The LTFLU (aka SZA263) reference zener diode circuit

[AG7CK]

I forgot to say: R3 adjusts the bias zener current through R2 and CR2 (in this case some 2.5+ mA - iirc it usually is about 3 mA). R15 and R16 sets the collector current of the refamp so that the tempco of the BE-diode and the zener cancel.

This circuit is NOT designed to deliver anything like 10v as mentioned in some posts over. Also - it does not tap the base voltage of the vref directly. The refamp transistor is "saturated" so that the collector is at virtual ground (because the base is grounded). Hence one gets -6.5 to -6.9 volt "zero tempco" voltage on the reference output (relative to ground).

[mimmus78]

I forgot to say: R3 adjusts the bias zener current through R2 and CR2 (in this case some 2.5+ mA - iirc it usually is about 3 mA). R15 and R16 sets the collector current of the refamp so that the tempco of the BE-diode and the zener cancel.

This circuit is NOT designed to deliver anything like 10v as mentioned in some posts over. Also - it does not tap the base voltage of the vref directly. The refamp transistor is "saturated" so that the collector is at virtual ground (because the base is grounded). Hence one gets -6.5 to -6.9 volt "zero tempco" voltage on the reference output (relative to ground).

Well after you make the math it's pretty obvious that total differential voltage is more than the usual 10V ... even by just looking at R3.

What I don't understand is why they used much more driving voltage than needed. Wasn't the module been more stable if those two resistors had a smaller ratio?

Inviato dal mio ONEPLUS A5010 utilizzando Tapatalk

[Kleinstein]

The higher voltage increases the amplification provided by the transistor. So the OP gets less important. This might be a relict from old times when an 741 was considered a good OP. If for some reason you have the +15 V available, there is no reason to not use the voltage available - the heat would otherwise end up in the OP and nothing is gained.

It depends on the details of the zener diode, whether CR2 is actually needed and if does some useful compensation. The first order TC compensation can be done without the diode as well, just through the Zener current and maybe transistor current. With an upper voltage at 10 V the effect of the diode is not very large anyway.

[zhtoor]

This circuit is similar to some older Fluke circuits and others using appr. +- 7v ref. I am sure you did not "find" it laying around somewhere, so maybe you could give information or link to where it originates.

hello,

the reference circuit is for Fluke 8520A 5-1/2 digit dmm.

regards.

-zia

[AG7CK]

Thanks to mimmus78 and zhtoor for providing and identifying this refamp-circuit from Fluke 8520A (5.5 digit DMM).



The topology is the same as that used in Fluke 8505A (6.5 digit DVM w/ claimed 24 bit discrete ADC and "7.5 digit performance and display in the native 10v range averaging mode").

I find it very interesting that the essentially same reference circuit can be improved an order of magnitude or more by selecting refamp (-set) and refining the circuit a bit.

[cellularmitosis]

I had the good fortune of finding 3 SZA chips on ebay (they are from hifi-szjxic, so I am hoping they are not fakes).

Last night I read through Lyzmex's excellent thread over on 38hot: http://bbs.38hot.net/thread-177-1-1.html

However, I noticed that all of the circuits in that thread, and all of the circuits in this thread (with the possible exception of the fluke 8520 circuit?) are true "refamp" circuits, in that they utilize the amplification (beta) of the transistor.

However most of these circuits suffer from having an "attenuation" of only 3 for the most sensitive resistor pair.

One of the big advantages of the classic bootstrapped zener circuit is that the resistors have very good attenuation (if the zener has good dynamic impedance).  For example, the LM399 has an impedance of about 1, so a 1k zener resistor will attenuate op amp voltage output errors by 1000.

I'm interested in seeing if we can gain the advantage of the bootstrap circuit, and get good resistor attenuation, by discarding the "amp" functionality from the "refamp".

Starting from the bootstrap circuit, the simplest thing to do is simply insert the Vbe junction into the circuit, ignoring the collector pin (see attached 1).

I think this would as expected if you simply threw a 2N3904 and a 5.6V zener onto a breadboard, but I suspect that this won't work with the SZA, as I suspect the transistor cannot handle the full zener current through its base.  Can anyone confirm that?  Is there a way I could find out how much current the transistor can handle, without sacrificing an SZA to find out?

Assuming the transistor can't handle the full zener current, we need to have independent resistors for the base and zener currents (another way of stating this is that we need to add an extra resistor to inject additional current into the zener).  This is circuit 2 (attached).

However, this isn't ideal, as my understanding is that the collector current is what gives you the ability to tweak the zero-tempco point.  So we can also add a third resistor to have independent control of the collector current.  (see circuit 3).

This is still the classic bootstrap circuit, and the base pin (the most stable point of the circuit) is what controls the op amp loop, so this should be as stable as we can get.

Does this approach work?  I plan on trying it out when my chips arrive.

