Ultra Precision Reference LTZ1000

[CaptnYellowShirt]

Did you happen to get an address that we could send a thank you card to?

No, I did not, but you can probably send it to the main office in Milpitas, CA and it will get to him.

I'm going to send him a thank you card from the [Volt-Nuts].

I think it'd be nice if he could see the international cross section of people that have benefited from his advice. PM me if you want your name or handle in the signed-by section. Include what you want to be called and a city and country if you don't mind.

I'll post a picture of the card before I send it.

[TiN]

Great, thanks!

[CaptnYellowShirt]

No need for the private message-- you could find out who I am anyway just from the domain name registration for DiligentMinds.com ...

The judge said I have to stop cyber-stalking people ... its part of my parole. 

Your name will be the first on the list.

[macfly]

@ DiligentMinds.com

Thanks a lot for sharing your informations 

[branadic]

19) The Zener in the LTFLU-1 is the exact same one as is in the LTZ1000.

Really? So what is different to LMx99?
I expected the LTFLU is completely similar to the LMx99, a 4 pin TO package with heater and zener diode.

[quantumvolt]

19) The Zener in the LTFLU-1 is the exact same one as is in the LTZ1000.

Really? So what is different to LMx99?
I expected the LTFLU is completely similar to the LMx99, a 4 pin TO package with heater and zener diode.

The LTFLU-1 and the earlier SZA263 have no heater. They are zener (avalanche) diodes nominally 6.2 V accompanied with a transistor for temperature compensation and reference voltage amplification.

The pins are Collector, Base, Emitter and Anode (Kathode is common with Emitter).

[Andreas]

Hello Ken,

thanks for sharing the information.

With best regards

Andreas

[Andreas]

Really? So what is different to LMx99?
I expected the LTFLU is completely similar to the LMx99, a 4 pin TO package with heater and zener diode.

Hello branadic,

a refamp comes usually as refamp-set with (perfectly) adjusted resistors.
See e.g. fluke 8840A handbook.
So a LTFLU without the adjusted resistor set will be only half of the truth.

With best regards

Andreas

[branadic]

The LTFLU-1 and the earlier SZA263 have no heater. They are zener (avalanche) diodes nominally 6.2 V accompanied with a transistor for temperature compensation and reference voltage amplification.

The pins are Collector, Base, Emitter and Anode (Kathode is common with Emitter).

Thanks for that, couldn't find such info yet and there is nothing mentioned about LTFLU in the "Current  Sources & Voltage References" by Linden T. Harrison.

Hello branadic,

a refamp comes usually as refamp-set with (perfectly) adjusted resistors.
See e.g. fluke 8840A handbook.
So a LTFLU without the adjusted resistor set will be only half of the truth.

Thanks, so it's no good idea to go for LTFLU unless you want to spend much time and money for the perfect resistors.

[quantumvolt]

Really? So what is different to LMx99?
I expected the LTFLU is completely similar to the LMx99, a 4 pin TO package with heater and zener diode.

Hello branadic,

a refamp comes usually as refamp-set with (perfectly) adjusted resistors.
See e.g. fluke 8840A handbook.
So a LTFLU without the adjusted resistor set will be only half of the truth.

With best regards

Andreas

I doubt that one single 'refamp'-type IC has been sold from any chip maker with any resistors. It is the instrument makers that sell matched sets in order to keep their boxes within tempo spec after repair. This is nothing special for the refamp. Even the EDC/Krohn Hite volt boxes (there is an EEVBlog video showing the reference) with a simple 10 USD 1N829A requires a very precise current setting.

There is no magic here. Anyone with good enough instruments to detect changes in net tempco for small changes in current can select resistor(s).

This technology is old and only based on the appr. +- 2 mV tempco of the avalanche diode and the PN-junction respectively. It was used by Hewlett Packard with the Motorola part SZA263 long before Fluke or Linear Technology came into the picture.

[Andreas]

There is no magic here. Anyone with good enough instruments to detect changes in net tempco for small changes in current can select resistor(s).

According to 8840A handbook there are 3 resistors in the ref amp set.
It seems that 2 resistors adjust the output voltage.
So the other resistor seems to adjust the "Zero TC" current.
Further there are the magical -50mV in the circuit diagram.

Since I cannot imagine that all the resistors do not influence each other:
Does anybody know a "adjustment specification" for the resistors with minimal effort for trimming?

