Ultra Precision Reference LTZ1000

[niner_007]

I mean why we have mix of THT and SMT parts? No kelvin connection for sense from LTZ package.

Hehe, the last SMD part on my PCB is a tantalum cap, these are mostly low ESR in SMD.

[Kleinstein]

The ring around the LTZ is there avoid uneven heat flow to different directions. It is not so much about heatflow to the inside, more about heat flow to the outside. The source of temperature differences between the pins 3 and 7 is more an asymmetric outside, not the chip itself. An additional wire can improve on this, but it is an odd part to place. Depending on the board used, the ring can be essentially for free.

The last layout from 3roomlab still has a mayor flaw with the ground connection: the output ground (pin 2 of J400) needs to be separate from the power ground (pin4 of J400) - at least to the star-ground, but much better directly to the LTZ1000.
Similar the +7 V output should ideally come directly from the LTZ1000.
The correct ground connection is way more important than any thermal details.

I see no problem mixing THT and SMD parts, especially not for low quantity and manual soldering.
The form factor and connectors for the A9 board may not be the best choice for a standalone reference, but it is also not that bad.
Having the reference in a similar position may still be a good idea (unless there is a really good reason against it), to really keep it exchangeable.

[floobydust]

I went through your calculations and had to stop part way. I just wanted to know what is the heat input to each pad, in order to calculate the temperature difference when people are not using symmetrical pcb traces to the Kelvin pins. The ring is a second aspect to look at - I feel cuts/slices in the copper scuttle the intent to equalize heat outflow.

Missing is the heat loss due to convection currents within the plastic cap and radiation, verses conduction losses to any foam insulation around the IC.
Missing is the bonding wire(s) heat transfer die to pin.
Missing are glass seals' thermal resistance; dimensions unknown. Schott has nice TO glass.

Just the Kovar pins:
Using 17.3W/m-K, 2mm long (shortest pins), 0.5mm dia pins=2mm^2.
Kovar Rt = L/(k*A) = (2mm/1000)/(17.3*(0.2mm^2/1000^2) = 578°C/W/pin. There's 8 of them for 72.3°C/W total.
Using 17.3W/m-K, 12.5mm long, 0.5mm dia pins=2mm^2.
Kovar Rt = L/(k*A) = (12.5mm/1000)/(17.3*(0.2mm^2/1000^2) = 3,613°C/W/pin. There's 8 of them for 451.6°C/W total.

Reality check, one of TiN's KX thermal cam pics showed a can temperature 46.93°C and PCB temp of around 34.53°C. The LTZ has long pins, say 12.5mm or 1/2".
LTZ1000 θJA=80°C/W value did not jive with the datasheet graphs    LTZ1000A θJA=400°C/W matches graphs OK though. θJA=175°C/W matches the (non-A) curves, LM723 is 165°C/W. That is reasonable for the measured temperature.

Using 0.13W (top of my head) to the die heater on an LTZ1000A, my pin calculations say the thermal resistance of the pins is too high for that temperature gradient so I must be wrong.

[Andreas]

Missing is the bonding wire(s) heat transfer die to pin.

And note: there are 3 bond wires on (heater) pin 1
and different lengths of 1 wire on the other pins.

with best regards

Andreas

[Noopy]

https://richis-lab.de/REF03.htm

 

I could try to measure the lengths but is that really relevant? 

[floobydust]

I think pin 7 has way more copper (trace thickness) and will run cooler than pin 3.
This was sort of leading me to try figure out and model the difference, by knowing heat input from the pins to pads.
For the bonding wire calc, my math:
32um (std size) AL wire x 2mm long Rt =0.002/(220*0.00080425mm^2) = 11,304°C/W each, or 1,130°C/W for all ten.
For the glass insulators, 0.5mm ID 2.2mm OD, 1.5mm L, Schott 8250
Rt = ln(2.2/0.5)/(2pi*1.5/1000*1.2) = 131°C/W each, or 16.4°C/W for all eight.
The glass is conducting the can's heat to the pins, glass is a much better insulator but also huge size compared to the bonding wires.

This ring stuff fools the mind so you think it's isothermal. The copper-pour border around the pcb edges are taking ambient air temp and conducting it under the cap. The copper slices leave thermal islands in the ring. We have no thermal imaging cam pics to verify. My thoughts anyway.

[Noopy]

I have taken some pictures of the "air impregnated polymer" die attach that is also used in the LTZ1000A.
I have posted the pictures here:

https://www.eevblog.com/forum/projects/ltz1000-nice-die-pictures/msg3369600/#msg3369600

 

[niner_007]

With connectors for the output I would more worry about heat flow around the connectors - these can also cause thermal EMF and there naturally is heat flow to the other side. It may make sense to have relatively thin traces at the connector for the critical signals.

The neat thing about the 3458A form factor, if you use it outside the 3458A, you don't have to use the connector, and use wires directly to the PCB. Very versatile.

[niner_007]

What could it be

[Haasje93]

WOW!
Your'e up to something.

