see below, plz.
Dick
----- Original Message -----
From: "Tony Duell" <ard_at_p850ug1.demon.co.uk>
To: <classiccmp_at_classiccmp.org>
Sent: Wednesday, June 13, 2001 2:54 PM
Subject: Re: Ebay horror ...
> > > If you know it's a fault on the serial port (for example), then you've
> > > done almost enough tests to determine what the fault really is. I am not
> > > talking about 'fault finding' that goes like : 'The modem doesn't work'
> > > 'Replace the serial card and see what happens' 'Oh, seems to be OK now'
> > > -- that's not fault-finding, that's why I object to board swapping.
> > >
> > Well, it certainly doesn't find the fault in the serial board, but it does
find
> > and fix the fault in the computer. Since serial boards cost less than the
10
>
> Not necessarily. You may think you've fixed the fault only to have it
> come back later. Maybe a part has gone temperature-sensitive. It cools
> off when you turn the system off to replace the serial board. Turn it
> back on (with a new serial board) and it all seems fine. Only for the
> fault to come back an hour later. Actually, the fault would seem to be
> cured no matter which board you replaced or even if you did nothing.
>
> Don't laugh, I've seen this happen. I've also had to sort out the
> resulting mess....
>
I've heard of such things, but never seen one myself.
>
> > minutes it takes to swap them, and it costs 100x that much to fix the serial
> > board, at a minimum, it seems to make sense to swap the board. That does
fix
>
> Are you seriously telling me it takes you 1000 minutes (or over 16 hours)
> to fix a serial board? The only time it could take me that long is if the
> 'serial board' was something like a DEC DMR11, and it had a nasty logic
> fault in the microcoded processor.
>
I'd never spend more than that 10 minutes I mentioned before on a serial port
board, because they only cost 50 cents or a dollar at the thrift store. If you
have a complicated serial board, i.e. one that doesn't talk the "normal"
asynchronous protocol, and that runs at a higher rate, e.g. 4.762 Gbps, and is
built using discrete GAs logic it might take a while. I generally don't use
stuff like that though, and it's only in cases like that I'd be likely to find
it justified. These boards with soldered-on ASICs aren't worth fixing, and the
ones with socketed components seldom break. It's only going to break if the
part is soldered down.
>
> > the computer, doesn't it?
>
> See above...
>
> > >
> > > And if you know where the fault really is, it doens't take that much
> > > longer to replace the fualty chip or whatever.
> > >
> > That's true if it's a serial board. However, if it's a mono video board,
also
> > costing less than a half-hour's time, and replacing it gets the server back
into
> > service, while tracking down and replacing the 100-pin fine-pitch pqfp that
>
> My mono video boards are all DIP chips -- 6845s, RAMs and TTL. But I digress.
>
> > > It's interesting you mention the serial port -- yes those do fail quite
> > > often. But most of the time _on all the machines I have_ it's the RS232
> > > drivers/receivers that fail, not the UART or other logic.
> > >
> > True, but not if you don't fiddle with the cables. Moreover, even the few
>
> Actually, a serial port meeting RS232 standards should be able to withstand
> hot-swapping of the cables. Shorting any 2 pins together, or to ground should
> do no damage.
>
Then why do they break? and why don't they socket the transmitters and
receivers?
>
> > minutes needed to desolder the transmitter, and that fails more frequently
than
> > the receiver, for some odd reason, in my experience, you've still spent more
>
> The driver is more likely to fail from short-circuits (overcurrenting the
> output transistors). The receivers are pretty solid.
>
> > than the entire replacement card costs.
>
> I wish I could get paid that sort of amount....
>
>
> > >
> > > Now, 95% (or so) of the machines I have -- minis, micros, etc -- use
> > > 1488s and 1489s for this. DEC, Philips, PCs, TRS-80s, all have them. The
> > > next most common chip is the MAX232. Then comes the oddball stuff.
>
Yes, they're common, and I have about a dozen and a half spares of each.