With best regards

Andreas
 

[Dr. Frank]

Hello fellow forum users,

I just finished with a telephone conversation with Bob Dobkin, and he was short on time, so I was not able to get all of the answers to every question I [and we] had, but here are the ones I was able to get:


3) The LTZ1000A has a different die attach than the LTZ1000, but they are both isolated quite a bit from external lead stress.  Both of them will respond a small amount to lead stress, but the 'A' version if far less susceptible to this.  Bob said that this can show up as a drift in output voltage with changes in barometric pressure.

8: All voltage references *do* age faster at higher temperatures.  So, since the LTZ1000 can be run about 10-deg-C lower than the 'A' version, it will have lower [about 1/2] long-term drift than the LTZ1000A under otherwise identical conditions.  [But, the 'A' version has *other* differences that are positive-- my words.  There are other sources of drift that the 'A' version is less susceptible to-- barometric pressure changes is one of them].

9) For the LM399 and both of the LTZ parts, he said a good burn-in routine would be to operate them at 125C [in an oven] for 2 weeks.  After that, you can cycle the power on and off 10 to 15 times at normal operating temperature, and this will get them to settle down.  This process should remove most of the initial drift that these devices exhibit.  [So, my initial guess for a burn-in cycle was pretty good-- there is nothing wrong with a burn-in 10 times longer than this.]

15) The TO packages are filled with *DRY AIR*-- not even nitrogen-only [so you would expect some degradation from this-- my words]-- but, Bob says that they don't have any parts that are degraded by this.

16) The LTZ1000 will use a lot more power than the LTZ1000A, so for battery operated circuits, it's best to go with the LTZ1000A, and insulate it to save power.

17) There is probably not a scenario where you can get zero-TC out of the LTZ-- he said you probably would have to operate Q1 at around 1uA, and that is not practical.

18) He said that it is not necessary to use the LTC2057, as the drift and noise of the Zener contribute far more than the LT1013.  Bob said that you could use almost any precision op-amp, and the drift would not be affected by much even by the less precise amps.  [I'm still going to use the LTC2057 anyway-- they are as cheap as popcorn, and have very little noise.  Since I *do* need a zero-drift amp in the follow-on boost circuitry, and it's more economical to keep the number of unique parts on the BOM as low as possible.]

19) The Zener in the LTFLU-1 is the exact same one as is in the LTZ1000.  He said that even though the exclusivity portion of the contract with Fluke is no longer in effect, they are not going sell these to the public because they don't want to anger a customer that buys millions of dollars in other parts from them.  Even if they did want to sell it to others, it has not been characterized and there is no data sheet-- so there would be a lot of work just to build that.  It's just not going to happen...

Regards,
Ken

Hello Ken,

thanks that you made the interview, sharing with us.

Actually, there were several myths being busted around the LTZ1000.

Concerning "Burn-In", he missed the point of strong hysteresis after a 125°C trip, which will definitely NOT be removed by simply switching off 10-15 times.

I feel confirmed, to keep the LTZ1000 PCB as simple as possible, at 45°C, with good thermal management/insulation.

Frank

[Dr. Frank]

I hope that this is not too far off-topic, but [I *think*] Hewlett-Packard make a model 735A "Transfer Standard"; and the Fluke 731B looks an awful lot like it.  What happened there-- did HP sell it's 735A technology to Fluke, or did Fluke make a "copycat" version of it?  Anybody know?

Sorry, but the HP735As schematic  does not ressemble the Fluke 731A at all. At least in the manuals, which are hosted on agilent site.

In the 735A, an ovenized, simple zener diode is used, no sign of a RefAmp device.

Anybody (Quantumvolt??) who can refer correctly to a HP device with a RefAmp, from the 60ties?

Did not find that, not in the 745A, and not in the 740A. All were based on a simple zener diode.


I've read ("Hewlett Packard, the early years") that  Packard and Fluke were close friends. They worked for GE and later both  joined Navy .

Later they made an arrangement to not interfere on certain businesses, so that Fluke made all those analog standards, but kept out of building DMMs and counters, I think. 

Frank

[quantumvolt]

I have also read stories about the persons behind HP & Fluke, but I do not know enough to comment on that.


The HP735A contains an ovenized 1N829A. Very high tech for 1966. And documentation in a class of its own:

http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1966-03.pdf
 
http://www.hpmemory.org/an/pdf/an_70.pdf


The HP3440A is described in 1963:

http://www.hpmemory.org/timeline/dave_cochran/hpj_nov63.htm.

