The LTFLU (aka SZA263) reference zener diode circuit

[BU508A]

Hi Andreas,

for me they look like fakes.
The printing orientation does not comply with the original ones.
(the pin 1 marker of the package in direction of the LT logo).

The more I'm thinking about it, the more I think you and plesa are right.
I did some checkings before (they weren't obviously not good enough),
but this little detail did not come to my mind.

  Pity me ...   

Ok, in Germany we would say "Lehrgeld bezahlt."

Maybe we can turn this little misfortune of mine into something useful.

What my plan was (and, well, still it is):
I have two Fluke 3330B, one operational, one for the parts.
Both are having a working 10V reference board, based on a SZA263.
Schematics: http://bama.edebris.com/download/fluke/3330b/3330b_partA.pdf look at page 11
Now I want to build some PCBs and compare the new LTFLU-1 with
these old but rock solid boards from Fluke and find out, what can be improved
in respect of noise, stability etc..

Here is my suggestion:
- do some measurements and publish the data of the 10V Fluke 3330B reference boards
- build some PCBs based on the design of the boards above (I'm aware that, for example,
  it is hard to nearly impossible to get the matching resistors from Fluke) but maybe some
  low tempco high precision resistors will do a good job as well
- do the measurements on the new (fake) LTFLU-1 based references (probably I'll get some
  day genuine ones  ) and publish the data
- identify some parameters which can be easily measured to determine, if it is fake or not
- open here a thread with information regarding fake units, how they can be identified and their sources
  to warn other ones

Any ideas, suggestions etc. from your side are highly welcome. 

Regarding this fake pieces I'll keep you updated (if you are interested in).

Regards,

BU508A

actual measurement gear:
- HP 34401A (calibrated)
- Keithley DMM7510 (calibrated)
- looking for a 3458A

[plesa]

For sure they are fake. All the same manufactured date? These could only be obtained as salvage surely?

Also, aren't the LTFLU normally matched with individually laser cut precision resistors as a set?

It is not manufactured in pairs, matching is later process step I suppose at Fluke. Probably after ageing procedure.
This set they are not selling,so it is more their internal set of matched components.

[Theboel]

@ BU508 your idea is very good but I have one problem. Some times a fake component mean "work" but far out of spec like fake sanken power transistor for audio or its not work at all like fake atmega uC now my question let say your parts is fake. Is it "work"

Sent from my GT-I9500 using Tapatalk

[quantumvolt]

'Any ideas, suggestions etc. from your side are highly welcome.'

BU508A, I think it would be interesting if you could get a REFAMP thread going. This one hardly went further than identifying the 4 pins ...

I have 2 old HP 'AMP. REF' original parts (HP used the SZA263 years before Fluke https://www.eevblog.com/forum/projects/pearls-of-the-past-%28only-for-the-voltaically-inclined%29/msg334139/#msg334139) - one running, and also a LTFLU-1 from eBay (same seller as the OP of this thread bought from). But I haven't done much trying to adjust current for near zero TC yet.

I have sold my 2 HP 3458A LTZ1000 boards because I found it very frustrating to try to 'characterize' them with LM399 based instruments (34401 and 2x 34970). Having a reference that is more stable than your measurement instruments is nice - but then : What do I do now? If you can buy 3 or more 3458A meters you can at least start to study relative drift and group stability. But I don't have the cash for that .

So I will use more time on studying electronic circuits. A first goal for a new thread could be to present the (5-10 different) REFAMP based Fluke schematics and try to understand how they work and how to stabilize them.

Just my 2 satang.

[Dr. Frank]

I have sold my 2 HP 3458A LTZ1000 boards because I found it very frustrating to try to 'characterize' them with LM399 based instruments (34401 and 2x 34970). Having a reference that is more stable than your measurement instruments is nice - but then : What do I do now? If you can buy 3 or more 3458A meters you can at least start to study relative drift and group stability. But I don't have the cash for that .

So I will use more time on studying electronic circuits. A first goal for a new thread could be to present the (5-10 different) REFAMP based Fluke schematics and try to understand how they work and how to stabilize them.

Just my 2 satang.

The sale of these 2 references was a big error..
Instead, building some additional LTZ or SZA based references and comparing all of them would have been the correct action.
It is NOT necessary to have a better reference, several ones being 'on par' also do the job (man-with-3-clocks-problem)


You also don't need a 3458A as a reference (especially as it is less stable than a well / better designed external reference), you only need a device which is able to do comparisons between different references.
That can be done by LM399 based devices also, when you use them as a difference voltmeter.


