Ultra Precision Reference LTZ1000

[branadic]

I propose a volt-nuts beer meeting at the Weinheimer UKW-Tagung ham festival (September 13th-15th).

Beer sounds good, but I'm outta town this time.

I will be there, Quarks is living less than 20 km from there, affordable trip for branadic and Dr. Frank too...

I'm sure for Andreas too.

[Andreas]

I'm sure for Andreas too.

The saturday after flea market in the morning sounds good to me.
But I would prefer some coffee.

Andreas

[Dr. Frank]

Sorry, I hope to travel to Guadalajara exactly at that time.

Another date, and I will fetch my equipment for that meeting.

Frank

[babysitter]

That is quite some response! 

I suggest a meeting at 10:30 in/at/near the "coffee tent" on the meadow in the center of the open-air part.
PM me for my phone number if you think you need a support hotline.
This should give everyone the possibility to hunt for nice stuff on the flea market early and also add just little delay for those who have to leave early.

Myself will stay a little longer, lingering on the camping site from saturday to sunday too. This will improve my Weinheim look & smell

Greetings

Hendrik

[branadic]

As announced by Andreas we wanted to find out what is best: "Using a solid pcb or spend some slots around the heat controlled voltage reference?" and furthermore: "Keep the leads as long or as short as possible?".
Therefore Andreas provided two LM399H (National Semiconductor) and me the pcb. The layout, similar to the one you can find in the world wide web for the LTZ1000, contains two equal trace arrangements but one of them has milled slots.
The test starts pretty simple, the two LM399 are soldered into the board keeping the leads as long as possible and with nearly equal lenght for both references. A 7k5 resistor for each reference as given in the datasheet and supplied with 15V from a bench PSU.
The pcb is arranged on a styrofoam with a window inside so that the IR cam can record the bottom of the pcb. To keep other IR sources away that can produce reflection on the tinned copper traces some foam was put around the references and another styrofoam box was put on top. At some places a black tape was necessary, to avoid reflection.
In the following pictures the references were running 30min minimum. To verify the results I always made a picture of the pcbs top side.

I used same temperature scale for all pictures, but as I have the raw data here I can change that if necessary.

So what can we see? With long leads and slots (left side) the temperature of the pcb seems to be higher compared to the version without slots (right side). However, in both versions all 4 pins seem to have the same temperature, no thermal gradient visible.
As expected the pcb gets more hot with short leads...

So let's start the discussion

[babysitter]

A very good experiment, good to see the IC itself spreads the heat evenly around itself. Thanks to HSG too?

Glass as main ingredient of PCB material is a known bad conductor for heat, the second ingredient epoxy resin is also not that good, and cooper a good known conductor (Yes, there is a rule that eletrical conductivity is often similar to heat conductivity..) , knowledge of those properties allows for some thermal designing. The cooper pour on my PCB was intended to spread the heat evenly.

What we not see, however, are the air currents I am afraid of, slots on the PCB allow them to circulate and I suppose they will be very erratic and turbulent in such an environment. Possibly I will study some liquid-to-liquid-turbulences at some time during the next few months, maybe I can put such a circuit in my test jig then ?

BTW, the Volt-Nut Meetup at Weinheim gains some traction !

Greetings
Hendrik

[branadic]

Thanks and yes, it's great having access to such equipment at weekend, one of the advantages working there. But I'm sure the cam is happy to have a sense and I also hope that you guys are happy that you can profit by this too.

About the turbulences I wouldn't care if the critical parts are put into foam, this can avoid air flow, see picture attached.

[Andreas]

So what can we see? With long leads and slots (left side) the temperature of the pcb seems to be higher compared to the version without slots (right side). However, in both versions all 4 pins seem to have the same temperature, no thermal gradient visible.
As expected the pcb gets more hot with short leads...

So let's start the discussion

Ok.

Mhm.

The pictures are really surprising me.

This will change my picture of the thermal world.

My favourite up to now was the short leads with the slotted pcb.
I would have expected all pins and the pcb around the pins at the same temperature.
And the large temperature gradient on the lines till the outer rings.
(Why should Datron have made them if they would not be necessary).
The thermal gradients are much more locally around the reference than I expected.

In my opinion we do not only have to regard wether all pins are equally but also the heat distribution within the solder junction.
The lead of the reference (in the middle of the solder junction) is made of Kovar which has around 40 uV/K against copper (at the edge of the solder junction).
Solder itself has around 3uV/K against copper. So lets assume that Kovar against solder has around 37uV/K.