The manual I have found on the web contains the "refamp" - which as HP part was called "Amp., Ref.". This edition contains pages that are revised in May 1970:

http://www.pa.msu.edu/~edmunds/DVM_HP_3440a/hp_3440a_dvm_manual.pdf


The HP3050A has a reference supply that is very similar to Fluke 731A, except that it uses a 'discrete' op amp. This manual also contains pages that are revised in May 1970. If I am not wrong, the Fluke 731A is from 1971.

http://cp.literature.agilent.com/litweb/pdf/03450-90007.pdf

[Dr. Frank]

Thanks for the hints!

The HP3420A (1971) also uses a RefAmp. Origin not determinable.

The HP3440A uses a 4JX19A519 from G.E.


Frank

[quantumvolt]

There is no magic here. Anyone with good enough instruments to detect changes in net tempco for small changes in current can select resistor(s).

According to 8840A handbook there are 3 resistors in the ref amp set.
It seems that 2 resistors adjust the output voltage.
So the other resistor seems to adjust the "Zero TC" current.
Further there are the magical -50mV in the circuit diagram.

Since I cannot imagine that all the resistors do not influence each other:
Does anybody know a "adjustment specification" for the resistors with minimal effort for trimming?

With best regards

Andreas

I am referring to http://assets.fluke.com/manuals/8840A___imeng0300.pdf Figure 5-6 p. 5.8.

Assume U702A is ideal. Then both inputs are at 0 volt. Hence the collector of U701 is at 0 volt. To avoid possible reverse bias C-B the base of U702 is biased at -0.05 volt (which gives 6.95 volt headroom for the REFAMP zener appr. 6.5 volt + Vbe appr. 0.45 volt).

Assuming first that the bias current of U702 U701 can be ignored, the bias divider Z701 is fixed in resistor ratio. Choose resistors that give a reasonable current (a few hundred uA or 1 mA or ? - it doesn't matter at the first try) and give appr. the wanted -0.05 V and -7 V.

Now adjust R701 (it will imo only affect bias for U702A) for minimum tempco in the -7 volt.

If TP701 has changed away from -7 volt, adjust U701 base voltage by changing the Z701 ratio marginally, and repeat. If necessary, choose a different Z701 current and start over.

---

I might be totally wrong  . But I could always ask on the forum for resistor values from a real box and start from there ...

[branadic]

It would have been helpful to ask Bob what is his opinion... does really make sense to preselect references before solder and age them? Would have been interesting if he could have shared any experience to that.

[branadic]

For the LM399, I'm guessing the soldering process doesn't hurt it as much as other references, because Fluke has the most stable 6.5-digit DMM, and they solder the LM399 *after* burn-in...

Thanks Ken for your response, but I doubt that the soldering process is without influence. At least one pin of the TO package is directly connected with the case and will heat up the die attach while soldering.

[macfly]

[Clever!]

Yes, it is!!

Many manufacturers offer a "super grade" version of their voltage references with guaranted initial tolerance and temperature drift. They are measured on 100% of chips. But how?

There's sometimes an advised "TEMP" output on voltage references (With a kind of linear dependance with temperature). This internal circuit is used to test chips after packaging (Internal die temperature is monitored through "TEMP" pin, and the chip is heated i.e. by injecting current through an ESD protection diode. An external control system can provide precise temperature regulation).

In this paper, such a system is diverted from its initial use to provide nearly 0ppm/K temperature drift, with a common plastic-case voltage reference chip :

http://electronicdesign.com/analog/internal-oven-provides-voltage-reference-less-1-ppmdegc-drift

How does this circuit work  ? IMO is could NOT!

1. There is only a positive supply voltage, so the output voltage of the opamp's can only be positive.
2. IC2B is connected as inverting amplifier. It gets a positive voltage from pin 1 of the reference, buffered by IC2A.
3. This should result in a negative output voltage of IC2B, but it could not, due to the lack of negative supply voltage.

Or I am a completely wrong ?

Regards

macfly

[fmaimon]

How does this circuit work  ? IMO is could NOT!

1. There is only a positive supply voltage, so the output voltage of the opamp's can only be positive.
2. IC2B is connected as inverting amplifier. It gets a positive voltage from pin 1 of the reference, buffered by IC2A.
3. This should result in a negative output voltage of IC2B, but it could not, due to the lack of negative supply voltage.

Or I am a completely wrong ?

Although it only have one supply, the voltage reference IC is being fed by a 5V zener, so the applied 12V, through the 10 ohm resistor is always going to foward bias the internal esd diodes.

