Compensating the temperature coefficient of a LTZ1000 voltage reference?

[TiN]

I don't see much benefit on using buffer opamps, instead of simple cheap transistor at the output. My goal was to enable +/-20mA at the output without degrading voltage output accuracy, to use as stable voltage source for bridge circuits for example. If that is not needed, transistors can be removed altogether as well.

RC network is to provide some isolation between LTZ ref output and chopper output stage. Noise caused by it is tiny compared to own LTZ output noise, so it can be ignored for typical application, where reference output is integrated/averaged over long time.

[imisaac]

Thanks, CalMachine and TiN for the explanation.

1. Would an additional low pass filter at the output side of the chopper amplifier be helpful to prevent the chopper noise going to the load (e.g. a capacitor across your R2)?

2. Was there a need to filter out the noise from the +/-15 V power supply before connecting them to the transistors? In other words, did the noise from the main power supplies show up at the load voltage?

Thanks in advance.

[Kleinstein]

The transistors are quite good in suppressing supply noise, so no trouble there. The +-15 V (if -15 V is needed at all - could be just -2 V) should be reasonable clean from a linear regulator, but I see no need to spend too much there.

I would be more worried about the supply current to the chopper OPs. So some isolation resistors/ferrites between the +-15 V and the ADA4522 can be a good idea, not to inject too much noise to the supplies.

[Dr. Frank]

In an early eevblog post, several members indeed have attempted to trim the temperature coefficient of a zener to nearly zero. However, the main part used there was LTFLU.

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

In one message (https://www.eevblog.com/forum/metrology/the-ltflu-(aka-sza263)-reference-zener-diode-circuit/msg1285987/#msg1285987), one member (Dr. Frank) mentioned that:

...Temperature stability is achieved by this zener / Ube combination, as the zener has a positive, Ube a negative characteristic.
The LTFLU / SZA obviously is better in this aspect, as  the LTZ1000, due to a smaller zener voltage, which matches better with the T.C. of a pn structure, which in the end allows for a really precise zero T.C. trimming.
The LTZ1000 sucks in this aspect, as it always has about +50ppm/°C w/o oven, and is not really trimmable to zero T.C., any further....

Does anyone know the reason why the +50 ppm/°C tempco of LTZ1000 w/o oven can not be trimmed further? For a 7 V output, this tempco corresponds to +0.35 mV/°C. Is he implying that finding a component with the negative tempco (i.e. -0.35 mV/°C) would be difficult?

Thanks in advance.

If you read my other post: https://www.eevblog.com/forum/metrology/eu-calclub/msg2420097/#msg2420097
, then you would understand, that this generic T.C. of the LTZ1000 is due to physics, or let's say due to the bad match between the diodes and the zeners T.C.s.
The LTFLU, the SZA263 and the other predecessors, are all optimized in this aspect, by simply having a lower reference voltage.

We already have discussed and tested the reduction of the T.C. (generic and composite one) of the LTZ in many, many posts, and there is simply no good way to reduce the generic T.C. (e.g. by adding a 22 Ohm resistor in series with the zener), and there is a really simple way to have near zero composite T.C. of the whole circuit, description e.g. here:
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg1336581/#msg1336581

Frankly speaking, you're .

Why do you want to re-invent the wheel, searching for a complicated, over-engineered solution,  instead of simply adapting the already described schematics and procedures?

Frank

[imisaac]

Thanks, Kleinstein as usual. By "isolation resistor", do you mean to put a highly ohmic resistor in series to the +/-15 V power supply outputs? If so, doesn't that limit the amount of current available for the ADA4522.

[imisaac]

Thanks, Dr. Frank for your detailed references.

1. I am trying to understand the physics limitation that you referred to:

There ist a deeper solid state physics reason/law, why ALL the LTZ1000 chips have about this T.C.
Any silicon diode with CC has quite exactly -2mV/°C.
The T.C. of zener diodes is always around zero @ 6.2V zener voltage, like 1N821.
It is predictably more and more positive at higher zener voltages.

Fluke has chosen and specified a Reference Amplifier voltage, so that the sum of both T.C.s nearly cancel, and that is about 6.9 V. The LTFLU is explicitly specified, that T.C. can be trimmed to zero by slight change of collector current.
As the LTZ 1000 has a zener voltage of more than 7.0V, its sum T.C. is  definitely higher, and predictibly at these +55ppm/K @ the 7.118V, which I measured.
Therefore, for physics reasons, the 3ppm/K must be wrong, or TiN has got an LTZ1000 with about 6.9V reference voltage.

If LT creates a new version of the LTZ1000, they should change the diffusion for a 6.9V ref. voltage.

Frank

The LTZ1000 is made of a transistor and a zener diode. I agree that the tempco of the two may not perfectly cancel each other and therefore leading to the 50 ppm/^oC without using the on die heater. However, it is not so clear that this tempco can not be trimmed further down by a special circuit topology. At least in the simulations that have been carried out in this thread, it is indeed possible that this tempco can be manipulated further. It looks like this approach has not been attempted so far in this forum (please correct me if I am wrong).