So from middle to edge of the solder junction it seems for me to be much more critical for the slotted board with short legs than the not slotted board with long legs.

So long legs and large thermal mass (perhaps the PCB material which diligent minds recommends) seem to be the best.
Of course you have to take more care keeping air currents away from long legs.

Other opinions?

@ branadic: with the raw data it should be more easy to evaluate the gradient within the pads than for me having difficulties with the colour gradients of the picture.

With best regards

Andreas

[Dr. Frank]

Branadic,

those are great measurements!

2 simple conclusions:

- short leads will increase the temperature of the PCB, long leads will give more thermal isolation between component and PCB => long leads preferred
- slots in the PCB give higher thermal isolation between solder joint area and outer PCB area, thereby reducing cooling of the solder joints through the PCB, and in turn increasing the local heating of the solder joints. Higher solder joint temperatures might cause temperature differences more easily and  will increase thermo voltages.

So I still do not understand the purpose of those slots, especially in the Datron PCB, where the LTZ1000 at a temperature of ~60°C is used.

I'm now very confident that 45°C and a PCB without slots is the most stable solution.

Thanks for sharing.

[Jay_Diddy_B]

Hi,
Just a little observation for you.

I noticed that your thermal camera is pointing at tinned traces on the PCB. These have a very low emissivity compared to the Fibreglass (FR4).

I suggest that you could try painting the traces with liquid paper, type correcting fluid, or give the board a quick spray with black paint (flat or matte).

Jay_Diddy_B

[quarks]

Very interesting, so long legs and no slots look thermally good.

But what about mechanical stress relief, what I though is the main reason for the slots?

[Dr. Frank]

Very interesting, so long legs and no slots look thermally good.

But what about mechanical stress relief, what I though is the main reason for the slots?

Yeah, what about??

Does "mechanical stress" cause any deviations in output voltage or decrease stability?
How does "mechanical stress" affect those electrical characteristics?
How is "mechanical stress" defined in this situation?

I still cannot see any mechanism, how that buzzword "mechanical stress" would have influence on the electrical output.

Especially, as all of the relevant components are leaded parts, not SMD parts, where you really would have mechanical forces on the solder joints, due to differences in thermal expansion between component and PCB.

Cold versus hot areas on the PCB may lead to bending / distortion of the PCB, but what would be the electrical effect of that, once again, when using leaded components?

Such mechanical effects would be relevant for oven temperatures of 90°C.. down to 60°C perhaps, but at 45°C, that's completely irrelevant.

Does anybody have any other ideas?

Sorry, but I am still searching for solid real physical effects and their explanations.

Frank

[eurofox]

Hi nV freaks,

I'm passionate too with this subject and follow this daily but I'm happy with 6 1/2 digits.

There is "the multimeter" you need on eBay with a failure but I suppose it should not be to bad to fix it.

http://www.benl.ebay.be/itm/261263662099?ssPageName=STRK:MESINDXX:IT&_trksid=p3984.m1436.l2649

eurofox

[branadic]

Hi,

as said the emission coefficient of metal is really worse, it's rather an infrared mirror reflecting all infrared sources from the surrounding.
The resolution of the cam is about 0,3K so with the sum of both you can't see a thermal gradient along the leads or the temperature of the solder junctions (they're also reflecting) directly. What you can do is cover the surface with a coating as pointed out by Jay, but I found that this wasn't necessary here.
The FR4 material in this case radiates enough to make things visible although glas is an infrared isolator (you can hide behind a glas but not behind a silicon wafer).
We can assume that the solder junctions are of the same temperature like the pcb material, so one can calculate the temperature gradient along the leads.
The heat conductance of the pcb material is worse but better compared to glas only. Because of the smaller thermal mass with slots the temperature is a bit higher, but not critical. The heated reference over-radiates the pcb temperature at bottom view, so the top view is more significant in this case.

About the stress with or without slots the only answer can give a measurement and the board exists so let's do some measurement and find out, shall we?

[quarks]

Yeah, what about??

Does "mechanical stress" cause any deviations in output voltage or decrease stability?
How does "mechanical stress" affect those electrical characteristics?
How is "mechanical stress" defined in this situation?

I still cannot see any mechanism, how that buzzword "mechanical stress" would have influence on the electrical output.

Especially, as all of the relevant components are leaded parts, not SMD parts, where you really would have mechanical forces on the solder joints, due to differences in thermal expansion between component and PCB.

Cold versus hot areas on the PCB may lead to bending / distortion of the PCB, but what would be the electrical effect of that, once again, when using leaded components?

