Intermittent faults...

[stevehertz]

The dreaded intermittent fault. The bane of the electronic engineer's life. Like a Will 'o the Wisp, one second it's there, the next it has gone..

Including my professional career as a test engineer in electronic automation equipment, I must have experienced hundreds of intermittent faults. I guess those who have worked on TVs can claim to have worked on thousands?!

But the worst one I ever had was an intermittent failure to record on one channel of an Aiwa AD-2000 cassette deck. The signs were not good from the start. I bought it new from Laskys in around 1980. It was one of those situations when you pay for your item then go to a collection point to pick it up. I had told the salesman that I wanted it in a sealed box. Sure enough, when I went to collect it, the box had been opened and re-taped . I pointed this out but they said that it was the last one.. . It all seemed to work well when I got it home but after a few days when playing back a recently recorded (on this machine) tape I noticed that one channel went missing after a few minutes. It was intermittently failing to record on the right hand channel. I returned it to Lasky's at great loss of time and trouble, I had to travel in to central Birmingham, park up, carry a large box through the streets etc etc. Then of course go back to pick it up when it was er, fixed. Soon after getting it home it became clear that it was not fixed. So I kinda gave up with Laskys. I suppose I should have demanded a replacement, but for whatever reason, I didn't.


So I got the service data from Aiwa's importer and set about finding the fault. There was a long, multi-way switch that was operated by a cord when engaging (I can't remember now) play or record. Sometimes - rarely - it seemed as if flexing this switch invoked the fault, but mostly it did not At those times, no matter how much I flexed it, bent it, it made no difference . So I bought a replacement switch from Aiwa. Made no difference. I tried lots of stuff. A friend of mine who is much better at fixing audio stuff tried to fix it. He could not. The big problem was invoking the fault. But whenever you put it back together, the fault would always re-appear at some point, not always straight away .

In the end I just used the machine for playback purposes and despite trying to find the fault on a number of occasions since, I never did find it. I finally sold the machine complete with fault, explaining its state and history to a guy who had come to buy something else but saw it and said he would have it off me. I'd love to know if he fixed it!!

So, that was my worst intermittent fault, it was never found!

[Mooly]

I always used to insist on sealed boxes too.

One of the most infuriating faults I have had is with my own Hung Chang 100Mhz scope which I acquired it for a very good price. It had been left with CPC for evaluation as to whether they wanted to stock the products, brand new etc... however that was back in the mid 90's.

It developed an annoying habit of having no trace at power up, a sharp tap to the case sometimes bringing it back. Slacken the screws to remove the covers and it would spring into life. It could go weeks and months and be OK and then the problem appeared.

A couple of years ago and I finally got to fix it after a determined effort. The cause were some components (more than one) that were not physically pushed through the board and soldered, the pads were tinned over but the leads didn't protrude through. The board has quite a few 'blank' pads, presumably for other options and models and so finding these was not straightforward.

Laskys... or Hardmans as it was back then in Preston. My parents bought a Sony ST515L digital tuner (mid 70's I guess) that had an intermitent fault. The digital display would jump to 108.00Mhz and no stations would tune in. Back it went... twice... and the intermittent fault would occasionally appear. It then went for many many years without trouble when again the problem struck. It was a dry, located on a crystal oscillator under a tin screening can.

Countless other tales work related but those were the personal ones.

[Radio Wrangler]

Intermittent fault, work related, electronic.... yeah.

At work the PCB shop had a whacking great NC drilling machine for making holes in reinted circuit boards. It had four heads spinning solid carbide drills at many thousands of RPM. Each head drilled through four stacked boards. So sixteen boards were made at a time. It was the days of early numerical control with a matching blue cabinet containing a paper tape reader and sundry gubbins.

It had an intermittent fault. It only happened when the drills were at full depth and the X-Y table would shoot off to back left. Douglas Adams would have said it was sulking. Sixteen boards and four drills written off in one go. Then it would work fine for a while before having a few fits on the trot. The people in that department didn't know what would happen.

So eventually they got funding for a new drill. An Excellon one with a table flying on air bearing pads on a synthetic granite slab. Only three heads but fast enough to more than make up for it. Great! and reliable.

