How to fix PCBs - A Beginners Guide

Whoa, you're into surface-mount there. Labeling & ID on those tiny things CAN get really strange. I'd guess that the majority of what's been discussed in this thread refers to thru-hole. Most 70s and 80s vintage PCBs will have little or no SMT stuff.

Edit: The one or two times I've had to delve into SMT crap, I found this page useful: http://www.marsport.org.uk/smd/mainframe.htm

yup yup, i figured a picture would be worth a thousand words,

C38/c31/c30 are thru mount but c26 and c33 are definitely Surface.
 
Don't forget ls244 and 18 pin sockets for ram.

I've got around 200 different ttl chips qtys ranging from 20-50 each.


Depends somewhat on the era and manufacturers you deal with mostly. My stuff is mostly late 70s to mid-80s; Atari skewed...YMMV. I keep the "LS" variety of TTL ICs on-hand; they're the most versitile. Only from time to time will you encounter a case where you NEED to match the original type (F, L, etc.).

I'd say (74LS): 00, 02, 04, 08, 16, 20, 32, 74, 75, 86, 139, 157, 161, 174, 175, 191, 245, 273, 367.

And if you want to get sockets too, the vast majority of TTL chips are 14 or 16 pin, with a smattering or 20 and 24 pin thrown in.
 
i am going to buy myself a logic probe,i have replaced a couple of ram chips on a invaders and a ic regulator on a power transformer so i think i am ok for removing and replacing chips on a board,just i am a novice with no experience at all with board diagnosing and repair,so i will have to learn like everyone else rather than have to send faulty boards off every time.

will use this thread as i am sure theres lots of good tips and pointers here to get me started.
:)
 
Just out of curiosity (and a stack of boards...). If you had to say the most common cause of watchdogging on older games that Z80 based, with 2114(48,49), etc etc (early 80's commons)... what would you have to say it is (in you experiences). Failing Z80? Bad RAM? etc etc.
 
On early boards, say pre 1984, the most common faults are physical issues, eg physical damage, corroded tracks, single wipe sockets that no longer make good contact with their resident chip, tarnished IC legs (socketed ICs only really), old style resistor networks that are incredibly fragile in their old age etc etc. People complain about the build quality of bootleg boards, but the early original boards are of a similar build quality to the bootlegs of later years.

Finding a board with a bad Z80 is relatively uncommon (if we ignore Capcom CPS1 boards), unless you have bought someones scrap board that they backfilled with bad parts to sell it as "complete but faulty" - which happens. Usually if the Z80s are dead then the board has met some electrical violence that has killed a lot of the ICs and the Z80s died with the rest of them.

RAM is probably the weakest part of any board but there are many many boards where the RAM is fine but the fault is something else. Going by what is most likely and focusing on that without getting there by troubleshooting the board logically is going to leave you with a dead board more times than not.

A common thing is for people to go round touching up all the solder joints, again this is almost certainly a waste of time unless you have isolated the fault to a specific location and are out of other options. Dry joints on thru hole componenty is extremely rare, especially after so many years. Any boards with truely bad joints died young, were fixed or scrapped. A faulty board from the early 1980s is not likely to have bad solder issues in 2011, so touching up all the joints on the board is a waste of time.

Basically you have to troubleshoot each board from the start : voltages, clock signal, /reset, /halt, address bus, data bus, /OE, /CE and so on.

All you can infer from a barking watchdog circuit is that there is a problem with the CPU, the system ROMs, the system RAMs, the address bus, the data bus and any address decoder logic, and support chippery, on arcade boards this is usually less than half of the entire system. If all that is ok the board should boot, even if the gfx are a mess, and the watchdog should be silent, beware on multiple CPU boards tho, the /Reset pins may be common, so the fault may be on a different CPU to the one you are looking at, ie the sound CPU has the above issues and it is reseting the main CPU which would run fine if it wasnt being constantly reset.
 
Last edited:
Basically you have to troubleshoot each board from the start : voltages, clock signal, /reset, /halt, address bus, data bus, /OE, /CE and so on.

All you can infer from a barking watchdog circuit is that there is a problem with the CPU, the system ROMs, the system RAMs, the address bus, the data bus and any address decoder logic, and support chippery, on arcade boards this is usually less than half of the entire system. If all that is ok the board should boot, even if the gfx are a mess, and the watchdog should be silent, beware on multiple CPU boards tho, the /Reset pins may be common, so the fault may be on a different CPU to the one you are looking at, ie the sound CPU has the above issues and it is reseting the main CPU which would run fine if it wasnt being constantly reset.