If you are doing experiments with the FLUKE RefAmp, your findings will be appreciated.
Frank

[quantumvolt]

Well ... I guess the judgement of error or not is in the eyes of the beholder. Most things in life - including precision voltage stuff - is in my opinion a matter of priorities and choices.

Both boards are sold to members here, so my guess is that the refs will get a new life in this forum.

The first one went to Australia. The new owner has a 34461 and a 3458. By good luck his measurements are as close to mine as possible to measure (10uV=1ppm on 10v range). I have been following the board more than a year with three Agilent LM399 based 6.5 digits meters. There has been no changes except from temperature variation, and I have data from 20degC to 30degC showing an average TC for all three instruments much less than 0.5ppm/degree (from 7.5 digits RS232 readings).

Even better - here in the tropical climate of the jungle of NE Thailand I have never seen the board go lower than 7.20880 or higher than 7.20884 (maximum change of around 40-50uv or ca. 6-7ppm for 7.2v) on any of the three Agilent meters for temperatures between 11degC and 39degC, and relative humidity from around 40-85%.

Given the fact that the HP LTZ1000 ref board in question is probably 20 years old and the three Agilent meters (34401 + 2x 34970) are 12-16 years old, and the only changes I have observed over more than a year are repeated cycles over temperature change, I do not expect to be able to detect any long time change at all (if there is one, it must be sub ppm).

So I decided that sitting year after year watching a heated avalanche diode (LTZ1000) is not my preferred use of time. I will do other electronic circuit experimenting stuff.

The new owner is happy. He is now more confident that the 34461 is not only within spec, but closer to real value. And I 'know' (very confident) that one of my meters is quite spot on for the 10v range (because it shows the same value as the new in calibration 34461).

The second 3458 board with the AD5791 evaluation board is on it's way to Paris, France. The new owner, also a member here, might do more with the 20bit DAC. That would probably be of interest for many here.

So all in all, I am happy that the LTZ1000's have got new homes. I am very pleased by deciding that my three Agilent LM399 based instruments (that I 'know' better than my wife and children) are the absolute and final limit for all my future precision experiments. Jedem gefällt das Seine.

I might take your advise and try to post some of my opinions and experiments with temperature compensated reference amplifiers.

[quantumvolt]

I thought it could be useful to present some diagrams that in my opinion are essential for understanding REFAMP devices.

The first picture is a generic diagram for a series regulator with negative feedback and error amplifier and an undefined reference. The second picture is a schematic for a discrete transistor implementation of such a regulator with zener diode reference (unregulated zener bias).

The third picture is the HP 3450 implementation of such a regulator where is added a differential amplifier (for higher error amplification and hence better regulation) between the error amplifier output and the series regulating element. The fourth picture is the Fluke 341 implementation of the same thing. Although these devices are from around 1970 or so, the principle of the modern Fluke 732B is the same as these old discrete transistor implementations.

[Dr. Frank]

Well ... I guess the judgement of error or not is in the eyes of the beholder. Most things in life - including precision voltage stuff - is in my opinion a matter of priorities and choices.

Hi, your original justification was quite different:

I have sold my 2 HP 3458A LTZ1000 boards because I found it very frustrating to try to 'characterize' them with LM399 based instruments (34401 and 2x 34970).

That reason is objectively wrong, as you can still compare 2 or more of these more stable references in differential mode.

But I can understand your latest reason, that you wanted to prioritize your researches differently.

A pity though, because there will be no comparison between these references possible, anymore.

Frank

[zlymex]

I thought it could be useful to plot the simplest circuit of 10V refamp both for reasons of understanding and for implementation, so here is my circuit:

The elimination of second divider makes Vc=Vb of Q1, which is actually a good thing.
The elimination of the output drive transistor makes the opamp a bit hotter but no big deal. The output is still short protected.
R1 and R2 should be matched pair of PWW or alike just like before.
R3 is for the temp-co adjustment, the value is varies according to the refamp. Weight is about 1/250, any reasonably good resistor will do.
(weight refers to the fraction of relative change of the output if R3 changes)
R4 is the current bias resistor with weight even less than that of R3.
C1 provide enough phase margin not only stabilize the whole circuit but also makes the output capacitor loadable.
U1 is OP97F just like that in 732B, where 11.5V supply is used. If a rail-rail output is used such as OPA188 or LTC2057, supply down to 10.5V is possible making the circuit 3-rechargeable-lithium-battery operational.

