-
If it's not the wafer, what "new materials supplier" is there? Who's putting the gun to their head to change suppliers on this sole-sourced part.
Can be the steel in the metal can, can be the kovar for the leads, maybe it is the isolation standoff. None of which should effect the performance too much.
Logically, continue making it with the old material until the new material gets approved.
Any quality/supply issues will cause big trouble for manufacturers dependent on it.
I'm skeptical. Mergers and acquisitions are never good for products.
AD will surely consolidate fab, remove redundant teams, "improve" things to lower costs. Increasing shareholder value is paramount and great art is never about that.
Dont be afraid for the LTZ1000. I'm sure that Agilent and Linear signed a contract for the availability, which is a bounding contract for Keys. and Analog.
Also, if they change something, they need time to verify it. This is not a digital gate, where you measure the rise time, and say that it is OK. Just making sure, that the drift for the first 2x1000 hours is OK: 3 months. With about 12 weeks extra as lead time.
Linear has a very different approach for the market. They make products, which sells in the 1000s and not the millions, and they have higher profit margin on it. Why sell microcontollers with billions of transistors in it, and have 5 cents of profit on each, when you can sell an opamp with a dollar of profit on it. The LT5400, which is "just" 4 resistor sells for almost the same as a Apple A8. -
I'm sure that Agilent and Linear signed a contract for the availability, which is a bounding contract for Keys. and Analog.
should be so.
anyways, this blog can *maybe* change their mind if they plan to do any hanky-panky.
viva la voltnutland!
regards.
-
I ordered a LTZ1000A 4 months ago directly from Linear's website.. still waiting.
-
Arrow has 14 LTZ1000CH#PBF right now. It's not the 'A' version, but it is an LTZ1000
-
I am pretty sure (99%) the LTZ1000 won't go into obsolescence in near future.
Ever heard of the mil./space version of LTZ1000 with a price tag of US$1500 -
might be DKK 1500 =US$ 300, pardon but I remember the number 1500 and not the sales channel, the number had me running in search for another source
What I do remember is the claim of "longe lead-time EOL sourcing", "One known fab only" (US as I remember), "narrowed wafer specs", "preconditioned" ...... and a documentation the size of the holy book
Any one ever thought of what it takes to produce and prove a product that stable that other people are willing to bet 10-11 digit US$ project on it ?
If you "own" that product you won't "kill" it
-
Chip1stop has 17 LTZ1000CH#PBF in stock.
-
I ordered a LTZ1000A 4 months ago directly from Linear's website.. still waiting.
Quoting myself to provide an update. I contacted Linear.. they said they'd be shipping me one today. So I guess my order just got forgotten somehow. Or they felt bad and are giving me one from a secret stash. -
One non A type from this 14 Arrows stock is ordered an on the way to me.
The sparrow in the hand, is better than pain in the b..., or so ...
who knows what else will happen
-
a documentation the size of the holy book
this forum would love to have access to *that* holy book
regards.
-
...
I have been noticing thatsomea very tiny percentage of the most recent LTZs have a lot of 1/f noise and popcorn noise. This is related to contamination of the junction layers with metals, (that are normally extracted during processing by infusion of phosphorous on the back side of the wafer, followed by a long bake at 900oC) and/or radioactive particle contamination. I suspect that something in the fab has gone Wr0nG, and they are trying to fix it. I wonder if this is related to the recent contamination in the atmosphere in the northern hemisphere with radioactive contaminants. Maybe it's time to coat the chip with super-pure silicone RTV before sealing the package?
I'm going to shoot a question at LTC and see if they will cough up the information...
-Ken
Radioactive contamination is not leading to popcorn noise. If at all one would get something like a few random errors in high densitiy DRAM of flash chips due to high energy alpha decay. Today most of those errors won't be visible due to chip internal error correction. So it is definitely not due to radioactive contamination.
The 1/f and popcorn noise is more due to chemical contamination's and maybe crystal defects. This might be hard to track, as it can be really trace amounts, both in the fab and in the wafers. Especially popcorn noise is something about a single defect (e.g. wrong atom) sitting at a sensitive place. More such defects at not so sensitive places give a source of 1/f noise.
