5 Oct
2005
5 Oct
'05
7:15 p.m.
Hi,
I have a control panel from a Univac mainframe that I hope to re-animate
someday. Unfortunately I know very little about Univac mainframes.
First I'll go for broke: does anybody recognize this unit and have
schematics for it?
http://www.saccade.com/writing/projects/UnivacPanel/UnivacPanel.html
No? OK, time for plan B, reverse engineering it. Most of the chips on the
panel are straightforward 74xxx TTL, however, a number of them appear to
have a seven digit part number instead of a regular 74xxx stamp. Is there a
translation guide between this seven digit number and regular 74xxx
numbers? Most of the 74xxx parts on the board also have these seven digit
numbers.
Any tips on how to guess the actual function would be most appreciated.
Thanks,
John Peterson
www.saccade.com
6 Oct
6 Oct
11:18 a.m.
No? OK, time for plan B, reverse engineering it.
Most of the chips on the
panel are straightforward 74xxx TTL, however, a number of them appear to
have a seven digit part number instead of a regular 74xxx stamp. Is there a
A number of manufacturers used house-coded ICs like this. HP are the
best-known (the 1820-xxxx numbers), but ICL, Xerox and I believe IBM did
it too. The main reason for this, apparently was that any IC with that
house code would work in the circuit. So there might be one number for a
generic 7493 and another for a specifically TI 7493 (I think that's the
chip were the TI one has a slight difference to others). TO have the
house code simplified manufacturing.
translation guide between this seven digit number and
regular 74xxx
numbers? Most of the 74xxx parts on the board also have these seven digit
numbers.
Well, with ICs marked with both, you can start to make the equivalents
list. If you are very lucky that will identify a few more ICs on the board.
Any tips on how to guess the actual function would be most appreciated.
Well, you can find the power connections, right? If you are lucky, there
will be a few ICs with odd power connections (i.e. not the corner pins)
which will be a start.
Then you might be abloe to identify some pins as definite inputs (driven
from totem-pole outputs on known ICs), others as
definite outputs (only
go to known inputs).
If all else fails you may have to desoler the IC, power it up on a
breadboard, find outputs with a logic probe, try the inputs, and so on.
-tony
12:05 p.m.
At 04:18 PM 10/6/05 +0100, you wrote:
> No? OK, time for plan B, reverse engineering it. Most of the chips on
the
> panel are straightforward 74xxx TTL, however, a
number of them appear to
> have a seven digit part number instead of a regular 74xxx stamp. Is
there a
A number of manufacturers used house-coded ICs like this. HP are the
best-known (the 1820-xxxx numbers), but ICL, Xerox and I believe IBM did
it too. The main reason for this, apparently was that any IC with that
house code would work in the circuit. So there might be one number for a
generic 7493 and another for a specifically TI 7493 (I think that's the
chip were the TI one has a slight difference to others). TO have the
house code simplified manufacturing.
If they're TTL or CMOS SSI ICs I have an IC tester that will id unknown
ICs. It's not 100% perfect since a lot of ICs have very similar functions
but it'll get you close enough that you can figure out the pinout and it's
basic functions.
translation guide between this seven digit number
and regular 74xxx
numbers? Most of the 74xxx parts on the board also have these seven digit
numbers.
Well, with ICs marked with both, you can start to make the equivalents
list. If you are very lucky that will identify a few more ICs on the board.
>
> Any tips on how to guess the actual function would be most appreciated.
Well, you can find the power connections, right? If you are lucky, there
will be a few ICs with odd power connections (i.e. not the corner pins)
which will be a start.
Then you might be abloe to identify some pins as definite inputs (driven
from totem-pole outputs on known ICs), others as definite outputs (only
go to known inputs).
If all else fails you may have to desoler the IC, power it up on a
breadboard, find outputs with a logic probe, try the inputs, and so on.
I used to have an article out of one of the electronics magazines that
told how to identify various pins on ICs using an ohm meter. Input pins,
output pins and power and ground pins all have different characteristics.
Unfortunatly I lost the article. Does anyone have any info on how to go
about doing this?
But all in all I like Tony's Suggestion of looking for the parts with
dual markings and make a cross reference list. When you get done, post it
on the net to benefit everyone. Spere has a good start on HP and Tektronix
cross reference on their website.
Joe
-tony
2:46 p.m.
Then you might
be abloe to identify some pins as definite inputs (driven
from totem-pole outputs on known ICs), others as definite outputs (only
go to known inputs).
If all else fails you may have to desoler the IC, power it up on a
breadboard, find outputs with a logic probe, try the inputs, and so on.
I used to have an article out of one of the electronics magazines that
told how to identify various pins on ICs using an ohm meter. Input pins,
output pins and power and ground pins all have different characteristics. Unfortunatly I lost the article. Does anyone have any
info on how to go
about doing this?
How about starting with a known IC of the family you're interested in and
recording the results for your meter between various combinations of the
pins?
