Internal Temperatures (SoC)

Black Hole

May contain traces of nut
Using a early unit with horizontal RF sockets, thermocouple on SoC IC, wire coming through a specially-punched 3.5mm hole and cover fitted properly, unit not recording or playing back during test, fan set to 50%, ambient temp 20-20.5C throughout, HDD temps as reported by drive, figures after things had stabilised for >20 minutes [Edit: That's after they's been stable for 20 minutes.]

Standard unit
Fan blowing out: HDD 39C, SoC 61C​
Fan blowing in: HDD 31C, SoC 62C​

With my slot-in-HDD-carrier mod
Fan blowing out: HDD 39C, SoC 61C​
Fan blowing in: HDD 31C, SoC 59C​

So there you have it - having the fan blow air in results an 8C cooler HDD as tested and has no effect on the SoC.

Cutting a slot in the HDD carrier and the fan blowing air in does result in a small drop in SoC temp but not as much as I'd hoped.

I've now done some experiments myself, using a 1TB HDR-FOX recently acquired (only 6½k hours on the HDD). My test probe is a thermocouple attachment for an ancient DMM resurrected for the purpose, which I was able to thread through the existing punch holes in the base and tape to the centre of the heatsink on the SoC (the SoC silicon will be substantially hotter than the heatsink, due to the thermal resistances (see spoiler) from the silicon to the heatsink through the encapsulation and any bonding, but without knowledge of the resistance and the actual power dissipation it is not possible to calculate the actual silicon temperature).

Heat moves in a similar way to a fluid or even electricity, from a source to a sink. The temperature difference between the source and the sink is analogous to pressure or voltage difference, and the rate at which heat is conducted is analogous to conductance (the inverse of resistance). Consequently, characteristics of the heat flow can be calculated using Kirchoff's Laws reapplied from electrical circuits to heat circuits, which (essentially) say that (in the steady state) the sum of the currents (or heat flow) into a circuit node is equal to the sum of the currents out of the node.

Thermal resistance is quoted in units of °C/W (degrees Celcius per Watt), in other words how great the temperature difference has to be to drive one Watt of thermal power through that item, and each component in the heat path can be stacked up in series or parallel (or combinations) just like resistors or water pipes. Usually we are only dealing with the series path from the heat source (the actual silicon of a chip), through the chip's encapsulation, to the heatsink.

The chip's datasheet will declare the thermal resistance from the chip to the surface of the encapsulation, and then the heatsink has a quoted (or calculated from standard formulae) thermal resistance from the contact area to free air and/or forced air. Add the two together and you get the total thermal resistance from the chip to free/forced air, and therefore the temperature difference between ambient and the silicon (presuming you know how many Watts the silicon is dissipating). There also needs to be a consideration for the thermal resistance of the bonding between the chip encapsulation and the heatsink – whether that be a shim of thermal grease (good conduction) or just touching contact (poor conduction).

In parallel with the heat path from silicon to heatsink, there is also the heat path from silicon to PCB. On a through-hole leaded component this path tends to have comparatively high thermal resistance, because the heat has to pass along the fine electrical bonding wires (or the high-resistance encapsulation) to the lead frame (as the metal pins are called), then through the lead frame to the solder pads on the PCB. Typically the component body is not in direct contact with the PCB. However, surface-mount components are in direct contact with the PCB, and high-power components frequently enhance that contact using metal thermal bonding soldered directly to the PCB under the package (by using industrial reflow soldering). Combined with a PCB deliberately designed to sink the heat away from the chip (by using the copper in a "ground plane" as the heatsink), this is why even high power devices can be so remarkably tiny when in surface-mount packages.

The aim is to keep the silicon below its maximum permitted operating temperature under all conditions of peak power and ambient temperature. Even below the maximum specified temperature, expected working life increases rapidly for every degree reduced below that. For the industrial production engineer, there is a trade-off to be struck between product life and the cost of adding more cooling (in terms of both parts cost and assembly cost).

How much is enough? Adding a fan (to provide forced air) is not only expensive, it is also an extra point of potential failure and production rejects. Given that I have three HDR-FOXes operating practically 24/7 for the last ten years, I would hazard a guess that Humax got it right (not necessarily from our point of view, but certainly theirs).

Note that the readings are uncalibrated and for indication only. The DMM/thermocouple recorded an ambient temperature of 20°C before I started, and it won't have changed much. fan* is set to 40%, HDD temperature readings are through sysmon*.

* Custom Firmware packages.

System load: Idling (BBC ONE HD or BBC TWO HD playing live).

Fan in suck mode (normal for V1):
HDD 44°C​
SoC 67°C​

Fan in blow mode (normal for RE):
HDD 37°C​
SoC 67°C​

With lid off:
SoC 57°C​

Adding load of iPlayer playback, or even two HiDef recordings while playing a third, didn't make a noticeable difference to the SoC temperature. Adding HiDef decryption (hardware assisted) to that took it up from 67 to 70°C.

With a CPU fan arranged very similar to this: https://hummy.tv/forum/threads/replacement-t2.10196/post-155156, the thermocouple temperature reading comes down from 67 to 52°C (which went up to 56°C while decrypting), a useful 15-degree reduction in the SoC operating temperature (which should track).
 