The principle can be now explained with ease: if for any reason the 10V is dropped, the divided voltage will also be dropped feeding to Q1 and the voltage of the collector will rise, U1 gives more thrust to compensate the drop thus remain the 10V stable.

[Kleinstein]

Having the amplification part directly at the reference amplifier was a big advantage in times when the OPs were not that good. As the OP in that circuit is not really critical - like R3 it only enters to a small part at the output. So you could use that circuit even with something like a CA3140 with hardly more drift.  An extra OP for amplification of a 7 V reference like LTZ1000 should be good quality, as the drift fully appears at the output - though this is not a big deal anymore with modern OPs like OP177 or LTC2057.

[zlymex]

@DiligentMinds.com :

Thanks very much!
That's also true for me. I do have quite some LTZ1000, but few LTFLU/263 for experiment only.
Indeed, resistor based boost circuits are troublesome, one of my friends here have successfully developed a PWM circuit that perform very nice and I hope I can get some of his board soon.

On the other hand, that refamp circuit may be regarded as a 7V source, just fetch it from the joint of R1 and R2. I have made all the alterations of my 732A and 732Bs leading-out their internal 7V outside for measurement.

[zlymex]

I'm looking forward to seeing how things went with your friend's PWM boost experience.  Are you going to post this on this forum?

Sure I'll post the test result when I get it. What I know is that he made a lot of versions and took long time to test for the noise, tempco, especially aging.

In the mean time, I have an idea of modifying the resistor-divider-based boosting circuit by employing a LTC1043 to reduce the sensitivity of R1/R2 to <5%, that is, more than 20 times better.

[BU508A]

Just a quick update:

I received today the 5 "LTFLU-1ACH", here are some pictures of them.

So, how to identify them, if they are fake?
I have here a reference board from a Fluke 3330B which runs with a SZA263 and my first idea was:
- solder out the original SZA263
- solder some wires to the solder points
- solder the new LTFLU-1ACH to the wires
- do some measurements and post the results here

But I do not really dare to solder out the SZA (which maybe would be a mistake) but I think using this board as a reference for comparison  would be helpful.

Any suggestions / ideas?

Cheers,

BU508A

 

[zlymex]

I suggest that you make a board of you own with sockets for LTFLU-1. The simplest one will do, or use boardless solder technique, to test the tempco, stability and of course to test whether they are fake or not.

[plesa]

Just a quick update:

I received today the 5 "LTFLU-1ACH", here are some pictures of them.

So, how to identify them, if they are fake?
I have here a reference board from a Fluke 3330B which runs with a SZA263 and my first idea was:
- solder out the original SZA263
- solder some wires to the solder points
- solder the new LTFLU-1ACH to the wires
- do some measurements and post the results here

But I do not really dare to solder out the SZA (which maybe would be a mistake) but I think using this board as a reference for comparison  would be helpful.

Any suggestions / ideas?

Cheers,

BU508A

Where did you purchase them?
I you want to check if they are genuine I can make teardown like we did with LTZ1000A from TiN or fake LTZ1000.
There will be signature on die for sure.
But this will damage the package, but we can have hire picture of LTFLU.

[Andreas]

So, how to identify them, if they are fake?

Any suggestions / ideas?

Hello,

on LT parts the pin 1 marker (nose) of the case is always on top side of printing.
(see also picture of 8842 on first side of the thread)
On your parts the pin 1 marker is at the right side.
So for me they look like fakes or at least re-stamped parts.

You cannot simply drop in a RefAmp into another cirquit.
You allways need to exchange the adjusted resistors of the complete RefAmp-Set together with the RefAmp.

with best regards

Andreas

[BU508A]

Hello Andreas,

thank you for your reply.

You cannot simply drop in a RefAmp into another cirquit.
You allways need to exchange the adjusted resistors of the complete RefAmp-Set together with the RefAmp.

Yes, I have suspected that, because Fluke states in the service manual of the 3330B that
the SZA263 (U2) and the resistors R2, R11 and R15 are selected parts and for a replacement they have
to be ordered together (they have together one part number).
I am just wondering, why they left out R12, which has to be ordered together with R13 as one bunch.

Do you think it could be of any use if I take an experimental PCB and assemble a similiar circuit based on the 3330B design but with standard components?

Best regards,

BU508A (Andreas, too  )

[Kleinstein]

Having a simple board with standard components would be one way to check if they at least are similar to the original.
You likely don't even need a printed board.

