Ultra Precision Reference LTZ1000

[Vgkid]

It was Mickle T. who transplanted the ltz1000 into the Solartron 7081.

[janaf]

LM399: This is the seller that was selling them at $4.60 recently. Closed for holidays now. Try in a couple of days. http://www.ebay.com/usr/polida2008

Here you have two for $10: http://www.ebay.com/itm/2pcs-LM399H-Precision-Reference-LM399-/181669101870?pt=LH_DefaultDomain_0&hash=item2a4c52812e

[paulie]

why anyone would pay more than this from an unknown buyer on the 'bay' is beyond understanding...

Shipping and handling where Fleabay still comes out ahead. There is that "fake" factor but as mentioned before I consider largely urban myth. Very popular romantic notion but in most cases turns out unjustified.

[paulie]

Here you have two for $10: http://www.ebay.com/itm/2pcs-LM399H-Precision-Reference-LM399-/181669101870?pt=LH_DefaultDomain_0&hash=item2a4c52812e

Actually $15 for two so like I said about twice what they used to go for. Polida has none now. I suspect surplus units getting rare as hens eggs.

FYI Polida is one of my favorite sellers. Very helpful and relatively honest. Over 1500 deals last couple years for work.

[paulie]

BTW are you guys aware "buried zeners" in LM399 and LTZ1000 do not even vaguely resemble a zener diode? Actually somewhat complicated integrated circuits with literally dozens of additional components. In recent conversation with an FAE was told there are even more undocumented and parasitic devices not shown in the internal diagrams.

So virtually zero chance of emulating with discrete components. Why can't life be simple?

[janaf]

Yes Polida still have them, just wait until vacations are over. I bought a dozen, all tested OK.

Last time I bought from LT it was US$ 45.98 shipping i.e. $65 for two LM399. Ebay: US$11 for two including shipping.

Dealbreaker if on a tight DIY budget. If you have a real budget and tight time-plan, of course go to a authorized distributor.

Linear Tech sells these direct:

LM399H#PBF :: (1-99): US$6.64 (100-999): US$5.47

LM399AH#PBF :: (1-99): US$9.37   (100-999): US$7.72

You can order by sending an email to orders@linear.com ....

So, why anyone would pay more than this from an unknown buyer on the 'bay' is beyond understanding...

The 'A' version is simply tested to better specs.  The non-'A' version does not have this extra testing [the DMM manufacturers like to do this themselves].

[Galaxyrise]

BTW are you guys aware "buried zeners" in LM399 and LTZ1000 do not even vaguely resemble a zener diode?

That's the LM399, and it's true pins 1 & 2 are not just a zener or even just a diode-compensated zener.  But D3 is still a "buried zener" diode.  And while that is called out in the datasheet, it is easy to miss, as evidenced by one of the discussions in the LM399 thread. The extra circuitry has a nice description in this old TI app note.

The LTZ1000 is much simpler inside.  Check out the die picture which was posted earlier in this thread.

Perhaps you should start your own thread for trying to squeeze out the best voltage reference for under a dollar? (or whatever price point it is that you've set yourself.)

[janaf]

I attached a picture of the setup I used to measure on the LTZ1000AHC.

The first version was a regular PCB with a couple of cheapo 1% resistor decades connected.

The second version is pictured.

• Eight 12 pos rotary switches for coarse setting of the five main resistors. Three spare rotary.
• One toggle switch for a low value "delta resistor" per coarse resistor. Typically value, 1% of the main resistor.
• A noise amplifier, AC coupled with a gain of 1000. Hardly used.
• Bottom side: 12 0.5% 25ppm thin film resistors per rotary switch, one resistor per toggle switch.

The small value toggle switch meant that the main resistor value could be changed by approximately 1%. The accuracy of the delta R would be 1/100*1/200 = 50ppm. The voltage changes would be in the order of 100-1000 times larger, ie 5000-50000 ppm. The voltage was read with a 7.5 digit LTZ1000 based dmm. As the transfer functions are quite linear (or even constant), I consider these to be quite feasible values.

The objective was to see if there where any sweet spots for resistor values. I was assuming the datasheet variation of output voltage, based on resistor variation, where uncertainties for the output. It turned out they where NOT, but actually dependencies, i.e. very predictable errors.

I re-did the measured two preliminary boards, once with the later, better, setup. In all, I used three different LTZ1000ACH. The measurements all show very similar results.

