The LTFLU (aka SZA263) reference zener diode circuit

[Kleinstein]

I agree that a circuit without an external diode has some advantage. Especially it does not depend on the diode and thermal coupling.

I see two possible reasons for the diode: one is having a little larger adjustment range - this may be needed with some ref Chips to reach the zero TC without to extreme currents.
A second reason could be the 2 nd order TC for a non oven use. The diode might reduce the 2nd order TC. In the extreme case one may even be able to also trim the 2 nd order TC.

[iMo]

Any idea what the 11 fuses on the LTFLU chip are good for?
PS: the chip (the source - branadic)

[razvan784]

The fuses are used to trim the transistor area by effectively connecting small transistors in parallel. This can be clearly seen on the high-res die photos. A bigger transistor has a slightly lower V_BE for a given I_C. A slightly different tempco as well. Trimming is probably done at the wafer level, before dicing and packaging - therefore I expect parts that are outside the trim range to be scrapped rather than packaged by LT.
Edit: they might also just be unused, as are the two on-die heater resistors that could possibly be bonded out in a different package but are not in the SZA-compatible one. Silicon revisions are expensive, better have some flexibility in the design...

[iMo]

..it is not a chip with a diode and a transistor wired to the pins, it seems It looks like there are 4 zeners in parallel (each with some small resistor in series)..

[dietert1]

Yes i can see the four zeners with their resistors.

The transistors appear to make something like an 8 bit binary DAC. The "bits" are present on individual pads with remnants of previous bonding/sampling. I labeled the corresponding pads by transistor size/weight. If the die were in a 18-pin package each individual bit could be brought out. In the LTFLU  we know (with four pins) the bits are hard wired by fuses. The pads were used for burning some fuses.
I understand there is nothing like "the LTFLU" and general wisdom will not tell you what kind of LTFLU you have. When you read free_electrons
"Fluke 732B DC Standard Teardown " thread and his description of a "transistor on top of a zener", reality is just a little more complex.

Regards, Dieter

[SilverSolder]

So when the chip was tested, they selected the best performing "bit pattern" and hard wired it? 

[iMo]

There are small "fuses" on the chip near the "programming pads". They blew the specific fuses (by applying a short current pulse) either while the chip was still on the wafer or after they bonded the chip to the package.

PS: would be interesting to reverse engineer the chip

[SilverSolder]

So basically, they would have been able to tailor the characteristics of the transistor to match (one of) the zener(s) and end up with an overall low temperature coefficient, eliminating or simplifying the need to temperature control the part?  Respect!  :-)

[iMo]

From the above information there are several interesting observations:
1. there are 4 buried zeners in parallel
2. there is an 8bit resolution fuse-programmable setting of either a current via transistor(s) or a voltage(?)
3. there are 2 large resistors (at the top and bottom of the picture) not wired, perhaps the heater
4. there are several other components on the chip (like some resistors, diodes).

[razvan784]

After some more thought, this is what I think about these fuses: at the circuit level, they adjust the effective transistor area, nothing more, nothing less. It makes sense perfect for the steps to be binary - wide trim range and fine steps with a small number of fuses. As a sidenote, Vishay foil resistors use a similar approach to trimming.
The low-level transistor equations feature current density rather than current, so essentially doubling a transistor's area is the same as halving its collector current. So in the classic refamp designs you always see the SZA and current-setting resistor listed as a matched set in the parts list - the resistor is hand-selected for a certain zero-TC current. With the newer LTFLU, the part can be trimmed at the factory and a constant current can be used, and more importantly a constant resistor value - a huge plus for manufacturability. The trimming is almost surely done on an ATE-type system at the wafer level, before dicing and packaging, and most probably at room temperature, probably based on voltage measurements at constant current. In support of this theory, please see the LT1236 teardown on this forum. It has fuse trim pads that are used for coarse trim before packaging, as well as bonded-out trim pins labeled Do Not Connect, used for post-packaging fine trim.
This is how I see it at least, only Fluke and LT\AD really know the details...

[iMo]

User zlymex did a teardown in past too:


https://www.eevblog.com/forum/metrology/the-ltflu-(aka-sza263)-reference-zener-diode-circuit/msg913137/#msg913137

PS: Btw, assuming the large resistors there are heaters (perhaps used during that calibration) where is a temp sensor then?

During the calibration they may heat the chip up to the desired temperature (as ordered by Fluke) and set the current (by blowing the fuses) to the zero TC..

[dietert1]

Yes, i noticed that zlymex had seen the binary trim before. Concerning the temperature sensor, i posted a schematic above showing how to extract chip temperature from Ube of the transistor. That seems to work very well.

Ah the other three fuses:
One is between two pads connected to the heater resistors. This must have been some option for selling the chip in a package with the heater pads bonded to pins. The remaining two fuses allow disconnection of base and collector of the largest transistor, while the eight trim fuses are between the zener cathode and the transistor emitters. Disconnecting the large transistor completely avoids leakage when running the chip at low collector current.
And by separating the large transistor, one arrives at something very similar to a LTZ1000, i mean a second transistor for temperature control.

Regards, Dieter

[Edwin G. Pettis]

Here's all the LTFLUs you could ask for:

https://www.jotrin.com/product/parts/LTFLU_1ACH

Have at it.