So the old big blue one got advertised as surplus, not working, and a dealer in second-hand such machines wanted it for spares or refurb. He came with his lorry to collect it, and asked where the main grounding strap was. "The what?". "It's a big wide copper braid that goes from the control cabinet to the main machine. If you don't fit one, the servos keep going crazy and break all your drills..."

I only heard about it because I'd had problems with losing some urgent prototype boards to that problem.

David

[G6Tanuki]

In times-past I supported a site that had thousands of IBM 3270-type "green screen" terminals, linked back to per-floor network-concentrators using RG62A/U 93-Ohm coax.

Intermittently and randomly, a bunch of these would misbehave in some very-odd ways. Screens locking, concentrators rebooting. IBM were stumped.

After some cable-tracing we found that a few months earlier a new paging-transmitter had been installed on the site. In the basement rack-room. The installers, seeing this nice thick bundle of coax cables running up one of the cable-risers to the comms room on the top floor, had decided to borrow an 'unused' one to take their hundred-watts-or-so of POCSAG on 150-odd MHz up to the roof mounted antenna.

RG62A/U is well-shielded, but not *that* well-shielded!

[Ian - G4JQT]

I completely cleaned and rebuilt a DAC90A. Re-aligned it, polished it, put in a bright reproduction dial It was a lovely perfectly working example for a couple of years.

Then it started making very intermittent loud crackling noises, reduced a bit with the volume down and much stronger on MW than LW so that gave me a clue. Eventually I took it apart to investigate, whereupon tapping almost any part made a loud click, but nothing more. I re-soldered all the joints around the RF and mixer sections and it was fine for a year or two.

It then required more investigations with similar results. I'd had another AC/DC set have similar sounding fault in the past which turned out to be a failing mains dropper. So I prodded and inspected the one in the DAC90A but every section measured correctly without any variation. But I did notice the clicks were slightly louder when I prodded components and wires around the RF section. I set the mains dropper and wound up a variac to give the whole set about 20% over voltage for an hour or so to see if any failing components showed themselves. None did and the set seemed fine again, for another six months.

After more frustrating investigation including using a hot air gun and freezer spray, I eventually tracked it down to failing insulation on one of the trimmer caps. Tapping in a particular spot exactly reproduced the noise. Non have any HT across them, but the loudness and severity of the noise certainly suggested that.

Anyway I removed and completely dismantled the trimmer caps and replaced the mica wafers - some of which were very frail.

After a thorough re-alignment all has been well for the past year or so...

[Skywave]

Many years ago, a friend delivered a Racal RA-17 to my door, asking "Can you do something with this, please?" Which I (correctly) interpreted to mean "Can you get it working?" The question was highly appropriate: it was in a terrible state, particularly mechanically. Started by replacing broken / missing mechanical parts, followed by a basic "What happens now if I connect it to the mains and an aerial?" type of test. Much to my surprise, most things seemed to work - more or less. So the usual overhaul for a RA-17 was completed and some serious measurement tests were made: adjustments as required. So far, so good: now a serious listening test. That's when the trouble began. Now and again - about every 5 minutes thereabouts - a loud burst of static-type noise (lasting about only a few seconds or less) would come crashing through the speaker. Hmm. Disconnect aerial: no difference. So began the typical 'What's causing that?" investigation. Eventually, after a lot of component substitution and many hours, culprit was identified: a small mica capacitor (100 pF or thereabouts) in the circuitry of the 1st. mixer. That capacitor had HT+ on one plate; the other plate was 'signal warm'. (I can't recall the exact detail now after all these years). Anyway, upon replacement and a subsequent extended 'soak test', all was well. Eventually returned the radio to its owner, whom several months later told me that it was working fine.

I've always wondered as to why that radio had been reduced to such a sorry state before it came into my hands. I've often speculated that the faulty capacitor which I (eventually) found had driven that previous owner crazy and that he'd given up on a repair and had then used it as a source of spares.

Al.

[Skywave]

This 'intermittent fault' thread reminds me of the old joke about a customer who had such a fault with his T.V. and called in a service technician. I dare say many of us have heard this before, but for those who haven't . . . .

"What's the problem?" asked the technician.
"Every couple of minutes, the picture rolls" came the reply. "Then it's O.K."

The technician looked at the screen and after a few minutes the fault become manifest. So, he went around the back of the set, paused, and delivered a substantial thump on the cabinet. He then went to the front and watched the screen. After 15 minutes, the fault had not appeared, so he wrote out the bill and presented it to the customer.

