LTZ1000 unstability Vs output caoacitance

[Armando]

I built a LTZ1000 based volgage reference. folowing the "oficial" schematic of LTZ1000 datasheet.
LTZ1000 shamples come  straight  from Linear Tech , so , they are genuine.
It seems to work properly: (7,02527V), measuring Vout with an HP3557A...BUT!...:
I noticed that Vout changes ,  up to  30uV depending on output load capacitance. That includes  shielding parasityc capacitance of the output cables. It est,  fev houndreds of pF.
With no shielded cables, effect is evidenced approaching hands to them.
Could some one help me to discover the reason ...and the sollution?

Thanks gentlemen

By the way, there is a diference in the schematic of Linear App Note 42, Fig 67,  and the proper LTZ1000 datasheet: In this last brochre there is a 1 MOhm resistor which not appears at the first one. I put this resistor in my card.

[Dr. Frank]

We have a very long thread about the LTZ1000 / LTZ1000A in the Projects, Design and Technical Stuff Forum.

Please visit that, perhaps you'll find answers.

Please, also post your schematic and where you apply the capacitor, and what is the exact value. Your description is far too unprecise.

I only can guess, that you put a capacitor directly on the output of the OpAmp.. don't do that, they are not built for that, and start oscillation.
If you want to reduce output noise, first check, what is the typical, achievable value  (see thread), and then, there are other circuitry how to filter noise.

Then, the original LTZ1000 circuitry is very sensitive to external AC signals, as unfiltered mains supply, or RF sources, or simply inductive disturbances (coffee machines, transformer switching, switch mode power supplies,). Those disturb the temperature regulator, and will also lead to low frequency kicks in the output voltage. 

Frank

[Armando]

Dears Diligent_M  and Dr Frank:

Thank you very yuch  to you both for your fast reaction!.

Following the sugestion of DM,  I conected my device to the oscilloscope ( Tkx 2465) trhough a  x 10 Probe. No traces of oscillation. Noise apperas at the limit of  scope sensibility.

I desconected the buitn in power supply (15V) and substitute it by an external lab PS .   Not improvement noticed.

I discover that the shift of 10 to 25 uV in VREFout  takes place when conecting a capacitor ( for instance 1nF) between +VREFout and general shield of the circuit, that is : the metalic box.

The same capacitor of 1nF between  VREFout and VREF 0V does no produce this unwantwed effect.

Unwanted effec becomes worse when connecting  LTZ1000 PCB ground  track to the shield.

Now I put a toroidal common mode choque (only aprox 100uH) in the four wire cable coming from the PS to the LTZ1000PCB and : Hoppp! the effect dessapeared.

That means I know ( I suppose...) HOW to fix the problema but I ignore really WAY. This is an unhealthy situation...

Finally I put a separate toroids for each couple of wires ( simply precaution): one for Heating pair and other for the zener pair . And only one point of connection with shielded box. Just at the power supply ground point.

In other hand, implementation of the LTZ 1000 seems to be quite carefull:

 I built an aluminium shielding case with separate room for the power supply. AC is filtered by Y capacitors + inductance. (Could be ground capacitor gilty of my problem?)

I manufacture  a couple of binding posts from pure copper (+tellurium?). They are  connected straigh to a PCB non plated copper pads.
LTC critical circuitry is covered in both sides by foam pieces.
LTZ1000 is placed in a "Svastic" PCB shape,  to low termal gradienst and mechanical stress....
I notice tan LTC1000CH is much lower in stabilizing than the LTZ1000ACH . The first is genuine LT The second...unknowwn mother coming from China.

...I am preparing some pictures to illustrate  the above writed.I will post them in a next message

Your sugestions and comments shouold alwais been very appreciated.

Armando

[Armando]

Sending some pictures about my implementation.
Schematic used is exactly that appears at the LTZ1000 datashet.Including 400K resistor as I have a LTZ 100CH.
 Do you have any sugestion to improve it?
Thanks again

[Armando]

More

[Armando]

Layout

[Dr. Frank]

Hello,

as already stated, there's that very long thread in the other (correct) category, where exactly such an effect is described.
The temperature sensing BE diode rectifies any AC disturbances and shifts the temperature stabilization point.
A few ppm as you observe, are typical.
Those disturbances can be mitigated by grounding the circuitry to the case, that is also the effect you see, as the capacitor obviously shorts RF signals.

Most probably your AC supply induces magnetic filed, and/or electrical fields into the LTZ circuitry.. The transformer might be too close.
Therefore, for testing remove  transformer, rectifier and stabilizer from the case, connect the LTZ1000 via a long shielded cable and additionally filter the DC supply.

Make sure, that there is no switch mode PSU near your LTZ1000 circuitry, best would be nowhere in the room the circuit is situated, and not in the AC mains grid.

Those PSUs extremely disturb the LTZ.

Then, leave the capacitor away and test some grounding schemes.

With some testing, you will quickly learn, what is the source of noise, and which is the undisturbed output volatge of the LTZ circuit.

The original circuitry from the datasheet can be used very successful after noise optimization, but it stays sensitive.

Andreas, from the thread I've mentioned, added several caps to the temperature sensing part of the circuit, and the whole circuitry got much more stable.

Btw.: As you are one of the few users of the non-A version, (besides myself), at which temperature do you stabilize the LTZ?
It seems to be 65-70°C (13.5k over 1k), that's a pity, too hot.. 45°C (12k over 1k) is totally sufficient, and gives a much better 1 year stability

Frank

[Armando]

"I discover that the shift of 10 to 25 uV in VREFout  takes place when conecting a capacitor ( for instance 1nF) between +VREFout and general shield of the circuit, that is : the metalic box."


MISTAKE. Just to clarify: The shift was 100 to 250 uV. Sorry by the mistake. I continue investigating....

Other question: Are there a trade- off between termal emf introduced by series elements of an output   LP filter and noise improvement? which shoulb de the frequecncy cut-off of shuch a filter?

[Armando]

Well! I'm ready to  try a zero DC offset low pas filter...

Frank: Do you think that  MAX280 or LTC1062 could be a chance?. At the AN there is howw to use them.

Other posibility is to implement an D element based LPF, but , in this case I have not an schemtic with aprox values or clear design info.

 What Should you suggest?. I'll appreciate very much your opinión.

Thanks a lot

[Armando]

I'm  refering to filter the output, of course.