18 Mar
2005
18 Mar
'05
3:42 p.m.
Hi All,
after a few days frustration, I have got a little further with my 11/45.
The power supplies checked out OK with respect to ripple and output, but a
couple have them have the odd VERY fast spike on them - is this normal? (I'm
using a 100MHz 'scope - I suspect they may not be visible on a slower one).
After reducing the machine to its basics, I found I could load an address
from the front panel, but not write or examine any
location successfully.
The address and data lines became active, but the processor
then goes into a
pause state. I found that the MSYN line was not being asserted, and it
appears that one ic on the Unibus controller (M8106) is not providing the
correct ouput - grounding the scope probe will force the line, and the
processor completes its bus cycle (Yes the inputs to the ic are correct!).
I have found a spare serviceable UBC, so I'll try that when I next get
chance, and if that fixexs the fault, I'll repair mine.
Jim.
Please see our website the " Vintage Communication Pages" at WWW.G1JBG.CO.UK
18 Mar
18 Mar
4:16 p.m.
Nice detective work Jim. Unless your "spare" UBC is known good and has
the same revision as your defective card, I would suggest replacing the
IC on your card and proceeding with the diagnosis. Otherwise, you may
just replace one set of symptoms with another.
--tom
At 08:42 PM 3/18/2005 +0000, you wrote:
Hi All,
after a few days frustration, I have got a little further with my 11/45.
The power supplies checked out OK with respect to ripple and output, but a
couple have them have the odd VERY fast spike on them - is this normal? (I'm
using a 100MHz 'scope - I suspect they may not be visible on a slower one).
After reducing the machine to its basics, I found I could load an address
from the front panel, but not write or examine any location successfully.
The address and data lines became active, but the processor then goes into a
pause state. I found that the MSYN line was not being asserted, and it
appears that one ic on the Unibus controller (M8106) is not providing the
correct ouput - grounding the scope probe will force the line, and the
processor completes its bus cycle (Yes the inputs to the ic are correct!).
I have found a spare serviceable UBC, so I'll try that when I next get
chance, and if that fixexs the fault, I'll repair mine.
Jim.
Please see our website the " Vintage Communication Pages" at WWW.G1JBG.CO.UK
4:38 p.m.
On Mar 18, 2005, at 1:42 PM, Jim Beacon wrote:
Hi All,
after a few days frustration, I have got a little further with my
11/45.
The power supplies checked out OK with respect to ripple and output,
but a
couple have them have the odd VERY fast spike on them - is this
normal? (I'm
using a 100MHz 'scope - I suspect they may not be visible on a slower
one).
After reducing the machine to its basics, I found I could load an
address
from the front panel, but not write or examine any location
successfully.
The address and data lines became active, but the processor then goes
into a
pause state. I found that the MSYN line was not being asserted, and it
appears that one ic on the Unibus controller (M8106) is not providing
the
correct ouput - grounding the scope probe will force the line, and the
processor completes its bus cycle (Yes the inputs to the ic are
correct!).
I have found a spare serviceable UBC, so I'll try that when I next get
chance, and if that fixexs the fault, I'll repair mine.
Jim.
Please see our website the " Vintage Communication Pages" at
WWW.G1JBG.CO.UK
If that fast spike looks like:
5v _|___|____
| |
| |
0v ________
... or similar, this can (will) cause intermittent problems when you
load the system
with usage like installation, compiles, etc. Especially when the spikes
bottom out at
less than 2.5v. We saw this cause very "odd" system crashes on 11/70's.
Components on
the h744 regulators started to degrade, `cause DEC pushed these to the
limit.
We replaced the two output electrolytic caps, the output (power)
transistor, and the
two disc caps that should be at the mate-n-lock that should go from 5v
to ground.
(These are an ECO from the late `70's.) If the spikes persisted, we
also replaced the
remaining components on the heat sink, the big filter cap, and the
bridge rectifier. This
procedure came from several months of test and experimentation. In the
field, we (AT&T
self maint. org.) generally replaced h744's with h7441's.
(I can't believe I can remember this! But then again, I can remember
the details of RP06
maintenance, too. Fun times!)
Mike
4:53 p.m.
Mike Cesari said:
If that fast spike looks like:
5v _|___|____
| |
| |
0v ________
... or similar, this can (will) cause intermittent problems when you
load the system
with usage like installation, compiles, etc. Especially when the spikes
bottom out at
less than 2.5v. We saw this cause very "odd" system crashes on 11/70's.