That's exactly what I'm dealing with... working on looking into my 36-pin Konami boards... I'm on MegaZone and RoadFighter right now... multiple CPUs... and they're kicking my butt... handsomely. /Resets are working as intended on the boards, so I guess it's just a matter of digging deeper and deeper. The customs aren't helping at all either... but I figure most of them are for misc functions that shouldn't related to, or cause, a watchdog boo-boo. Just gotta keep digging... ugg.

FWIW, I have a Gyruss and a Tutankham that I were fixed by running my hand over the back of the boards. Wasn't looking for cold solder joints or anything like that, just hoping that the voltage from my body would trigger something in a line that wasn't acting right. Ended up working well, both were bad RAM and as soon as my hand would hit the bad lines, there would be a change in the game. Both cases were bad graphics (jibberish on some lines), when I hit the RAM the lines would go away (on Gyruss), and the blocks on Tutankham would clear up. So... there is merit to running your hands on the back of the boards, but out of the piles of boards I have here, those are the ONLY 2 that that has worked on.

To be honest, if anyone was local to me that can repair boards well, I'd probably give my left, right, and middle nuts to watch them work with commentary and be able to ask questions. Maybe in the future I'll be able to afford a road trip for just that.

Thanks for the thread and the pointers. I'll keep tinkering as much as I can... but I don't want to risk the boards by shotgunning too much. In then end, I'll probably have to just put them back down and wait until I can afford some lessons.
 
The watchdog is a simple circuit that if not perodically accessed by the CPU will smack said CPU over the head by activating the reset line. The idea is that if the CPU doesn't boot up properly to reset it until it does. Well, sometimes it never comes up so the watchdog is smacking it like a little bitch that never does what it's supposed to. It's the chip version of Machismo. ;)

With that said...

+1 on the physical damage. It's the biggest cause of issues on the old boards IMHO.

Look for broken caps or resistors in the reset circuit. These will cause the reset circuit to be totally dead though - not watchdogging.

Corrosion on chip legs, corrosion in sockets, liquid damage, rodent fecal/urine damage, gouged traces, leaking batteries, bent pins on socketed ICs, bent pins on surface mount chips, popped pins on surface mount chips from board flexing, and the list goes on.

Beyond that, +1 on physically failed components. Crappy sockets, crappy resistor packs, and failed capacitors that have dried out or split their tops are clear problems.

Whenever I work on a new board the first thing I do is to check for physically failed items.

Next, look for heat damage. Discolored tops on chips, cracked tops, chunks missing from audio amplifier chips, and any chips that burn your finger when touched. The CXK5814 SRAMs on the Neo Geo boards have a tendency to burn your finger when dead... and they have dark spots in the middle too which give you 2 clear indicators as to which one is the bad one.

Once all that is done and the game is still in watchdog the next step is to pull the ROMs, read them in a programmer, and compare the 32 bit checksums to what is posted in MAWS. Replace any bad ones and check again.

Next is to start poking around with the logic probe to see what is going on. What is the reset line doing, what are the address/data busses doing, and what are the CPU control lines doing? What are they doing on the EPROMs? What are they doing on the Work RAM? If the enable lines never activate on the program ROMs then you have an address decoding problem. If the address or data bus lines look trashed then you may have a CPU or interface chip problem with the address bus buffers or data line latches.

RJ
 
Ah Konami boards...

Those can be nightmares to work on. I had a Hyper Olympic '84 that I ended up replacing over a dozen bad LS logic chips on to fix. What a bitch.

Then there's the Hyper Sport that had a bad custom. It was stuck in watchdog because of it. I forget if that was a 501 or a 504 custom chip. The best thing to do for those Konami boards is to have a spare so you can narrow down the problem to the board... or if on both at least to have a board you can hook the bad one to for fixing it. It's hard to fix both bad boards at once.
 
One question I get asked is where go I go learn more about logic gates?

If you really want to learn, it'll take quite a bit of reading to understand the digital logic that you are looking at with the logic probe. A great primer can be found online here.

Be sure to review the section on numeration systems as you DO want to be able to count/convert binary and hexadecimal to decimal and vice versa. This will help you to understand what address lines are/are not active when accessing areas of memory such as a "ROM at $D000" or what address lines are active/not active when the memory location was written as "FF" and read as "FE" to find the errored data line.

Enjoy!

RJ
 
Be sure to review the section on numeration systems as you DO want to be able to count/convert binary and hexadecimal to decimal and vice versa. This will help you to understand what address lines are/are not active when accessing areas of memory such as a "ROM at $D000" or what address lines are active/not active when the memory location was written as "FF" and read as "FE" to find the errored data line.

