Maker Pro
Maker Pro

DAC failure mode

R

Robert Latest

Jan 1, 1970
0
Hi folks,

currently I'm looking at a piece of equipment (a PC32 DSP board from
Innovative Integrations) that has an AD669 D/A converter on board. Instead
of the desired sweep from -10V to +10V, I get an odd assortment of linear
pieces that looks like this:

output
^
+10 | / /
| / /
| / /
+5 |/ /
|
|
|
0---------------------> command (0-FFFF)
/ /
/ /
/ /
-5 / /

Or, in words: When the output should be -10V (input 0000), it is -5V and
rises linearly. On the 3FFF->4000 transition, the output jumps back to -5V
and then rises properly to 0. On 7FFF->8000, I get a jump to +5V and a ramp
to +10V, at BFFF->C000 back to +5V, and another ramp to +10V which is
reached at FFFF.

Question: How can such an error arise? Is this a typical DAC failure mode?
Is it possible to cause it through ESD at one (buffered) analog output?
Stupid question of course. Anything can be caused by ESD. I wouldn't have
expected a commercial product to be that susceptible though.

The background is as follows: I've been asked by a colleague to look into a
problem that he seemed to have with some circuit connected to the DAC output
in question. To test that other thing (which turned out to be healthy) I
connected it to the PC32's output, found weird behavior, and then found the
problem on the PC32 itself. Now I'm being accused of having destroyed the
PC32 by touching its outputs. I won't go into the details of the
accompanying issues here, but up to this point I was the only one who didn't
have severe problems with this socially weird fellow, who is now convinced
that I have destroyed his card and expects me to repair it. Personally I
suspect that the card had been shot before and he'd just not noticed it in
his measurements (which isn't all too plausible either). Malevolence on his
part, weirdness aside, can be excluded.

robert
 
J

Jeroen Belleman

Jan 1, 1970
0
Robert said:
Hi folks,

currently I'm looking at a piece of equipment (a PC32 DSP board from
Innovative Integrations) that has an AD669 D/A converter on board. Instead
of the desired sweep from -10V to +10V, I get an odd assortment of linear
pieces that looks like this:

output
^
+10 | / /
| / /
| / /
+5 |/ /
|
|
|
0---------------------> command (0-FFFF)
/ /
/ /
/ /
-5 / /

Or, in words: When the output should be -10V (input 0000), it is -5V and
rises linearly. On the 3FFF->4000 transition, the output jumps back to -5V
and then rises properly to 0. On 7FFF->8000, I get a jump to +5V and a ramp
to +10V, at BFFF->C000 back to +5V, and another ramp to +10V which is
reached at FFFF.

Question: How can such an error arise? [...]

From the looks of it, I'd say the bit with weight 4000 is stuck on. If
that isn't visible at the DAC input bus, then we must conclude that
something inside the chip is shot.

Regards,
Jeroen Belleman
 
W

Winfield Hill

Jan 1, 1970
0
Hi folks,

currently I'm looking at a piece of equipment (a PC32 DSP board from
Innovative Integrations) that has an AD669 D/A converter on board.
Instead of the desired sweep from -10V to +10V, I get an odd
assortment of linear pieces that looks like this:

output
^
+10 | / /
| / /
| / /
+5 |/ /
|
|
|
0---------------------> command (0-FFFF)
/ /
/ /
/ /
-5 / /

It looks like the dac's 2nd to the MSB bit is stuck high
all the time, perhaps a problem on the PCB? At any rate,
that's an issue on the input side of the dac, not on the
output side, so "touching" the card couldn't be a cause.
 
R

Robert Latest

Jan 1, 1970
0
Winfield said:
It looks like the dac's 2nd to the MSB bit is stuck high
all the time, perhaps a problem on the PCB? At any rate,
that's an issue on the input side of the dac, not on the
output side, so "touching" the card couldn't be a cause.

Yeah, looks like it. Now I've got to find an ISA raiser card to look into
that problem, and get my head ripped off again if caught doing it.

robert
 
J

John Larkin

Jan 1, 1970
0
Yeah, looks like it. Now I've got to find an ISA raiser card to look into
that problem, and get my head ripped off again if caught doing it.

robert

Inspect the soldering. Most problems like this turn out to be visible.

