Source for low current zener in China?

[Joerg]

Hello Frank,

No care for the existing and substantial customers, no care for the new
and small customers... the world is coming to an end.

The bottom line of the delay you experience is that they don't have a
solution for you. A "Yes-no problem" answer always arrives quick.
For that "No-we're sorry" answer, you always have to wait a bit longer.

Do we have a plan B ?

Yes, working on plan B, plan C and plan D

Plan B: CentralSemi has stock of their CMHZ series but the data sheet
isn't detailed enough. However, they promised to have an answer this
afternoon from Engineering with more data. What the "big players" have
to realize is that, typically, after plans B, C and/or D work out the
plug gets pulled on them. One of them in your country doesn't even know
how much biz they have lost. Water under the bridge though because now
they've auctioned most of it off.

 

[Frank Bemelman]

Yes, working on plan B, plan C and plan D

I hope plan D are two tickets to Hawaii.

Plan B: CentralSemi has stock of their CMHZ series but the data sheet
isn't detailed enough. However, they promised to have an answer this
afternoon from Engineering with more data. What the "big players" have to
realize is that, typically, after plans B, C and/or D work out the plug
gets pulled on them. One of them in your country doesn't even know how
much biz they have lost. Water under the bridge though because now they've
auctioned most of it off.

They couldn't care less. Suppliers experiment with their level of
service, we search for the cheapest ones around until we get bitten,
and hop from one to another after too much frustration. I don't think
it will change things a bit.

But then there's always plan D

 

[Joerg]

Hello Frank,

I hope plan D are two tickets to Hawaii.

No, it's a blender full of margarita, heavy on the Tequila

They couldn't care less. Suppliers experiment with their level of
service, we search for the cheapest ones around until we get bitten,
and hop from one to another after too much frustration. I don't think
it will change things a bit.

Not really, although the pricing has to click, of course. The where
many, many times where I found "the" perfect part at Philips or Infineon
only to find out that I couldn't get any here in the US. US companies
are usually better although TI recently lost a design-in because they
failed (after a couple reminders) to furnish more data on the 16bit ADC
in the MSP430F2013. I did what I knew would work, designed the whole
chebang analog again. Like usual. Once a design is done, it's done. IOW
a finished and signed ECO is like a hardened batch of cement. You can't
really move it anymore.

But then there's always plan D

About six hours from now it'll be off to plan D...

 

[Tony Williams]

Plan B: CentralSemi has stock of their CMHZ series but the data
sheet isn't detailed enough. However, they promised to have an
answer this afternoon from Engineering with more data........

A quick search of Diodes Incorporated turned up their
DDZ9690-7, which seems to be in stock at Digi-Key and
Mouser.

<http://www.diodes.com>

 

[Robert Baer]

Hello Robert,





Yes, they do. CMHZ4626, but non-stock everywhere I checked. I found some
others as well but lead times are as high as 18wks and we need them
pretty much right now :-(

Then get a sample of the same part withthe higher current spec and
check on the curvetracer.
Maybe you will get lucky and find the high current part behaves at
the low end.
Some parts get a different partnumber if pass into a different bin.

 

[Joerg]

Hello Tony,

A quick search of Diodes Incorporated turned up their
DDZ9690-7, which seems to be in stock at Digi-Key and
Mouser.

<http://www.diodes.com>

Yes, but in the breakdown graphs (fig 4) they have left out that one
type, the 9690. Interpolating between the 9688 and 9692 it looks like it
becomes rather mushy around 200uA, and even more so at the rated 50uA
current level. It doesn't look like a true low current zener.

 

[Joerg]

Hello Robert,

Then get a sample of the same part withthe higher current spec and
check on the curvetracer.
Maybe you will get lucky and find the high current part behaves at the
low end.
Some parts get a different partnumber if pass into a different bin.

That would be hand selecting. Expensive and very difficult to do since
it's automated assembly with reel-mount machines. We would have to break
the whole reel because you never know whether they'd be all the same.
Also, operating in the knee renders the whole scheme shaky because a
couple volts drop in the 9V battery could already send it out of range.
We need a true (smaller die) low current zener in this case.

