Ultra Precision Reference LTZ1000

[MK]

A while ago I made myself a spice circuit of the LTZ1000 using some 2n5088/2n5089 parts for the small Q1 and Q2, just used a btz6.2 zener as the core zener just for modelling how it behaved, it seems prone to oscillation almost no matter what I did, could it all be oscillating quietly?

[Andreas]

Hello,

@andreas, i have same "pop corn" problem before, and i didnt quite know the way to solve it.

if i tap on the SOT23 lightly with a insulator stick, it can trigger/reproduce the noise.

Just for noise testing even a more normal resistor would be OK to test a different current. Its allways the possiblity to have a certain current or temperature  the zener diode does not like. So it might be interesting to do a slow temperature ramp (e.g. 1 degree), while doing a longer noise measurement.

I did several tests on LTZ#5:
Knocking on critical components -> no additional events.

Changed current through Zener: 1K resistor in parallel to R1=120 R -> no change in behaviour.
Additionally changed temperature setpoint: 22k resistor in parallel to R5 = 1K -> the first measurements showed no popcorn noise events. Then I did my "last noise measurement". Result: extreme number of popcorn noise within this measurement.
The following measurements showed again numerous popcorn noise events.
So the problem was not gone. (only a "statistical" improvement).

I decided to built the 2 further references immediately to check wether the chopper (LTC2057) is the guilty.
So the 2nd device with LTC2057 (LTZ#6) showed no popcorn noise.
Whereas on LTZ#4 (with LT1013A) showed 1 single (small 0.35uV event = within 3 sigma limits) during 15*100 sec measurement time.

I tried to differentiate between OP-Amp and zener by measuring noise at the OP-Amp Output (the gate of the BF245C FET).
The interesting result is that the noise at this point is a factor 2.9-7.6 higher than at the zener.
The average relative amplitude on LTZ#5 between noise is factor 2.9
The average relative amplitude of the popcorn noise jump heigth is factor 2.0
So the probability is higher that the Zener is the guilty.

But anyway: I think I will try exchanging the current regulator OP first. (since it is cheaper).

With best regards

Andreas

[Kleinstein]

The transconductance of the BF245 is rather low compared to the load impedance, which is only the zener and 120 Ohms. So the sourcefollower will have an AC  gain of only about 1/3 to 1/2. This is why noise is higher at the OPs output.

However this hardly can explain more than a factor of 3. There might be excessive LF noise from the FET: the transistor in the LTZ and the OP will compensate for that - thus the noise of the FET would appear at the gate, which is the OPs output. Still I don't think the noise of the FET will be a problem as it will hardly make it to the output.

[Galaxyrise]

So a possible explanation would be that the hfe of Q1 is significantly higher than that of Q2 for the active cirquit.

How would one safely measure hfe of Q1 in a LTZ1000?  My attempt worked for my mock ltz but gave a garbage value for the real thing.  (I wonder if I've killed another one, now.)

[Galaxyrise]

Why can't you just trust the data-sheet value of 200?  Anything you measure will be for your LTZ anyway-- and since there are always process variations, your value of hfe will not necessarily be the same as the typical data-sheet value, or as someone else's LTZ.

Gives me confidence that I understand what's going on and everything is as it should be.  Having gotten a garbage value demonstrates that at least one of those two is not correct.

[3roomlab]

i have some OT Qn. did mickle T have a thread about how he modded his solartron with LTZ1k?

[Vgkid]

i have some OT Qn. did mickle T have a thread about how he modded his solartron with LTZ1k?

I believe he modded his 7081. I have a document about it, look on here.

[ManateeMafia]

Try here http://www.ko4bb.com/manuals/index.php

Search for 7081. The file is ~100MB.

[Vgkid]

^^^That looks like it, a very large djvu file.

[Andreas]

So the probability is higher that the Zener is the guilty.

But anyway: I think I will try exchanging the current regulator OP first. (since it is cheaper).

Hello,

I did exchange the current regulator OP (LTC2057) and the FET (BF245C).
-> no change. Still having the popcorn noise.
So before I exchange now the LTZ I will try if the noise cures when doing a run-in phase
with a higher temperature setpoint for some weeks.

Has someone experiences in this direction?