Therefore, where is the solid state physics limitation coming from again? Thanks in advance.

2. It has been very informational and helpful to be able to discuss with the experts like you and many others in this thread. I think part of this fruitful learning is due to the fact that discussion has been tailored for a not so advanced person. It is unclear if this learning process can ben deemed as "reinventing the wheel"......It would be interesting to see the measurement results from this unattempted circuit scheme and compare with existing LTZ1000 based reference circuits.
 

[Kleinstein]

The way the LTZ1000 is build, there just is no good way to compensate the unheated TC. It is possble to use the 22 Ohms resistor, but this resistor would than have more influence than the savings from a better TC before the heater: one resistor with 1 ppm/K (the 22 Ohms) tends to be more expensive than 2 resistors with 3 ppm/K (the 1 K / 14 K divider). Things could change if the LTZ would start with a lower TC, but the way the LTZ are made is like it is.  No need to repeat this discussion  .

I would very much like an LTZ1001 starting with some 6.9 V and low unheated TC, so that the collector current (e.g. R2) could be used to adjust the unheated TC to zero. However this would be a different chip. I doubt one could get such out of tolerance chips.

The "isolation resistor" in the supply is a common technique used if RF circuits - get the supply to circuit parts through a resistor or inductor, so that local decoupling of the supply is really effective. So for the ADA4522 with some 2 mA supply current (ideally not much load current added) one could use some 100-500 Ohms with only some 0.2 -1 V lost for the supply. Modern OPs don't need a super stable supply, it's more improtant to keep out RF trouble.

[imisaac]

Thanks, Kleinstein.

Is there a need for a bypass capacitor right at the power supply pin of the ADA4522 as well? If so, then the capacitor and the isolation resistor form a low pass filter, right?

To facilitate discussion, an effective circuit containing various suggestions collected so far in this thread is attached.

[MiDi]

U7: Inputs flipped

I guess the bjt buffers Q2, Q3 want to be inside nfb of op amps.

Not shure about C4: phase margin, leakage

To protect op amps, inputs should be current limited.

Edit:
R13, R14 make only sense with bypass capacitor.

[chuckb]

Pin 7 of U5 (LTZ1000) needs to be grounded for Q1 to work correctly.

[TiN]

In my reference design power is also separated, LTZ reference is powered by +11V from LT3042 LDO, while output amplifiers are powered directly from +/-12V input voltage. As result I do not have R13/R14 resistors, only decoupling caps at the opamp power pins.

[imisaac]

Thanks, all!

There were many mistakes picked out. Here is the revised version. In particular, the outputs are now in the control loops of the chopper amplifiers and the emitter of the LTZ1000 Q1 transistor is correctly grounded.

Question:
Is it ok if 100 nF decoupling capacitors to GND are added to all the bases, collectors, and emitters of the buffer NPN and PNP transistors at the output? The hope is to reduce the chopper noise at high frequency appearing at the load.

[TiN]

I recommend adding low ohm resistor on low side output opamp between Low post and inverting input to have room with EMF battle.

Also what is the purpose of R8?

[Andreas]

Hello,

I recommend adding low ohm resistor on low side output opamp between Low post and inverting input to have room with EMF battle.

It is absolutely necessary to limit the input current to maximum 10mA. Otherwise the clamping diodes between +/- input will be destroyed when the negative output is forced to +10/15V.
I would also compensate the output against capacitive loading and balance the input resistors.

But there are at least two further errors in the cirquit ....
Who will find them?

with best regards

Andreas

[TiN]

Missing cap around U6, and not a kelvin connection/wrong reference for low side output opamp?
Ah, Q3 as well

[branadic]

and there is a really simple way to have near zero composite T.C. of the whole circuit, description e.g. here:
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg1336581/#msg1336581

Frank, did you compensate the zener circuit only or the overall circuit including the t.c. of the 10V boost stage?

-branadic-

[imisaac]

Thanks, TiN. I need some help in the choice of several resistors in your circuit.

1. A 1Ω resistor is now added in the negative feedback path of the "Low" ADA4452 (revised figure attached). Other resistors are also copied from your original boosting stage design. By the way, how was the 1Ω value determined? Was it calculated from the biasing parameters of the ADA4522, e.g. Vos, Ios...etc? Or was it related to the noise of the opamp?

2. The base current of the BJT transistors are very small because β is presumably large. Why are R17 and R18 needed and how are their values derived?
 
3. The input bias current, multiplied by the R16, will give a small voltage offset at the non-inverting input of the ADA4522. This small offset will show up at P2 (output Low pin). How is this desirable?

4. I am not sure the functionality of R8 myself. I thought it was for protection (when combined with the diode D1) or some sort. Maybe imo or some other experts can explain better....sorry.

5.