John
 
R

Robert Latest

Jan 1, 1970
0
John said:
Inspect the soldering. Most problems like this turn out to be visible.

In fact I did, and it looked promising: The DAC712 (not the AD669, contrary
to the board's datasheet), had obviously been reworked in the past. It was
hand-soldered and came from a different batch than the other two. Its D14
pin (being in a corner of the package) was easily inspectable under a
microscope and looked suspicious, and indeed, I could get a scalpel into
where there should have been solid solder. Rather than taking the chance of
a pad that might have been damaged during the rework, I directly soldered a
wire to the D14 pin of the neighboring DAC712 (both on the same data bus).

Seldom had I been more confident of having found the right fix, so I was
quite disappointed when the problem remained (intermittent, by the way, as
I've found out in the meantime. Sometimes the darn thing works).

Lacking proper logic analysis equipment, I clipped a scope to my wire and
played with the DAC's inputs (unfortunately I can't set them directly but
have to mouse-move a slider on the application that talks to the board). The
D14 line sits frozen at 0V, although it should of course cycle high and low
at times when it should be set high for this particular DAC.

Then I started the part of the application that ramps the other DACs (all of
which work flawlessly). Once this is running, the D14 line starts hopping
merrily between 2.5 and 5V (!!!). When I stop the scanning process, the D14
line sits back at ground. Capacitive coupling through a faulty trace? Why
the three-level thingy, then? Could it be a measurement thing -- because I
don't have an ISA riser board, I couldn't fit a proper scope probe onto the
board but had to use clips and banana-plug cables.

More headache. I'm actually not hell-bent on fixing this; having it repaired
externally is no problem. The problem is the time factor involved in
shipping this thing back and forth overseas, and my piece of wire would have
been just so beautiful.

robert
 
F

Fred Bartoli

Jan 1, 1970
0
Robert Latest a écrit :
In fact I did, and it looked promising: The DAC712 (not the AD669, contrary
to the board's datasheet), had obviously been reworked in the past. It was
hand-soldered and came from a different batch than the other two. Its D14
pin (being in a corner of the package) was easily inspectable under a
microscope and looked suspicious, and indeed, I could get a scalpel into
where there should have been solid solder. Rather than taking the chance of
a pad that might have been damaged during the rework, I directly soldered a
wire to the D14 pin of the neighboring DAC712 (both on the same data bus).

Seldom had I been more confident of having found the right fix, so I was
quite disappointed when the problem remained (intermittent, by the way, as
I've found out in the meantime. Sometimes the darn thing works).

Lacking proper logic analysis equipment, I clipped a scope to my wire and
played with the DAC's inputs (unfortunately I can't set them directly but
have to mouse-move a slider on the application that talks to the board). The
D14 line sits frozen at 0V, although it should of course cycle high and low
at times when it should be set high for this particular DAC.

Then I started the part of the application that ramps the other DACs (all of
which work flawlessly). Once this is running, the D14 line starts hopping
merrily between 2.5 and 5V (!!!). When I stop the scanning process, the D14
line sits back at ground. Capacitive coupling through a faulty trace? Why
the three-level thingy, then? Could it be a measurement thing -- because I
don't have an ISA riser board, I couldn't fit a proper scope probe onto the
board but had to use clips and banana-plug cables.

Not capacitive. Just a dead short between 2 outputs, and the drivers'
RDSon divide according to their status.

Go hunting for another data/signal line showing the same pattern and
you're almost done.
 
J

Joerg

Jan 1, 1970
0
Fred said:
Robert Latest a écrit :


Not capacitive. Just a dead short between 2 outputs, and the drivers'
RDSon divide according to their status.

Or (but more unlikely) something going into tri-state.

Go hunting for another data/signal line showing the same pattern and
you're almost done.

Or trace it back to the respective bus driver chip and see if that looks
healthy. Quite possibly a solder joint has come off at that chip.
 
F

Fred Bartoli

Jan 1, 1970
0
Joerg a écrit :
Or (but more unlikely) something going into tri-state.



