Ultra Precision Reference LTZ1000

[Andreas]

Hello,

I did some changes to LTZ#3 which had a too high amount of T.C. (-80ppb/K).

see also:
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg881955/#msg881955

now it is nearly in final state:
- slot between regulator and LTZ
- TEKO shield soldered
- and: I tried to reduce the T.C. by "half shortening the legs"

from LTZ#4 I know that mounting the reference directly on the pcb makes a difference of around +0.1ppm/K against long legs.
So I did not want to "overcompensate" the T.C. and tried a shortening of around 5 mm on the legs.

The result:

The T.C. regression curve has changed from -80ppb/K to around -40ppb/K.
Which is now "within spec". (50ppm/K of datasheet)

Now the question is: shall I do further shortening the legs leaving a rest of 2 mm to obtain "zero T.C." or not?

With best regards

Andreas

[zlymex]

Now the question is: shall I do further shortening the legs leaving a rest of 2 mm to obtain "zero T.C." or not?

That's entirely up to to your requirement.
My suggestion is to leave it untouched, and compensate this little T.C. in later stages such as 7V to 10V step-up since there will be additional TC introduced which will void the zero T.C. you may have achieved.
Also, I don't recommend adding a critical component to solve a none critical problems since the stability is the number one priority than T.C. in most cases.

[Andreas]

I forgot where I read it [maybe in the LTZ's datasheet?], but Linear-Tech recommends 1mm spacing between the bottom of the TO99 package and the PC board. 

Hello,

I could not find this in the LTZ datasheet.
Has anyone a hint where this info can be found?

But if I think about it then I did a error when shortening the pins of LTZ#4 to zero.
On the other side the T.C. on #4 was very widely off without this measure.

This cathode resistor should be very stable [Edwin's PWW with stabilization or VPG foil with PMO].

That is not my way of adjusting the T.C. You need a additional very stable component. (and my PCB layout is not prepared for this).
And finally the copper traces on the layout determine a large amount of the T.C. of that additional component.
In a small range I can adjust the T.C. by populating R9 (which is already in Layout). And here I can use a standard 1% metal film resistor.
For a working point of 50-55 deg C the compensation is around +0.04 ppm/K for each 1 Meg resistor.

That's entirely up to to your requirement.

I think it was more or less a rhetorical question. As a volt nut I have at least to try it.

The 10 V output is still far away.
I think I will have to use a DAC and a NTC to compensate for the temperature changes
in order not to worse the 0.05 ppm/K spec of the reference.

With best regards

Andreas

[plesa]

I forgot where I read it [maybe in the LTZ's datasheet?], but Linear-Tech recommends 1mm spacing between the bottom of the TO99 package and the PC board. 

I could not find this in the LTZ datasheet.
Has anyone a hint where this info can be found?

There is info about the lead diameter only, same nor similar to LM399.
LEAD DIAMETER IS UNCONTROLLED BETWEEN THE REFERENCE PLANE
AND THE SEATING PLANE (0.254mm – 1.143mm)
One of my LTZ was almost impossible to insert close than 1mm due to the leads angles.

[Andreas]

Hello,

further shortening of leads on LTZ#3
See also:
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg924793/#msg924793

now measuring 7.5 mm from top LTZ to PCB.
(should have been 0.5-1.0 mm shorter but its difficult to keep exact height when the TEKO shield is already soldered).

Result: reduction from -42 ppb/K to -22 ppb/K for the regression curve.
Still not "zero T.C." but a value I can live with.

Further I changed LTZ#6 with R9 = 1M8.
Regression curve changed from -28ppb/K to -15ppb/K.

Had expected to be closer to Zero T.C. with the formula +40ppb/K for each 1 Meg resistor (or +22ppb/K for 1M8),
 but obviously this is not always exact.

With best regards

Andreas

[Kleinstein]

The compensation via R9 is nonlinear, as the heater voltage is nonlinear depending on the temperature, so the compensation might only work for a limited temperature range and is likely also influenced by the thermal setup (e.g. isolation cap, cutouts around the LTZ1000) and the length of the wires: Shorter wires need more heating and thus R9 has more effect.  With a relatively small value for R9 one might even have to make sure that R9 itself is stable enough.

[doktor pyta]

Progress:
-inox front and rear panel done (photos). As You can see it is designed to be stand-alone voltage reference or to be an OEM module to be built into some instrument.

-top and bottom LTZ1000 covering cup made of SUNDPLAST SP163.

-4 units assembled. Banana terminals are Pomona low thermal EMF type.

