>
> Has anyone got any useful tips for tracing out schematics? The SMPSU in
> this 'ere NCR Tower is toast, but it's a complex beastie as it's pretty
> huge and spread over two large boards.
>
> Is copying down component locations (but in their schematic form) to a
> sheet of paper the same sort of size as the PCB a good place to start?
> (Which can then be tweaked into a more sensible-looking schematic later)
I wouldn't do it that way. I ahve an aversion to copying things out --
firstly it takes up time that could be better spent hacking, and secondly
it's a good place to introduce errors.
>
> What about making a list of every component on the board and how many
> tracks leave it at each terminal, which can then be used for sanity
> checking?
>
> I'm not used to drawing out schematics for something this complex, so on
> smaller items I normally just muddle through somehow :) At least this
> one's only 2 layer...
OK, some tips...
Ask if anyone has the schematic :-). Seriously, you don't want to spend
time drawing out a diagram to find somebody has already done it. Of
course you need to compare the time taken in drawing out the diagram with
the time taken in getting a copy of one already drawn (in my case, it
would often be quicker to do the former!).
Make a list of at least the main components on the board. It's probably
not necessary to list every resistor (at least not when you've got used
to drawing out schemaitcs), but list every chip (or section of a chip),
the main transistors, etc. Use manufacturers designations (U1, Q7, etc)
if they're silk-screended on the board. If not, then either label things
(You can get little numbered stickly labels about 8mm in diameter from
stationery chops) on the board/components, or make a sketch (photo?) of
the board and label that.
You'll get a list like :
U1 a b c d 74LS00
U2 Z80A-CPU
U3 2764
U4 74LS138
U5 a b c d e f 74LS04
and so on. When you've drawn a component, cross it off the list. That way
you don't keep on drwaing out the same thing...
Identify (at least) the ground and prefereably power rails. Several times
I've found unexplained connections between what should be unrelated
circuits which turn out to be extra (low noise, etc) power rails!. In the
case of SMPUs, remeber there will be 2 grounds -- the 0V on the output
side and the 'hot ground' -- the -ve side of the mains smoothing
capacitor (normally). Use different symbols and don't confuse them.
Think what should be on the board, and try to identify the main stages
before you start drawing things out. This might involve making rough
sketches of part of the circuit. Identify external connections as far as
possible
You _must_ have data on all the components. Well, 'must' is a strong
word, and I've drawn out boards containing custom chips that I had no
data on. But working without data -- at least pinouts and prefereably
application circuits -- is hard. The application circuit is handy, if,
e.g. you have a chopper control IC that has an external RC network to set
the frequency. It's a good bet that network is in your supply too -- find
it.
'Open circuit' all components that would look like a dead short on your
meter. This probably means removing all transformers and inductors (the
ones in SMPSUs have resistances well less than 1 ohm!), sense resistors,
fuses, swithces, relays, and so on. You will have to desolder the
transformers and remove them totally, things like resistors are best
lifted at one end only so you don't forget where they go!.
Remove any daughterboarrds and handle them separately. You should be
able to figure out the signals to the pins, for example. This, again, may
involve desoldering.
Transformers -- particularly multi-tapped ones -- are hard. You really
want to indentify the sequence of pins on each winding. Sometimes
carefully inspecting the component will help. Sometimes you have to
deduce it based on your understanding of the circuit when you've drawn it!
Complex components (e.g. microprocessors) are a lot easier than simple
ones (e.g. resistors). The former are only used in one way. The latter
could be almost anything.
Assinging signal names is hard. You will want to name signals that go
off-sheet, and maybe signals on the same sheet if they're key signals in
the design. But to give a signal a name you really have to know what it's
for. That means you have to understand the whole thing, And you should
put titles on the main blocks of the schematic too, but again this
means you understand what's going on.
OK, ready to start. I would start at the mains input. Trace where it goes
(probably through a filter, then a bridge, then to smoothing capacitors).
You do the tracing with a _good_ ohmmeter, combined with visual
inspection (where does the track seem to be going!) and knowledge (the
emitter of the chopper transsitor probably goes to the 'hot ground' via a
sense resistor). Find the chopper control IC, find how it's powered, how
it drives the chopper, and so on. Then find the feedback loop (votlage
regulation loop). You will _never_ successfully trace a schematic you
don't understnad.
You will find some suprises. I've had circuits that seem to make no sense
at all, but do so after a lot of thought. If you get something like this,
check and check again that you've not missed something...
Personally I'd not pick an SMPSU as my first serious schematic to trace,
but you can't always choose what you have to do :-)
>
>
> Aside: the PSU doesn't blow fuses, but is totally dead and doesn't even
> attempt to start. I found a dead 5W 10 ohm resistor (open circuit, no
> discolouration) in the live feed upstream of the bridge, which seemed
> like an odd component failure. I had a spare in the junk pile so
> replaced it, and that one immediately went the same way.
This is presumably on the mains side of the PSU. Can you find any dead
shorts with an ohmmeter? Is there a dead short across the output of the
bridge, for example.
-tony
Received on Thu Nov 18 2004 - 18:19:37 GMT
This archive was generated by hypermail 2.3.0
: Fri Oct 10 2014 - 23:37:17 BST