Thermometers, Thermocouples, and Thermal Readings

[robincnn]

Well ,yes ..
I agree ..
I prefer large diameter fans with low rpm and low power dissipation.
But the heat sink has to be chosen in accordance with that kind of fans ..
Meaning thick and short fins ,with quite a space between them ,in order to minimise
flow resistance (thus pressure drop ) ,as large rotor /low rpm /low power fans tend to stall ,
under high flow resistance .

This kind of active cooling is way different in many ways ,than HSF systems of
small diameter fans with high rpm ...
Those fans work best with another design of heat sinks ...

Regarding the mass and size of the heat sink ,the large diameter /low rpm HSF system ,
falls somewhere in between passive cooling and small diameter / high rpm active HSF system .
The former needs large size/mass heatsink ,while the latter utilises smallest mass/size heat sink possible ...
The former is totally quite ,the latter quite noisy ...
The former is fail-proof ,the latter if the fan fails ...bye-bye COBs ...

My taste ,is somewhere in between ...

I agree the heatsink choice is important along with the fan size and type. Fan blades design and static pressure also important.
A fan with high static pressure may do better under high flow resistance.
And a fan with higher CFM but lower static pressure may do better under low flow resistance.

Seems you prefer an active cooled design with heatsink design such that NO 'Bye bye COB' if the fan fails. Thick short fins with wide space in between or
Medium size Pin type with pins length cut to 30-40 mm and a fan should work well in active cooling and allow sufficient passive cooling if fan fails.

 

[stardustsailor]

I agree the heatsink choice is important along with the fan size and type. Fan blades design and static pressure also important.
A fan with high static pressure may do better under high flow resistance.
And a fan with higher CFM but lower static pressure may do better under low flow resistance.

Seems you prefer an active cooled design with heatsink design such that NO 'Bye bye COB' if the fan fails. Thick short fins with wide space in between or
Medium size Pin type with pins length cut to 30-40 mm and a fan should work well in active cooling and allow sufficient passive cooling if fan fails.
View attachment 3596466

Yes ..But pin type heat sinks are way expensive ...
That's what I think ,at least ...
And not that easy to find them in appropriate sizes ,to fit a large COB ...

 

[Abiqua]

http://www.qats.com/Qpedia/Qpedia_v8_apr14.ashx

The junction to case thermal resistance [Junction-to-Case J-C] of a semiconductor package is not uniquely defined, but depends on the cooling condition at the package boundary (case). Computer simulations predict differences up to 31% between J-C values computed with constant and with floating case temperature boundary conditions...


Measured values of J-C are 15% to 58% higher than those predicted by the corresponding simulation with constant case temperature boundary condition. This is a consequence of the fact, that thermocouple measurements almost always over-estimate the Junction to Case. It is very difficult to ensure that he thermocouple actually measures the case temperature (Tc) of the package and not the temperature of the cold plate or some average value between the two. Also, different set-ups are likely to produce deviant (j-c) values.
Transient dual interface measurements, which are accurate to 15%, are so far the only known way to produce more reliable measurement results....


The boundary conditions must always be stated to enable a fair comparison between the (J-C) values obtained through different methods. In order to have a realistic common base for simulated and measured values, a heat transfer co-efficient boundary condition [in the operating range of a liquid-cooled cold plate} should be defined and applied in all simulations.

While there is a considerable difference between the (J-C) values for constant and floating case temperature boundary conditions, this difference is of little practical importance, since (J-C) constitutes usually only a small part of the total junction-to-ambient thermal resistance.


???????

 

[Abiqua]

brain melt

Transient dual interface measurement of the rth-jc of power semiconductor packages
http://www.electronics-cooling.com/2010/09/transient-dual-interface-measurement-of-the-rth-jc-of-power-semiconductor-packages/

 

[AquariusPanta]

brain melt

Transient dual interface measurement of the rth-jc of power semiconductor packages
http://www.electronics-cooling.com/2010/09/transient-dual-interface-measurement-of-the-rth-jc-of-power-semiconductor-packages/

This article was way easier to trek through in comparison to the previous article you provided. Both are good reads.

(From the latter article) "Due to heat-spreading, the heat-flux decreases quickly with increasing distance from the chip surface."

