AD587LQ

[GigaJoe]

the back has a Philippines and date-stamp and code similar to AD

Serious ?  are you believe in repackage soic to ceramic ?

[Andreas]

Hello,

in the mean time I think: ALL is possible.
I have here "genuine" MAX6350MJA
The numbers are printed (and not laser marked as those from MAXIM).

Bottom markings laser marked (PHILIPPINES and some N-Numbers typical for LT-parts)

The sealing between the ceramic discs is not opaque like glass.
It is grey like cement.

Noise is a factor higher than on a buried zener. So more like a band gap reference.
On Pin 3 I measure a temperature dependant signal like on a band gap.

My best guess: some Ref02 re-marked/re-packaged to a MAX6350.

with best regards

Andreas

[iMo]

IMO, nobody in the world will re-package chips from a package A to an different B package, and sell them for a few bugs on ebay. It may easily happen the already packaged chips were stolen directly at the manufacturer's site before the labeling process, and then re-labeled in someone's cellar..
Another option is to steal the entire Si wafer and to package the chips at a different manufacturing site (or well equipped cellar), and then label them, sure..

[GigaJoe]

ok .... another questions:

If I'd like to averaging from a 4-8 chips, what would be a reasonable way to adjust output ?
like to all chips hook per var. resistor - adj. every one ; and for one only as external ?
(all broad inside thermostat )

and cap for noise canceling, 1uf or 100uf - is it a reasonably efficient ?

[Andreas]

Based on the above, I could not recommend buying these. The AD586JQ through a trusted supply chain, and selection for tempco seems a better bet.

Hello,

I fear you are right. It seems that the "good parts" are mostly gone.
I got my running order (a batch of 10) from 11.11. this week end.

A quick check.

visual inspection:
9 devices with date code 9702 -> smells like fakes: previous batches were each with different date code.
1 device with date code 9731
hermetically sealing is light grey (like cement) and not opaque like glass as on the previous batches (the one in the socket) -> not a good sign too

electrical tests (voltages and noise)
3 devices were completely dead: no current consumption, no backward diode between pin 4 (GND) and pin 2 (VCC).
From those 1 had a broken leg on pin 4 (Gnd). And on one device a (internal) short between Pin 1 + Pin 2.
1 device measured around 5 V instead of 10 V. (so it seems they have packaged a AD586 into the AD587 package).

From the remaining 6 devices:
1 with excessive 1/f-noise (90 - 100 uVpp instead of typical 4 uVpp) additionally a broken leg on pin 5 (adjust)
1 with excessive and frequent (all few seconds) popcorn noise of nearly 20 uV
1 with around 10 uV (still in evaluation) popcorn noise all 1-5 minutes.

so three out of 10 seem to behave normal regarding output voltage and noise.
But I have not tested T.C. and ageing drift up to now...

with best regards

Andreas

[Andreas]

ok .... another questions:

If I'd like to averaging from a 4-8 chips, what would be a reasonable way to adjust output ?
like to all chips hook per var. resistor - adj. every one ; and for one only as external ?
(all broad inside thermostat )

and cap for noise canceling, 1uf or 100uf - is it a reasonably efficient ?

do not have experience with averaging. But I would keep the individual cell voltages equal to keep equalizing currents near zero.

100 uF as foil capacitor or as electrolytic (which has leakage currents and thus generates additional noise).
And where do you want to put the cap for noise canceling.
If you want to filter noise below 10 Hz (1/f noise) then it gets really hard.
So the only thing you can effectively filter is wideband noise (above 10 Hz).

with best regards

Andreas

[MiDi]

Same here, order from 2018-10-23, all 10 same datecodes and on bottom different markings.
So it seems these are not genuine anymore, details follow...

[GigaJoe]

Interesting ... a diff source but the same black dye relabeling,   it seems a single source for all diff. sellers on Ali. 

Usually, if such thing happens, I simply wipe by acetone 2 chips, make a picture  of sanding traces, fake relabeling , and open complain for a full refunding ... it usually works.

[syau]

Found hermetically sealing in my batch also light grey (like cement) but the last row of the marking (ie the row below the date code, always start with alpha follow with few digits) match the one on the bottom.

[MiDi]

Top gives exact same datecode and number.
The marking is not consistent with older order.
The cases are flat on both sides, the former ones have bevel on both sides.

New


Old


Left side is older order, right side new with different markings.
So obvious they are fakes.

[Andreas]

it seems a single source for all diff. sellers on Ali. 

I fear you are right.

I did a acetone test on 2 parts.
The black topping easily was removed.
On the first view I did not recognize that the parts have been black topped and remarked.

with best regards

Andreas

[iMo]

Open a fake package and have a look at the chip die..

[Andreas]

Hello,

as there are not much options for a 10V buried zener,
I guess most of the parts are re-labeled from JQ to LQ.
So you will not see it by the chip itself since the grades are selected by the manufacturer.
The ongoing T.C. measurements also indicate JQ devices.

Except the 5V part and eventually the noisy 10V part.
But it could also be that the noise results from a mis-treatment.

with best regards

Andreas

[GigaJoe]

seems true.   another thing are fully tin coated pins,   so chips simply dumped in a liquid tin, for how long ... unknown ...