Or trace it back to the respective bus driver chip and see if that looks
healthy. Quite possibly a solder joint has come off at that chip.

Tssttt... With a sig line hoping between 2.5V and 5V?
Not impossible. But very unlikely.
 
J

Joerg

Jan 1, 1970
0
Fred said:
Joerg a écrit :


Tssttt... With a sig line hoping between 2.5V and 5V?
Not impossible. But very unlikely.

It depends on what else is hanging on that line. For example, you can
have that very scenario with a thevenin termination. High -> 5V or
whatever VCC is, tri-state -> 2.5V or whatever the thevenin ratio has
been set to.
 
F

Fred Bartoli

Jan 1, 1970
0
Joerg a écrit :
It depends on what else is hanging on that line. For example, you can
have that very scenario with a thevenin termination. High -> 5V or
whatever VCC is, tri-state -> 2.5V or whatever the thevenin ratio has
been set to.

OK, you almost win. Almost, because it's an ISA card and I don't see
Thevenin terms used there :)
 
J

Joerg

Jan 1, 1970
0
Fred said:
Joerg a écrit :


OK, you almost win. Almost, because it's an ISA card and I don't see
Thevenin terms used there :)

Well, sans Thevenin you theory about two outputs fighting each other
looks more likely. That should be quite easily detectable via touching
the chips. When it goes "Ouch!", then that's one of 'em ;-)
 
L

Lionel

Jan 1, 1970
0
Then I started the part of the application that ramps the other DACs (all of
which work flawlessly). Once this is running, the D14 line starts hopping
merrily between 2.5 and 5V (!!!). When I stop the scanning process, the D14
line sits back at ground. Capacitive coupling through a faulty trace?

[I'm assuming that your DAC is a high-density package, with a zillion
pins on 4 sides.]

The first thing I'd look for is a short (likely a solder whisker, from
your description) between the D14 pin & the next pin over. If so, you
should be able to detect it (powered down, for preference) with a
multimeter.
It's a very typical kind of fault on boards with high density chip
packages (eg; QIPs, etc), especially ones that've ben banged around by
service staff, because they pins are so close that a tiny bump can
bend them just enough to touch. I fix those by sliding the point of a
scalpel between the pins to separate them.

Another common one is an invisible dry joint *under* the 'L' part of
the pin, which typicially shows up as an intermittant fault that gets
progressively worse over time, or with heat or vibration. The scalpel
trick will often reveal that problem as well.

The easy repair is to lift the pin with a blade or fine needle,
solder-wick & tin the pad, do the same for the pin, then re-tin the
pin, bend it down, & dab it onto the pad with a *freshly cleaned* fine
iron. Test for shorts between the pin & the ones on either side, & you
should be all fixed.
 
L

Lionel

Jan 1, 1970
0
Tssttt... With a sig line hoping between 2.5V and 5V?
Not impossible. But very unlikely.

The 2.5V will be because it's shorted to another line, & you're
getting 2.5V when one is trying to go high, but the other is trying to
go low. Betcha the chip driving the short are both running hot. ;)
 
R

Robert Latest

Jan 1, 1970
0
Fred said:
Not capacitive. Just a dead short between 2 outputs, and the drivers'
RDSon divide according to their status.

Go hunting for another data/signal line showing the same pattern and
you're almost done.

Right. Sounds reasonable. At the moment I don't have access to the card so
I need to speculate a bit (an interestingly nested problem here):

1. If this short exists, shouldn't all four DACs (or even more parts; I
don't know how many share this particular segment of the data bus) show
weird behavior?

2. The 712's logic threshold is somewhere between 0.8 and 2 volts. Yet it
only works properly while I see all the 2.5-5V stuff going on on the D14
line. When D14 is solid low, the 712 behaves as if it were stuck on high. Of
course a lot could be clarified here if I could just have a quiet sit-down
at the bench, triggering on CS and whatnot. Suffice it to say that this is
not possible at the moment.

3. The fact that this particular 712 had been replaced before (by whom I
don't know -- the card has a somewhat muddy history) suggests that this
problem had been there earlier, and that it was erroneously traced to the
712 when in fact it was shorted outputs or some other cause.