Such mechanical effects would be relevant for oven temperatures of 90°C.. down to 60°C perhaps, but at 45°C, that's completely irrelevant.

Does anybody have any other ideas?

Sorry, but I am still searching for solid real physical effects and their explanations.

Frank

I do not know the answers, but my thoughts (until now) are influenced by the fact that the Datron/Wavetek design is with the slots.
Also the gear that uses this design (like 4808, 4910 and 1281 to name a few) is representing and rated top notch.
So my guess is, there probably is a good reason for slots.

In the LT Application Note AN82
http://cds.linear.com/docs/en/application-note/an82f.pdf
there is a statement (page 6) that Metal Packages are largely immune to board stress
(DiligendMinds also stated this before, especially for the LTZ1000ACH),
but negative effects are described/shown there (see att.).

Also on http://www.amplifier.cd/Technische_Berichte/Spannungsreferenzen/Spannungsreferenz.html
(sorry forgot the translation to english)
http://translate.google.de/translate?sl=de&tl=en&js=n&prev=_t&hl=de&ie=UTF-8&u=http%3A%2F%2Fwww.amplifier.cd%2FTechnische_Berichte%2FSpannungsreferenzen%2FSpannungsreferenz.html&act=url
there are some findings about mechanical stress.

At least for me that is good evidence to think there probably must be something behind it.
But if it is not worth the effort to implement it, than of course that is good news, because things get a lot easier.

bye
quarks

[babysitter]

I will have AT our zwick AT work for pushing and pulling on pcbs. sent from mobile.

[Andreas]

But what about mechanical stress relief, what I though is the main reason for the slots?

Ok that would be another thing of consideration.

I still cannot see any mechanism, how that buzzword "mechanical stress" would have influence on the electrical output.

As is stated in many datasheets of precision voltage references mechanical stress applied to the PCB shifts output voltage.
And you can simply measure it.
Usually I do this with any of my "ADCs" and try to keep the influence to below 5uV drift due to mechanical bending of the PCB. With every 5V reference that I tried up to now I got many 100uV shift when all pins where soldered to the PCB. So up to now this ended always in this way that only the GND-Pin is soldered to the PCB and the other PINS are wired with a thin VERO wire. This is a bit tricky for SMD devices but it works.

I was very surprised when testing the VRE3050AS reference which is in a hermetically sealed package with gull wing leads. In this case I thought that the influence of the PCB to the chip would be negligible. But this was not the case. I measured values from -16 to +186 uV against untwisted PCB.
Other references have other values depending on construction and individual make.
A LT1236ACS8 gave around 400uV between minimum and maximum value.

Unfortunately I did not do a test with my LTZ1000 boards. And now I don´t know if I shall risk a new ageing cycle.

On the other side a LTZ1000A should have a better behaviour than a LTZ1000 because of the different die attach.

A similar test like the LTZ1000 could be done with the LM399-Board of branadic.

About the stress with or without slots the only answer can give a measurement and the board exists so let's do some measurement and find out, should we?

Yes of course.

With best regards

Andreas

[branadic]

We have a "Zwick" too and also a second push/pull machine with temperatur control, but I guess this is breaking a butterfly on a wheel. A simple alternating force an top of the voltage reference is more than enough.

[Andreas]

We have a "Zwick" too and also a second push/pull machine with temperatur control, but I guess this is breaking a butterfly on a wheel. A simple alternating force an top of the voltage reference is more than enough.

I do not put force on the top of my references (because this would also change temperature)
I simply bend the board by applying forces to the edges of the board.

With best regards

Andreas

[Dr. Frank]

I do not know the answers, but my thoughts (until now) are influenced by the fact that the Datron/Wavetek design is with the slots.
Also the gear that uses this design (like 4808, 4910 and 1281 to name a few) is representing and rated top notch.
So my guess is, there probably is a good reason for slots.

Yep, and that's exactly, what's always done during Copy-And-Paste designs ... i.e. copying possible errors or nonsense also.

Many people in this or in the Asian community simply copy the LTZ1000 'A' version, available from China, because it's sitting in the venerable 3458A, including the 95°C operating temperature, without thinking.

Perhaps, Datron did not calculate anything also, and simply introduced those slots by similar hand waving arguments also.
All those instruments you mention, base on the same reference module (afaik), so that's no argument, either.

And there exist also several successful  designs without any slots, I think in the 732B, the ceramic substrate doesn't have those, either.