I think the biggest disadvantages of this approach are that you only get a 7V output, and it is a high impedance output.  However, for me, that's ok.

Any feedback / thoughts are very appreciated!   

[Andreas]

Starting from the bootstrap circuit, the simplest thing to do is simply insert the Vbe junction into the circuit, ignoring the collector pin (see attached 1).

Hello,

if you short the base + collector you will get a better "ideal" diode (the base emitter resistance cancels out by beta).
This will also help to increase zener current.

with best regards

Andreas

[Kleinstein]

Using the amplification of the transistor makes absolute sense - it reduces the effect of the OP. It is not that important anymore with modern OPs, but the extra amplification of the error signal essentially comes for free (no extra costs and not much (if any) noise).

The simple circuit to give a roughly 7 V reference should not have a high sensitivity to resistor drift. Those to output a higher voltage naturally have the more critical resistors as these set the gain. AFAIK these resistor attenuation is a little less than with the LTZ1000, but not that much.

Using a 5.6 V zener and a small transistor can be a first approximate circuit for tests.

Even if the transistor can handle a high base current, there usually is extra series resistance, that gets more important when the base current is high. So the stability will likely suffer if the main part of the transistor current is not flowing through the collector as intended.

The LM399 has internal resistors to set the actual zener current. So one can not directly compare the differential resistance to bare zener refs - there are critical resistors inside the LM399. So the drift is not avoided the resistors are just in the case.

It is a good idea to do a simulation (e.g. LTSpice)  of the circuit to check both the sensitivity to resistors and also to check the loop stability. Due to the extra gain from the transistor the loop stability is usually not that simple and usually needs at least one extra capacitor.

[cellularmitosis]

Ah, I hadn't considered that the same resistor pair could have different attenuation, depending upon where you "tap" it (the 10V node or the 7V node of the divider).

I'll play around in LTSpice this evening and breadboard up a dummy circuit while waiting for the chips to arrive.

Andreas, in the situation where the base and collector are shorted, does the overall transistor current then allow for tweaking the zero-tempco point?

[Kleinstein]

With base tied to collector the transistor current can still be used to adjust the TK.

The normal circuit is not that much different: quite often the base and collector will still have the same potential, just with the control loop taken from the collector to use an amplified signal to control the rest of the circuit to make the the difference zero.


The Fluke 8520 circuit is kind of typical for a circuit without extra voltage gain. The hight of the auxiliary voltage (here the positive reference) will set the gain of the transistor. 

[eurofox]

I had the good fortune of finding 3 SZA chips on ebay (they are from hifi-szjxic, so I am hoping they are not fakes).

Last night I read through Lyzmex's excellent thread over on 38hot: http://bbs.38hot.net/thread-177-1-1.html

However, I noticed that all of the circuits in that thread, and all of the circuits in this thread (with the possible exception of the fluke 8520 circuit?) are true "refamp" circuits, in that they utilize the amplification (beta) of the transistor.

However most of these circuits suffer from having an "attenuation" of only 3 for the most sensitive resistor pair.

One of the big advantages of the classic bootstrapped zener circuit is that the resistors have very good attenuation (if the zener has good dynamic impedance).  For example, the LM399 has an impedance of about 1, so a 1k zener resistor will attenuate op amp voltage output errors by 1000.

I'm interested in seeing if we can gain the advantage of the bootstrap circuit, and get good resistor attenuation, by discarding the "amp" functionality from the "refamp".

Starting from the bootstrap circuit, the simplest thing to do is simply insert the Vbe junction into the circuit, ignoring the collector pin (see attached 1).

I think this would as expected if you simply threw a 2N3904 and a 5.6V zener onto a breadboard, but I suspect that this won't work with the SZA, as I suspect the transistor cannot handle the full zener current through its base.  Can anyone confirm that?  Is there a way I could find out how much current the transistor can handle, without sacrificing an SZA to find out?

Assuming the transistor can't handle the full zener current, we need to have independent resistors for the base and zener currents (another way of stating this is that we need to add an extra resistor to inject additional current into the zener).  This is circuit 2 (attached).

However, this isn't ideal, as my understanding is that the collector current is what gives you the ability to tweak the zero-tempco point.  So we can also add a third resistor to have independent control of the collector current.  (see circuit 3).

This is still the classic bootstrap circuit, and the base pin (the most stable point of the circuit) is what controls the op amp loop, so this should be as stable as we can get.

Does this approach work?  I plan on trying it out when my chips arrive.

I think the biggest disadvantages of this approach are that you only get a 7V output, and it is a high impedance output.  However, for me, that's ok.

Any feedback / thoughts are very appreciated!   

I got a LTFLU in my Fluke 8842A multimeter   

[AG7CK]

Hi plesa,

Let me know if you know cheap source of LTFU (including not working). Die image can follow

I've bought here today 5 pieces of LTFLU-1ACH. Price per piece: U$ 25.-- plus shipping plus Paypal fee.
http://goo.gl/aD205s  (Link points to de.aliexpress.com)

Cheers,

BU508A

This is a photo I got from Barry (seller at this webshop)

EDIT: I guess I quoted the wrong post. This is the photo from post #87 https://www.eevblog.com/forum/metrology/the-ltflu-(aka-sza263)-reference-zener-diode-circuit/msg901493/#msg901493:




Was it ever established if this was the real thing or not? There is now a seller with 2000 (?) pcs with the same date code 0625. 2pcs costs some 55ish USD. Wondering if I should buy 2 pcs for fun.

I emailed him and he sent me this picture:



https://www.ebay.com/itm/2PCS-X-LTFLU-1ACH-CAN4-LT/332495632007?hash=item4d6a48b687:g:QCkAAOSwFyhaPL0K

[eurofox]

Hi plesa,

Let me know if you know cheap source of LTFU (including not working). Die image can follow

I've bought here today 5 pieces of LTFLU-1ACH. Price per piece: U$ 25.-- plus shipping plus Paypal fee.
http://goo.gl/aD205s  (Link points to de.aliexpress.com)

Cheers,

BU508A

This is a photo I got from Barry (seller at this webshop)

Was it ever established if this was the real thing or not? There is now a seller with 2000 (?) pcs with the same date code 0625. 2pcs costs some 55ish USD. Wondering if I should buy 2 pcs for fun.

I emailed him and he sent me this picture:

I already order from those little devils 

[Edwin G. Pettis]

You guys do realize these FLUs are genuine fakes, outside of maybe a very few being removed from broken equipment, there will not be any real, brand new genuine Fluke LTFLU-1AHs on the open market.  LT knows where everyone of those chips goes and who bought them, just like the LTZ1000/As and there are never any quantity of genuine 'surplus' chips, especially in any large quantities on the open market.  It isn't that hard to put a transistor and zener diode on a chip, just about anyone can do that, what they can't do is produce a genuine part.  By all means test the little buggers but they aren't going to be the same.  Linear Tech is the sole source of these, they're not like the old Motorola SZA263s, they are similar of course but not the same.

[eurofox]

You guys do realize these FLUs are genuine fakes, outside of maybe a very few being removed from broken equipment, there will not be any real, brand new genuine Fluke LTFLU-1AHs on the open market.  LT knows where everyone of those chips goes and who bought them, just like the LTZ1000/As and there are never any quantity of genuine 'surplus' chips, especially in any large quantities on the open market.  It isn't that hard to put a transistor and zener diode on a chip, just about anyone can do that, what they can't do is produce a genuine part.  By all means test the little buggers but they aren't going to be the same.  Linear Tech is the sole source of these, they're not like the old Motorola SZA263s, they are similar of course but not the same.

I got that in mind when I order it, I add in my comments of the order "Must be genuine"

I got in mind as well that Fluke probably assemble production in China like many other instruments producers and there could be some leftover ...

I will test them and we will see ...

A least I have a genuine LTFLU in my Fluke multimeter 

[Pipelie]

even this is the genuine one, it's probably a remark, refurbished, and pins are probably ’re-connected‘. 

[eurofox]

I got my LTFLU's today, based on the logo it look genuine compared to the logo from the company.

Just  the date code is suspicious : they all got 0625, mayby all the same batch ...

I check the presence of a zener between pin 1 & 4.

[gamalot]

HELP! Could anyone give me the pinout diagram of LTFLU-1ACH?

THANK YOU!

[gamalot]

HELP! Could anyone give me the pinout diagram of LTFLU-1ACH?

THANK YOU!

I can't believe that I've made such a STUPID mistake!!! 

[eplpwr]

Hi!

I've built the circuit from Dr Frank, post #155: https://www.eevblog.com/forum/metrology/the-ltflu-(aka-sza263)-reference-zener-diode-circuit/msg1066470/#msg1066470

I've bought - from two different chinese sellers - the LTFLU-1ACH with the printing "wrong" (LT-logo not at the tab) and the very same datecode "0625"; 10 + 10 pcs for a total of 20 pcs. What I don't understand is if the reference needs to be heated, I am seeing voltage drift by just getting close to the LTFLU, and quite obviously I do not have zero tempco. A 3D-printed "hood" for the LTFLU slows down the short-term voltage-changes, but long-term (minutes, hours) it drifts according to room temperature. I'm measuring the circuit with a DMM7510 (which uses a heated LTFLU i think).

Is it possible to design the circuit for (near) zero tempco, or is it neccesary to keep the reference heated at a constant temperature to keep the output stable? Or - worst case - is the component totaly fake and not worth the effort/time?

If someone wants to cut up one of my LTFLU:s and watch them in a microscope, I'm all in. Probably it's not needed since I seem to have components from the same batch as others, regarding the rotated print and the same datecode.

[Kleinstein]

If the part performs reasonably well, especially low noise, chances are the part is good.  Only way to have a working fake i could imagine would be a selected 2DW232 reference with a transistor - but even this would show up as a rather low voltage.
Even a well working Chinese copy might be worth a circuit to do the real long term test.

So the suitable test would be a simple unheated test-circuit with a roughly adjusted current and than a noise measurement. This could not tell the difference between an LTFLU and SZA263 (old Motorola), but between  and   .

The LTFLU can be used heated or not heated.  The non heated circuit would need carefully chosen resistors for the given unit to get a low TC. AFAIK Fluke has these as a set in there BOM. At least the linear TC can be adjusted, some square law part will likely remain. So this might work well over a limited temperature range. AFAIK the reference in the Fluke 8046 and  DMM7510 is not heated - but likely uses extra numerical corrections.

The other option is to use less accurately set current (and thus a larger TC) and than add temperature control, usually for the whole reference circuit including scaling to 10 V or whatever. This is used in the Fluke 732 etc.  Depending on the quality of the current adjustment the temperature does not need to be as stable as the LM399 or LTZ1000, as the TC to start with can be smaller (e.g. around 5 ppm/K vs 50 ppm/K for the LTZ). Fine tuning (to get near zero linear TC near the set temperature) could be done with the temperature or current if really needed.

[Echo88]

Fluke 5500A also doesnt use a thermostat for the LTFLU: https://www.eevblog.com/forum/metrology/fluke-5500a/msg1432755/#msg1432755 Picture A6_2

[gamalot]

If the part performs reasonably well, especially low noise, chances are the part is good.  Only way to have a working fake i could imagine would be a selected 2DW232 reference with a transistor - but even this would show up as a rather low voltage.
Even a well working Chinese copy might be worth a circuit to do the real long term test.

So the suitable test would be a simple unheated test-circuit with a roughly adjusted current and than a noise measurement. This could not tell the difference between an LTFLU and SZA263 (old Motorola), but between  and   .

The LTFLU can be used heated or not heated.  The non heated circuit would need carefully chosen resistors for the given unit to get a low TC. AFAIK Fluke has these as a set in there BOM. At least the linear TC can be adjusted, some square law part will likely remain. So this might work well over a limited temperature range. AFAIK the reference in the Fluke 8046 and  DMM7510 is not heated - but likely uses extra numerical corrections.

The other option is to use less accurately set current (and thus a larger TC) and than add temperature control, usually for the whole reference circuit including scaling to 10 V or whatever. This is used in the Fluke 732 etc.  Depending on the quality of the current adjustment the temperature does not need to be as stable as the LM399 or LTZ1000, as the TC to start with can be smaller (e.g. around 5 ppm/K vs 50 ppm/K for the LTZ). Fine tuning (to get near zero linear TC near the set temperature) could be done with the temperature or current if really needed.

I think the LTFLU in DMM7510 is heated.

[branadic]

So Andreas left some LTFLUs here on Metrology Meeting 2019 for x-ray analysis.
Since I couldn't resist, yes I've asked Andreas before, I dirty hacked together some components on a breadboard/veroboard. The circuit is based on the schematic presented by Frank here:

https://www.eevblog.com/forum/metrology/the-ltflu-(aka-sza263)-reference-zener-diode-circuit/msg1066470/#msg1066470

I used what I had at hand: LT1006, simple 1% through hole resistors and a precision socket for the reference. R15 is made by 9.1k||180k, R7A simply by paralleling 2x 10k, R7B is 10k and instead of R13 I connected my GenRad 1434-G.
The output voltage connected to my R6581D and the circuit powered with 15V from a lab power supply I used my Ersa icon set to 150°C and touched the top of the TO package to slightly heat up the reference and read the change of output voltage in particular its direction when warming up. This way I was able to determine a resistor value of ~23k needed to have almost no change in output voltage, thus zero t.c.
I then started overnight measurement and went to bed. This morning I stopped the measurement, checked the readings and continued measuring. During the day the reference stabilized, obviously the reference had to relax after my heat treatment.
The veroboard is clamped in a bench vise hanging in free air, so results are not to bad considering that everything is hacked together with cheap components, not packaged and not ovensized.
Need to prepare a temperature sensor and mount it close to the reference so that we can get some temperature readings too.
And yes, output voltage is not trimmed to 10.00000V, but that wasn't the purpose of this quick look onto the part that where called to be LTFLU fakes.

-branadic-

[Kleinstein]

The settling / dirft of the reference circuit could very well be to a large part from the resistors used for the 7 to 10 V part. To really check the reference one could measure the 6.9 V point at the base of the reference transistor (a series resistor for safety would be a good idea here).

Still I would not complain about such a reference.