Quick Repair Guide

[pablofg]

Hi,

I have a lot of PCBs and although I've been able to repair some, I still have lots to be repaired.

It would be nice to have some kind of list of what to test in order to find the broken part. The goal is to quickly finding the problematic part, and very specially avoid desoldering.

These are my two cents:

* Visual inspection for cut tracks
* Visual inspection for blown, sweting elements
* I have no thermal camera, so I just power it up and start touching all chips. A hot one, means in most cases BAD.
* Push the chips one by one and see if it boots, that may rule out a bad solder.
* If the board boots up, try the test menu or use the information provided on display.
* If there is a processor, check the Clock signal and if it works, look for activity in the address and data bus. No activity, normally means BAD. Check RESET, CE, HALT signals, etc. in the processor to make sure the lack of activity is not comming from somewhere else.
* Check the Eproms /Roms /Rams and see if there is activity. If you have a programmer, extract the Eproms and verify them. Rams unfortunately normally require esoldering.

I haven't found a way to test 74s without desoldering other than mentioned above.
Also from Dave's video, checking the resistance between +5V or +3.3 and GND should give valuable information. Can someone elaborate on what are the expected values?

Any suggestion / corrections are very welcome.

@Dave we need more repairing tutorials. We love them.

[Tita]

Some very interesting videos:
https://www.youtube.com/playlist?list=PL958FF32927823D12

https://www.youtube.com/channel/UCf-EZoze_1mWNMi0XOf5amQ

[c4757p]

* If there is a processor, check the Clock signal and if it works, look for activity in the address and data bus. No activity, normally means BAD.

Not always all-caps BAD. The older-style circuits that actually have external address and data buses often have many reasonably benign things that could shut them down. For example, I recently repaired a multimeter that used an external power-on reset circuit for the main processor, and this POR circuit had a tantalum timing capacitor. If that capacitor failed short (somewhat common for tantalums), the processor would be held in reset and the symptoms would resemble a dead CPU unless you thought to check the reset line.

[c4757p]

Also from Dave's video, checking the resistance between +5V or +3.3 and GND should give valuable information. Can someone elaborate on what are the expected values?

Dead short = bad. It's hard to say, otherwise. I remember working on a board in my oscilloscope that had ECL logic, with multiple divider-type terminators. This ended up in the board having just under 50 ohms straight across the 5V rail. Other boards will read well in excess of 100k, some "infinity".

[pablofg]

* If there is a processor, check the Clock signal and if it works, look for activity in the address and data bus. No activity, normally means BAD.

Not always all-caps BAD. The older-style circuits that actually have external address and data buses often have many reasonably benign things that could shut them down. For example, I recently repaired a multimeter that used an external power-on reset circuit for the main processor, and this POR circuit had a tantalum timing capacitor. If that capacitor failed short (somewhat common for tantalums), the processor would be held in reset and the symptoms would resemble a dead CPU unless you thought to check the reset line.

OK. I'll add to the list. Check RESET, CE, HALT signals, etc. in the processor.

[PA4TIM]

mabey a bit unusual and extreme but there is another way. Used for a very long time and still very helpfull.
It involves the use of a multimeter and a scope.
I know it is a shock, but these things even are to be made for this purpose.

My list:
- study schematics
- optical inspection for problems
- pre-startup work if it is long not used (like reforming caps, cleaning switches, connectors,  etc)
- power up behind a variac and isolation transformer and test it's functions.
- Note faults
- measure powersupply rails for correct voltage and ripple. Also vary input mains voltage to check regulation. This is the most important step. As long as the psu is not 100% you have a problem. (the cause can be outside the psu so sometimes that means disconnecting a rail and adding a load to it for testing.
- note faults and study schematic again. look for clues, which circuits/boards/modules could be involved.
- decide what and where to measure and calculate values upfront in schematics.
- make notes and pictures (for ease of assembly)
- follow signals when possible. For instance in a radio, put RF op the input and follow it with a scope , or if a scope has no sweep start at the sawtooth generator to see where you lose it..
- remove the component that looks to be the problem after measurements
- test it if possible outside the circuit.
- replace it if it was indeed dead.
- After every progress I check powersupply rails again. A bad component can pull a line out of regulation.

Tools I use for every repair:
- several mulimeters, I like to monitor more rails at once
- 1 or two Scopes with 10X, 100X, HV and diff probes
- LCR meter/bridge
Besides that I use a lot of other gear depending on the job. For a SA repair you need more gear as for an audio amplifier.
- SA
- Function generator
- signal generator
- counter
- VNA
- curve tracer
- external powersupplys (to replace an internal powerrail as test)
- simulator software, nice way to try to reproduce a complex fault by changing components into defective components and see the result.

[Spunky]

I stopped doing PC repair in 2005 so I'm a little out of date (hence why I come here to learn a bit). First step was always identify the faulty peripheral or board (diagnosis by substitution), i.e. have power supplies, boards monitors and peripherals on the shelf. Same logic goes for any other system.

After that what the guy in the video says about asking questions saves a lot of time. Diagnosing totally 'blind' takes longer. Questions like when (thunder storm, hot weather), did they spill coffee on it, were they plugging/unplugging it, did it crack/flash/smoke/hiss, did their partner through it through the window just before it 'died', were they doing a firmware update that didn't finish, and in fact did it ever work or did they buy it as is from a boot sale?

Next is knowing (i guess from experience) what tends to go on certain things (common points of failure).
Power supplies are often fuses, bridge rectifiers, diodes, followed by voltage regulators and caps.

Motherboards used to have a lot of trouble with the caps next to the power connector (modern ones seem better), and really old ones (386/486) used to fail on cache ram, oscillators, keyboard serial decoders, and RTC modules.

Monitors (I can only talk about the old CRTs), were power switches about a third of the time, colour driver chip if it's lost 1 colour (usually LM1203N on the end of the tube), X2 capacitors near the flyback if it's got frame/field collapse, or the flyback itself if it's hissing. Also watch out for those fuses that look nothing like fuses (axial lead or small cylinders). You would need a high voltage prove and a high voltage discharge lead before tackling these things (26kv is fairly common at the anode).

Inkjet printers are usually power supply faults, laser printers usually just need cleaning or have failed optical sensors.

Modems were usually lightning strikes, and hard drives either head crashes or stalled motors (you can kick start them for data recovery by applying power then holding sideways and thumping it into the palm of your other hand). A stethoscope is handy for diagnosing HDDs as a rapid shortcut to the fault.

Most of the above requires little real understanding of electronics, from there you move onto more detailed diagnosis. I often found a logic probe as useful as a scope.

Perhaps the best use of time is to spend 5 minutes just looking at the board first. Probably half of the common faults show up visually. Of course if these boards haven't come straight from the consumer then the easy stuff has probably all been done, leaving you with the hard work.

Occasionally I'd have time for a 'proper' diagnosis, checking a board right through for the fault, but in IT retail the product is worth sod all and the customer only has £30 in his pocket anyway, so if I couldn't find the fault in 20 minutes I would swap the board. Hopefully you're not constrained by that sort of thing.