hic!

[dinitro]

I used to have a problem and it was intermittent.  It gradually got worse and I correlated the problem to
a couple of days AFTER it rained.

Unfortunately it meant physically examining each sensor, in my case I found out one exterior sensor had corrosion
in its tails.  As soon as I discarded it everything was fine.

I recently found a new problem, however this time I get a no read error on one sensor in logtemp... but the rest of the sensors are fine.

There is now a new version of Logtemp, unfortunately there is no relay control for the hobbyboards relay board!

dinitro

[wookey]

I've had intermittent network breakage too. A power cycle of the computer+1-wire driver fixes it. Unfortunately I can't power-cycle the 1-wire driver on it's own which would be useful to determine if it's actually software or hardware getting sulky.

I have an 8-channel 1-wire multiplexer chip which makes star wiring easy (and hopefully more reliable). I traced most of my issues to hard-water leaking from my cyclinder and dripping onto sensors and corroding the boards. I've conformally coated them now (and fixed the leak!)

Better 1-wire software diagnostics would be useful. There are loads of magic values in the depths of the software (including various sort of error counts) which may help indicate what sort of problem one has. I've not yet spent enough time to work out if any of that stuff actually helps or not.

[EccentricAnomaly]

Do you have access to an oscilloscope? Quite easy to see what's wrong with one.

Trying to make up my mind if I really need one of these or if they're just shiny:

http://www.coolcomponents.co.uk/catalog/product_info.php?cPath=44&products_id=582
http://www.coolcomponents.co.uk/catalog/product_info.php?cPath=44&products_id=748

[ericw]

If you have a PC with a parallel port you might look at http://jwasys.home.xs4all.nl/old/diy2.html  I used this to debug some of the early halfbee problems.
Debugging 1 wire signals with a standard scope is not easy unless you can generate a suitable trigger from the data on the bus.
 

[Wyleu]

What's a parallel port?

[EccentricAnomaly]

The trouble with logic analysers is that they only analyse logic. When things go wrong electricity isn't logical in the sense that the abstraction of an electrical voltage as either 0 or 1 breaks down. E.g., a problem, particularly an intermittent problem, with a 1-wire bus is likely to be something like that a signal is getting through but is not quite high enough voltage to operate the device properly or the edges are rounded off in way which results in certain bit combinations causing timing errors or something. Looking just at notional 0 or 1 signal levels won't help; it'll just be confusing.

I have one of these which has been very helpful for a couple of projects: Pico ADC-10. I wrote my own software for it in VB and C as theirs was fine but didn't do exactly what I wanted; IIRC, I wanted some weird triggering modes. Haven't used it in a while, not least because it needs a parallel port (I remember those, he says in a croaky old voice). Did wonder about connecting it to an Arduino - I think its ADC is quicker (though lower resolution) than the one in the Atmel chip but I've not checked. Also, it's better protected against out-of-range voltages.

Those Cool Components scopes I referenced are cute but it does seem a bit wasteful to me for them to have their own displays when they're likely to be used on a bench with a laptop or tablet to hand. There are plenty of commercial USB scopes around but for reasonable performance they're often not very cheap - seems like something for the open-source hardware community to have a go at. Hmm...

[MN]

I found the leg that was erroring

Wonder if it is this?

[SimonHa]

I found the leg that was erroring
Wonder if it is this?

I think there's a lot to be said for using CAT5 with RJ45 throughout for 1-wire, even though it sounds like overkill. Doing this so far I've only had errors where wires have been pushed into a breadboard(!). Also if each device (or set of devices) is connected to the network via two RJ45 sockets to can easily remove it from the circuit using a RJ45 coupler (v cheap on fleabay).

As Wookey mentioned I think "conformal coating" of circuit boards would be a good idea in damp environments (I'll start a separate post for this).