It hadn't occurred to me before but the distance between the case temperature (Tc) measuring points for the Vero 29 and CXB3590 share different distances from the LES or the chip surface. Surely that affects the reading of temperatures at some degree with the use of a thermocouple, with the CXB3590's Tc measuring point being more than twice the distance from it's LES when compared to that of the Vero 29's LES/TCMP distance.

Keep in mind that heat rises.

 

[Abiqua]

This article was way easier to trek through in comparison to the previous article you provided. Both are good reads.

(From the latter article) "Due to heat-spreading, the heat-flux decreases quickly with increasing distance from the chip surface."

It hadn't occurred to me before but the distance between the case temperature (Tc) measuring points for the Vero 29 and CXB3590 share different distances from the LES or the chip surface. Surely that affects the reading of temperatures at some degree with the use of a thermocouple, with the CXB3590's Tc measuring point being more than twice the distance from it's LES when compared to that of the the Vero 29's LES/TCMP distance.

Yes! I noticed this too and hadn't thought of it again until you said something....maybe only 1-2mm of actual Distance farther away from LES, but on CREE, it seems the LES has much more pronounced "lip" around LES surface and the TC point actually sits below the LES instead of nearly level with Vero LES.......maybe the Vero has more difficulties to get an accurate Tc reading just by a simple measure......

but playing Devil's advocate here, Tc is important for electronics design, you would think someobody said "Hey, what about the Tc design and its measuring limitations?" or did someone just say "Eh, slap the fucker on there, Peace" You would .....think....that Bridgelux is aware of Tc monitoring limitations, as well as Cree, at least with a thermocouple, before dropping millions? on tooling dies......you would think.....


and I haven't actually gotten a chance to read the 2nd article completely, maybe in a hour or two

 

[guod]

this is a standard method.
from philips-lumileds

Measuring Rth J-C
The junction-to-case thermal resistance, Rth J-C, is a measure of how well heat can be dissipated from the die to the package case from which heat is extracted by placing it in thermal contact with an external heat sink. For single emitters, the case location is defined as the back of the LED package at the center of the thermal pad (see Figure 1). For emitter arrays where the LEDs are attached to a printed circuit board (PCB) or metal core printed circuit board (MCPCB), the case is defined as the center of the Light Emitting Surface (LES) on the opposite side of the LED substrate (see Figure 2).
Lumileds uses the transient dual interface method, which is described in great detail in JDEC Standard JESD51-14 [1], to determine Rth J-C. This method measures the transient cooling curve for the same power device twice, with thermal interface materials of differing thermal conductivity between the device and the heat sink. As heat travels from the die towards the heat-sink, the two transient cooling curves will overlap until a point in time where the curves diverge due to the difference in thermal interface between the device and the heat-sink. The benefits of using transient measurements of the junction temperature as opposed to the thermocouple method described in MIL Standard 833 [2] is that the transient method does not require determination of the case temperature, which is difficult to measure both accurately and repeatably between different measurement set-ups.

source: http://www.lumileds.com/uploads/568/WP23-pdf


one of the tools for this

MicReD tools T3Ster & TERALED
T3Ster - the thermal transient tester is designed with the needs of the semiconductor, electronic appliance and LED industries and R&D laboratories in mind. T3Ster is the fastest and most accurate thermal transient tester on the market. Its related software T3Ster-Master fully supports the latest thermal testing standard JEDEC JESD51-14 for junction-to-case thermal resistance measurement. TERALED is a CIE 127:2007 compliant total flux measurement system with temperature control. When used together with T3Ster for high-power LEDs, it performs self-consuistent thermal and radiometric/photometric characterization LEDs - forming a comprehensive LED testing station. Full automation of the measurements with TERALED allows extremely fast operation: the LED under test can be characterized in over 100 operating points (forward current and temperature combinations) in about 2 hours.


and if you want real numbers read this...guilty for all cobs!!!

http://www.cree.com/~/media/Files/Cree/LED Components and Modules/XLamp/XLamp Application Notes/Solder_Point_Temp.pdf

 

[stardustsailor]

This article was way easier to trek through in comparison to the previous article you provided. Both are good reads.

(From the latter article) "Due to heat-spreading, the heat-flux decreases quickly with increasing distance from the chip surface."

It hadn't occurred to me before but the distance between the case temperature (Tc) measuring points for the Vero 29 and CXB3590 share different distances from the LES or the chip surface. Surely that affects the reading of temperatures at some degree with the use of a thermocouple, with the CXB3590's Tc measuring point being more than twice the distance from it's LES when compared to that of the Vero 29's LES/TCMP distance.