With best regards

Andreas

[Dr. Frank]

Hello,

I did exchange the current regulator OP (LTC2057) and the FET (BF245C).
-> no change. Still having the popcorn noise.
So before I exchange now the LTZ I will try if the noise cures when doing a run-in phase
with a higher temperature setpoint for some weeks.

Has someone experiences in this direction?

With best regards

Andreas

Andreas, don't do it! 
Yep, I have some -bad- experience in treating these LTZs with higher temperatures..

You will definitely have hysteresis, and after removing hysteresis, you will have bigger drifts the next 1-2 years, compared to a LTZ w/o excessive temperature increase.

Frank

[Andreas]

Hello,

@Frank: too late: I already put a 5K resistor between J6 Pin 10 + 11 to gently increase the setpoint.
I will not go above 15:1 (like HP) so dont worry.
It´s just trying to ride a dead horse. (I think I will have to exchange the LTZ anyway).

I have read somewhere that the ageing rate of zeners (probably 1N82x) is correlated to the noise.
I think that "popcorn noise" is meant. So this candidate will have anyway a higher ageing rate.
I will try to automate this: e.g. 10 sec high setpoint to 10 sec low setpoint.
I hope that if there is a fault in the die attach that then this may be cured by temperature cycling.

@3roomlab
No certainly not.
warm up time is around 500 ms to (near) steady state.
The popcorn noise events are 1 per minute in average.
(some minutes no event, other minutes up to 4 events).

After setting the setpoint 9 degrees higher there were no events for some minutes.

See also:
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg837667/#msg837667

With best regards

Andreas

[Kleinstein]

If it is the noise from the zener diode, I think chances are better to try a even higher current than higher temperature. Even if chances are not that good a lower current might be worth a try too, as this is a low risk try. Just for noise tests the current setting resistor does not have to be high stabilitiy. Just in case the popcorn noise is comming from the transistor inside the LTZ1000, one might try using the transistor with a slightly higher collector voltage (e.g. add a resistor (100 K range) from the regulated 7 V to the inverting input of the OP).

It is rather unlikely that high temperature burn in will change much. Though chances are there if you go to a range where significant dift occurs. However this irreversible - so it's more like a last chance.

[branadic]

Thanks to our small volt nut meeting last saturday I'm now owner of one of Andreas LTZ1047B boards. So I started building up the circuit as far as I could.
Therefor I used our 3D printer to print a small spacer (material: thermoset) for the LTZ1000 (with 0.7mm holes) that keeps away airflow at the pins and increases the distance between pcb and reference to support long wires at the reference.

The Edwin G. Pettis LTZ resistor set is already placed and thermally coupled with some copper tape.

There are still some components missing, that I need to order as quick as possible to get the beast finished. 

[TiN]

That Jamicon capacitor hurt's my precision eye  .
Nice job, I see you used jelly-bean 2057 too. Don't you love that little guy?
Fresh LTZ too.

What's your plan on usage for this module?

[Andreas]

Hello,

That Jamicon capacitor hurt's my precision eye  .

never mind it´s only to calm down the on board (low drop) voltage regulator.
It will do it´s job.

@Branadic: nice stand off.
Do you have any data sheet about the thermoset material?

But I never expected that you would divide up the 400K resistor into 2 parts.
I would use this guy (which is within a 1% tolerance).
http://www.reichelt.de/1-4W-1-100-k-Ohm-976-k-Ohm/METALL-402K/3/index.html?ACTION=3&GROUPID=3080&ARTICLE=11793&SEARCH=402k&OFFSET=16&WKID=0&

But I personally think that the tolerance of this resistor is not critical.
(except when you adapt it especially on each individual LTZ).

With best regards

Andreas

[Kleinstein]

The suitable resistor for the "400 K" resistor will rather much depend on the LTZ1000 unit, possbly the OP used and also on the thermal design and likely the temperature setpoint (would expect higher resistors at lower temperature) used.  It is a try to compensate the residual temperature coefficient / imperfection of temperature control. So the right value can change a lot - should be more something in the 200 K - 5 M range depending on the individual unit and thermal setup. So I would not expect to need a tight tolerance and also no super low drift - especially with just the first guess of 400 K is used.

The more interesting question is to find a procedure to find (measure) the best value for a give setup - might need to measure temperature coefficient and optimize. So this step might need quite some time, especially without good procedures to follow.

[branadic]

@Branadic: nice stand off.
Do you have any data sheet about the thermoset material?