-TC measured (I don't use 400k resistors for temperature compensation):
Unit#1 with Vishay Z201 resistors T.C.=+0,03ppm/C
Unit#2 with precision wirewound resistors T.C.=-0,14ppm/C
Unit#3 with precision wirewound resistors T.C.=-0,08ppm/C
Unit#4 with precision wirewound resistors T.C.=-0,11ppm/C

By the way, I've bought some G.R. 8G16D wirewound resistors and some of them have T.C. as high as 7...8ppm/C instead of declared max. +/-5ppm/C. Does anyone else notice similar problem ?


To do:
-laser engraving of text on front and rear panels

-making better temperature chamber with PID controller

-redesign PCB to allow testing of CERN or PREMA solution with a JFET. My goal is to develop an EVAL board that allows testing different configurations.

-perform LF noise testing

P.S. Is someone of colleagues interested in putting one of my units into T.C. testing?
I could borrow one to compare the results (please send a PM).

[Andreas]

Hello,

Frank had a issue with one value (120R) of the econistors:
https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg624931/#msg624931

For the T.C.: is it measured on the direct zener output or after some buffer/divider?
(I am still missing for your schematics).
Over which temperature range did you test. (ramp or fixed points with how long rest time at the temperature?)

I would try to compensate the T.C. by R9 = 400K on Unit#2+#4 and perhaps around 500-680K on unit #3.

With best regards

Andreas

[doktor pyta]

Hello Andreas,

Frank had a issue with one value (120R) of the econistors:
https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg624931/#msg624931

Thanks, I haven't seen that before. This agrees with my observations. I measured T.C. close to 8ppm/C for 120 ohm econistors.
So this explains exactly the meaning of this "eco" prefix. You get what You pay for. All in all they are good for the 70k collector resisitors

For the T.C.: is it measured on the direct zener output or after some buffer/divider?

I use typical  LT application LTZ1000ACH + LT1013 and additionally a follower made of OPA2277.
I'm measuring the voltage at the buffer's output.

(I am still missing for your schematics).

I will show them before the next iteration of PCBs.

Over which temperature range did you test. (ramp or fixed points with how long rest time at the temperature?)

I tested the reference in three points: 15'C ; 22'C ; 30'C because my setup allowed only that points. I gave 4 hours (before each measurement) to thermically stabilize. Now I'm working on better PID controlled chamber.

[Dr. Frank]

Frank had a issue with one value (120R) of the econistors:
https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg624931/#msg624931

Thanks, I haven't seen that before. This agrees with my observations. I measured T.C. close to 8ppm/C for 120 ohm econistors.
So this explains exactly the meaning of this "eco" prefix. You get what You pay for. All in all they are good for the 70k collector resisitors

I cannot confirm, that these econistors are really a bargain.

I confirm, that certain values have higher T.C.s than expected.
I got replacement for these 120 Ohm types, which were in spec, i.e. around +3..5ppm/K
Keep in mind, that the 120 Ohm is less critical, like both 70k resistors.

The 12k all were very good, around +1ppm/K or less.
1k were between +3..5ppm/K.

Therefore, these latter resistors numerically give around 0.04 ppm/K in the completed LTZ circuit. Shall I say 'only'?

I got the feeling, that the overall T.C. in the end simply has to be measured, and other T.C.  compensation been added (length of leads, 200k .. 1M of "T.C. compensation resistor, and so on).

I think, that it's not useful to strive for the best possible (towards zero T.C.) parts, as even the best selected resistors may give much higher T.C.s, than expected.

Frank

[Mickle T.]

How about Datron LTZ1000 TC corrections? Datron 4000, 4700, 1271 e.t.c. have one.

[Andreas]

How about Datron LTZ1000 TC corrections? Datron 4000, 4700, 1271 e.t.c. have one.

Hello,

Could be a interesting method for compensation of the 7 -> 10V Divider.

Attached I have a overview of the 4 LTZ1000A references #3-#6 (7V) with all tweaking and measures done.

T.C. is trimmed from up to 230 ppb/K down below 25 ppb/K on all devices. Either per trimming lead length or by adjusting with suitable R9.

PSRR is reduced from up to 0.15ppm/V down to 0.05-0.06 ppm/V by slots between the voltage regulator and the LTZ.
The soldering of the teko inner housing has also contributed a part.

The slots and TEKO also help against voltage shifts by tilting the reference. (now less than 2uV or 0.15ppm).

R9 and shortening the legs have no visible impact on 1/f noise of the zener.
The popcorn noise on LTZ#5 is still there.
I decided not to exchange the zener by another LTZ1000A because the tempco of this device is near zero without shortening the legs / using R9.

Next steps:
- record ageing drift of the devices (against my pre-aged references #1 + #2 with around -1 and -2 ppm/year drift).
  here I hope to see wether the (popcorn-) noise has a correlation to ageing.

With best regards

Andreas

[Mickle T.]