[quantumvolt]

The comparator is referenced to VREF = 2.5 V. Nothing goes negative relative to the potential at the GND level.

The sensor voltage at pin 1 falls as the chip cools. When it is 0.53 volt LOWER than VREF (which means the chip has cooled below 75 degrees), the buffer/inverter will have converted the marginal fall to a rise on the noninverting comparator input which then goes positive relative to VREF, and a new heating impulse is initiated.

Since VREF is the only voltage stable enough to use for comparison - the main task for this circuit is to convert the passing of the sensor pin below VREF-0.53 volt to a trig signal for the VREF-referenced comparator. In effect it is an AC-signal at a few hundred hertz.

[Andreas]

For the LM399, I'm guessing the soldering process doesn't hurt it as much as other references, because Fluke has the most stable 6.5-digit DMM, and they solder the LM399 *after* burn-in...

Thanks Ken for your response, but I doubt that the soldering process is without influence. At least one pin of the TO package is directly connected with the case and will heat up the die attach while soldering.

The pictures of HP34401A reference that I find in web show a LM399 within a socket (horror).   
I do not know if this is a unmodified unit but with my bad experiences of socket related voltage shifts I would not do that.



With best regards

Andreas

[rf-design]

4) It's been 20 years, so he didn't know the exact numbers, but Q1 is only rated to a few hundred micro-amps of collector current.  [This means my single-resistor idea will *not* work! -- back to the original circuit from the ap-note!]  Q2 will have a max rating of around 1mA, and the Zener can take quite a bit more.

5) They *could* make a model for LTspice, but the temperature effects are very difficult to simulate, and the market need is not large enough to support the development effort [my words].  He said the best way is to build the circuit and test it out in real life.

17) There is probably not a scenario where you can get zero-TC out of the LTZ-- he said you probably would have to operate Q1 at around 1uA, and that is not practical.

Refering to the chip photo posted early in this thread I could see 6 heater ring resistors, 4 equal sized diffusion isolated npns, 2 equal sized diffusion resistors and 2 possible lateral pnps or diodes. All in a very old single metal bipolar process. Is this one still in production? Or was it all time before shutting down the process?

Did anyone check the corrospondence between the schematic in the appnote and the circuit seen in the photo?

Based on the layout dimensions operating current ratios of mA to uA for optimum temperature stabilisation seams unlikely. The main target for the circuit operation would be the die temperature. For better understanding it would be very helpful to have a full schematic for modelling purposes which must include thermal coupling of the devices. Did anyone tried such a model?

One aspect of the thermal regulator is that if the interlayer between the die bottom and package is not even the thermal surface plot would get a gradient which is not foreseen in the bipolar layout placement and configuration. To be immune it should circular interdigitized which is not. Further the 4 bipolars are not crosscoupled. So a thermal conductivity gradient in the axis connecting the two electrical groups would give also a gradient between the bipolars vbe. So I see this as a design flaw. But back to the model that could be integrated in thermal coupling resistors as a mounting imperfection.

If the temperature cycle amplitude is less than 20K the hysteris will be closed. So there is no aging but another unidentified process. Is there any idea?

[Andreas]

4 equal sized diffusion isolated npns, 2 equal sized diffusion resistors and 2 possible lateral pnps or diodes.

If the temperature cycle amplitude is less than 20K the hysteris will be closed. So there is no aging but another unidentified process. Is there any idea?

Hello,

https://www.eevblog.com/forum/projects/ultra-precision-reference-ltz1000/msg279145/#msg279145

Sorry I see 4 large transistors (horizontal and vertical) connected all to each other building Q1.
And 4 smaller transistors (45 degrees interleaved) building Q2.
The emitters on the outside of the transistors are all connected together.

Hysteresis is usually related to the package of the device.
I.e. the die attach consisting usually of silver filled epoxy resin.
(and perhaps hollow glass filled (low density) resin for the A-device)

With best regards

Andreas

[macfly]

The comparator is referenced to VREF = 2.5 V. Nothing goes negative relative to the potential at the GND level.

No!
I refer to figure 1 of the article.
The complete OP amp's supply are referenced to ground. The supply of the reference also.
The voltage on R1C can only be positive above ground. But a positive voltage feed
into RN2A (buffered by IC2A) would  (with a negative supply voltage for the OP) result in a negative output voltage of IC2B.
But without negative supply, the output voltage will always stay near zero volts or an erroneous phase-reversal occurs.