Keep in mind that heat rises.

Yes! I noticed this too and hadn't thought of it again until you said something....maybe only 1-2mm of actual Distance farther away from LES, but on CREE, it seems the LES has much more pronounced "lip" around LES surface and the TC point actually sits below the LES instead of nearly level with Vero LES.......maybe the Vero has more difficulties to get an accurate Tc reading just by a simple measure......

but playing Devil's advocate here, Tc is important for electronics design, you would think someobody said "Hey, what about the Tc design and its measuring limitations?" or did someone just say "Eh, slap the fucker on there, Peace" You would .....think....that Bridgelux is aware of Tc monitoring limitations, as well as Cree, at least with a thermocouple, before dropping millions? on tooling dies......you would think.....


and I haven't actually gotten a chance to read the 2nd article completely, maybe in a hour or two

Gentlemen ...
Take under consideration ,that those two COBs have some big design differences between them ,
that might be justifying the fact that they have their Tc measuring point ,located in different spots ...

1) CXx COBs have a case made of Ceramic material and no plastic case over them.
Ceramics do not share the same thermal properties with metal substrates ..
(Different Thermal air convection ,emissivity ,conduction & thermal capacitance )

2) Vero COBs have aluminium case ,with a round plastic " jacket/holder " over their square case


My guess is that ceramic cases have a more uniform / even temperature across them,
than aluminium cases ..
Aluminium cases most probably ,do have a more pronounced " temperature difference gradient " across them, as they have less thermal capacity than ceramic materials and higher thermal conductivity & air convection.

Thus ,at Vero series Tc measuring point has to be closer in the chips ,
while at CXx series that Tc measuring point can be further away from the chips ...

My guess ..

Cheers.

 

[frica]

Well ,yes ..
I agree ..
I prefer large diameter fans with low rpm and low power dissipation.
But the heat sink has to be chosen in accordance with that kind of fans ..
Meaning thick and short fins ,with quite a space between them ,in order to minimise
flow resistance (thus pressure drop ) ,as large rotor /low rpm /low power fans tend to stall ,
under high flow resistance .

This kind of active cooling is way different in many ways ,than HSF systems of
small diameter fans with high rpm ...
Those fans work best with another design of heat sinks ...

Regarding the mass and size of the heat sink ,the large diameter /low rpm HSF system ,
falls somewhere in between passive cooling and small diameter / high rpm active HSF system .
The former needs large size/mass heatsink ,while the latter utilises smallest mass/size heat sink possible ...
The former is totally quite ,the latter quite noisy ...
The former is fail-proof ,the latter if the fan fails ...bye-bye COBs ...

My taste ,is somewhere in between ...

I do wonder if that would really happen.

According to the Cree datasheet the max case temp for the CXB3590 72V at 700mA (which a lot of people drive them at from what I've seen) is slightly above 125 degrees celcius.
That's a pretty high temp to reach with a 50 watt 50% efficient light.
I think as long as the defect fan is noticed within 24 hours it probably won't bring any long term COB damage but there are a lot of factors that will influence and damage like amperage or room temp

 

[AquariusPanta]

Gentlemen ...
Take under consideration ,that those two COBs have some big design differences between them ,
that might be justifying the fact that they have their Tc measuring point ,located in different spots ...

1) CXx COBs have a case made of Ceramic material and no plastic case over them.
Ceramics do not share the same thermal properties with metal substrates ..
(Different Thermal air convection ,emissivity ,conduction & thermal capacitance )

2) Vero COBs have aluminium case ,with a round plastic " jacket/holder " over their square case


My guess is that ceramic cases have a more uniform / even temperature across them,
than aluminium cases ..
Aluminium cases most probably ,do have a more pronounced " temperature difference gradient " across them, as they have less thermal capacity than ceramic materials and higher thermal conductivity & air convection.

Thus ,at Vero series Tc measuring point has to be closer in the chips ,
while at CXx series that Tc measuring point can be further away from the chips ...

My guess ..

Cheers.

That's a good point to bring up, that the housing materials for the CXB and Vero vary; it wouldn't make sense for Bridgelux to place the Tc measuring point anywhere besides onto the aluminum case of the LES, as the plastic is merely there for fastening down the entire piece and wires.