Bottom line: The output voltage changes, caused by variation of each resistor, is very predictable within limits!

i.e by monitoring / measuring the drift of resistor values it should be possible to predict the output voltage drift caused by the resistors (but of course not by the drift of the zener itself).

I never intended to do absolute voltage measurements. It would be meaningless as the LTZ1000AHC is rated 5% anyway.

I also did not intend to produce results down to the ultimate decimals. In the end, the resistor specs are uncertainties, ie not very predictable, unless you somehow map them individually and match them or correct the output of the circuit.

[MK]

Hi Janaf,

just take an 1N829 run at approx 7.5 mA find the zero tempco current of your sample, age for a year or so and you will have about a 6.5 digit capable reference of your own, at room temps that is.

[paulie]

That's the LM399, and it's true pins 1 & 2 are not just a zener or even just a diode-compensated zener.  But D3 is still a "buried zener" diode.

The LTZ1000 is much simpler inside.  Check out the die picture which was posted earlier in this thread.

So we agree there is no actual zener available in either device. And thanks for the TI link which was best description of that diode circuit inside LM399/LM329 so far. I'm no semi engineer but it looks like many many junctions in the LTZ photo so it may be simpler but not by much. I would love to find out what is really inside that IC that's being called a zener.

Perhaps you should start your own thread for trying to squeeze out the best voltage reference for under a dollar? (or whatever price point it is that you've set yourself.)

Good idea. I've already gotten a couple offline inquiries so will almost certainly follow your advice. My original post here was specific to the LZT1000 and I suspect if that had been in another thread there would have been very few useful replies. This was the goto place for what I needed and I'm very happy Diligent and the others were on tap. Google sucks.

ps. In response to your comment about my price point, ATM it's around 9 cents for a complete reference module including things like tl431 or pair of zeners, noise filter, low TC resistor, and something to mount it all on.

[janaf]

So we agree the is no actual zener available in either device.

No, both have a buried zener inside but other stuff too.
LM399 has more because it has the complete temperature compensation, heater and regulation in the chip.

[paulie]

We were basically referring to just the "zener" part and the conclusion was no actual diode available to the user. In both cases part of a more complicated circuit. I'm not sure if the heater "resistor" they refer to is actually a resistor or, like the zener, part of a collection of components too. I would love to see a diagram of the actual internal LTZ circuit and also a description similar to the one in the TI app note.

[paulie]

I posted a schematic of just the 399 "zener" couple pages back but it was at the end of the page so in case you missed it  here it is again:



Not a diode, at least not available to the outside world.

[Dr. Frank]

I posted a schematic of just the 399 "zener" couple pages back but it was at the end of the page so in case you missed it  here it is again:



Not a diode, at least not available to the outside world.

Definitely.. the Buried Zener is D3, and together with Q13, they even form sort of a Reference Amplifier, like in the SZA263/LTFLU.
The rest of the circuitry consists of some cc sources and amplifiers, similar to the external circuits of the LTZ1000.

There's a detailed description in the original AN-161 from National Sem., which I'd like to upload also.

So, as National explicitly speaks of a buried zener, D3, I don't understand, why you negate that statement?

Frank

[paulie]

I think this is obviously a language issue and no need for further discussion. I'm very pleased to have learned a bit about the LTZ/LM399 references that was not discussed previously in this thread and also zeners in general. So moo-chow-grassy-ass guys. And thanks too for the motivation for another Worlds Cheapest thread.

[janaf]

The big LM399 thread: https://www.eevblog.com/forum/projects/lm399-based-10-v-reference/

[paulie]

Yes, I've gone through this thread twice but 3 times for that one. Mainly because it's only 1/3 as long but also that part is cheaper. In both cases an enormous learning opportunity. Not just for the target devices but many (way) off topic discussions and science in general.

[janaf]

I post this again to ask that someone repeats one of my tests on the LTZ1000 circuit.
It's simple, only requires a 10K 1% resistor and a 6.5 digit DMM or better.

• Measure the output from a LTZ1000 circuit.
• Hook up the 10K resistor in parallel with the 120R. Its decreases by 1.18%
• Measure the output voltage again

• If the voltage increases by 0.108mV it matches my measurements.
• If the voltage changes by 0.83mV it matches the data in the datasheet.

I'd really appreciate if someone could do this! Please 

[Andreas]

Hello,

already did a similar measurement on my LTZ#2 by increasing setpoint +9.6 deg C with 24K7 resistors.
Result is a 3.73 mV change of output or around 50ppm/K.