Problem is, when I detect excessive impedance and ask the mfgs about it
they seem to become silent. Hmm....

 

[Robert Baer]

Hello Tony,


Yes, but in the breakdown graphs (fig 4) they have left out that one
type, the 9690. Interpolating between the 9688 and 9692 it looks like it
becomes rather mushy around 200uA, and even more so at the rated 50uA
current level. It doesn't look like a true low current zener.

You cannot depend on mfg data sheets...you need to do a reality check
with anactual part.

 

[Robert Baer]

Hello Robert,




That would be hand selecting. Expensive and very difficult to do since
it's automated assembly with reel-mount machines. We would have to break
the whole reel because you never know whether they'd be all the same.
Also, operating in the knee renders the whole scheme shaky because a
couple volts drop in the 9V battery could already send it out of range.
We need a true (smaller die) low current zener in this case.

Problem is, when I detect excessive impedance and ask the mfgs about it
they seem to become silent. Hmm....

*NOT* true; like i said, test one sample part to see if it behaves.
They all will act in a similar manner - even from lotcode to lotcode
(until they change their process, which they will not tell you even if
the contract sez so and you do millions of dollars of business).

If you do not mind using an oddball part in an un-documented and
unspeced manner, try the SOT-363 package dual PNP by Rohm: the UMT1NTN;
use the E-B junction as a zener (somewhere in the 7-8V region). Carried
by DigiKey.
If i remember correctly, it is has a well-behaved knee in the nanoamp
region and has no negative resistance or oscillations to somewhere in
the milliamp region.

 

[Tony Williams]

Yes, but in the breakdown graphs (fig 4) they have left out that
one type, the 9690. Interpolating between the 9688 and 9692 it
looks like it becomes rather mushy around 200uA, and even more
so at the rated 50uA current level. It doesn't look like a true
low current zener.

Look at Fig. 10 though. It shows the dV/dI slope
of a 5v6 device being about 2000 ohms at 50uA,
dropping to 350 ohms at 250uA.

 

[Eeyore]

Look at Fig. 10 though. It shows the dV/dI slope
of a 5v6 device being about 2000 ohms at 50uA,
dropping to 350 ohms at 250uA.

That's surely what you'd expect ?

Graham

 

[Robert Latest]

On Wed, 27 Sep 2006 17:13:02 GMT,

Problem is, when I detect excessive impedance and ask the mfgs about it
they seem to become silent. Hmm....

So how come this thing has worked in the past? (You said it was a legacy
design so I assume it has been successfully in production for a while.)
Were those real "low-current zeners" that all of a sudden have gone to
the land of unobtainium, or did the finished product work only by sheer
luck in the high-impedance knee region of whatever diodes were designed
in back then?

robert

 

[Eeyore]

So how come this thing has worked in the past? (You said it was a legacy
design so I assume it has been successfully in production for a while.)
Were those real "low-current zeners" that all of a sudden have gone to
the land of unobtainium, or did the finished product work only by sheer
luck in the high-impedance knee region of whatever diodes were designed
in back then?

My thoughts as well. Maybe someone else was being 'economical with the truth'
?

Graham

 

[Winfield Hill]

Tony Williams wrote...

Look at Fig. 10 though. It shows the dV/dI slope
of a 5v6 device being about 2000 ohms at 50uA,
dropping to 350 ohms at 250uA.

A good way to evaluate the relative die size of two zeners
is to look at their capacitance. Figure 7 of ON Semi's
'4690 datasheet shows about 110pF at 1V bias. This matches
the 100pF value in figure 10 of the 1n5232 datasheet. The
'5232 is the classic 500mW 20mA axial-lead zener, which has
notoriously-poor regulation characteristics at low currents.
They also have identical peak-power ratings at 1ms.

We can speculate that these two zener families have similar,
if not identical, dies and are made the same way. However,
they should be graded at their respective test currents.

 

[Joerg]

Hello Robert,

You cannot depend on mfg data sheets...you need to do a reality check
with anactual part.

You cannot rely on either alone. The guaranteed values in a data sheet
is what's important for production. Testing supplements that but if you
test, say, in September then the die for December might come from a
wafer where the process had veered substantially. That's what seems to
have happened in this case (and the vendor is surprisingly
non-responsive about it...).

