Slot holes around LTZ1000(A)

[Flinstone]

Hello Voltnuts,

                      it is more or less a rhetoric question (for me) to all of you using slot holes around the LTZ1000A ... ?
Can anyone provide the rationale why board designers are copying these VooDoo holes - do people have scientific facts why these should be in place ... ?
Why would one avoid these slot holes using in particular in conjunction with the LTZ1000A ... ? Slot-holes can be effective but are they useful in conjunction with LTZ1000A ... ?

Best regards
Fred Flinstone


Yaba Daba Doo !

Fred Flinstone
55 Cobblestone Road
Bedrock

[Synthtech]

Would it be to reduce mechanical stress due to any board flexing?

[branadic]

I once investigated LM399 with short and long leads with and without slots.

-branadic-

[TiN]

They look cooler and lure new baby volt-nuts better than usual boring looking board .

I don't do slots. 

[Andreas]

Hello,

My opinion
Slots or not is a question of the whole thermal concept.

You have always to pay attention wether a thermal measure increases or decreases thermal balance.
Keep in mind that 1K temperature difference between the KOVAR leads of 2 pins
of the LTZ and the copper on your PCB creates 40uV (6 ppm) EMF.

So the concept may be different wether you use a 1 Layer or a 4 Layer board.
In all cases you may want to have as much thermal isolation around the pins as possible and keep air drafts away.

My personal opinion (from branadics thermal measurements) is that slots which are very close to the pins
increase thermal gradient in the area between the pins,
so you will need more thermal isolation (at least with short leads) to compensate for it.

On the other side: slots at a certain distance (>1-2 cm from the center of the LTZ) help
to keep termal gradients from non constant heat sources away from the pins.

I have definitely reduced the PSRR error from around 150 ppb/V down to 50-60ppb/V
by introducing a slot between the LT1763 voltage regulator and the LTZ pins.
So slots introduced at the right point definitely increase stability and are no voodo.

https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg875274/#msg875274

datron goofed up (no offence intended to Pickering)

no bashing please if you have no facts.
according to specs the Datron references are on par with Fluke 732B see summary here.
https://www.eevblog.com/forum/metrology/teardown-voltage-standards/msg902849/#msg902849

with best regards

Andreas

[SvanGool]

I searched again, but can't find it anymore:

I once saw a very old video (I think it was a Linear video on Youtube, but couldn't find on their channel anymore) where they scientifically tested/compared the "slot technology" thermal behavior with normal pcb-plates and concluded that it did make a difference. At least they convinced me with that video.

[Edwin G. Pettis]

According to the folks at Linear Tech, that video wasn't put up by them and that is probably why you can't find it.  Once again, from the source who brought you the LTZ1000/a, "they don't like slots because it creates TOO MUCH airflow on the board and close to the device."

There you have it, any slots on a LTZ board is not sanctioned by Linear Tech, they have found that slots cause more problems than they may solve.  The longest running continuous board design, the 3458A's reference module, does not and never has used slots.

[David Hess]

The slots decrease thermal conductivity and reduce strain on the package however as mentioned, increased air flow needs to be taken into account.

Strain on the package is more of a problem with surface mount parts and you will sometimes find precision operational amplifier with slots cut around them.  Analog ICs in plastic packages use a special encapsulation within the package to isolate the silicon IC from the package material.  For digital ICs this is irrelevant.

[splin]

The longest running continuous board design, the 3458A's reference module, does not and never has used slots.

Good point. The 3458A would never have sold for as long as it has if they'd designed the reference board with slots. 

[Andreas]

But why does Keysight then use slots on the (successor) 34470A reference???

[janekm]

But why does Keysight then use slots on the (successor) 34470A reference???

Your second picture shows the plastic baffle they are using to restrict airflow, so it seems they get the best of both worlds with this design.

[Andreas]

Hello,

You have not understood how to maximise profit:

They will not do a re-design of the existing board.
But they will now keep back the best reference boards
with the lowest noise + ageing rate for the new 34480A.

with best regards

Andreas

[SvanGool]

@Flinstone:

Thanks !