[GigaJoe]

quite redonkelous quaestio:

Assuming i will feed AD587 by meander in 10 sec period.  Will it help "make it stronger"  , or a wasting time?

[Andreas]

Hello,

what exact do you mean with "make it stronger".

10 sec period is not enough time to heat up and cool down the AD587.

What I did with a AD586 (5V-version of AD587) is moving the ageing faster to its final state (< 2 ppm/year)
by loading the output with 15 mA over night with a supply of 14V at the input.
(so heating the AD586 with 15 mA * (14V-5V)  = 135 mW additionally to self heating periodically).
After some weeks the initial ageing of 3.4 ppm/sqrt(kHr)/ new ageing from day 420 with 2.4 ppm/sqrt(khr) suddenly got slower.

light blue: initial ageing from day zero
pink: ageing from beginning with 15mA load over night (from day 420)

with best regards

Andreas

[GigaJoe]

ok ...
but the load as a current flow, has no affection on zener, just a heat, as site effect of load, I assume.
means: if my chip in a heater, that would be the same effect.

... OK .. i get it ... the best way to cycle the heater ...

[Noopy]

I opened the parts MiDi had ordered:

https://richis-lab.de/REF06.htm


Surprise: They are all genuine parts:






Nice resistor tuning track. It gives a big change at the beginning and a small change at the end of the line. 




Perhaps that´s the buried zener... 




Six dies are showing nothing special but on three dies there is dirt! 
It looks like the cement holding the package together.
More Pictures on my website: https://richis-lab.de/REF06.htm
I assume all the AD587 are rejected parts. With that in mind you have to fear that "good" AD587 out of this charge are also not really good...

[magic]

For completeness, this was shown in older datasheets:



No idea what ground force/sense is
Maybe just separate grounds for the precision bits and the buffer opamp. I think it may be this, the "sense" connection seems to go solely to the suspected zeners and some thin film resistor...

The output may have actual force/sense connections.

[magic]

I had a second look. Despite two metal layers and low resolution, after some digital enlargement the image is surprisingly readable.

The right output pad appears to be "force" and the left "sense". The resistor to the left of output sense is the feedback resistor, the (trimmed) resistor below it at the bottom of the die is the gain resistor. Their junction goes towards those six transistors laid out symmetrically in the center. This looks like a µA741-style input stage, this time fully symmetric and without built-in offset trickery. That's in agreement with the simplified schematic.

The test pads below the output pads appear to be connected right across the die to the corresponding output pads

The pads on the left go somewhere into the reference, including the square trimmed resistor, but contrary to the datasheet show no signs of being bonded to the package.

[Noopy]

I had a second look. Despite two metal layers and low resolution, after some digital enlargement the image is surprisingly readable.

Need more resolution?

https://richis-lab.de/temp/AD587LQ.JPG

[magic]

A little schematic of the reference section. This is a comparatively simple circuit with no provisions for permanent adjustment other than the laser trim.



Some remarks:

R1 is trimmed to mirror a controlled fraction of R2 voltage.
Q1 base current is cancelled by Q2 and the mirror.
Q1 base voltage is subtracted from the output by Q3.
Q4 performs similar subtraction and generates constant current in R4. R3 current drops with temperature, increasing Q4 current
Q4 and Q5 are twice the area of Q3 and Q6, R5 is half as long as R6. Not sure what's the point. Not sure what R7 is doing.

Q10 and the rest hopefully play no role other than starting up the circuit.

tp1, tp2 and tp3 are the test pads. I don't count tp0, which is just a dummy resistor. They could be exploited for external TC nulling, but don't seem to be connected.

[miklos2]

Lars wrote  <https://www.eevblog.com/forum/metrology/best-out-of-the-box-10v-reference/100/
"Joe used the LQ and I think all had the date code 0045. Most of the LQ Joe had were around +2ppm/C at 25C so the simple add-on with a NTC+ resistor worked well."
I purchased three of the legendary AD587LQ with date code 0045 from Canada, Mississauga in 2021 april. Component Electronics, stock# AD587LQ-AD1.
They seems new, nothing suspicious. I made fast temperature sweep test with next results:
AD587LQ (1) at 25°C 10V+2,01mV; ∆U(20°C-40°C) 14ppm;  histeresis<1ppm
AD587LQ (2) at 25°C 10V+2,01mV; ∆U(20°C-40°C) 18ppm;  histeresis 2ppm (32°C-38°C), 1ppm <32°C
AD587LQ (3) at 25°C 10V+1,63mV; ∆U(20°C-40°C) 8ppm;  histeresis 1ppm (34°C-38°C), <1ppm <34°C
The test started at 40°C, then cooling to 20°C (15 minutes) and heating to 40°C (15 minutes).
Devices: a piece of copper and Peltier element. Voltage measurements with Keithley DMM6500 (1 measure/sec), temperature measurements with another synchronized voltmeter (hp3455A) and LM335 fixed on the copper.

[miklos2]

The histeresis of the three AD587LQ date code 0045. Signed with arrow the cooling side of diagram.

[eplpwr]

I have some chips marked "AD587LQ/+", they measure according to spec. Does anyone know what the /+ suffix means?