But all in all I like Tony's Suggestion of looking for the parts with
dual markings and make a cross reference list. When you get done, post it
on the net to benefit everyone. Spere has a good start on HP and Tektronix
cross reference on their website.
I know for a fact that some of the IC's on that HP list were indetified
by finding dual-marked ones on HP boards. And others were indentified by
finding inputs and outputs and indentifying the functions. I know this,
because I did it :-)
There was a good HP equivalent list published _by HP_ in Bench Briefs at
one point. It was half a dozen pages, 3 columns, fairly small font. I
haev it, but I know it wouldn't stand scanning or copying, and I don't
feel like typing it all in.
-tony
3:14 p.m.
At 07:46 PM 10/6/05 +0100, you wrote:
Be warned that this sort of test depends critically on the particular
logic family, and the type of ohmmeter in use.
As you hace the chip on a circuit board, you can easily find pins that
are tied to +5V or ground (you _must_ be able to find those 2 rails on
the board, if nothing else by starting from a chip you know the
number and pinout of). OK, there may be pins other than the Vcc and Gnd
pins that are connected to those points on the unknown chip (e.g. unused
inputs), but I think you should be able to indentify the actual power and
ground pins fairly easily.
They also got a good number of the crosses from instrument service
manuals. Unlike the calculator manuals, the instrument manuals usually gave
the HP PN, manufacturer code and manufacturer's PN. I supplied a lot of the
instrument part number listings to Sphere also some of the numbers from the
dual marked parts.
Then you might be abloe to identify some pins as
definite inputs (driven
from totem-pole outputs on known ICs), others as definite outputs (only
go to known inputs).
If all else fails you may have to desoler the IC, power it up on a
breadboard, find outputs with a logic probe, try the inputs, and so on.
I used to have an article out of one of the electronics magazines that
told how to identify various pins on ICs using an ohm meter. Input pins,
output pins and power and ground pins all have different characteristics. Unfortunatly I lost the article. Does anyone have
any info on how to go
about doing this?
How about starting with a known IC of the family you're interested in and
recording the results for your meter between various combinations of the
pins?
But all in all I like Tony's Suggestion of looking for the parts with
dual markings and make a cross reference list. When you get done, post it
on the net to benefit everyone. Spere has a good start on HP and Tektronix
cross reference on their website.
I know for a fact that some of the IC's on that HP list were indetified
by finding dual-marked ones on HP boards. And others were indentified by
finding inputs and outputs and indentifying the functions. I know this,
because I did it :-)
There was a good HP equivalent list published _by HP_ in Bench Briefs at
one point.
Bench Briefs published a series of issues with cross references. I don't
remember if each one carried a different range of PNs or if they broke them
up by equipmnet useage or what but I know the cross reference was spread
over a good number of issues. I had a fair number of BBs but never had the
whole cross reference.
Joe
It was half a dozen pages, 3 columns, fairly small font. I
haev it, but I know it wouldn't stand scanning or
copying, and I don't
feel like typing it all in.
-tony
12:32 p.m.
Don't rely on CMB to have the datasheets for those mystery chips, even if they seem to
say that they do. Often what you'll get is a datasheet that says "No data
available".
You know, there's another (piece) of a Univac panel on eBay:
http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=8702180075
I speculate that someone's replaced the indicator lamp covers to read 1-2-4, instead
of the more logical 4-2-1 order.
Cheers,
Chuck
2:26 p.m.
You know, there's another (piece) of a Univac
panel on eBay:
I suspect that is actually a manufacturing jig of some sort, as the card
machines on 9000 series machines do not have panels (they are integrated
on the main CPU panel). STill it is real Univac - the lights are
distictive.
William Donzelli
aw288 at osfn.org
5:12 p.m.
(One possibility would be to look on bitsavers for UNIVAC manual/schematics
from the same period, in the hopes it might list the
part numbers you're
looking for.)
(And while it doesn't help much in this case with the ICs soldered in,
house-numbered ICs sometimes have the 'real' part number stamped on the
underside of the IC, sometimes with an apostrophe replacing the prefix, e.g. a
7404 might have on the underside: '04 , along with some other gobbledy-gook
letters.)
If you intend to reverse engineer it I would suggest proceeding to do so, even
without knowing what the unknown ICs are.
Leave the unknown ICs as black boxes in your schematic. There is a high
probabilty you can figure out the function of the unknowns with a little
analysis of the schematic. For the majority of circuits, inputs and outputs can
be identified with the application of a little logic such as:
- if unknown-pin is connected only to input pins of known devices,
then unknown-pin must be an output.
- if unknown-pin is connected (perhaps with inputs on other devices)
to an output pin of a known device,
then unknown-pin must be an input.
(The general rule being that one-and-only-one output can drive one-or-more inputs.)
Once the inputs and outputs have been identified a little more observation of
the schematic and the functionality may well be discernable. Of course, keep a
TTL (in this case) reference manual at hand to help correlate pinouts. Some TI
manuals (e.g. TI TTL Data Book/2nd Ed./1981) tend to be good for this because they
have all the pinouts collected within a few pages.