Usually there is no way to get rid of trace contamination - it's all about starting clean and keep it clean. A common practice is removing some of the surface by etching to get rid of damages from cutting the wafers.
In the early days they has a story about yield being much lower after lunch. The anecdote is they got the yield back up, when they offered free lunch, as they traced the problem back to copper contamination from the money the workers used to pay the lunch. -
I have been noticing that
somea very tiny percentage of the most recent LTZs have a lot of 1/f noise and popcorn noise.
hello Ken,
can you expound on "a lot of noise", ie; some measurement details? how far off-spec or in-spec?
best regards.
-
a documentation the size of the holy book
this forum would love to have access to *that* holy book
regards.
I will try to retrace my browser history
-
I was asking this too.I have been noticing that
somea very tiny percentage of the most recent LTZs have a lot of 1/f noise and popcorn noise.
hello Ken,
can you expound on "a lot of noise", ie; some measurement details? how far off-spec or in-spec?
best regards.
Inviato dal mio Nexus 6P utilizzando Tapatalk
-
What is the source of "Some recent parts?"? How many are we talking about and what date codes?? Are these actual real parts from LT or authorized distributor? Was this on a known good PCB design?
Normally it is -very- rare to see a real LT factory checked LTZ out of spec on a properly built board, and we've seen about zero of what you're talking about on recent batches this year. I think over the several hundred I can count the bad LTZ's on one or two fingers, and LT replaced them since they came directly from LT (They will also do that if they come from a real authorized distributor, but not if you bought them from some broker or eBay, etc).
It is common to see the noise on these parts from the shady parts brokers / dealers trying to peddle the "floor sweeping" rejected / fake parts that somehow got legs and walked out of the fab, or sometimes someone got hold of some reject die and repackaged them. You can usually tell when the parts aren't sealed in the proper tray with the LT logo.
Those "used" LTZ parts on eBay are a bit suspect also - you never know what blowtorch was used to get the parts demounted. You might get lucky and they are fine, and maybe not.
We've always ordered direct from LT and never had any real problem.
-
Hello,
noise testing is not cheap.
So I guess that LT is only measuring wideband noise for a short time and not the 1/f noise in excess of 10 seconds during production.
The closer you look the more noise you will find.
I have also the impression that newer buried zeners (AD586/AD587) have more (popcorn) noise but I will have to verify this with older stock devices.
My LTZ#5 DC 0945 via official distributor has also a lot of popcorn noise in excess of 2uV.
If you look at the "right" 10 seconds measurement duration you will see only the typical 1.2uVpp noise.
With best regards
Andreas
-
@Ken (?Dobkin), @MisterDiodes
does anybody know the yield percentage on the LTZ1000 die?
regards.
-
Andreas,
...that's why the datasheet clearly states "about 1.2uV noise". 2uV p-p is listed as max, and that data looks still in the ballpark of that. You also have take your measurement uncertainty into account and look at how far your experiment drifts in the same time frame, as well as the overall noise floor of your setup. Are you scanning other known Vrefs simultaneously to compare? When you're down at uV levels that helps to distinguish noise on a -single- device from your experiment setup noise. The earlier part of your data is quieter for sure, did something get switched on nearby? -
did something get switched on nearby?
No,
its clearly popcorn noise. And whole setup including scope is battery supplied.
As I measure with a filter amplifier (HP = 0.1Hz) you see only the rising and falling edge as triangle pulse
when the duration of the jump is longer than about 1 second.
Noise floor is typical 100-120uVpp so far away from the 1.2uVpp of the LTZ. (see noise measurement thread).
with best regards
Andreas -
I would expect the LTZ1000 to be tested for a little longer than 10 seconds. It already needs some time to get a stable temperature - so no way to do a super fast test anyway. Popcorn noise is likely one of the more important reasons to do the checks and sort out failing units.