Google's calculator is pretty handy for converting between hex/deicmal/binary. For example:
http://www.google.com/search?q=0xd000 in binary
http://www.google.com/search?q=0xFE in decimal
http://www.google.com/search?q=170 in hexadecimal

Just type in the the number you want to convert (preceed hex numbers with "0x", binary with "0b"), then " in " and finally the form you want the result (binary, decimal or hexadecimal).
 
Google's calculator is pretty handy for converting between hex/deicmal/binary. For example:
http://www.google.com/search?q=0xd000 in binary
http://www.google.com/search?q=0xFE in decimal
http://www.google.com/search?q=170 in hexadecimal

Just type in the the number you want to convert (preceed hex numbers with "0x", binary with "0b"), then " in " and finally the form you want the result (binary, decimal or hexadecimal).

Windows Calculator does the same thing... Just use Scientific or Programmer modes.
 
Quick Checks for testing logic levels?

Is there anyway to do real quick tests on IC's to see if there are any floating or stuck pins?

Will a logic analyzer sound an alarm or highlight a floating pin?

If I have a clip setup for sixteen pins and I clip onto the different 16 pin chips around the board when attached to a logic analyzer will I be able to see problems and what would I look for?

Can you tell if a pin is stuck low or high by just looking at it on a scope or do you need to always pulse it inputs?

I have tried this but I never see any problems and I don't know if that is because there aren't any problems or the analyzer is pulling things up or down or I am not sure what to look for.

Thanks
Russ
 
Checking for floating pins is about all I use my logic probe for these days, but a floating pin is NOT always an indication of a fault. You need to look at the datasheet/pinout for the chip you are looking at.

If the floating pin is an output pin, and the chip is not in high impedance mode on its output pins then it is probably a fault. Even if the gate is unused the output should be high or low.

If its an input pin then either the board does not have that pin wired up (many logic chips are often only partially used) then it either isnt a fault, or (if the pin is wired up) then the fault is at the upstream chip on that track (of that chip is in high impedance mode for some reason).

Just looking for floating pins is likely to find a tonne of false leads, and not all floating pins are actually silent, the track can be driven by some other chip that time-shares it, the logic probe or scope will show that pin happily buzzing away at the chip even tho it is not that chip that is driving the line.

Fujitsu TTLs are the most common candidates for floating pins, other manufacturers chips very rarely have that fault.

No idea about logic analysers, never used one, I dont really see their benefit to board repair as such, board and system design yes, but repair in the absence of any documentation (which is usually the way with arcade boards) I don't see the utility. All you can do is read the datasheet for that chip and see if the outputs are doing what they should based on the inputs (noting whether the outputs are in high impedance mode if it is a tri-state chip).
 
Last edited:
Thank you.

Well that saved me months of trial and error, thank you. Is there anything that can be learned from observation of a chips signals on a scope? Is there anything that screams "FAULT" if you were to probe each pin or is this always going to need documents and study?

The frustration in this business is when you have to get tons of things done and you come to a brick wall and you have to pour over details and documents chasing problems down. I am always looking for quicker ways to troubleshoot and test things.

Here is a page from a Wavetek Signal Generator relating to troubleshooting that I wanted to post. It gives some quick tests on come components. Do you find this kind of info helpful?

6.4 TROUBLESHOOTING INDIVIDUAL
COMPONENTS
6.4.1 Transistor
1. A transistor is defective if more than one volt is
measured across its base-emitter junction in the
forward direction.
2. A transistor when used as a switch may have a
few volts reverse bias voltage across base emitter
junction.
3. If the collector and emitter voltages are the
same, but the base emitter voltage is less than
500mV forward voltage (or reversed bias), the
transistor is defective.
4. A transistor, when used as a linear amplifier, is
defective if its base current is larger than 10%
of its emitter current (calculate currents from
voltage across the base and emitter series
resistors).
5. In a transistor differential pair (common emitter
stages), either their base voltages are the same
in normal operating condition, or the one with
less forward voltage across its base emitter
junction should be off (no collector current);
otherwise, one of the transistors is defective.
6.4.2 Diode
A diode (except a zener) is defective if there is greater
than one volt (typically 0.7 volt) forward voltage
across it.

6.4.4 FET Transistor
1. No gate current should be drawn by the gate of a
FET transistor. If so, the transistor is defective.
2. The gate-to-source voltage for a junction FET
(JFET) is always reverse biased under a normal
operating condition; e.g., the source voltage is
more positive than the gate voltage for a
N-Channel JFET such as 2N5485, and the
source voltage is more negative than gate
voltage for a P-Channel JFET such as 2N5462.
Otherwise, the FET is defective.

6.4.5 Capacitor
1. Shorted capacitors have zero volts across their
terminals.
2. Opened capacitor can be located (but not
always) by using a good capacitor connected in
parallel with the capacitor under test and
observing the resulting effect.