A classic Datron schematics in the small form-factor (another DMM modding).
TC 20-40 ppb/K without any adjustment or compensation (R9=0).

[acbern]

Looks very interesting! Can you elaborate a little bit about it (parts used, circuitry details..).

[Mickle T.]

It's a simplificated Datron circuit without guarding and Vishay custom resistors network.
R2, R3 - hermetic foil type S5-61, R1 - hermetic foil type R2-67, R4-R5 - foil Alpha Electronics FLC.

[Mickle T.]

A copper chamber with the all sides heating. Was taken from the metrological level standard cells group H488/1.

[acbern]

It's a simplificated Datron circuit without guarding and Vishay custom resistors network.
R2, R3 - hermetic foil type S5-61, R1 - hermetic foil type R2-67, R4-R5 - foil Alpha Electronics FLC.

What values for R16 and R18 did you use? R9=0ohms, right. R7 and R15 are 160 ohms?
I would think one can also modify this to work on non-symmetric voltages (12V), with all signals referenced to 0V being AC-coupled.

[Mickle T.]

R16+R18 is needed for DMM mod only. The exact value (~3.4k) is individually adjusted for each ref module.

[doktor pyta]

Hi, Mickle T.
What are the advantages of the temperature control section proposed by DATRON?
Is Linear approach somehow worse ?
What are Your observations ?

[Kleinstein]

The Datron temperature regulation uses a more simple compensation (one less series resistor) and a different output stage. Both changes are rather independent.
The current controlling output stage does not look good to me, as it has less gain at low currents / power and thus makes the nonlinear heater curve even more nonlinear. So I don't see this as an advantage, but more a not so good decision.

The removed resistor in series to the feedback cap reduces the gain in an intermediate frequency range. It depends on the thermal properties on how important this is for regulation. AFAIK LT uses the same for both the LTZ1000 and LTZ1000a despite of different thermal characteristics. Also the heater curve is a square law, so loop gain is lower at high environmental temperature or lower set-point. So the circuit is not really tuned for a very fast response, which is not needed nor best for low noise. To really judge on the regulation one would need to measure the properties of the thermal system.

It depends on the environment if more gain is better or worse. If there is not much external variations (e.g. well shielded, not much  uncontrolled turbulence) the lower gain can be a small advantage. In a more turbulent environment higher gain would be better as in this case noise of the temperature sensor is lower than thermal variations.  Anyway the difference will be small as the thermal variations of the sensor itself should not contribute significant to reference noise in the relevant frequency range (e.g. 0.1 .. 10 Hz). So it may be just economics to use smaller caps, omit one resistor and save a few cents.

[doktor pyta]

I'm placing final photos and the schematic diagram of my version.
The temperature controller is taken from the LT app note.
The Zener section combines Datron and LT schematics, so You can test both configurations.
Future revision C will have minor changes, but I'm satisfied with current revision B.
Using Z201 resistors in the temperature setpoint divider and Zener current sense + Econistors in the rest of the circuit I'm obtaining T.C. in +/-0,06...+/-0,03ppm/K range without additional compensation (initial tests). That's fine for me and shows that there are no big errors in the (thermal) layout.
Still waiting to finish better temperature chamber. Then I will publish my measurements.

[TiN]

Very nice, one of best looking assembled LTZ units (if not the best) Looking forward for testing and performance results.

[Andreas]

I'm placing final photos and the schematic diagram of my version.

Still waiting to finish better temperature chamber.

Hello,

thanks for the schematics.
There are some ideas where I still have not thought of:
- guarding
- using ferrites

What do you plan for the temperature chamber?
I think you already know my low cost chamber from the T.C. measurements thread.
https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg462298/#msg462298

I simply use a peltier cooling box for cooling. (which is running continously for the cooling phase).
And 2 heater foils mounted on a aluminium sheet which is used as heat spreader.
Controlling is done by a FET power stage via RS232-signals (RTS/CTS) for the heater.
And by a SCR-Relays via RS232-signals for the 230V peltier cooling box.
The controller runs on PC. Simply switching on/off  the heater every second.
(with a small p-controller part generating some kind of PWM within 0.3 deg C of the setpoint).

With best regards

Andreas

[TiN]

I'm parsing data and got retested HP 3458A's reference module (A9 PCBA) in thermal box.



Having hard time to find out tempco of it, no visible change over 15°C delta.

[doktor pyta]

Below  mesurements of first assembled unit.
I used Fluke 732A + Fluke 720A + Fluke 845AR. The test setup is quite good, but I think I'm observing also the tempco of my measuring setup (for example 720A KVD has max. +/-0.1ppm/K tempco according to the manual). Any ideas how to solve this ? Maybe comparing results at different room temperatures would be OK, but it would take very long time.