However, buying a handful of spares, i.e. 3 of each, at the local surplus dealer
costs more than a half dozen serial boards cost. I think I had to pay $0.65 for
the 4-bit transmitters and receivers (MC1488/1489) last time I bought them, and
that was some years back. I scrounge MAX 232's either as samples when I feel
like it, or off circuit boards where the necessary capacitors also are avaiable,
so I don't even have to look in my parts bins.
>
> > several of my old S-100 boards use the 75154 and other somewhat odd parts.
I
> > hand-wired a mezzanine card that plugs into the sockets occupied by two
8-pin
> > transmitter parts and one 75154 and replaces them with a 1488 and a 1489,
just
> > because of that. Apparently there were supply problems or cost issues that
> > caused CCS to design with the TI parts.
>
> FWIW, I think the TI parts are still available if you need them. If not,
> then the daughterboard is the obvious solution.
>
Yes, they can still be had, but not at what I like to pay. The ones I like the
best are the 8-pin 75155, which, BTW, is available at Radio Shack.com, though
they're not free either. That's a 1-bit transmitter/receiver pair that requires
external supplies. However, for simple setups one can use + and - 5. I like
sticking that part and a '2691 in a 32-pin x 0.300" socket. to make a complete
serial port, uart, transmitter/receiver, baud rate generator, one bit wide.
With a single-chipper, you can leave out the uart and brg, of course.
>
> > >
> > > Now, I do enough designing/prototyping here that I keep 1488s, 1489s and
> > > MAX232s 'in stock'. It is actually easier for me to find a replacement
> > > chip than a complete serial board. It also makes a lot more sense to me
> > > to keep small parts that can be used to repair many different machines
> > > than complete boards that only fit one machine.
> > >
> > In recent years I've found it MUCH easier to find a replacement serial board
for
> > a PC than a replacement 1488/89. The fact they're not so popular as they
were
> > back when they weren't on the motherboard makes a difference. They cost
> > typically $1 for two at the local thrift stores.
>
> Over here 'charity shops' (==thrift stores) don't generally sell
> computer parts. They may sell complete computers if you are lucky (and
> can find a volunteer to do electrical safety tests). But I've never seen
> a serial card or anything like it in one.
>
The common PC at a thrift shop has had the "interesting" boards removed, along
with most of the RAM and drives, and normally has just a PSU, simple video
board, keyboard, and monitor. They ask $5-10 for the computer. The monitor is
often separately sold for $5-10. They always have the basic stuff in them,
though the memory and hard disk often are absent. However, if the fan in the
PSU works, it's worth buying the entire box for the fan. In the US, an AT-type
replacement PSU fan costs $40. The PSU, fan and all, costs $30, and a case with
PSU and fan costs $22. At the thrift store, you get a used fan with the PSU,
but it will get you through a few days until you can order and take delivery of
a new case.
>
> So, I either go to the local PC shop and find they no longer sell serial
> cards (and then try to sell me a replacement motherboard with serial
> ports built in), and then go to more and more PC shops trying to find the
> board I need, or I go straight to the local components shop and pick up a
> 1488 for <\pounds 1.00. And if said shop is out of stock, I know a couple
> more I can easily get to. There is the added benefit that the electronics
> component shop sells other things I can use, the PC shop doesn't. I can't
> remember the last time I bought something in a PC shop.
>
Me too. PC shops only sell what's in fashion this week. I like to shop the
web.
>
> That's assuming, of course that I don't have a 1488 in stock. Or that I
> can't borrow one from another machine until I next go to the components
> shop (which I generally do at least twice a week).
>
I'm getting too old to replace those teensy 0.25" pin-pitch so packaged parts,
barely the size of a pinhead. When the serial ports on the motherboard die, I
disable the ones on the motherboard and plug in one of those old ISA I/O boards.
I don't know what I'll do when they no longer have ISA slots ... <sigh> ... the
last motherboard I bid on on eBay, just this week, has six PCI slots, but no
ISA. We'll see ...
>
> There are very few machines round here that I could borrow a complete
> serial board from to fit in a different machine...
>
> > I'm sure that with all your experience, the replacement of a faulty IC is
> > thoroughly executed and produces a reliable result. For me, though, since
the
> > electricity to desolder and resolder a faulty component often costs more
than a
>
> What sort of soldering iron are you using????
