Re: What's the hi-side current sense chip du jour?

On May 29, 8:30 pm, John Larkin
<jjlar...@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
On Fri, 29 May 2009 20:12:20 -0700 (PDT), "m...@xxxxxxxxx"



<m...@xxxxxxxxx> wrote:
On May 29, 6:54 pm, John Larkin
<jjlar...@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
On Fri, 29 May 2009 18:42:21 -0700 (PDT), "m...@xxxxxxxxx"

<m...@xxxxxxxxx> wrote:
On May 29, 12:05 pm, Joerg <notthisjoerg...@xxxxxxxxxxxxxxxxxxxxx>
wrote:
m...@xxxxxxxxx wrote:
On May 29, 9:26 am, Joerg <notthisjoerg...@xxxxxxxxxxxxxxxxxxxxx>
wrote:
m...@xxxxxxxxx wrote:
On May 28, 8:57 am, Joerg <notthisjoerg...@xxxxxxxxxxxxxxxxxxxxx>
wrote:
MooseFET wrote:
On May 27, 8:40 am, Joerg <notthisjoerg...@xxxxxxxxxxxxxxxxxxxxx>
wrote:
MooseFET wrote:
On May 26, 9:01 am, Joerg <notthisjoerg...@xxxxxxxxxxxxxxxxxxxxx>
wrote:
MooseFET wrote:
[...]
That's a bit iffy. Burp mode is no problem and I can get nearly any chip
to do that even if that is not a native mode. But the sensing should be
a bit more precise.
The one I did wasn't "burp mode" because the switching speed slowed
down smoothly.  The frequency just fell and the on time stayed at the
minimum
I usually try to do that as well, less audio noise. People don't like it
if stuff hisses ;-)
They like a high pitch whine even worse.
Or the occasional whiff of high amperage smell ;-)
[....]
I wouldn't use a Maxim part either.  I designed in one of their parts
many years ago.  I had to design it back out.
Probably because of the typical unobtanium effect :-)
Since more than 20 years "design-out" is all I ever did with Maxim
parts. Mostly because their production planning seems skewed or hardly
existing. OTOH that has brought me some design work ;-)
Their switcher driver chips (I forget the number) also had a problem
when the gate of the MOSFET yanked the gate drive pin below ground.  A
little capacitively coupled glitch and the chip lost its mind for a
few microseconds.
Whoops, a wimpy substrate structure?
I am currently wrestling with "unorthodoxies" in LTSpice. Gate drive
remaining at 1.8V for 50nsec after FET turn-off, causing humongous
virtual losses. It's a totally flat line, looks like a model fault or
something and kind of thwarts any meaningful efficiency analysis.
--
Regards, Joerg
http://www.analogconsultants.com/
"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
While it is possible the Maxim part in question is defective, I assure
you every design is benched and flogged by applications as well as
design. This is true at any semi I worked at. Nobody wants a hardware
recall.
More likely than not, the customer PCB is not designed well. [AKA
pilot error.] Most companies sell (OK, give away if you look serious)
demo boards with functioning circuits.
It is probably no surprise that most switchers are not tested in
circuit on the handler. The inductance of the contactor makes it hard
to run a switcher. [Charge pumps don't have such problems, so they are
tested in the actual circuit.] There are usually test modes that allow
the test engineer to measure the power fet on resistance, comparator
trip points, etc.
I have tested a few customer boards over the years. In every situation
but one, it was a component error, component defect, or layout
problem. [OK, one was a bug on my part, but the part was being used in
a funny manner nobody anticipated.]
So far I can't remember a Maxim part ever being flawed (can't say the
same about some other mfgs). I believe the engineers at Maxim are
actually quite good. The problem seems to be more in the upper
management. 100% of my design-out cases happened because Maxim was
unable to deliver quantities in a timely and predictable manner. If that
issue hasn't been recognized by upper management in decades I am not
sure it ever will be :-(

It's quite similar with some (large) European companies. Excellent
engineering combined with serioulsy flawed S&M departments. Consequently
I avoid their parts if possible.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

What really happens is the small companies don't get serviced well.
You buy a few hundred K of parts, you get serviced. You buy a few
hundred parts, you get samples but not follow on. ...