If any of you have held a Vero 29 and CXB3590 close to one another, you might eventually come to the conclusion that the available surface area for heat transfer between the COB and the mating surface (heat sink) is slightly greater for the CXB3590 than the Vero 29, as well as being slightly thicker but not by much.

@guod

(from the last link you provided)

"Cree recommends using type T thermocouples to take measurements on CX family LEDs. The CX Family Design Guide provides information on attaching a thermocouple and temperature measurement for CX family LEDs. For other XLamp® LEDs, Cree recommends using type K thermocouples that are insulated with Teflon® (FEP) or wrapping the thermocouple leads that are on the printed‑circuit board (PCB) with three layers of white Teflon tape.3 Doing so blocks any direct light on the thermocouple during testing."

Any idea on why they might recommend using type T thermocouples instead of type K??

 

[stardustsailor]

That's a good point to bring up, that the housing materials for the CXB and Vero vary; it wouldn't make sense for Bridgelux to place the Tc measuring point anywhere besides onto the aluminum case of the LES, as the plastic is merely there for fastening down the entire piece and wires.

If any of you have held a Vero 29 and CXB3590 close to one another, you might eventually come to the conclusion that the available surface area for heat transfer between the COB and the mating surface (heat sink) is slightly greater for the CXB3590 than the Vero 29, as well as being slightly thicker but not by much.

@guod

(from the last link you provided)

"Cree recommends using type T thermocouples to take measurements on CX family LEDs. The CX Family Design Guide provides information on attaching a thermocouple and temperature measurement for CX family LEDs. For other XLamp® LEDs, Cree recommends using type K thermocouples that are insulated with Teflon® (FEP) or wrapping the thermocouple leads that are on the printed‑circuit board (PCB) with three layers of white Teflon tape.3 Doing so blocks any direct light on the thermocouple during testing."

Any idea on why they might recommend using type T thermocouples instead of type K??

You can't solder type K thermocouples at the Tc solder point ....

 

[AquariusPanta]

You can't solder type K thermocouples at the Tc solder point ....

Yeah I tried with the CXB3590, twas a no go.

 

[Abiqua]

That's a good point to bring up, that the housing materials for the CXB and Vero vary; it wouldn't make sense for Bridgelux to place the Tc measuring point anywhere besides onto the aluminum case of the LES, as the plastic is merely there for fastening down the entire piece and wires.

If any of you have held a Vero 29 and CXB3590 close to one another, you might eventually come to the conclusion that the available surface area for heat transfer between the COB and the mating surface (heat sink) is slightly greater for the CXB3590 than the Vero 29, as well as being slightly thicker but not by much.

@guod

(from the last link you provided)

"Cree recommends using type T thermocouples to take measurements on CX family LEDs. The CX Family Design Guide provides information on attaching a thermocouple and temperature measurement for CX family LEDs. For other XLamp® LEDs, Cree recommends using type K thermocouples that are insulated with Teflon® (FEP) or wrapping the thermocouple leads that are on the printed‑circuit board (PCB) with three layers of white Teflon tape.3 Doing so blocks any direct light on the thermocouple during testing."

Any idea on why they might recommend using type T thermocouples instead of type K??

Yes, thank you Guod....very nice!


To me this is the more interesting part of the quote....

For other XLamp® LEDs, Cree recommends using type K thermocouples that are insulated with Teflon® (FEP) or wrapping the thermocouple leads that are on the printed‑circuit board (PCB) with three layers of white Teflon tape.3 Doing so blocks any direct light on the thermocouple during testing."


And also the Cree is using Tsp [solder point] and vero is referring to Tc........

Also those fuckers, Mentor Graphics are not 10 miles down the road from me .....I cannot afford their dual transient measuring equipment cripes....although it would be killer to have!

 

[Abiqua]

sorry for the delay all....like I mentioned before, slipped a disk [again] and its been a doozy....Did manage to mount some 3500BB 3070's onto the Mecha sinks....Tsp is covered by BJB holders and haven't fucked with it yet, and I don't have a T Couple either, although there is plenty of room Under the holder to mount to the Tsp......

So I am just sticking with my trusty K type Lutron handheld for the most rudimentary of measurements...

Driver is Meanwell 120-1050A
with 5v Fan Supply

24-27 C Ta


The mecha are nice to get a thermo reading right from the heatsink itself....



Need to convert the math, but reading is in Ft/min

12v Power


5v



Its working!