However with popcorn noise is can be difficult to detect - especially that type of noise with more infrequent jumps. Chances are one would not notice this in a short interval of maybe 30 seconds. Those very low frequency popcorn cases are the really bad ones, as one never knows and longer averaging also does not help that much. With the more frequent and symmetric jumps, like in TIN's jummpy sample from Ebay one has in theory the chance to detect and even correct for this. -
I would expect the LTZ1000 to be tested for a little longer than 10 seconds. It already needs some time to get a stable temperature - so no way to do a super fast test anyway. Popcorn noise is likely one of the more important reasons to do the checks and sort out failing units.
However with popcorn noise is can be difficult to detect - especially that type of noise with more infrequent jumps. Chances are one would not notice this in a short interval of maybe 30 seconds. Those very low frequency popcorn cases are the really bad ones, as one never knows and longer averaging also does not help that much. With the more frequent and symmetric jumps, like in TIN's jummpy sample from Ebay one has in theory the chance to detect and even correct for this.
hello,
looks like the biggest bottleneck in the production of LTZ1000 would be effective noise testing.
regards.
-
I still fail to see the big LTZ datasheet spec "Oops" in Andreas' data. That shows a part still within the realm of datasheet specs it looks to me: It's showing clearly a P-P value of ~2uV (equipment / experiment noise -floor- limits are not known) which is the datasheet value. Of course the longer you measure, the more likely you'll see more noise. Especially as 1/f creeps in. Double the measure time and you get another octave's worth of 1/f possibilities. If there is -some- popcorn noise, it will probably still be able to pass the LT quality check.
The problem we saw before on just one or two parts was an excursion more like 6 or 8uV P-P across 10 sec - clearly out of spec. Those got replaced.
DigilentMinds: Were you thinking about a bigger "oops" deviation from datasheet value? That's what I was curious about. How many parts and what date codes are we talking about??
-
I still fail to see the big LTZ datasheet spec "Oops" in Andreas' data. That shows a part still within the realm of datasheet specs it looks to me: It's showing clearly a P-P value of ~2uV (equipment / experiment noise -floor- limits are not known) which is the datasheet value. Of course the longer you measure, the more likely you'll see more noise. Especially as 1/f creeps in. Double the measure time and you get another octave's worth of 1/f possibilities. If there is -some- popcorn noise, it will probably still be able to pass the LT quality check.
The problem we saw before on just one or two parts was an excursion more like 6 or 8uV P-P across 10 sec - clearly out of spec. Those got replaced.
DigilentKen: Were you thinking about a bigger "oops" deviation from datasheet value? That's what I was curious about. How many parts and what date codes are we talking about??
hello,
what *is* the LTZ1000 go/nogo quality check with respect to noise specifically?
(maybe another thread about LTZ1000 noise testing is in order)
regards.
-
It already needs some time to get a stable temperature
RTFM
Note 2: All testing is done at 25°C. Pulse testing is used for LTZ1000A to
minimize temperature rise during testing.
And against other data sheets like LT1021 or LT1236 there is no 100% noise testing specified.
The newer data sheet of the newer "3rd generation buried zener" LT1027 already specifies a method with reduced testing time
Note 5: RMS noise is measured with an 8-pole bandpass filter with a
center frequency of 30Hz and a Q of 1.5. The filter output is then rectified
and integrated for a fixed time period, resulting in an average, as opposed
to RMS voltage. A correction factor is used to convert average to RMS.
This value is then used to obtain RMS noise voltage in the 10Hz to 1000Hz
frequency band. This test also screens for low frequency "popcorn" noise
within the bandwidth of the filter. Consult factory for 100% 0.1Hz to 10Hz
noise testing.
So the LT1027 is tested from around 22 -41 Hz (with minimum 50 ms testing time).
(no 100% test specified).
In the LTZ1000 datasheet there is no noise testing specified.
(it is always more essential to know what is missing in a data sheet as that what is specced).
with best regards
Andreas
-
RTFM
hello andreas,
your observation is very correct:-
TNFM (There is No ****ing Manual with respect to noise testing of LTZ1000)
best regards.
