Author Topic: Temperature time constant of chip capsule epoxy?  (Read 1572 times)

0 Members and 1 Guest are viewing this topic.

Offline MTTopic starter

  • Super Contributor
  • ***
  • Posts: 1687
  • Country: aq
Temperature time constant of chip capsule epoxy?
« on: August 20, 2019, 02:12:13 am »
Whats the temperature time constant of chip capsule epoxy? Particularly SOT323 and SC70 ie SOT23-6?
« Last Edit: August 20, 2019, 02:45:39 am by MT »
 

Offline Circlotron

  • Super Contributor
  • ***
  • Posts: 3362
  • Country: au
Re: Time constant of chip capsule epoxy?
« Reply #1 on: August 20, 2019, 02:45:16 am »
"Time constant" is not the right term. Doesn't make sense. Can you describe it differently?
 

Offline MTTopic starter

  • Super Contributor
  • ***
  • Posts: 1687
  • Country: aq
Re: Temperature time constant of chip capsule epoxy?
« Reply #2 on: August 20, 2019, 02:46:41 am »
Updated!  :)
« Last Edit: August 20, 2019, 02:58:20 am by MT »
 

Offline DaJMasta

  • Super Contributor
  • ***
  • Posts: 2393
  • Country: us
    • medpants.com
Re: Temperature time constant of chip capsule epoxy?
« Reply #3 on: August 20, 2019, 03:12:38 am »
Thermal conductivity/resistance?  Dielectric constant?

If it's the thermal resistance, such as junction to ambient, it's usually specified in the chip's datasheet, as it will differ somewhat with the die and construction - TI's app note SLVA937 outlines how differing internal constructions can play a big difference in effective heat dissipation.

If it's dielectric constant, this is already factored into the specifications of the device and only very rarely has significant interactions with external parts, but there are some guidelines floating around for approximate/likely values.  Perhaps another free reference available online (which also addresses some thermal resistance properties of materials used) is Intel's "Physical Constants of IC Package Materials" which is chapter 5 of the 2000 Intel Packaging Databook.
 
The following users thanked this post: MT

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Temperature time constant of chip capsule epoxy?
« Reply #4 on: August 20, 2019, 03:13:37 am »
Ballpark 10s to 1000s of milliseconds.  Yup, see the thermal transient response curves. :)

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 
The following users thanked this post: MT

Offline MTTopic starter

  • Super Contributor
  • ***
  • Posts: 1687
  • Country: aq
Re: Temperature time constant of chip capsule epoxy?
« Reply #5 on: August 20, 2019, 11:52:17 pm »
Thermal conductivity/resistance?  Dielectric constant?

If it's the thermal resistance, such as junction to ambient, it's usually specified in the chip's datasheet, as it will differ somewhat with the die and construction - TI's app note SLVA937 outlines how differing internal constructions can play a big difference in effective heat dissipation.

If it's dielectric constant, this is already factored into the specifications of the device and only very rarely has significant interactions with external parts, but there are some guidelines floating around for approximate/likely values.  Perhaps another free reference available online (which also addresses some thermal resistance properties of materials used) is Intel's "Physical Constants of IC Package Materials" which is chapter 5 of the 2000 Intel Packaging Databook.

Thanks for suggestions. I have looked them up.

Thermal conductivity, but i doubt package epoxy place major role since legs so (case SOT363) short they carry the heat in or out of die, i ponder if there really is a 0603 NTC (cheapness solution) with similar temp change rate that can instantly compensate (current sink exponanting pair) temp drift there will be a distance from legs to 0603 even if NTC bridged over a leg and epoxiblobbying the whole package. 
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Temperature time constant of chip capsule epoxy?
« Reply #6 on: August 21, 2019, 12:26:25 am »
Thermal conductivity of encapsulant isn't terrible; it's filled with a whole lot of stuff, not just epoxy.  It's mostly silica fume IIRC.

Anything board level isn't going to be great though: heat has to flow through the leads, traces and board laminate to reach another component.  Time constant ballpark one second.