Components on
the h744 regulators started to degrade, `cause DEC pushed these to the
limit.
We replaced the two output electrolytic caps, the output (power)
transistor, and the
two disc caps that should be at the mate-n-lock that should go from 5v
to ground.
(These are an ECO from the late `70's.)
etc.....
What does ECO mean?
Ashley
5:32 p.m.
On Fri, 18 Mar 2005, Ashley Carder wrote:
What does ECO mean?
Engineering Change Order - used to make a modification to a device that's
already in production.... cut and lift U20/pin 6, run a wire from R51 to
U33/pin4, add a 100K 1/8W resistor from the junction of Q7/E to the + side
of C141... stuff like that.
Cheers
John
5:42 p.m.
On Fri, 2005-03-18 at 16:53 -0500, Ashley Carder wrote:
What does ECO mean?
ECO = Engineering Change Order
It basically means that after the product was released, there were
changes that need to be applied and ECO's are used to track the changes.
Each ECO is numbered (in some way). ECO's can be fairly trivial such as
changing some writing on a label, others can be fairly involved.
If you're a software type, think "patch".
When you see a system, it may have tags that indicate what ECO's have
been applied. Some ECO's can be applied in the field, some have to be
done at the factory. There is usually a fair amount of documentation
surrounding ECO's.
--
TTFN - Guy
7:50 p.m.
ECO = Engineering Change Order
It basically means that after the product was released, there were
changes that need to be applied and ECO's are used to track the changes.
Each ECO is numbered (in some way). ECO's can be fairly trivial such as
changing some writing on a label, others can be fairly involved.
If you're a software type, think "patch".
When you see a system, it may have tags that indicate what ECO's have
been applied. Some ECO's can be applied in the field, some have to be
done at the factory. There is usually a fair amount of documentation
surrounding ECO's.
Thanks for the explanation.
A former DEC employee told me the other day that the RK05 was
DEC's "most ECO'd" product.
Ashley
19 Mar
19 Mar
4:37 p.m.
On Fri, 18 Mar 2005 16:53:44 -0500
"Ashley Carder" <wacarder at usit.net> wrote:
What does ECO mean?
Ashley
It can also mean Expensive Change Order, i.e. you've got 500 heart
monitors out there in use at 247 clinics worldwide, and a shock hazard
is identified that requires a retrofit.
It almost always involves sucky meetings, and staff being pulled off
worthwhile projects to fix old stuff.
And prodigeous amounts of paperwork. (I used to joke about initiating
an ECO that just incrememts all part-numbers company-wide by one).
8:30 p.m.
Scott wrote:
I used to joke about initiating
an ECO that just incrememts all part-numbers company-wide by one.
I've seen cases where releasing a new firmware revision requires nearly
fifty new part numbers assigned, and over five hundred signatures on
all the resulting engineering documents (including but not limited
to ECOs).
The firmware release has to be signed off, which might take five
signatures. Suppose there are eight EPROMs involved. Each one gets a
new part number assigned for the new revision, and each of those has an
ECO that needs seven signature. The EPROMs go on a daughterboard, so
the BOM and parts placement diagrams have to be revised, each of which
needs another seven signatures. The motherboard into which the
daughterboard plugs in needs its parts list and BOM revised, for another
seven. The parent card cage assembly needs its parts list and BOM
revised, for another seven. The chassis parts list and BOM are revised...
the system parts list and BOM is revised... the ship list is revised...
the customer documentation is revised... and I think I missed a bunch of
other things in the chain as well. Oh yeah, there was paperwork for the
part number assignment for the new revisions of all the subassemblies
and assemblies. And updated test specifications and procedures.
And that's for a product that was not for use in medical or life-
critical applications, and that didn't need any certifications other
than the usual UL and FCC Part 15.
It turned out that there was *much* less paperwork involved in releasing
a software update that shipped as a separate "product", to be installed
by the customer or an FE.
At one time I thought that perhaps a higher level assembly parts list
and BOM should refer to the "most recent revision" of a subassembly,
but after a little more thought I realized that doing it that way
would be disasterous because you'd completely lose traceability.
Eric
18 Mar
18 Mar
5:18 p.m.
From: "Mike Cesari" <mcesari at comcast.net>
5v _|___|____
| |
| |
0v ________
... or similar, this can (will) cause intermittent problems when you
load the system
with usage like installation, compiles, etc. Especially when the spikes
bottom out at
less than 2.5v. We saw this cause very "odd" system crashes on 11/70's.