Yes, there are always exceptions to the above such as open-collector or
tri-state circuits, paralleled buffers, some devices may not be 7400 series:
perhaps specialised IC or 75000 series interface IC, etc., but considering the
nature of what you are dealing with (a front panel, in which you already have a
lot of known inputs & outputs (switch & lamps), along with known ICs), I'd
say
your chances of figuring everything out are good.
Sometimes one can be mislead when a connection to a pin is actually just a
'way' for the trace, but considering the fineness of the PCB traces, that may
be less likely to be an issue in this instance that it is on earlier boards.
It all makes for a rather fun puzzle actually, once you get past the tedium of
tracing traces.
(Qualification for making such a suggestion: Not to toot my own horn too much,
but just so you know I'm not blowing wind and suggesting 'Oh, it could in
principle be done like this', I have reverse engineered entire machines (mostly
early calculators) containing up to 250 SSI/MSI ICs, and in which all the
ICs were black-boxes (no info or cross-reference available), and managed to
produce the complete logic schematic, followed by successfully building a
logic-gate-level simulation of the machine, which, for the most part, 'proves'
the schematic.) (e.g. http://www.cs.ubc.ca/~hilpert/eec/ics/M5300.html)
"J. Peterson" wrote:
Hi,
I have a control panel from a Univac mainframe that I hope to re-animate
someday. Unfortunately I know very little about Univac mainframes.
First I'll go for broke: does anybody recognize this unit and have
schematics for it?
http://www.saccade.com/writing/projects/UnivacPanel/UnivacPanel.html
No? OK, time for plan B, reverse engineering it. Most of the chips on the
panel are straightforward 74xxx TTL, however, a number of them appear to
have a seven digit part number instead of a regular 74xxx stamp. Is there a
translation guide between this seven digit number and regular 74xxx
numbers? Most of the 74xxx parts on the board also have these seven digit
numbers.
Any tips on how to guess the actual function would be most appreciated.
Thanks,
John Peterson
www.saccade.com
3:14 p.m.
(One possibility would be to look on bitsavers for UNIVAC manual/schematics
from the same period, in the hopes it might list the part numbers you're
looking for.)
HP were often particularly kind about this, their manuals, if they
contained full parts lists, did give the equivalents. Alas most HP
desktop computer and calculator manuals are boardswapper guides, but...
(And while it doesn't help much in this case with the ICs soldered in,
house-numbered ICs sometimes have the 'real' part number stamped on the
underside of the IC, sometimes with an apostrophe replacing the prefix, e.g. a
7404 might have on the underside: '04 , along with some other gobbledy-gook
letters.)
If all else fails, I've even been known to desolder the IC to firstly
look at the underside, secondly to look at the PCB (I once found the
normal uumber in the silk-screen under the IC -- this was in a device
where all the IC numbers had been sanded off to keep me guessing for at
least 5 minutes!), and thirdly to test the IC on a breadboard.
Once the inputs and outputs have been identified a
little more observation of
the schematic and the functionality may well be discernable. Of course, keep a
TTL (in this case) reference manual at hand to help correlate pinouts. Some TI
Even if like me, you've got the common TTL pinouts hard-wired into your
brain :-)
It all makes for a rather fun puzzle actually, once
you get past the tedium of
tracing traces.
Having done this to a fair few devices, I would agree with you.
(Qualification for making such a suggestion: Not to toot my own horn too much,
but just so you know I'm not blowing wind and suggesting 'Oh, it could in
principle be done like this', I have reverse engineered entire machines (mostly
early calculators) containing up to 250 SSI/MSI ICs, and in which all the
The most complicated calcualtor I ever did only contained 8 ICs. And
those only had 8 pins each. What made it interesting were the several
hundred transistors, the thousand-or-so diodes, core memory,
core-on-a-rope ROM, and a PCB ROM. No, I haven't extracted the contents
of the ROMs yet.... (HP9100B if you're wondering).
Fortunately, for the later HP desktops, I had the equivalents list, so
most of the non-custom ICs were known. Figuring out details of the custom
parts was one interesting bit. As, of course, was working out how the
whole machine was supposed to work.
-tony
9 Oct
9 Oct
2:08 a.m.
appear to have a seven digit part number instead of a
regular 74xxx
stamp. Is there a translation guide between this seven digit number and
regular 74xxx numbers? Most of the 74xxx parts on the board also have
these seven digit numbers.
Years ago (1978), I picked up a chip cross reference print from
Honeywell Surplus in Phoenix AZ. Looks like the numbers just might cross
correctly for you. The list can be downloaded from my archive at:
http://www.bytecollector.com/archive/info/
Enjoy!
-Bryan
6866
days inactive
6870
days old
9 comments
7 participants
participants (7)
-
ard＠p850ug1.demon.co.uk -
aw288＠osfn.org -
cclist＠sydex.com -
hilpert＠cs.ubc.ca -
oldcomp＠cox.net -
pdp11＠saccade.com -
rigdonj＠cfl.rr.com