With best regards

Andreas


23.12.2010 LTZ1047 #2
=================
Bestimmung Tempco
Setpoint 50.5  = 12K5 + 1K                = 533,3  mV
Setpoint 60.14 = 12K5 + (1K||(22K+2K7))   = 514,06 mV -> 19,27mV -> 9,64 Grad

Setpoint 50.5 (22:31-22:35)
ADC4 -> 3592.48702 = 7185.02534
ADC8 -> 3592.48546 = 7185.00504

22:35 22K gesteckt
Setpoint 60.14 (22:41-22:45)
ADC4 -> 3594.35358 = 7188.75846
ADC8 -> 3594.35320 = 7188.74052

ADC4: 3594.35358 - 3592.48702 = 2*1.86656 = 3.73312 mV / 9,64K = 387uV/K = 53.9 ppm/K
ADC8: 3594.35320 - 3592.48546 = 2*1.86774 = 3.73548 mV / 9,64K = 387uV/K = 53.9 ppm/K

22:55 22K gezogen

Edit: was LTZ#2 not LTZ#1

Data for LTZ#1 is similar with 22K

22.10.2010 LTZ1000A #1
==================
Bestimmung Tempco
Setpoint 50.5  = 12K5 + 1K
Setpoint 61.25 = 12K5 + (1K||22K)

Setpoint 50.5  -> 3571.8537
Setpoint 61.25 -> 3573.7066 -> 2*1.8529 mV = 3.7058mV/10.75K = 345uV/K = 48ppm/K

[JohnnyBerg]

already did a similar measurement on my LTZ#1 by increasing setpoint +9.6 deg C with 24K7 resistors.
Result is a 3.73 mV change of output or around 50ppm/K.

Fascinating.
Stability from the LTZ1000 is the stability from the temperature control, would be my (maybe to simple) conclusion

[janaf]

Andreas: If I get you right:

You changed from 1:12.5 to 1:13 ratio for the heater, which gave 3.733mV output change?

That would be a 4% resistor change versus approximately 0.0519% voltage change, ie 0.0125 ppmU/ppmR.

(Or as given in the datasheet 1.29ppm change in voltage for 100ppm resistor ratio, while in the datasheet says 1.0ppm/100ppm) My measurements are close to yours, I got around 1.25ppmU/100ppmR

ie the V/R ratio you get matches well the datasheet and even better with my measurements! Nice to know!

Yet, for R1, 120R, my measured values are very different from whats given in the datasheet, therefore I'm still very interested in some "R1 results".

Jan

[Andreas]

Hello Jan,

sorry I mixed up the 120R with the 1K resistor.
On the 120R resistor I have no possibility to measure without removing all thermal isolation.
(the temperature setpoint pin is accessible on a sub-board connector to have the possibility for "degaussing").
And the pin itself is also like the temperature setpoint very sensitive to RF injection.
So you have to solder with very short wires to the 120 R resistor.

With best regards

Andreas

[janaf]

Thanks Andreas, I hope someone else can do it!

[babysitter]

Hi,

I came across some liquid at work with interesting behaviour which might be of interest for those who want to keep moistore away from precious resistors:

Novec 7100 by 3M, which is a H-FKW. Nonconducting, not aggressive to most plastics, not corrosive to metals, due to quite big molecules it doesn't enter too much into most plastics at all. Aggressive to PTFE and other polymers that contain a amount of F. Water is almost insoluble in it. I am not testing its electric behaviour, but it has almost no surface tension, leaks out of about everything - I am used to keep syringes with water without a cap to store them for a while, H-FKW will leak.

Could be a (quite cheap) alternative to silicone and mineral oils for those who want to put their reference in hermetically sealed containers.

Other news: My LTZ1000A is seeing power supply again, and I found that one of my low thermal emf cable had a bad soldering point: variable between 200-600 miliohm by twisting the cable, after reflowing its back to 86 miliohm just like its same-length cousin.

The "cant look at my reference" period seems to be over, finally.

[janaf]

?

Stability from the LTZ1000 is the stability from the temperature control, would be my (maybe to simple) conclusion

Stability to surrounding temperature is absolutely dependent on temperature control.

I have never seen any explanation on why the LTZ1000 has better aging characteristics than other buried zeners. Is it simply a characteristic of the devise or does the circuit somehow compensate for a "known" aging process? I've never seen any suggestion on this either, just wild speculation....