I've done it before for a protection feature -- a bunch of TVSs absorbing fault energy, with copper pours on their leads wrapping around a thermistor.  This worked partly because the TVS has a high temperature rating (it'll function and survive at 200°C) and because the thermistor is set for a fairly low threshold (something like 60-80°C) so it trips early through the thermal-time-constant slope.

For analog thermal matching purposes, you're pretty much stuck either dealing with it, or using a more mainstream/modern solution.

You don't need any matching at all for low accuracy.  For high accuracy, you might use a cascode, or series resistors, to help equalize power dissipation.  The problem with a compensating thermistor is, its response is delayed, so it can account for ambient temperature drift, but cannot respond to low frequency AC changes (say 0.1-10Hz), which will be mainly driven by convection, for which some thermal insulation or mass might be helpful.  For high accuracy, simply not using one at all is preferable, in which case you might synthesize the equivalent function in the digital domain (FPGA or MCU).  Or at least, until Zeloof nails down his process well enough to make custom proto ICs... :-DD

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 
The following users thanked this post: MT

Offline MTTopic starter

  • Super Contributor
  • ***
  • Posts: 1687
  • Country: aq
Re: Temperature time constant of chip capsule epoxy?
« Reply #7 on: August 21, 2019, 11:35:31 pm »
Thermal conductivity of encapsulant isn't terrible; it's filled with a whole lot of stuff, not just epoxy.  It's mostly silica fume IIRC.

Silica fume+ink+binding agent?

Quote
Anything board level isn't going to be great though: heat has to flow through the leads, traces and board laminate to reach another component.  Time constant ballpark one second.

My point heating/cooling tests shows close to a second for SC70/SOT23-6 probably itty bitty quicker for SOT363 so yep.

Any idea about a 0603 NTC that have 0.9-1sec response time? I actually just need compensation from 15 to 30 degC.
So far i have only found a through-hole that have 1,2sec response time but thats in oil. Vishay NTCLE100E3103JB0
E.g the TDK/EPCOS B57330V2103J260 have 4sec cooling time constant and others much worse.

Quote
I've done it before for a protection feature -- a bunch of TVSs absorbing fault energy, with copper pours on their leads wrapping around a thermistor.  This worked partly because the TVS has a high temperature rating (it'll function and survive at 200°C) and because the thermistor is set for a fairly low threshold (something like 60-80°C) so it trips early through the thermal-time-constant slope.

Quote
For analog thermal matching purposes, you're pretty much stuck either dealing with it, or using a more mainstream/modern solution.
Whats a "moderns solution" to exponating a current whos cost equaling 2 trannies and a NTC/PTC and some scaling  resistors set aside Zeloof's garage home cooked integrated solutions?

Quote
You don't need any matching at all for low accuracy.  For high accuracy, you might use a cascode, or series resistors, to help equalize power dissipation.  The problem with a compensating thermistor is, its response is delayed, so it can account for ambient temperature drift, but cannot respond to low frequency AC changes (say 0.1-10Hz), which will be mainly driven by convection, for which some thermal insulation or mass might be helpful. 

Yep i know, back in the day we just slabbed a KT81 PTC over matched pair (DIL or Sil) epoxied or just thermal compounded it all and that was it. I only need to compensate for a specific voltage (current in the pair) in the given temp range, everything beyond that is just a bonus.

Quote
For high accuracy, simply not using one at all is preferable, in which case you might synthesize the equivalent function in the digital domain (FPGA or MCU).  Or at least, until Zeloof nails down his process well enough to make custom proto ICs... :-DD

The current sink's is (are several but dispersed over PCB) controlled by a MCU, 16bit DAC but there is no room for temp change feedback loop traces nor ADC channels to the MCU nor a linear (switched current ranges) current sink , which i would have preferred.
« Last Edit: August 21, 2019, 11:46:42 pm by MT »
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Temperature time constant of chip capsule epoxy?
« Reply #8 on: August 22, 2019, 12:32:17 am »
Any idea about a 0603 NTC that have 0.9-1sec response time?