4. All this is moot because the card will be sent out for repair for two
reasons:
a) I'm currently being blamed for having caused the defect (which
is ludicrous because I don't have telekinetic capacities). If I were to
try my own hands at a repair, one of the following three things will
happen:
- I manage to fix it but I won't gain anything because in the eyes of the
beholder I just righted my own wrong as I damn well should.
- If I can't fix it it will still have to be sent out.
- If I break the thing altogether I'll take even more blame.
In short: I can't win but can only lose.
b) I neither have the schematics to this (6-layer) board, nor the proper
measurement outfit nor SMD rework equipment.

But thanks for bearing with me. The shorted-output idea nicely fits the D14
behaviour on the scope, but not the proper operation of the other devices on
the bus.

robert
 
R

Robert Latest

Jan 1, 1970
0
Fred said:
OK, you almost win. Almost, because it's an ISA card and I don't see
Thevenin terms used there :)

This signal isn't on the ISA bus but on the data bus of the TMS320 on the
board. Or at least the part of the bus that feeds the DACs; there may be
drivers in between (I don't have any schematics).

robert
 
R

Robert Latest

Jan 1, 1970
0
Lionel said:
[I'm assuming that your DAC is a high-density package, with a zillion
pins on 4 sides.]

It's a big 28-pin SOIC (1.27mm lead spacing)
Another common one is an invisible dry joint *under* the 'L' part of
the pin, which typicially shows up as an intermittant fault that gets
progressively worse over time, or with heat or vibration.

That's exactly the way in which the D14 pin of that chip looked suspicious.
I fixed it (with a proper piece of wire, not just reflowing that pin) but to
no avail).
The scalpel
trick will often reveal that problem as well.

Yeah, it seemed to. I mentioned it in an earlier post.

robert
 
L

Lionel

Jan 1, 1970
0
That's exactly the way in which the D14 pin of that chip looked suspicious.
I fixed it (with a proper piece of wire, not just reflowing that pin) but to
no avail).

Jumpering it is not that good a way to fix that kind of fault. You'll
be more likely to get a reliable repair via the method I described (or
any other method that ensures that the pin is soldered back where it
should be).
Yeah, it seemed to. I mentioned it in an earlier post.

Seriously, ditch the jumper & redo the pin properly. You might be
surprised what you find under it when you lift the pin under a
magnifier & a really bright light.
 
L

Lionel

Jan 1, 1970
0
4. All this is moot because the card will be sent out for repair for two
reasons:
a) I'm currently being blamed for having caused the defect (which
is ludicrous because I don't have telekinetic capacities).

Oh boy, I can really relate to your situation. I've recently parted
company with an obnoxious & rabidly clueless client who repeatedly
blamed his network (that I built) for faults that he'd caused himself.
His most recent self+foot+gun disaster was when he crashed the servers
& switches by scheduling builders to perform structural work in his
office - in the middle of the working day - without informing me, or
taking any precautions, or at least shutting down all the electronics.
He then put the cherry on top by letting the builders plug their saws,
grinders & **ARC WELDER** into the same power circuit supplying the
server, etc. Needless to say, the moron blamed the failure on me.
If I were to
try my own hands at a repair, one of the following three things will
happen:
- I manage to fix it but I won't gain anything because in the eyes of the
beholder I just righted my own wrong as I damn well should.
- If I can't fix it it will still have to be sent out.
- If I break the thing altogether I'll take even more blame.
In short: I can't win but can only lose.
b) I neither have the schematics to this (6-layer) board, nor the proper
measurement outfit nor SMD rework equipment.


Sorry Robert, I hadn't seen this part of the thread when I wrote my
posts. Obviously, it makes more sense for you to wash your hands of
the whole thing. ;)
 
R

Robert Latest

Jan 1, 1970
0
Lionel said:
Sorry Robert, I hadn't seen this part of the thread when I wrote my
posts. Obviously, it makes more sense for you to wash your hands of
the whole thing. ;)

Never mind. I learned a great deal in this thread. Thanks, everybody, for
contributing.

Now I have a little electronics problem that's more domestic. See another
thread.

robert
 
Top