I saw these videos/pictures, while trying to dig up the video I meant, but those were not the ones. It was an old kind of "sixties style" digitalisation of a video=tape-recording, it showed a comparison of quite big pcb-panels that had segments having slots and others having no slots around multiple heated central points. I searched again but can't find it anymore.

@Edwin G. Pettis:
I appreciate you saying that the video was not from Linear, but do you (or your contacts at linear Tech) know then which video it was and where to find it?
You are probably right, that there is more to it then just thermal stress.

It might be that the video was not specific LTZ1000 related, but it was specifically for those slotted hole structures on a pcb and thermal gradients/stress. It also might be that it was only a part of a larger video, that is why I probably can't find it.

[David Hess]

You have not understood how to maximise profit:

They will not do a re-design of the existing board.
But they will now keep back the best reference boards
with the lowest noise + ageing rate for the new 34480A.

Tektronix did the same thing which was well documented.  Practically the only time they re-engineered part of an old design was if parts availability became a problem or significant flaws were found.

[Andreas]

Practically the only time they re-engineered part of an old design was if parts availability became a problem or significant flaws were found.

Ups why did they not simply use the old existing boards where they have thousands in their ageing boxes?

[David Hess]

Practically the only time they re-engineered part of an old design was if parts availability became a problem or significant flaws were found.

Ups why did they not simply use the old existing boards where they have thousands in their ageing boxes?

I am not sure what you mean but the puzzling thing to me is why Tektronix did not reused designs between different products.  For instance every 100MHz 76xx oscilloscope mainframe seems to have a different but similar horizontal and vertical amplifier assembly when they could have all shared.

[Andreas]

Hello,

what I mean is the re-use of the 3458A reference board which could have been used in the 34470A.
But of course it is a different form-factor. So probably there are space limits.

with best regards

Andreas

[3roomlab]

Practically the only time they re-engineered part of an old design was if parts availability became a problem or significant flaws were found.

Ups why did they not simply use the old existing boards where they have thousands in their ageing boxes?

jus a rough run down of costs

Code: [Select]

3458a qty 34470a qty
$60.00 $20.00 Z foil 3 $30.00 $10.00 Z1206 3
$1.20 $0.20 Ra 6 $3.50 $0.50 R0603 7
$6.00 $2.00 Rb 3 $0.30 $0.10 E1/2/3 3
$0.75 $0.25 C 3 $0.84 $0.14 C0603 6
$0.14 $0.07 D 2 $0.30 $0.30 CR1 1
$0.14 $0.14 Q 1 $0.03 $0.03 TO92 1
$1.00 $0.50 5pin 2 $0.50 $0.50 8pin 1
$40.00 $40.00 LTZ 1 $40.00 $40.00 LTZ 1
$2.30 $2.30 DIP8 1 $2.10 $2.10 SOIC 1
$0.50 $0.50 PCB 1 $0.50 $0.50 PCB 1
$112.03 sum $78.07 sum

come to think of it, i am not sure if the SMD zfoil cost is $10, but does anybody think 34470a ref board is more expensive than 3458a?

assumption 1 - i think 3458a is a huge 12kg thermal mass. 732a is also 12kg, 5xxx calibrators are all over 20kg. the speed at which these things heat up (mitigated by its own mass?) might be a factor in long term stability. from andreas TC measurements, i suppose that precision resistors do not like fast warm up? am i right?

in addition, mashing up some numbers in excel, i find that slots improve thermal isolation only if the traces are small enough (<10mil = more than 5% bonus). so i would assume the way all major heat sources are spread out and isolated with slots make for a slower temperature ramp. plausible?

im not the slot police trying to protect its use or lecture anybody, but as a real noob, i think we need more reasons to call something voodoo

1oz 6mil wide 12mm long Cu trace ~ 5900C/W
1.6mm thick 5mm wide FR4 12mm long ~ 6200C/W
as we can see, when R(Cu trace) is high enough, R(FR4) becomes a significant player. so its there, not magic voodoo.

so they are/could-be trying to reduce the speed of temperature change to "save" the lousier non-LTZ precision parts from "thermal shock". shock might be too large a word, but isnt SMD parts more likely to go out of spec faster than THT parts under similar thermal cycling conditions?