6.4.6 Digital TTL IC's (e.g. 7400 Series)
1. The device is operating correctly if the output
high state is > + 2.4V and low state is < + 0.5V.
2. The input must show the same two levels as in
step 1. If the levels are between + 0.8V and
+ 2.0V, the connection to the driving circuit output
is open.

6.4.3 Operational Amplifier
1. The " + " and "-" inputs of an operational
amplifier will generally have less than 15 mV
voltage difference when operating under normal
conditions except when used as a comparator
2. When the output of the amplifier is connected to
the" - " input (voltage follower connection), the
output should be the same voltage as the" + "
input voltage; otherwise, the operational
amplifier is defective.
3. If the output voltage stays at maximum positive,
the " + " input voltage should be more positive
than "-" input voltage, or vice versa; otherwise,
the operational amplifier is defective.


Thanks again,
Russ
 
No idea about logic analysers, never used one, I dont really see their benefit to board repair as such, board and system design yes, but repair in the absence of any documentation (which is usually the way with arcade boards) I don't see the utility. All you can do is read the datasheet for that chip and see if the outputs are doing what they should based on the inputs (noting whether the outputs are in high impedance mode if it is a tri-state chip).

Often you can't see if the outputs are doing what they should based on the inputs when the inputs are toggling too fast. There is usually no way to set inputs to a steady, known state in order to check a device against its truth table. A logic analyzer is great for this though. Let's suppose you have an AND gate that you suspect is bad. Using a logic probe, you can see that the inputs are toggling, but the output is always low. Is the output stuck low because the device is bad? Or simply because all the inputs just don't happen to go high at the same time? A logic analyzer will let you see exactly what's happening.

It's also handy when you want to look at logic states with respect to a clock or enable signal. For example, when looking at a tri-state device on a common data bus. A logic probe only tells you the data lines are being toggled by some device somewhere on the bus. A logic analyzer will tell you what a particular device is outputting onto the bus when its 'Output enable' signal is active, or when a falling edge of a clock occurs, etc.

Granted, a comparator would be the first thing to use many situations like this, but sometimes you don't have a reference IC for a particular device, the suspect chip is too big and won't fit in the comparator clip, or the comparator gives false positives due the slight timing differences between brands of ICs and you still want to verify it.
 
Yes but you will have to interpret the output from the analyser based on the datasheet, the OP was hoping he could connect the analyser up and have it tell him if any of the pins were bad, ie an easy way to troubleshoot boards, using an analyser is the epitomy of the hard way. He will have to dig through the datasheet timing charts and see if the chip on the board is in spec with the timings.

In the cases you mention where an actual comparator is no help it is just far far simpler to desolder the chip (if you have the tools to do it easily) and drop it into any modern USB eprom reader that can test TTL chips. For a 20 pin TTL I can have it off the board and tested within less time than it would take to even start working out whether the inputs outputs match the inputs. Tho of course that requires a desolder station and a modern eprom burner, which are both rather fundamental tools for arcade repair. If you have no ability to read the ROMs then you could spend days looking for a hardware fault when the issue is software.

You could do board repairs with a logic analyser, but the 20 year old board you are trying to fix is likely to be a 30 year old board by the time you have worked out where the faults lie with an analyser.
 
Last edited:
While it was tedious to set up...I used a Logic Analyzer to find a problem. I even made a nice vid of it and sadly I think I lost it when my phone went belly up earlier this year.

Long story short the logic probe showed a chip had activity the way it should, but something told me it wasn't right (I forget the specifics now). Since there were a handful of whatever chip I suspected on the board, I set up the logic analyzer and used a test clip so I could just run through and check all the same number chips.

The suspect chip had a bad signal in plain sight...you could tell...it wasn't a clean pattern, in fact it was very random and wasn't supposed to be (I believe this was a clock/sync circuit). An o-scope would prob show the same thing but in this case I could test all the outputs at one time and honestly the analyzer is easier to read sometimes.

That said it was the only time I used it so far.
 
Keep in mind that stuck pins are not always stuck pins... I have seen many cases where the output signal change duration is so short that the logic probe would not pick it up - even on Mem mode. A logic analyzer or oscilloscope is the only way to see these very short pulses.

Also, if you don't have a logic analyzer but do have a dual trace scope you can use it to check both inputs on the and gate. ;) The only issue there is that you are limited to the # of channels on your scope. Mine has 2... they commonly make 1/2/3/4 input scopes.

RJ
 
Would you recommend I get a logic prob if I already have an oscope? Is the audio feedback enough to make it necessary?
 
Back
Top Bottom