[Free_WiFi]

Just a little bit of topic ...
On YouTube i saw a video where was clearly showed that's ltz1000 tend to heat-up intermittently .... like a blinking led.
Since the integrated heating oven of the ltz1000 is not linear,because it's tend to make some pwmming,thus i'm asking my self if
- the LTZ 1000 packaging is not already obsolete for our decade? (2021 and so on....).

[Kleinstein]

If build correctly the heater control is a linear PI regulator and does not oscillate. The control is slightly nonlinear (as the heater is a resistor) and the regulator usually works on the heater voltage (with a series diode). This is not ideal (the regulator gain changes with powert), but it can still work, as the thermal system has a rather short dead time. In theroy one could imporve on this, but usually no need for it. So why fix it if not broken.  It the regulator is oscillating there would be something broken to fix.
There is some special circuits to intentionally add some oscillation to remove hysteretic effects - though more like a very special case and only for the initial phase.
 

There is nothing wrong with the packaging: just a simple card box with aluminum for ESD shielding around the parts. The hermetic sealed parts need no special humidity control.

[ramon]

There could be some kind of package that allows hermetic sealing without needed Kovar leads.

And if Kovar is absolutely needed, someone would prefer SMT to allow mounting directly on a ceramic or special substrate.

Unless you air wire everything, and then you really connect the 'star point' directly into the leads instead of the PCB. But then you will need to do extra work reducing effect of air currents.

Current packaging need too many considerations (tricks?) regarding thermal, mechanical, electrical, etc ... efects. As a new package would have too.

[dietert1]

If somebody wants to improve on thermal EMF, they can use the LTZ1000 in an oven. Then you can keep the opamps for zener current control and for temperature measurement at constant temperature as well. And maybe the 7V to 10V gain stage, too. That one will in general need an oven, anyway. Except you hand-select resistor sets for zero TC.
As far as i understand, one weak point of the LTZ1000 package is the missing 4-wire connection for the zener cathode. There should have been two pins with two separate bonding wires. One pin for current feed, the other for voltage sense. The star wiring should be on-chip, not behind a bond wire with its two wire bonds.

Regards, DIeter

[antintedo]

There could be some kind of package that allows hermetic sealing without needed Kovar leads.

Copper to glass seals are possible, Vishay uses them for many hermetic resistors.

[Andreas]

There could be some kind of package that allows hermetic sealing without needed Kovar leads.

Copper to glass seals are possible, Vishay uses them for many hermetic resistors.

Not directly there is always some Kovar involved.

From the VHZ datasheet:
"These parts are made with glass to metal seal enclosures employing Kovar eyelets which allow the copper leads to pass through the enclosure"

So basically this fact already proofs tht's LTZ1000 packaging is more like : Meeeeeh ..... VS Very Good!

So you like the "elephant feet" of the vishay resistors?
The LTZ package would double in size with those.
And heater power would also increase for a factor.

with best regards

Andreas

[DenCraw]

the amount of knowledge here overwelmd me  a lot.
I want to ask if it is a good idea to put the reference voltage box inside a vacuum box with a Peltier cell
with a pwm controller with temperature sensors inside and outside the external box
in order to control the ambient temperature to 25 degrees.

[BU508A]

the amount of knowledge here overwelmd me  a lot.
I want to ask if it is a good idea to put the reference voltage box inside a vacuum box with a Peltier cell
with a pwm controller with temperature sensors inside and outside the external box
in order to control the ambient temperature to 25 degrees.

TiN already had the idea with a Dewar flask

https://en.wikipedia.org/wiki/Vacuum_flask

[DenCraw]

.. that was fast...

[3roomlab]

put the reference voltage box inside a vacuum box with a Peltier cell

the INRIM "travelling" std 2017

[DenCraw]

Excellent!!!
that is what i was looking for!.
Here is my lousy experiment but with a LM399 and a
TEC1-12706,40mm x 40mm x 4mm,0~15.2V DC and 0~6A

[cowasaki]

I have my reference in a plastic box wrapped in bubble wrap with a small heater pad which keeps the temperature between 34 and 36C constantly.

Connected to my Keithley 2100 I have been seeing 10.00000 for six months but today it dropped to 9.99999 and now 9.99998 I know that its just 20uV but as it was totally stable and suddenly dropping today I was wondering what sort of level of variability it should have with a 6.5 digit Keithley 2100?

My LTZ1000 board is powered from AC via a 12-0-12 transformer across both sides so 24v into a 7415 with a 2200uF/25V cap on the 15V side.  The 15V is good for 15.000

[Andreas]

Hello,

when reading the spec: the uncertainity is 38 + 6 ppm in the 10V range so +/- 440uV for a 10 V reading.
So 20uV is a good result.

By the way: my K2000 has a tempco of 0.55 ppm/K environment temperature.
Besides this it suffers from some "pocorn" noise reading sometimes higher or lower around 1 ppm even at the same temperature.
But I fear: you will need much more DMMs to decide wether the DMM drifts or the reference.

with best regards

Andreas

[TiN]

LTZ at -260°C ? Coming soon  .

[DenCraw]

what if you use also a vacuum chamber...