Mike,
thats the little beastie!
I figured there must be a couple of ceramic caps to catch them somewhere.
I'll dig around and see if I can change them.
Thanks
Jim.
4:45 p.m.
Jim wrote:
after a few days frustration, I have got a little
further with my 11/45.
The power supplies checked out OK with respect to ripple and output, but a
couple have them have the odd VERY fast spike on them - is this normal?
(I'm using a 100MHz 'scope - I suspect they may not be visible on a
They are switching power supplies, so there will be some high frequency
noise. I'm not sure what the typical switching frequency is, though I'd
guess that it would be in the 100-200 KHz range; maybe Tony Duell knows.
Anyhow, one would expect some very high frequency spikes occurring at
the interval of the switching frequency. But the spikes shouldn't have
very high amplitude; if they do, it's possible that you have some bad
filter capacitors.
I'm out of my element on analog stuff, so hopefully someone else can
help you more.
Eric
5:46 p.m.
Eric wrote:
Jim wrote:
The switching frequency of these supplies is in the neighborhood of 10KHz,
which I'm guessing at based on my memory of the last time I listened to
one whine/whistle.
Are you sure that you don't have a ground loop in your scope setup? Have
you isolated your ground plug wire on your scope? What is the amplitude
of the spike?
--tom
after a few days frustration, I have got a little
further with my 11/45.
The power supplies checked out OK with respect to ripple and output, but a
couple have them have the odd VERY fast spike on them - is this normal?
(I'm using a 100MHz 'scope - I suspect they may not be visible on a
They are switching power supplies, so there will be some high frequency
noise. I'm not sure what the typical switching frequency is, though I'd
guess that it would be in the 100-200 KHz range; maybe Tony Duell knows.
Anyhow, one would expect some very high frequency spikes occurring at
the interval of the switching frequency. But the spikes shouldn't have
very high amplitude; if they do, it's possible that you have some bad
filter capacitors.
6:30 p.m.
From: "Tom Uban" <uban at ubanproductions.com>>
The switching frequency of these supplies is in the
neighborhood of 10KHz,
which I'm guessing at based on my memory of the last time I listened to
one whine/whistle.
Are you sure that you don't have a ground loop in your scope setup? Have
you isolated your ground plug wire on your scope? What is the amplitude
of the spike?
--tom
Tom,
it's not ground loop type noise - it is not present on all of the 5V
supplies. I suspect that a couple of the supplies are faulty, but I wasn't
sure if the spikes were a normal function of loading the supplies. I found
that the lightly loaded supplies have more ripple than the heavily loaded
units, and wondered if the spikes were apparent only under certain loading
conditions, but from Mike's reply, it would seem that it is a known fault
condition.
Jim.
7:02 p.m.
Jim wrote:
I don't, but I think it's lower than that. 50kHz?
after a few days frustration, I have got a little
further with my 11/45.
The power supplies checked out OK with respect to ripple and output, but a
couple have them have the odd VERY fast spike on them - is this normal?
(I'm using a 100MHz 'scope - I suspect they may not be visible on a
They are switching power supplies, so there will be some high frequency
noise. I'm not sure what the typical switching frequency is, though I'd
guess that it would be in the 100-200 KHz range; maybe Tony Duell knows. Anyhow, one would expect some very high frequency
spikes occurring at
the interval of the switching frequency. But the spikes shouldn't have
very high amplitude; if they do, it's possible that you have some bad
filter capacitors.
I would suspsect the output filter capacitors. If those go high ESR, yuo
will get spikes on the output, maybe even enough to trip the crowbar.
Oh, and WRT the UDC card. IIRC, there are at least 2 revisions, and
there's a backplane mod you have to do (adding another clock signal) if
you replace an early one with a late one. Personally, I'd stick with the
card I know has worked in that machine, and change the defective chip.
-tony
7059
days inactive
7061
days old
14 comments
10 participants
participants (10)
-
ard＠p850ug1.demon.co.uk -
chenmel＠earthlink.net -
eric＠brouhaha.com -
ggs＠shiresoft.com -
jfoust＠threedee.com -
jim＠g1jbg.co.uk -
jpl15＠panix.com -
mcesari＠comcast.net -
uban＠ubanproductions.com -
wacarder＠usit.net