[wraper]

Crappy thermal image done under angle. Does not tell much.

Have a look to the PCB below - it is not perfect, much headroom for circuit improvements - but at least the PCB designer understood his job ...

Yeah, mechanical stress and temperature drop over kovar leads to the maximum  .

[3roomlab]

excluding the older versions of simulation i have done before. for this post, i did FEMM in a curent/amperes method to see temperature gradients.

much more interesting than slots i think is conduction. or iso-thermal rings? a few months ago i think i simulated a scenario where a mini iso thermal pad can be used to try and equalize the pin temperature at LTZ base.
https://3.bp.blogspot.com/-svBNjYX9wvI/Wvq4VsaX5BI/AAAAAAAADAE/o8SBdL46xHUMQPBWDEN9MFENVUIaVyuoQCLcBGAs/s1600/031.gif
https://3.bp.blogspot.com/-JIeiXx3xyxk/Wvq4VISXhDI/AAAAAAAADAA/DpHRSbvTNOwW5aP5XguQwHjZSDGrmDyUQCLcBGAs/s1600/033.gif

in this round, i flipped 1 to the outer edge of the PCB.

no outer iso-ring, no slot, normal Cu inner iso-ring (per iso line estimate = 0.1C)
https://1.bp.blogspot.com/-iC7SY9ebk8U/WvqoXSJQOsI/AAAAAAAAC_g/OFLNs5uJGqE_1yycYYf4ym43W6bMq8ldACLcBGAs/s1600/038.gif

+slots, +beefy equiv outer iso rings, +1 large trace inside PCB
https://3.bp.blogspot.com/-zllZwYd9MZQ/WvqnHhIuM2I/AAAAAAAAC_M/zNeM0vPf3s8ZC2_Qii5qr9xJ9UZQC76EwCLcBGAs/s1600/036.gif

it simply also mean that it is possible to use conduction to eliminate temperature gradients in passive areas w/o a heat source (see the top left of this PCB)
https://4.bp.blogspot.com/-R1vCAhz7vIo/Wvqu3WGIeoI/AAAAAAAAC_w/LMA6UtYl73wF94QrLwfVScZ4M_Rsw1oigCLcBGAs/s1600/039.gif

while a small ring outside a LTZ has some effect in trying to normalize temperature, the real temperature gradient killer appears to be a beefy outer ring.
and since the simulation is done in a VA manner, it is like putting extra capacitor to reduce AC ripples

slotting for reduction in thermal RC or conduction increase in thermal RC, i guess we can call it thermal guarding? just like op amp input guarding?

everything said above is now "old", but still might be informative
because i tried to do a "calibration" on FEMM. ie : apply 1W on a 10C/W material, but it doesnt result in 10C rise unless parameters are changed. this problem appears in both current mode and heat mode. weird isnt it?

ie : either you re-scale your materials or you rescale the input watts/heat. while reverse theta works in current mode, the air doesnt seem to behave normally, but heat simulation mode, the air appears to be normal.

but still after numerous simulation, the "trick" remains the same, air isolation slot enhances thermal resistance only if the path/Cu/bottleneck is small enough.

[floobydust]

The FEMM work is very interesting.
Can I back up a step before going to the use of copper strips as heat dams and an equalizer.

We know slots lessen conduction heat loss from the ref IC pads through the PCB. This would lower heater demand, and another slot can prevent the heater transistor from warming up other parts on the board. It also can make a layout more compact instead of all that real estate of FR-4 around the LTZ to deal with the gradient.

Second, slots isolate any PCB flex/bending moment from applying strain to the ref. IC through its leads. Ideally the slots cover both X, Y planes, leading us to voodoo patterns and crop circles. Voodoo slots may carry over from a designer's experience with parts shifting after any flex, during assembly/manufacturing.


The 34470a ref pcb I see slots around the LTZ pads to lower heat loss (except at the spokes); the LTZ is on a thermal island. But there's no bottom-side cover, so fan airflow cooling the bottom, leading to extra heat loss there.

The 3458a ref pcb has a cover on both sides of the LTZ, so minimum convection heat loss but no slots so extra (conduction) heat loss there.


Note the 34470a ref pcb extra pads on pins 3, 7. I think if these were brought out to a big split circle under the IC, it would be better, with no slit in the copper, to keep these pins isothermal.

You seem to have two LTZ pins with greater heat compared to the other six?

[3roomlab]

The 3458a ref pcb has a cover on both sides of the LTZ, so minimum convection heat loss but no slots so extra (conduction) heat loss there.
You seem to have two LTZ pins with greater heat compared to the other six?

i think LTZ pins are never equal in heat output, so i purposely try to show that.

[Andreas]

Hello,

there is one pin with 3 bond-wires, is that also in your model?

with best regards

Andreas

[floobydust]

The FEMM simulation seems to be showing almost the same temperature (gradient) a fixed distance away- whether FR-4 or air. I don't think that's correct if air has 1/10x the heat transfer compared to FR-4. What numbers are you using.

Thermal Conductivity (Tc) of Cu  ~ 355 W/m·K
Thermal Conductivity (Tc) of FR4 ~ 0.25 W/m·K
Thermal Conductivity (Tc) of Solder Mask ~ 0.21 W/m·K
Thermal Conductivity (Tc) of Air ~ 0.0275 W/m·K

https://www.engineeringtoolbox.com/thermal-conductivity-d_429.html


I see three bonding wires on heater pin 1 (?) It may be a tiny bit more heat transfer but the glass and Kovar I'd think would swamp that.

[gamalot]

I'm learning English, so I tried to translate it. 

[floobydust]

That pic came from a search engine caching amosbbs, it has pin numbers so I repost it here. Thanks for the translation, it seems fine.
OT, but I see six rings of heater resistors, two are used normally and four go to special factory test pads. A seventh thick ring might be for thermal equalizing.
Looks like two parallel transistors for each. Such lovely art.

Earlier discussions on LTZ thread page 74 here saying... slots are useless, the HP3458a does not use them and is proven. LTZ airflow and vibration can be worse with slots The 34470a ref is not metrology-grade so take it lightly. etc. etc. This topic has probably been talked about before.

I think slots have merit but maybe not. I'm going to try FEMM and see what I can learn modeling slots/pads on a pcb.

[z01z]

however 1 thing i dont understand, where is the 120ohm in this Vref?

There's a 100 Ohms resistor (R95?) beside the 1k one (R5), that could be it.

Regarding the model, shouldn't LTZcap connected to temp C?

[3roomlab]

however 1 thing i dont understand, where is the 120ohm in this Vref?

There's a 100 Ohms resistor (R95?) beside the 1k one (R5), that could be it.

Regarding the model, shouldn't LTZcap connected to temp C?

oh yes indeed

[Magnificent Bastard]

The slots attempt to accomplish 2 things:

• Reduce heater power required (important on battery, less important with the 'A' version).
• Keep the reference's heat out of the rest of the circuitry.

It's far more important to:

• Keep the LTZ1000's leads at the same temperature.
• Exclude air drafts from the LTZ1000's leads.

G10/FR4 board substrate material has a low thermal conductance, which is why the Wavetek 7001 has a ring but no slots.  The signal leads traces are on the next layer down with very little epoxy-glass material between the leads signal traces and the ring.  This helps maintain the same temperature.  In either case, an air-draft excluder (which can be a plastic pipe cap or even a copper pipe cap) is always necessary.

[floobydust]

As an amateur I tried FEMM for some slot temperature modelling. It's limited as 2D software, and exact models would be tedious to build.

I encountered many assumptions and errors in the modelling, so it's wrong- although shows some tendencies.

1. Each pad+pcb trace act as a tiny heatsink to the LTZ pin- which does not have infinite heat available at each lead. You'd have to 3D model it with the die+attach+bonding wire+glass+Kovar. Here I crudely set each pad equal to 35C but expect pad temps to drop, esp. with thick pcb traces. People feel the IC's leads have different temperatures, not all equal.

2. If in FEMM you select a material i.e. copper, it is solid through the plane thickness (i.e. 1.6mm PCB) which is not realistic for PCB traces. I tried the "smearing" approach where you average the conductivity based on volume (of Cu and FR4). In the end it made no big difference as FR-4 is way more of an insulator.

3. The slot's air temperature is awkward if you have no forced airflow as there is convection and radiation mechanisms at work.

4.  For pads, solid copper verses solder and Kovar and copper not a big difference as metals are pretty good conductors of heat. It's the FR-4 and air that really drive things.


My observations, that may be wrong:

The LTZ pcb center area and pads are pretty much isothermal, NOT including the thermal resistance of the die/attach/bonding wires/glass/leads, which may make effects much worse. I'd have to model more and see if the imbalance is more of an issue.

Slots:
- good for preventing mechanical strain on the LTZ due to PCB flex. If you have the LTZ located on a cantilever at one end of the ref PCB, it's not important.

- not ideal as you have to have FR-4 & Cu "spokes" for the pcb trace runout which means the heat loss surrounding the IC pads is not symmetrical. A slot close to a pin means the pad runs hotter.

This all would be better if the lead's thermal model was accurate.

[floobydust]

The LTZ1000 thermal modelling I looked at on eevblog does not include the glass on the Kovar leads.
Example Schott 8250 borosilicate is 1.2W/m-K so a good insulator.

I expect the LTZ leads would be supplied heat through that high thermal resistance (TO-5 base-glass-lead), meaning PCB pad temperatures even more sensitive than models suggest, to slots, airflow and copper traces.

I've stopped working on this as others have done 3D modelling, compared to this limited 2D work which can be misleading.


edit: about slots on SMT packaged references from AN82 Understanding and Applying Voltage References:

Board Stress
Burn-in can help “relax” a stuffed board, but additional mechanical stress may be introduced when the board is mounted into the product. Stress has a directly measurable effect on reference output. If the stress changes over a period of time, it may manifest itself as unacceptable long-term drift. Circuit boards are not perfectly elastic, so bending forces may cause permanent deformation and a permanent step-change in reference output  voltage.
Devices in metal (TO-5 and TO-46) packages are largely immune to board stress, owing to the rigidity of the package and the flexibility of the leads. Plastic and surface mount packages are another matter.
Board stress effects are easily observed by monitoring the output of a reference while applying a bending force to the board. A controlled experiment was performed to measure  the effect of board stress on an LT1460CS8-2.5 surface  mount  reference; see pic

[3roomlab]

there was a secondary question nagging in the back of my mind for some time regarding the effect of the shape of slots. i wasnt specifically sure how to simulate this problem/question, but then i came up with this idea (see pic). 1 of the questions which i sort of answered myself using simulations is that, i understood small widths and long thermal paths help to isolate the inside from the outside. but since PCB is not infinite size, will thermal isolation be different if we squeeze the size into a square? also for the sake of fitting square things on a square PCB.

*updated, a few more versions is done for this simulation, up to where the square and spiral leak are similar, but the square appear to require a slightly larger area, and have isolines spewing guts over a larger area. it would be pointless to expand the square further so that it performs "better" than the spiral but waste more PCB area to contain the heat. i guess that somewhat satisfies my curiosity about shapes if you use a rule to compare the sizes on screen, the spiral advantage is just about missing the square corner bits. now for reality, who would sit a VREF in the middle of a spiral? but in any case, this spiral/square thing is quite a surprise and fun to simulate.

[floobydust]

I think you need symmetry around the LTZ pads to ensure equal heat loss at each pad. Equal area of FR-4, copper, air gaps etc. as best is possible.

Using slots, notice some pads have lots of FR-4 surrounding them, others have little FR-4 and more insulated because they are close to air (slots). So some pads would be cooler or hotter than others which we do not want.

I'd think the copper path is dominant for heat transfer but FEMM simulations aren't really showing a difference in pad temperature due to a PCB trace cooling the pad.
A PCB trace is thin copper foil on FR-4 laminate, I was not sure what numbers to use 'volume weighted, effective thermal conductivity, in the thermal modeling of a printed circuit board' numbers in various papers are all over the place.

I see mistakes in LTZ thermal simulations, i.e missing the glass/Kovar interface, so the heat output from a LTZ can lead is less than we are assuming (infinite), and plotting flux vs isotherms, no convection or radiation losses, etc.

So it's hard to tell how much upset slots can cause to the pad temperatures.