But I never expected that you would divide up the 400K resistor into 2 parts.
I would use this guy (which is within a 1% tolerance).
http://www.reichelt.de/1-4W-1-100-k-Ohm-976-k-Ohm/METALL-402K/3/index.html?ACTION=3&GROUPID=3080&ARTICLE=11793&SEARCH=402k&OFFSET=16&WKID=0&

But I personally think that the tolerance of this resistor is not critical.
(except when you adapt it especially on each individual LTZ).

With best regards

Andreas

As I sad, I used what I had lying around, but as I have to order some parts the 400k made of two 200k resistors will be replaced by 402k. However, 400k in value is somewhat uncommon. I wonder why the used that special value instead of something available in the e series, if the it is not that important. 
The datasheet for the material is confidential, but as said it's a thermoset.

[Andreas]

Hello,

again some measurements.
PSRR on LTZ#3. (With LT1013A)

For this I have supplied the LTZ at the output of the voltage regulator. (so without stabilisation).
Yes, the LT1763 is one of the few regulators which can be used this way without damage.

From 8.75V to 18.25V the change is around -13uV at ADC input (after 2:1 divider)
 so -26 uV or less than 4 ppm over a 9.5V input range.
Giving -0.4ppm/V for the LTZ cirquit.

below 8.3V the output voltage is in dropout.
Until 8.3V it is stable.
 (ok should have waited some more time to stabilize between switching on
and starting measurement from 9V downto 8.3V)

with best regards

Andreas

[Andreas]

Hello,

next measurements: PSSR on LTZ#5
(with LTC2057 which has Rail to Rail output)

Same setup as on LTZ#3

From 8.5 to 18.5V the change is around -11uV  at ADC input
so 22uV or around 3 ppm over a 10V span.
Giving -0.28 ppm/V as regression curve.

Dropout starts (as expected due to Rail/Rail output)
at the lower value of around 7700mV
Until 7750mV the unbuffered output is stable.

with best regards

Andreas

[ltz2000]

PSRR on LTZ#3. (With LT1013A)

Many thanks!

[Andreas]

PSRR on LTZ#3. (With LT1013A)

Many thanks!

But be aware that I have a additional BF245C
against the original cirquit for the current regulator.
Without this the drop out voltage will be much higher.

With best regards

Andreas

[Andreas]

Hello,

added automated temerature setpoint increasment cirquit
on LTZ#5:
A timer (PIC) switches 2 4K7 resistors via small signal MOSFET
between J6 pin 10+11.
(which is in series to a 2K7 and all in parallel to the 1K setpoint resistor).
So every 10 minutes the setpoint is switched between 12.5 / 1 and 15 / 1.
(roughly 40 deg C temperature change each time).

So lets have a look if this has a influence on the annoying popcorn noise.

With best regards

Andreas

[Vgkid]

That is a pretty interesting idea.

[Andreas]

That is a pretty interesting idea.

Hello,

See Pickering patend.
But in my case I only increase the setpoint.

Some measurement values from the setpoint excitation on LTZ#5 with LTZ1000A:

first:
Current for the whole unit measured between battery and the voltage regulator.
After changeing setpoint it needs the full 10 minutes to finally stabilize for the current.
When setting the low setpoint it needs several seconds (around 10-15s) until the heater is switched on again.
(The rest of the cirquit consumes around 7mA).

The difference between high (around 90 deg C) and Low (around 50 deg C) temperature is around 8mA in current in steady state.
So having around 0.2 mA / deg C. (most probably not linear).

2nd:
Zener output voltage (unbuffered).
On heating the output voltage rises quickly (within 2 seconds until less than around 2ppm of final value).
On cooling down we have the 10-15 sec delay until the zener output is near stability.
Voltage rises from 7204 mV to 7220 mV (around 2220 ppm) or 54 ppm/deg C.

By the way: the "step" within the curves is due to the fact that the 2 resistors are switched on/off with around 4-5 sec delay one after the other.
The X-axis of the voltage measurements is number of measurements with 10 NPLC+offset compensation (0.4 sec/measurement).

3rd:
Setpoint voltage (measured over 1K Resistor).
Setpoint voltage behaves inversely with temperature.
At ~50 deg C we have 533.3 mV
At ~90 deg C we have 451.4 mV
Again the cooling phase needs longer whereas the heating phase goes "instantly".

With best regards

Andreas