>
> In my case, a Weller TCP. Lets call it 50W (it's actually a bit less).
> That means I could run it for 20 hours on 1kWh of electricity. Even if
> the transformer in the soldering iron PSU is 50% efficient (it's a lot
> better than that), I still get 10 hours of use for 1kWh. 1kWh costs about
> \pounds 0.07 here. And it's going to take me about 5 minutes to desolder
> and replace a 14 pin chip at most. Which means that it costs me \pounds
> 0.0006 or thereabouts for the electricity to run the soldering iron while
> I change that chip. I doubt you can get a serial card for that little :-).
>
last week I got 35 boards for $1.99 at one of the local thrift stores. 3 of
them were fully functional 10/100 mbps ethernet boards, capable of either rate,
and at full duplex at that. There were three SCSI boards, too. Half a dozen or
so were multi I/O boards with serial, parallel, FDC, and IDE. A couple were VLB
multi-I/O types with a separate ROM so they can serve as additional IDE ports,
not that it will help much, and a couple or three others are VLB video boards.
I'll probably scrounge the VRAM from the latter. I haven't even looked at most
of them, but the rest are just common multi-I/O boards, some with disk I/O and
some without. I also got a bag of shielded external SCSI cables with various
combinations of connectors for about $6 for the dozen or so.
>
> I have to conclude that either you're soldering with an arc-welder or you
> are taking months to change a single chip.
>
It's not months, but electricity isn't that cheap here. My soldering iron uses
butane, by the way, but still, it's much cheaper to replace the entire board.
If I'm bored and want to do some mindless soldering, I'll take and remove all
the 1488's and 1489's from a half-dozen I/O boards and replace them with
sockets. Then I put the 1488's in a tester and subsequently put the good ones
in the sockets. Then I test the 1489's and do likewise. I won't do that,
however, until after I've gotten a new glass catcher tube for my
solder-sucker/soldering iron, as I dropped and broke the %$#_at_! thing. I bought
that station thinking I'd use it a lot, but I generally use the mechanical one
instead. It doesn't take as long since it doesn't have to warm up.
>
> > There was a time when I loved isolating and replacing faulty IC's. Sockets
made
> > it easier, by far, and quick enough that it was worth doing. However, those
> > same sockets made the board more costly and less reliable. The sockets some
>
> For 'static' equipment (i.e. not stuff that's carried about, mounted in
> vehicles, etc), turned pin sockets (machined pin sockets) seem to be
> reliable. I always use them on my prototype boards, not because it makes
> it easier to replace chips, rather that it's easier for me to pull chips
> and force signals high or low when tracing design bugs.
>
Yes, but they cost like the devil, and the pins on the wire-wrap types are too
long to suit me. For soldertails, I like the ones made by Burndy and Augat,
among others, that really pinch the pins against the plastic or against the
other half of the VERY firm opposing contact. They require you to use a
screwdriver to remove the parts, but they don't fall out in transit.
>
> Once I've got the design worked out, I socket (unless there are good
> reasons not to) :
>
The sockets take up considerable space perhaps better used for air flow. That's
a reason not to.
>
> Any I/O buffer chips (they fail easily, it's nice to be able to replace them)
>
> Programmed parts (microcontrollers, EPROMs, etc) (It makes it easier to
> update them)
>
yes, that's true, though the down-wind part adjacent to a socketed part seems to
get less benefit from air movement than it should.
>
> Anything delicate or expensive
>
> I don't generally bother to socket TTL, 4000 series CMOS, common
> microprocessors and RAMs, etc on boards where I know the design is good.
>
> I never use cheap folded-metal-contact sockets. I've had far too many
> problems with those.
>
The lifetime supply of those folded-metal (the gas-tight variety) that I have is
on big memory boards having 288 or 192 memory sockets on each, along with about
two dozen sockets for the schottky drivers for the memory addresses and data.
These were reputed to be extremely reliable error-correcting memory boards with
discrete logic to do the SECDED using parity gates and random logic with the aid
of the host bit-slice processor.
>
> -tony
>
>
Received on Thu Jun 14 2001 - 00:44:20 BST