If they are fakes, chances are they are not even close in function - more like getting To99 versions of µA741 or µA723.

[Macbeth]

I had the ridiculous pair of fake LTZ1000A - for some reason I was ridiculed by a troll for trying to test it before actually using it in a very expensive board with all the support circuitry etc. Simply measuring the heater was enough to prove they were fakes. Also, lopping off the can and looking at the die - Just like with TiNs real LTZ, plesa made a scan of my utter fake.

However this fake is impressive in that the leads are looking gold plated, etc. The giveaway are the exact same date codes and the tab orientation being wrong.

If you haven't okayed this AliExpress item I suggest lopping a sacrificial one and comparing the die with the published ones on here. In my LTZ case it was obviously fake.

Despite my protests and evidence to AliExpress the same seller carries on selling fake LTZ1000's. I did of course get a full refund. Strangely there are customers who are happy with them and upvote the seller! 

[Andreas]

Do you think it could be of any use if I take an experimental PCB and assemble a similiar circuit based on the 3330B design but with standard components?

Hello Andreas,

should be possible.
There are enough infos in the tread how to adjust the T.C.

With best regards

Andreas

[lowimpedance]

However this fake is impressive in that the leads are looking gold plated, etc. The giveaway are the exact same date codes and the tab orientation being wrong.

Are gold leads an indicator of fakes !, I have a number of LTZ1000's which do not have gold leads and TIN now has 4 of them for further testing, (note the heater test showed approx 270 ohms, and they work in circuit!).
 But as for these REF amps a mock up test circuit is the first step then if still in doubt hopefuly a sample will be sent to Plesa for some 'close ups'.

[Jay_Diddy_B]

Hi,

I found this on Alibaba:





What's the chance of these been fake, given the price and the ability to supply 1E27 parts per day!!!

(Given the world population is about 7.4E9)

Giving the price to 6 decimal places, is obviously an attempt to appeal to 'Volt nuts'.


Given the nature of this part, it was only used by Fluke, the chance of finding large quantities of New Old Stock, NOS, is very slim.

Regards,

Jay_Diddy_B

[Macbeth]

However this fake is impressive in that the leads are looking gold plated, etc. The giveaway are the exact same date codes and the tab orientation being wrong.

Are gold leads an indicator of fakes !, I have a number of LTZ1000's which do not have gold leads and TIN now has 4 of them for further testing, (note the heater test showed approx 270 ohms, and they work in circuit!).

They should have the gold plated leads, but if they have been salvaged (by literally setting fire to the donor boards) then the chinese salvagers will spot weld new plain leads onto the stumps as part of the refurbishment. I can only imagine this causes havoc with the internally aged characteristics, etc, not to mention the now introduced thermal noise with the differing metal junctions.

[Andreas]

They should have the gold plated leads,

The LTZ1000(A) can be ordered with gold plated leads (LEAD FREE) and with tinned leads according to datasheet.
So this is no indicator for the LTZ1000.

With best regards

Andreas

[quantumvolt]

...
You cannot simply drop in a RefAmp into another cirquit.
You allways need to exchange the adjusted resistors of the complete RefAmp-Set together with the RefAmp.
...

This statement is true only as far as Fluke boxes are concerned AND if you expect a similarly low TC of the box after changing the refamp.

The refamp is only a transistor and a zener (avalanche mode) configured so as to trying to have the (current dependent) positive TC of the zener being cancelled by the negative TC of the transistor base-emitter junction (sometimes called the '+/-2mV effect').



This effect has been used in power supplies long before Fluke made their first voltage standard.



ANY discrete NPN transistor and avalanche mode operating diode that is thermally connected will to some degree benefit from this cancelling effect. What Fluke has done is to characterize every refamp and supply it with resistors for it's optimal working current.

In HP3450A Hewlett Packard chose to use a SZA263 'drop-in' circuit with a cooling assembly. No resistors or other component are selected to match the refamp. May be they found it more convenient to reduce the TC of the box further by using a constant temperature chamber for the refamp in stead of 'nulling' the TC by supplying matched resistors for the refamp.



In the Fluke 732 series and other boxes one will find both temperature chamber (in that case an oven) and optimized resistors for minimum TC. This way the temperature dependence of the box is minimized by both low TC for the reference AND a constant temperature for the reference part of the box.

 --- ooo ---

 '+/-2mV effect': pp.38 in http://www.onsemi.com/pub_link/Collateral/HBD854-D.PDF