"£4 one-and-six!" exclaimed the customer. "Huh? How do you get that price?"
"The one-and-six" is for thumping the set, said the service technician.
"And the £4?" enquired the customer.
"For knowing where to thump it" came the reply.

Al.

[AC/HL]

It was 10 Guineas when I heard it first, ten bob for the thump.

[Radio Wrangler]

That one turns up in every trade. I wondered about the Romans or the ancient Greeks, but I reckon it may be a dead heat between the Egyptians and the Chinese if the amoeba didn't beat them to it.

But is it older than "When I nod my head, you hit it"

David

[peter_scott]

Originally Posted by Ian - G4JQT

After more frustrating investigation including using a hot air gun and freezer spray, I eventually tracked it down to failing insulation on one of the trimmer caps. Tapping in a particular spot exactly reproduced the noise.

As stated heat, cold or rough handling or tapping usually helps to track them down.

Peter

[ionburn]

I think it is probably little use in demanding goods in a sealed box in many cases.

When at college I got a summer job with a well known telecommunications company. They had the contract for processing items returned under warrenty from a well known High Street chain (no longer trading under that name but the group still exists). I remember that items had to be repaired to literally 'as new', even cleaning flux residue from print. The devices (answerphones and call loggers) were then re-packaged in brand new boxes and returned to the company. Whether these were just used for warrenty return exchange or new stock, I don't know. The equipment was well repaired (at least the ones I did and others I saw) but it really made me rather cynical.

In the TV trade I have come across a number of intermittant faults. I think, though, that the one which really got me was the particular set (from one of the Japanese manufacturers if I remember rightly) which had to have all the connections on the main board re-soldered if the problem occurred. This always sorted the problem. Luckily we only had one or two of the model go wrong and it was never me who had to sort. I always thought it was too much to have to go that far with a warrenty repair and they should have been returned to the supplier.

[stevehertz]

Originally Posted by ionburn

I think it is probably little use in demanding goods in a sealed box in many cases.

When at college I got a summer job with a well known telecommunications company. They had the contract for processing items returned under warrenty from a well known High Street chain (no longer trading under that name but the group still exists). I remember that items had to be repaired to literally 'as new', even cleaning flux residue from print. The devices (answerphones and call loggers) were then re-packaged in brand new boxes and returned to the company. Whether these were just used for warranty return exchange or new stock, I don't know. The equipment was well repaired (at least the ones I did and others I saw) but it really made me rather cynical.

I don't want this thread to get drawn off topic on to the subject of 'sealed boxes', but I strongly disagree, I think there's every reason to ask for a your item to be in a sealed box unless you get a discount, an explanation/reason for the opened box, a full guarantee and are happy to accept the deal. I think what you describe is a pretty unique set of circumstances, I doubt very much if UK importers (most stuff comes from abroad) have the facility to put refurbished items into brand new 'sealed' boxes. Proof of that is the fact that you can buy lots of stuff from auction sites that states clearly that it is refurbished and may not be in the original box or that the box has been opened etc.

Further, just a couple of weeks ago I bought a new pair of speakers. Having spoken to them on the phone to check availability, before setting off the store called me back to say that they only had one, boxed and sealed pair in stock and so would NOT be able to demo them for me. Reason: they only sell stuff in sealed boxes and they didn't have a demo pair in at the time. Anyway, I bought them and all was fine. No, always a sealed box for me now, you just don't know what has been going on if the box has been opened and the item put back. It could have just been demo'd, or it could have been returned for a niggling (intermittent!) problem, you just don't know..

['LIVEWIRE?']

The joke posted by Al (Skywave) is a variation on one published in Radio Constructor many years ago, in which a TV repair man squited some servisol into the channel change switches, then, when asked how much, replied 'tuppence to you, Major for the servisol and a tenner for knowing where to squirt it!

[Peter.N.]

On old valve radios crackles were some times caused by dry or corroded joints inside the IF cans, another problem can be HT leakage on the wavechange switch, if you can catch it doing it you can check for DC volts where there shouldn't be any.

Peter

[G6Tanuki]

One of my strangest "Intermittent" faults was actually caused by components that were not there!

Early-1980s Solid State dataloggers: basically a box with 64Kbytes of CMOS RAM, a crystal-controlled date/time clock, an analog/digital converter and some CMOS control-logic. all powered by a pack of "AA" cells and fitted inside an IP67-rated alloy case, with waterproof-to-50-metres plastic/rubber ITT-Cannon plugs/sockets for external connections.

These could be set to sample all sorts of things [individual line-transitions, output of the A/D] and store the results in the CMOS along with a very accurate timestamp, for later playback on a minicomputer (a PDP11). Used for things like counting fish going up a weir, electronic rain-gauges, peak traffic-flows of the punters at the Isle of White festival, rate and number-of-rowing-strokes by oarsmen in the Boat Race... anywhere you needed robust accurate storage of 'events'.

Problem was, some of them were failing with flat batteries (nd loss of data) after only a few weeks of deployment. The manufacturers and hirers were pulling their hair out because of irate customers who'd paid loads of money for a service which then wasn't delivered. They tried everything - different batches/suppliers of CMOS chips, different batteries, different PCB materials/processes, purging the sealed cases with dry Nitrogen, nothing seemed to fix it. A box that had failed once may go on to work perfectly, whilst another that had been rented-out several times without issue might suddenly fail.

I got called-in and spent a few days talking to the people concerned. Thay were all dedicated digital-electronics types, with one poor production-engineer to deal with the packaging/assembly side.

Looking at the PCB for the control-logic board I spotted a number of PCB tracks and associated pads that didn't have any components fitted.

"Oh, they were for a prototype Mk.1 version with additional inputs - we asked customers if they needed these inputs but nobody seemed interested so we deleted the components from the production versions" was the reply.

Alas, these unused PCB tracks connected to the inputs of eight CMOS gates (shared across three chips whose other gates still did necesary things)

In the original design there were pull-down resistors to stop these inputs floating. The "customers not interested so we deleted them" version had these resistors deleted, but there were still several inches of PCB track connected to what were, in essence, infinitely-high-input-impedance gates.

Random drift of electrons was sometimes enough that the charge building up on these floating inputs was biasing the associated gates into their linear operating-region - at which point they began to pull several milliamps per gate from the battery-pack.

Depending on the number of gates randomly activated, and for how long, the battery life was being seriously or not-so-seriously compromised.

Putting pull-down resistors on these gates solved the problem entirely! A lovely case of how components-not-actually-there can cause an intermittent fault, and a rather-lucrative bit of weekend work for me.

[G4_Pete]

I guess we have all got stories so here is mine.

For many years I looked after an industrial computer one of a pair which never had the percentage up time of its partner. Eventually the day came to turn it off for good and dismantle it. Deep in the PSU rack one of the dual feeders pulled out black and charred on the end. All those years of intermittent crashes and this was probably the cause, if only it had failed completely and not been part of a dual power system. I just sat on the floor and stared at it for some time thinking of all the times I had been woken at 2 am and not found the fault! In fact you could not even see these particular terminals. In later years deep memory storage scopes and event triggered mains monitors etc. were part of our standard first check fault finding kit, sometimes left on for weeks, but not back then.

Pete

[Refugee]

I can remember being asked to repair an AR88.
The reported fault was intermittent dial lights causing various random faults.
Various resistors had been changed in the IF stages and even capacitors has been disconnected and new ones hung under the chassis. These were the dual grey square ones on the power rails under and between the IF cans.
I got the fault to show and yes indeed the power was cycling up and down causing the grid coupling capacitors to charge up and drive the valves hard on.
The IF stages just needed fixing again and the dial lamp wires were nut burned enough to make me feel that they were the cause.
There appeared to be a problem on the AC side. On going through the AC side I found that the mains voltage selector was full right up with char coal.
The previous owner before the one I was fixing it for should have been able to trace it easily from the smell. They didn't and it was down to me to hide a chocolate block under the transformer.

[Station X]

My favourite is this one, a connection in a Racal 9087 signal Generator which had never been soldered.

[julie_m]

G6Tanuki --That's similar to the legend of the prototype ARM-1 silicon apparently working -- albeit not 100% reliably -- without the power rails connected, by scavenging power from any logic 1 inputs via the input protection diodes.

[Jon_G4MDC]

Nice one StationX.

Racal managed that in my RA117 on the 40Mc/s filter. Silver plated wired wrapped on a silver plated trimmer tag, applying solder optional. Until it had spent a few years at sea presumably.