If there's a physical effect, it should be obvious, or should be measurable in an appropriate experiment.


The cited AN or the experiment in the CD forum deal with a different effect, than directly on the PCB, I think!

It's well known, that molded IC are sensitive to bending, due to their internal construction.
The mold compound might imply pressure on the chip directly, and the base lead frame, where the chip is epoxied onto, is flexible, because it's a thin sheet of copper only.

So, if you bend or distort the legs of such molded IC, you will easily bend the chip itself.

But that is perhaps not the case for the hermetical TO case. There's no mold compound, it has long legs, and the chip carrier is more solid.
And I do not see an advantage of those slots, if you want to avoid flexing of the PCB.
In contrary, slots might intensify bending forces on the legs, because slots can move, but a solid PCB area can not by that amount.
 
Sorry, no, I still can not identify any causality in there.

Frank

[Dr. Frank]

As is stated in many datasheets of precision voltage references mechanical stress applied to the PCB shifts output voltage.
And you can simply measure it.
...
I was very surprised when testing the VRE3050AS reference which is in a hermetically sealed package with gull wing leads.

Andreas, I have read that also, but always in conjunction with molded components only.
And that mechanism I can understand, due to the mechanical construction.

Unfortunately I did not do a test with my LTZ1000 boards.

Well, that would have been very interesting, as the TO case is a totally different mechanical construction.

On the other side a LTZ1000A should have a better behaviour than a LTZ1000 because of the different die attach.

Why? In the LT datasheet, the different die attach is reasoned only for better thermal insulation, not for better mechanical stability.
The only effect is that the required heater power is decreased a lot. (What other mechanical effect should exist there comparing the different die attaches??)
And that is important (only) if you run the LTZ1000 on higher temperatures (95°C), as in the HP3458A, or in the Keithley DMM.

Other instruments at lower temperature (Datron 49xx, Fluke 7001) simply use the LTZ1000 version.

 

A similar test like the LTZ1000 could be done with the LM399-Board of branadic.



With best regards

Andreas

Yes, that's right.
But I could imagine, that there's an effect in the solder joint only, i.e. that thermal voltages are depending on mechanical forces also. (But I could not name or cite anything around such an effect.)

Regards Frank

[Andreas]

On the other side a LTZ1000A should have a better behaviour than a LTZ1000 because of the different die attach.

Why? In the LT datasheet, the different die attach is reasoned only for better thermal insulation, not for better mechanical stability.
The only effect is that the required heater power is decreased a lot. (What other mechanical effect should exist there comparing the different die attaches??)

The reason is simple:
AFAIK normally the die is directly attached (by some silver filled epoxy resin) to the Kovar bottom plate of the TO-99 package.


If you do some thermal isolation it is very likely that it will be something that is not so stiff (ideally some kind of foam or maybe low density glass filled epoxy)  than the bottom plate. Thus introducing less forces to the chip.

With best regards

Andreas

[branadic]

Okay, okay... we have the Kovar package with gold finish but what are the bond wires made of? Obviously the chip metallization is not gold, so I expect this is not an aluminium bond wire?

[Dr. Frank]

Andreas, Quarks,

sorry that's all not convincing.

The speculation about the consistency of the isolation material does not bring any new information.
Additionally, I don't see at all, how bending forces on the PCB could propagate into the chip. Definitely not over the bond wires  , and hardly via the very stiff chip carrier.
 
In the AN 82 from LT, the robustness of the TO package against mechanical stress, compared to molded components  is pointed out definitely.

The diagram from AN 82, of a slotted vs. a stiff PCB, is made on a molded component (LT1460)  in SO8 package, definitely!
To use this as an indicator, that there possibly might be some appreciable effect on the LTZ1000 also, is more than far-fetched.

Sorry again, up to now, there is no document or experiment yet, supporting this mechanical stress theory on the LTZ1000 package.

The contrary is the case, to my impression.

Frank

[Dr. Frank]

Okay, okay... we have the Kovar package with gold finish but what are the bond wires made of? Obviously the chip metallization is not gold, so I expect this is not an aluminium bond wire?

bond wires used to be pure gold, but now were changed in the whole electronics industry to pure copper.

aluminium is always used for the top chip metallization, and the bond landing pattern .

On that photograph, obviously LT uses aluminium bond wires also.

Afaik, this material is used in power semiconductors, and for low bond temperatures, therefore low resistance may have been the argument here.
http://heraeus-contactmaterials.com/en/products/aldr/productpage_smt_adhesives_3.aspx

Frank