[iMo]

Imagine you buy 10pcs of the LTFLU chips (random prod date). How do you know for which temperature they set the zero TC?
I still think the chips are "preset" for a "zero TC" at a specific temperature (not the ambient one). Small fine-tuning is done by the external resistor then.

[dietert1]

The measurements of my four LTFLU-1ACH 0625 were taken in the basic circuit at 3 mA zener current. All four chips work well, with low noise (< 0,2 ppm or so). I have not seen any unexpected drift yet, but that's a little early to say considering the age of those parts. TC measurements are at ambient temperatures and accurate to about 0,1 ppm/K. The characterization indicates that i got two different kinds of LTFLU reference.

Two of the parts have TC = 7 ppm/K. Their larger Ube indicates they are configured for lower collector currents. Their TC nulls at about 40 uA collector current. According to Dr. Frank this is an acceptable value. Then they exhibit a TC curve very similar to the curve presented by branadic, with a flat maximum at around 30 to 40 °C.

The other two references appear somewhat different, with a large negative TC and with lower Ube. In order to null their TC at room temperature, they would need about 1 mA of collector current. I can null their TC at 0,08 mA collector current using the temperature measurement and compensation circuit proposed above. Then the nonlinear TC contribution exhibits a minimum at some 30 to 40 °C.

Regards, Dieter

[iMo]

My suspicion comes more less from the general information the chips are delivered to certain vendors only (based on their specific order), and the chips are not offered on the free market, and none from those vendors is using the chip at the ambient temperature (but inside a box at say 35-40C or inside an oven somewhere between 50-60C, or..).

[dietert1]

When i look at the temperature curves of my LTFLUs, they are so flat that i will never see the nonlinear contribution (minimum or maximum) unless i first trim TC to about zero. Then the parabola fit of the maximum is like +0 / -1ppm within +/- 3 °C. In the end, i also want to have the reference inside a precise oven. free_electrons 723B thread gives some insight on how one can do that.

Regards, Dieter

[iMo]

LTFLU RevEng v0.001
A naive attempt so far

[iMo]

And the v0.002 with TC.
With 8 emitters 

PS:  1ppm within 7C..

[iMo]

dT=20degC
EDIT: dV<+/-1ppm

[dietert1]

So we should not only try and tune TC to zero, but we should also learn how to find the largest possible range of small TC. I have no idea how you did that.
The best i could get until now is about +/- 6 °C for a 1 ppm change below maximum resp. +/- 2 °C for a 0.1 ppm change.

Regards, Dieter

[Kleinstein]

For a larger range of low TC it would take a way to not only adjust / trim the linear TC, but also to trim the 2nd order TC (square contribution).  There are a few different possible variations that also change the 2nd order effects: a different zener current, the hight of the effective voltage driving the zener current. A possible diode in series with the resistor to set the zener current.

For a version with oven, there is no real need to trim the 2nd order TC: with regulation square law contributions are very low. It only needs to get the linear TC at the oven temperature below a certain limit.

The extra trim may be needed for a non heated version. As there can be also effects from additional resistors in the circuit it does not make sense to only look at the reference chip. The resistors can also contribute higher order effects.

The simulations have only limited value, since AFAiK there is no reliable model for the buried zener including the thermal effects. The simulation may still help to get the number on how much change is expected from changes in some resistors. 

[iMo]

So we should not only try and tune TC to zero, but we should also learn how to find the largest possible range of small TC. I have no idea how you did that.
The best i could get until now is about +/- 6 °C for a 1 ppm change below maximum resp. +/- 2 °C for a 0.1 ppm change.

Regards, Dieter

Here is a 23.6degC range span below +/-1ppm ("S" shaped).
Mind it is just a naive simulation (not sure it reflects the chip wiring properly). Pretty sensitive to the external resistors values (the external resistors and the opamp with TC=0 here).

[dietert1]

Your simulation is very interesting. By the way the resistors R2, R4, R6 and R8 should be smaller. As far as i remember, differential resistance of the LTFLU zener as a whole was measured by janaf to be about 8 Ohm at 3 mA and 7.4 Ohm at 10 mA. It has always been said the LTFLU zener should get 3 mA, but the 8842A circuit runs at 4,2 mA = (7+0,5 V) / 1887 Ohm. Collector current is 43 uA = 7 V / 162K.

As far as i understand you got a wider temperature range by increasing the zener current and by adding 16 more transistors. Maybe the transistor configuration isn't really about having a preadjusted collector current (in order to use a fixed resistor), but for a wide flat range. This could explain why many LTFLU applications still have those selected resistors of unspecified value in the collector path.

Regards, Dieter

[iMo]

The zeners - the structure around the zeners looks more complex, imho. The question is how is the wiring done in reality.. 

The resistors - you may get the values by counting the number of squares the resistor is made of, and multiply by a resistance per square (I've guessed 100ohm/square).

But again, it would be great an expert in reading the analog silicon would analyze the area around the zeners..

PS: Zeners - here we did a similar exercise while I had used two additional tempco parameters for the zener - https://www.eevblog.com/forum/metrology/compensating-the-temperature-coefficient-of-a-voltage-reference/msg2387187/#msg2387187
The two parameters were not used in this simulation, you may try to add them into the zener's model.

PPS: below
a. with zener's TRS1 and TRS1 tempco parameters, RZ=700ohm, and Rcollector from 15k to 16k step 50ohm
b. with Rc=15.7k you get 24degC span within 1ppm