Such a rating would be pointless: the thermal mass of a chip component doesn't much depend on the material it's made from.

A smaller component altogether will have a shorter time.

It's not really a thermal time constant, it's more about thermal diffusion through the board material and traces; it's a bit messy of a process.  So you wouldn't necessarily expect an 0603 to have half the time constant of a 1206, despite being half the length and width.


Quote
Whats a "moderns solution" to exponating a current whos cost equaling 2 trannies and a NTC/PTC and some scaling  resistors set aside Zeloof's garage home cooked integrated solutions?

Beats the hell outta me, what are you exponentiating and why?

Point being, the solution lies not in solving superficial problem Y, but underlying problem X.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 
The following users thanked this post: MT

Offline DaJMasta

  • Super Contributor
  • ***
  • Posts: 2393
  • Country: us
    • medpants.com
Re: Temperature time constant of chip capsule epoxy?
« Reply #9 on: August 22, 2019, 03:05:39 am »
Just brainstorming, don't actually know if they'd apply to your problem or if they'd be worth the effort, but you could try:

Measuring current into the device to infer temperature spikes with a slower temperature sensor for ambient changes and a micro.

A noncontact (ir) thermometer could give you very quick monitoring with no added thermal mass or interference with natural cooling.... but it would be expensive and I don't know of a side-firing module that would make it easy to monitor a part on a PCB on the same PCB.

Ovenization.  If you want your device to be thermally stable, you can turn up the heat to beyond it would normally get to and then just hold it there and not have to compensate.
 
The following users thanked this post: MT

Online David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: Temperature time constant of chip capsule epoxy?
« Reply #10 on: August 22, 2019, 10:29:53 am »
Whats a "moderns solution" to exponating a current whos cost equaling 2 trannies and a NTC/PTC and some scaling  resistors set aside Zeloof's garage home cooked integrated solutions?

For low changes in current, the change in power dissipation is minimal so all that is required is good thermal coupling.  This goes double when the external signal conditioning holds Vce to a very low value.

The common alternative is to use an integrated transistor array with one or two transistors operating as active elements, one transistor operating as a temperature sensor, and one transistor operating as a heater.  This reduces the thermal time constant to milliseconds or lower.
 
The following users thanked this post: MT

Offline MTTopic starter

  • Super Contributor
  • ***
  • Posts: 1687
  • Country: aq
Re: Temperature time constant of chip capsule epoxy?
« Reply #11 on: August 23, 2019, 01:22:15 am »
Whats a "moderns solution" to exponating a current whos cost equaling 2 trannies and a NTC/PTC and some scaling  resistors set aside Zeloof's garage home cooked integrated solutions?

For low changes in current, the change in power dissipation is minimal so all that is required is good thermal coupling.  This goes double when the external signal conditioning holds Vce to a very low value.

The common alternative is to use an integrated transistor array with one or two transistors operating as active elements, one transistor operating as a temperature sensor, and one transistor operating as a heater.  This reduces the thermal time constant to milliseconds or lower.

Im perfectly aware of sensor/oven solution e.g LM3046 (used it many times) for instance but thats one are obsoleted (thanks TI) not to mention uber obsolete Ua723 , no not the regulator.

My problem , and i apologize if not explained in clear is board space and component count to fix a thermal drift
in the case of expo and range issue in the case of linear. The linear one requires redesign and more control lines
and eats more components thereof the expo type, both gives different problems to what has to be done by the actual circuit.

Other exotic solutions are available such as double expo's/multipliers one counteract the other tempco but thats only reasonable integrated and what was once available in 1980 is obsolete since then. Also a temp sensor in each current sink (6 of) feedback to MCU as mentioned previous, or linear current sink with switchable ranges, all cost more in all aspects of the initial design thats why the NTC/PTC request for that tiny fickle current that shift just so tiny much.

Quote
Quote
Any idea about a 0603 NTC that have 0.9-1sec response time?

Quote
Such a rating would be pointless: the thermal mass of a chip component doesn't much depend on the material it's made from.