That's a huge mistake. I only buy a handful but the design decisions I
make can have six-digit or more impact on the bottomline. Companies like
AD or LTC understand this, others do not. And they lose out, big time.

                                             ... I haven't been
there in a decade, so I don't know how its run now, but many of the
same fools are still in management

Just MHO: It's time for major personnel changes up top. But I am not the
boss, don't own stock in it, never will.

My point really was regarding parts out the chute not working, ans the
other poster claimed. Now every semi buys and benches the competitions
parts. Internally, they have reports of their warts, so nobody gets
blindsided. Every part has some issue, but more or less meets the
function and certainly passes the electricals in the data***.

Looks like John had a few. I never bothered to look much because the
client requests were usually "We can't get these, could you find a
solution?". The last encounter with chips from there (existing design
thrown over my fence) was 10 days ago. Designed them all out.

DC/DC chips all have one critical node, namely the pin connected to
the external inductor. Make the trace to long, it rings, radiates, and
things get ugly. Some chips (OK, my designs) use a one shot designed
to blank out any "decisions" while the LX pin moves. Who knows if the
reference is wiggling, current is coupled capacitively into inputs,
etc. Blanking decisions for a short period of time won't effect the
control if you set the timing right. This is one of those things that
doesn't show up in the data*** block diagram.

I know you like to roll your own, but a good DC/DC is full of
protection circuitry you never see. I've gone as far as adding
watchdog timers to chips so if the idiot customer shorts out pins that
effect control, the part can reset itself.

True. One has to know all the possible stresses a circuit could
encounter. One difference to John's circuit is that mine were usually
current-mode controlled. That takes care of the major phhhst ...
*KABLAM* scenarios.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

It may be a matter of logistics, but if you claim a part doesn't work,
the company wants to hear about it. It might be a way to get a free
eval board (hint hint).

When the MAX9690s started to fail, Maxim sampled me 3500 pieces of the
9691. But replacing 3500 units in the field was sort of a hassle. We
had to make a baby-board adapter, too, since the 9691 isn't a drop-in
replacement.

ftp://jjlarkin.lmi.net/Break2.jpg

ftp://jjlarkin.lmi.net/OnBoard.jpg

John

What exactly failed on the part? Was it a packaging problem?

Maxim subbed the fab to some shop in Minnesota that apparently had a
dirty oxide problem or something. Over time, some carriers would
diffuse somewhere and pinch off a resistor, at which time the
comparator would get erratic. The parts failed at high temperature,
and over time - like a year maybe - the "fail" temp would creep down
towards room temp. The weird thing was that you could bake the parts
and raise that "fail" temp back up for a while. All this was very
confusing.

They discontinued the part without warning, in fact when shipments
were long overdue. They then announced their "drop-in replacement"
- the MAX9691 - which is the only comparator I've ever heard of that
has BACK-TO-BACK DIODES BETWEEN ITS INPUTS. Imagine what that did to a
capacitor-linear-ramp/dac time delay trimmer circuit.

grrrrrr.

John

I can see this happening. My original comment was about parts being
non-functional right out the chute.

But the 9690 was a bipolar part. Hard to believe oxide has that
effect.

I've only used bipolar inputs in bicmos for ASICs. The clamping under
those conditional is not that unusual, then again, I'm setting up the
operating conditions. If you yank a differential pair far apart, two
ugly things can happen. One could be a polarity reversal in the
comparator, which depends on a lot of things. The other more sneaky
problem is the offset of the differential pair shifts. The whole
reason you went through the pain of using bipolar inputs (speaking
from a BICMOS point of view) is the low untrimmed offset off the
differential pair.

That was a Gain Technology design. Those guys spent way too much time
in the Arizona sun and fried their brains. While you end users may
think you are the judge of component quality, your best source of how
well a part works is usually the test engineer that handles it. Trust
me, the test engineers didn't like the Gain Tech designs. The term
cowboy was used often.



.

Loading