Missing cap around U6, and not a kelvin connection/wrong reference for low side output opamp?
Ah, Q3 as well

Where exactly in U6 and Q3 to add a capacitor?

[imisaac]

Thanks, Andreas.

But there are at least two further errors in the circuit ....
Who will find them?

Could you provide more hints?

[Andreas]

Missing cap around U6, and not a kelvin connection/wrong reference for low side output opamp?
Ah, Q3 as well

I only meant Q3 which is in reverse mode.

and of course C8 which gives instability of the output.
(should be between Pin 1 + 2 of the OP-Amp and could be reduced to some nF)

Thanks, TiN. I need some help in the choice of several resistors in your circuit.

R15+R16 in sum have to limit the current across the input pins to less than 10-20 mA. (in case of short cirquit of the -output to 10V or 15V).

I also fear that you need some load from the negative output to the positive, otherwise you will have instable output.
(you could also connect R2 to the negative output instead of gnd).

Do you really want to use VDD + VFF as extra supply? (if yes why is the high current output connected to the zener supply)

with best regards

Andreas

[imisaac]

Thanks, Andreas.

1. Q3 polarity is now corrected.

2. A capacitor is added between pin 1 and pin 2 of the "high" arm ADA4522.  Is the purpose of this capacitor shorting the chopper noise back to the op amp instead of the load? If so, why not short the chopper noise directly to the ground?

3. Why is the output unstable in the open circuit condition (and hence needing a dummy load)?

4. If I understand you correctly, you suggest to connect R2 to the P2 such that any drift in P2 will not change the voltage difference across P1 and P2, right?

5. I do not understand your question about the power supply. Which "zener supply" are you referring to?

The revised circuit is attached for further examination.

[TiN]

Sorry imisaac, it sounds like you would be much better off actually building circuit on breadboard and playing around it with DMM, as it feels like we design your ref for you. While it's all good and fun for all involved, we have no idea if our understanding of the project matches what you want out of all this. This already went quite a bit out of topic, which was about LTZ circuit itself. Boosting 7V to X V is separate (much more challenging than LTZ ref itself!) topic.

This is just one example:

I am not sure the functionality of R8 myself. I thought it was for protection (when combined with the diode D1) or some sort. Maybe imo or some other experts can explain better....sorry.

If you don't know what R8 is doing (neither I am), why you add it into circuit?

You don't need LTZ reference to test out the output stage, you can take LM399 which just need power and one resistor (PTF56 will be fine) and a DMM. Then by experimenting you can answer all of your questions (and surely will also get much more questions that might need attention instead). You will see from actual results if the outcome fits your needs or not, from which point you can go refining the experimental setup. Many of the high-end metrology circuits are counter-intuitive (e.g. adding more "parasitic" parts that does not do anything or make more noise to balance out/mitigate worse issues like thermal EMFs), and trying to take all of these hacks you see on board A or board B without knowing the impact will only confuse you more. Heck, the heated debates about two little capacitors in LTZ circuit here on forums already caused few people moderating themselves out and unwanted personal attacks. 

Again, I'm not trying to discourage you from posting or discussing things on forum. My intention is the opposite, have alternative approach and play with the circuit on the bench, as no theory or simulation can account for all real-world complications. Otherwise we collectively just tearing your design apart, everybody pulling resistors and signals direction they see best fit (out of best intentions, not trying to harm). I designed and implemented my output stage because it fit my project needs well enough. Is it most stable possible output stage? I'm sure it is not, but it works for me. 

[imisaac]

Thanks, TiN for the advice.

Regarding the boosting stage, it does seem that the boosting stage is trickier than the LTZ reference and all the questions that was raised so far related to the circuit was to understand the circuit before building so that the component quality can be decided. That is, I agree that the function of R8 should be understood before being placed on the real circuit.

I think you are suggesting to have a new thread for this boosting stage because it is off topic from this original thread topic, right? Yes, we shall do that. The reason that the boosting stage was discussed here was that the tempco of the LTZ1000 tempco setting resistors can actually correct for the tempco of the boosting stage circuit as well.....so not completely off topic.

In addition, I agree that building a real circuit is better than simulation when it comes to circuit evaluation. However, some simulation is still invaluable at the beginning stage to have a rough picture of the circuit behaviour.

Sorry imisaac, it sounds like you would be much better off actually building circuit on breadboard and playing around it with DMM, as it feels like we design your ref for you. While it's all good and fun for all involved, we have no idea if our understanding of the project matches what you want out of all this.

You are right that this circuit is being designed collectively by all of us in this forum. The original goal was to explore how the tempco of LTZ1000 can be compensated by a transistor so that a better stability may be reached. Therefore, many theoretical questions about the circuit limitations were raised. I do not understand why you think this circuit is being designed just for me. That is in contradiction to the open nature of the forum.....maybe that is not what you meant. Anyway, the most important point of doing any discussion is that

...it's all good and fun for all involved...

quote by TiN

In summary, I agree that we should move the discussion of the boosting stage to a different thread after some initial test results are available.