Sorry dont understand your point, if the NTC are made of concrete it would ofcourse react very different then what a NTC usually are made of would. I aslo see package reaction times between DIP14 and SOT23-6 beeing quicker.
So if NTC/PTC have one time constant and the current sink another it not going to work right?
Also there is tolerances on NTC B and PTC t.c.r who usually is 10% some 5% but anyway, while el cheapo NTC can be had at 1%.

Quote
A smaller component altogether will have a shorter time.
yes, thats my point! I thought!

Quote
It's not really a thermal time constant, it's more about thermal diffusion through the board material and traces; it's a bit messy of a process.  So you wouldn't necessarily expect an 0603 to have half the time constant of a 1206, despite being half the length and width.

Well, looking at the NTC/PTC time constants specs are all over the place, some manufacturer dont even specify,
some manufacturers of chip termistors also not mention self heating.Some even say chip types is not on par with wirewounds while they have their problems as well.

Quote
Beats the hell outta me, what are you exponentiating and why?
A tiny control current point that needs to be compensated due to environmental temp change.

Quote
Point being, the solution lies not in solving superficial problem Y, but underlying problem X.
Tim

Uhum yes, and thats what im trying to do and i dont have many options then a PTC or NTC.

So there are no 0603 or 0805 NTC 3300ppm i can epoxy blobbing together to fix temp drift?
Are there any linear current sink/source , low pin count chip that either have automatic current
range switching or can be modified to have switchable current ranges?
« Last Edit: August 23, 2019, 01:32:47 am by MT »
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: Temperature time constant of chip capsule epoxy?
« Reply #12 on: August 23, 2019, 10:04:43 pm »
Whats a "moderns solution" to exponating a current whos cost equaling 2 trannies and a NTC/PTC and some scaling  resistors set aside Zeloof's garage home cooked integrated solutions?

For low changes in current, the change in power dissipation is minimal so all that is required is good thermal coupling.  This goes double when the external signal conditioning holds Vce to a very low value.

The common alternative is to use an integrated transistor array with one or two transistors operating as active elements, one transistor operating as a temperature sensor, and one transistor operating as a heater.  This reduces the thermal time constant to milliseconds or lower.

Im perfectly aware of sensor/oven solution e.g LM3046 (used it many times) for instance but thats one are obsoleted (thanks TI) not to mention uber obsolete Ua723 , no not the regulator.

My problem , and i apologize if not explained in clear is board space and component count to fix a thermal drift
in the case of expo and range issue in the case of linear. The linear one requires redesign and more control lines
and eats more components thereof the expo type, both gives different problems to what has to be done by the actual circuit.

Other exotic solutions are available such as double expo's/multipliers one counteract the other tempco but thats only reasonable integrated and what was once available in 1980 is obsolete since then. Also a temp sensor in each current sink (6 of) feedback to MCU as mentioned previous, or linear current sink with switchable ranges, all cost more in all aspects of the initial design thats why the NTC/PTC request for that tiny fickle current that shift just so tiny much.

Another method which might apply to your situation is to use a PTC as a temperature regulator and heater.  They used to make little PTC ovens which fit on TO-99 cans and worked very well to make very low drift operational amplifiers.
 
The following users thanked this post: MT

Offline ajb

  • Super Contributor
  • ***
  • Posts: 2785
  • Country: us
Re: Temperature time constant of chip capsule epoxy?
« Reply #13 on: August 23, 2019, 10:37:30 pm »
You could certainly stick an NTC to an IC or transistor package to get the thermal parameters of the board out of the equation.  Time constant is then dependent on the mass of the NTC and the total thermal resistance between the die and the NTC.  Would that get you a fast enough response? I dunno.  I guess that if you wanted an even faster response you could weld a thermocouple onto a lead, that would give you minimal additional mass and intermediate thermal resistance.  Then you just have to deal with rejecting whatever signal is present on that particular lead from the TC signal (plus all the normal annoyances of using thermocouples!)
 
The following users thanked this post: MT


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf