Inverter welders

They're "amazingly reliable" for the first few years, but I can tell you from very extensive first hand experience that vintage electronics age just the same as everything else. And the more complex they are, the harder it becomes to trace the fault and fix it - and that may become next to impossible when the manufacturer has used custom chips had short production runs.

BULLSHIT.

BULLSHIT.

More bullshit with accidents

BULLSHIT.

For a lot longer than that.

Haven't seen that with any electronics except caps.

Wrong with stuff that is complex enough to have its own diagnostics.

Doesn't happen much with cars.

If you read enough IBM documentation, this value is "100,000 hours".

We did not use such a metric in our fab. The electromigration frequency was just "X", and there was no qualifying "for 100,000 hour life". Whatever our number was, it wasn't in print. Maybe our reference point was in some separate generic reliability document I never saw. IBM likes to put that reference figure in documentation.

But this is a reference point, for technology choices. It basically says, the target most of the time, is

The customer designing integrated circuits in that fab, does not desire the chip to die precisely at 100,000 hours.

If you run a CMOS chip at 135C, the dopant migration might make it last 100,000 hours, but the datasheet tells you to run it no higher than 100C die temp. And then Arrhenius (curve fit), tells you how many million hours it will last.

A chip can last a lot longer than that.

If loose chlorine pollution happens to be inside the molded plastic IC package, the IC might only last a year or two. IBM traced a case like this, to the usage of a marker pen to mark bad die during wafer-sort. (That's where you do an ICCC test while probing die sites before the wafer gets scribed.)

It's even possible they don't do wafer sort any more, and just package it all, and that's where bags of 555 timers with a dead short between the legs, come from :-) If there was a wafer sort for quiescent current, it would have tossed those die. But the ICCC test was only really a great idea, when CMOS was 3u. At 65nm, there is enough leakage, the quiescent current is not a good enough indicator of chip quality. You might have to run test vectors instead. And packaged parts are likely easier to cool, for the high power ones. (Like the ones with a thermal slug on the bottom.)

There is one processor, which is rated for 300C. But for only 1000 hours :-) This is a processor sent down boreholes, like for an oil exploratory hole. They characterize at that temperature, because a major customer asked them to. There is no industry standard for 300C integrated circuits. Mil spec is the most stringent "widely seen" point. I would think your little printed circuit board would cook to a dark colour at that temp :-) Like your morning toast.

Imagine how long such a processor would last at 25C. Why... forever...

Paul

Do the circuit analysis, and spot the weakness.

That circuit should have some filtering before the full wave bridge, to stop switching noise from going back onto mains.

Where the design has an advantage, is the transformer does not have to be as big, at the switching frequency used.

That circuit has three snubber networks, and it has series resistors on the switching transistor gates to prevent "punchthru". Transistors with 3000pF gate capacitance, a couple amperes of current flows from the pre-drive, into the gate, and this can burn a hole in the gate, if the peak current is too high. The circuit has the necessary series resistors to reduce the peak current value.

The dot on one of the transformer windings , is the wrong way round :-)

There are some circuit functions, that are notorious for failure, such as variable speed motor drive. But I don't know the reason for that.

Paul

You see it with a lot of 80 era microcomputers. The NMOS process used on many results in failures with age, and there were many custom ICs and application specific gate arrays used. So they are getting increasingly hard to repair.

On the more well known names you are starting to see quite decent supply of re-engineered chips using off the shelf EPLDs and FPGAs mounted on DIL shaped PCBs as drop in replacements.

Vintage electronics are, by definition, not very complex. Custom chips only appeared in the 90s really.

I can confirm that this is exactly how electronics is.

In general modern resistors, non electrolytic capacitors and coils/chokes do not age. Chip, especially chips run hot, do. So do fans. Running chips cool should net you 30+ years of life.

I suggest you watch some of the earlier Youtube "Rainman Ray" videos when he worked for a franchise tyre/oil/repair business. Many of his videos show him diagnosing faults by way of the multitude of the electronic sensors to a point "where he didn't have to fire the parts cannon". His videos do show a (great) level of skill in knowing how a car works and him not just blindly following what the computer says. Not only does he explain what he is doing and why he ignores the red herring routes to diagnostics but also shows the fix to the car(s).

For the past 3 months or so many of his videos have been about setting up his own business with less content showing him using professional OBD tools to diagnose faults.

In my experience modern cars (of any age) are a lot more reliable than the cars of the past and in the main just as easy to work on if things do go wrong. In fact these days you can probably easily find a "how to" video to fix items on a car and a many forums on the web where you can discuss the problem, how to fix and the possible pitfalls - much better than the Haynes manuals. The following possibly shows how frustrating it was to work on an old car

"Diagnose Dan" is the man who really knows his computers.

Watching both of them has given me a very good general understanding of engine management system, at least so far as the number of sensors involved and what happens when they fail, and a horror of modern European cars with canbus and half a dozen computers to run the suspension, the brakes, the climate control the engine,m the transmission, the door locks and airbags, Arrgghh!

Unfortunately it is all beyond the average grease monkey's intellect, so they just replace parts at random.

Which is why I got a basic OBD2 scanner. I reckon I will be better at diagnosing 'warning light; issues than the person I take it to

Yes. I remember my first electronic ignition fuel injected car, I bought to commute to Brussels. An Opel Manta 2000. It simply never went wrong. No blocked carb jets, because you had to have fuel filters in an injected car. etc.

The following possibly shows how frustrating it

Depends on application. Not sure when Ferranti started producing ULA but I believe in the 1970's they were widely used in Cameras.

In home computers, in the 1980's the Ferranti ULA was used in the BBC B, Sinclair Spectrum and QL which can make repairing these challenging.

This technology led to the development of the ARM CPU as the experience gained designing the BBC B enable the team to understand high level logic design....

Dave

Pretty much every 80s era micro had a level of custom chippery... Many used applications specific mask programmed gate arrays (usually made by Ferranti) to massively cut down the amount of glue logic for address decoding (even the ZX81 had one - eliminating large numbers of 74 series logic chips from the ZX80 (overall chip count fell from something like

Commodore owned a complete semiconductor fab company by virtue of buying MOS Technology[1] and so made heavy use of custom chips for their video controllers, sound chips, IO chips, and even custom CPUs (like 6510 in C64, 8502 in C128).

By the time you got to the mid 80s, and the start of the 16 bit era, machines like the Amiga were festooned with very sophisticated custom chips that handled every aspect of the machines performance and gave it performance levels (particularly graphics, sound, and IO) far in excess of what you could expect by looking at the CPU specs alone.

By the mid 80s you also had Acorn designing their own custom CPU (and we all probably now own a bunch of their descendants!)

Then you've never been involved in chip design. I have.

Yep, you didn't say bullshit to that. Are you feeling ok?

I have a friend working on this field and some modules last hours in their environment, long enough to fulfil their purpose.

Some of the automotive temperature specs are pretty tough. Why some ECUs use an alumina substrate. Not sure if they still are?

When I have time and in the mood I do watch his videos. I also attend a couple of forums and it's a common story where a garage may or may not have fault codes and the parts cannon gets used, where each time the original fault returns.

Most of these failures were difficult to trace. My point was the simplicity of older cars. Where say an alternator is a single entity, and not controlled by an ECU. EGRs are another common failure. Carburation was by a cheap single component.

All the new fangled parts are there for good reason, to improve economy and emissions, however they add complexity and reduce reliability to suit.

In terms of access, I could get an engine and gearbox out of a Ford Cortina in 40 minutes, I suspect a pro in much less time. The book time for a 15 year old Mondeo clutch change is 8 hours. It took me 2 days. As a result the Mondeo becomes more valuable as scrap for the average guy.

It's one reason I like Fords, as Forscan is mainly free and is a complex scanner than can also do key coding.

As a car ages, some scanner compatible with the car will generally become available, if only a knock-off from China.

LOL but even then it was largely down to having the right tools. They have come down in price in real terms and and the cost can easily be justified in terms of saving compared to garage labour costs.

Vintage electronics is an oxy-moron. The only vintage electronics in one vehicle I have is the radio!

Are you sure? The transistors are wired in series with the transformer, it's not a push-pull affair as there's no centre tap or full or half bridge. I assumed to halve the voltage across transistor when it's not conducting?

I can't really figure out what they're doing there.

As it would seem if you just had the low side drive, that would be enough, and the low side has the feedback to the flyback controller (presumably for some sort of current monitoring).

The diode at the top, shows a zener symbol but it's a fast rectifier diode with a 600 volt rating. Clamping the low side drive to no more than a diode drop above positive rail.

When I did a flyback here, it was push-pull so maybe that's why I'm looking for push-pull.

Paul