 

[Joerg]

Hello Win,

A good way to evaluate the relative die size of two zeners
is to look at their capacitance. Figure 7 of ON Semi's
'4690 datasheet shows about 110pF at 1V bias. This matches
the 100pF value in figure 10 of the 1n5232 datasheet. The
'5232 is the classic 500mW 20mA axial-lead zener, which has
notoriously-poor regulation characteristics at low currents.
They also have identical peak-power ratings at 1ms.

Many data sheets omit capacitance :-(

But when you measure they seem to be all the same :-(((

We can speculate that these two zener families have similar,
if not identical, dies and are made the same way. However,
they should be graded at their respective test currents.

That's exactly what seems to be the problem, a "one size fits all"
approach. Doesn't work in real circuits as much as it doesn't work for
socks and T-shirts. What they often (maybe always?) do is tune the
process so they get the right Vz range at the current they want it to
be, regardless of how far it's into the knee.

Maybe the market for low current zener isn't sufficient. Otherwise I
wouldn't understand why they would use the same die size and get only
100,000 gross off a 5" wafer where they could harvest more than twice as
much.

 

[Joerg]

Hello Robert,

So how come this thing has worked in the past? (You said it was a legacy
design so I assume it has been successfully in production for a while.)

Well over a decade without problems.

Were those real "low-current zeners" that all of a sudden have gone to
the land of unobtainium, or did the finished product work only by sheer
luck in the high-impedance knee region of whatever diodes were designed
in back then?

Not by luck. The design is within data sheet limits. The low cutoff is
what matters. Data sheet says it's between 5.32V and 5.88V at 50uA. We
run at 200uA and had a lot of them below our cutoff limit of 5.1V. Oops!
So I wrote, asking about whether it could be a lot glitch. No call back
so far...

 

[Joerg]

Hello Robert,

*NOT* true; like i said, test one sample part to see if it behaves.

Have test results on 50 samples now. However the behavior of samples
from these two lots is not enough for a release. We need hard data the
mfg stands by. Which I got yesterday from one and may be getting next
week from another that would then become the 2nd source.

They all will act in a similar manner - even from lotcode to lotcode
(until they change their process, which they will not tell you even if
the contract sez so and you do millions of dollars of business).

Ok, I guess they deserve credit here now: Centralsemi did exactly that.
They were very open with me about process details and I appreciate that.

If you do not mind using an oddball part in an un-documented and
unspeced manner, try the SOT-363 package dual PNP by Rohm: the UMT1NTN;
use the E-B junction as a zener (somewhere in the 7-8V region). Carried
by DigiKey.
If i remember correctly, it is has a well-behaved knee in the nanoamp
region and has no negative resistance or oscillations to somewhere in
the milliamp region.

Thnaks. However, 7V to 8V wouldn't work here and a new package means a
re-layout. In that case we'd do it right and move to the TLV431.

 

[Jim Thompson]

On Thu, 28 Sep 2006 16:24:11 GMT, Joerg

[snip]

Thanks. However, 7V to 8V wouldn't work here and a new package means a
re-layout. In that case we'd do it right and move to the TLV431.

Face it, and do it. Otherwise the next lot will be out of the "spec"
you need, and the client will be unhappy. Can't you make a little
"teensy-weeny" daughter board in there with a TLV431 ?

...Jim Thompson

 

[Joerg]

Hello Jim,

Face it, and do it. Otherwise the next lot will be out of the "spec"
you need, and the client will be unhappy. Can't you make a little
"teensy-weeny" daughter board in there with a TLV431 ?

It's already on the suggestion list for that client. Depends on how long
this design is going to remain in production. Daughterboards aren't
really feasible because it's small and there is more than one zener.
Plus it is automated mass production.

Out of spec on this one only happened once. I am still wondering how it
could happen, considering that our spec is way more loose than what the
mfg states in the min-max table. However, with smaller suppliers you can
negotiate a deal where they make sure that every shipped reel is within
spec. The small ones really care. I guess they know the old saying in
sales "If you don't take take of the little guy, someone else will". And
today's little guys have a habit of becoming tomorrow's big guys