Last edited:
As the Humaxs get older there is another possible problem to consider, that the heatsink in question is not hot, but the SoC below it is, there has already been a report of a heatsink falling off completely (only one as fas as I know), but there is always the possibility that all this adding better heatsinks / heatsink fans will amount to nothing if the thermal bond between the SoC and it's heatsink has gone and that this is the actual problem we are facing
 
The fact that the heatsink itself is running 47°C above ambient implies (in the test case) the thermal bond between the heatsink and the chip is relatively low resistance compared with the resistance from heatsink to free air. If the bond were compromised the measured temperature would go down (presuming it hasn't gone down already!).

I propose that measuring the heatsink temperature could be diagnostic of bond failure.

I had a look at the mechanical mounting on a faulty one, but the fixing appeared strong and I didn't (at that occasion) go so far as to force the issue. If a safe means could be devised to remove the heatsink, it should be a simple matter to re-fit it (potentially with better thermal bonding).
 
there has already been a report of a heatsink falling off completely (only one as fas as I know)
I discovered (when one fell off recently) that the heat-sink in the centre of the HD-Fox T2 main board is actually fixed at the four corners with what appeared to be some industrial double-sided tape but might just have been glue baked for nearly a decade. The lack of a heat-sink didn't have any short-term effect, but possibly there are many units operating marginally with poor contact between the chip (with its layer of heat-sink compound) and its heat-sink.
From what I can see, the "glue" is indeed some kind of bonding tape with a specified thickness. So there was actually some thermal grease between the heatsink and the chip?

Any kind of glued fixing is liable to ageing and failure, mechanical fixing is preferred for extreme longevity. Perhaps I should develop something to include in a fan kit.
 
https://hummy.tv/forum/threads/hdr-fox-t2-stability.8950/page-2#post-143204
SOC is held on by double sided heatsink adhesive (less than 0.5mm thickness).
It was easy to slice the tape with the dental tape method without damaging the SOC.
The remaining tape can be peeled off and sticky residue can be removed with isopropyl alcohol.
Note it was stuck on very well on this SOC, so I haven't bothered to remove it on other HDRs I have.
 

Attachments

  • humax-soc-heatsink1.jpg
    humax-soc-heatsink1.jpg
    166.3 KB · Views: 20
  • humax-soc-heatsink2.jpg
    humax-soc-heatsink2.jpg
    121.9 KB · Views: 22
Last edited:
SOC is held on by double sided heatsink adhesive (less than 0.5mm thinkness).
It was easy to slice the tape with the dental tape method without damaging the SOC.
The remaining tape can be removed by peeling and any sticky residue can be removed with isopropyl alcohol.
Note it was stuck on very well on this SOC, so I haven't bothered to remove it on other HDRs I have.
I don't understand that. The ones I have looked at have a thick mounting tape around the edge, specifically the corners, which provide the bulk of the stability. It might have some thermal bond tape on the chip, I won't know until I dare to take one off.

You can just about make it out here, as an off-white layer between the heatsink and the PCB:

AAA4EABA-0EE8-41D6-BC72-5B9B83C176C0.jpeg
 
The fact that the heatsink itself is running 47°C above ambient implies (in the test case) the thermal bond between the heatsink and the chip is relatively low resistance
Agreed, but in the test case you are fixing a SoC heat problem that doesn't exist, no heat reduction is needed in 'good' units, so we need to work on failing units, not easy I admit
 
To be honest, I don't think the SoC heatsink is an issue in any of my units. That's why I haven't bothered to remove the heatsink on them. They are all held firmly on the 3-4 units I have come across. I practiced on a non working unit.
Besides what is the temperature range/spec for the SoC?
As a guess if it is anything like Raspberry Pi SoC, then it may be up to 70c.
I've always concentrated cooling on my drives and waft the overflow air to the SoC heatsink area.
 
... a report of a heatsink falling off completely ...
The problem HD was the one that I carry around in a bag for use in several premises, and for signal testing in lofts. Static, horizontally placed units wouldn't be stressed as much.

I wonder if Wilko imitation Bostik is a good substitute for the original tape ...
 
I wonder if Wilko imitation Bostik is a good substitute for the original tape ...
I would say it probably is yes, I originally thought that the tape / glue whatever needed to be thermally conductive in order to transmit heat from die to heatsink, however it looks like thermal paste between the die and centre of the heatsink is transferring the heat and the 'glue' just holds the heatsink in place
 
I don't understand that.
Simple enough - heat spreader stuck to the BGA package substrate with double-sided tape, possibly a die-cut shape with a hole in the middle for the die, and some kind of thermal transfer stuff on the die.
[Edit] That's what I thought BT's post indicated. But having reviewed it I'm now not at all sure as we don't have a shot of the underside of the removed heat spreader before cleaning.
 
Last edited:
It seems like I will have to sacrifice one to the cause.
I hadn't spotted BT's post of last year. The reason I'd not tried to shift the heatsink/spreader is sometimes they're stuck to the die with an adhesive that sets rock hard and even with decent underfill between die and substrate I could see the HS/S adhesive winning.
 
Pictures after removing the very sticky double sided heatsink adhesive.
The dental floss method works well for initial separation of the adhesive tape. Then it's gentle srubbing using fingernails or plastic spatula type tool.
As the adhesive was still very strong, I doubt it has suffered overheating.
Overheating will usually make the adhesive tape brittle - which this wasn't at all.
The centre die is raised by approximately 1mm.
Looking at the residue on underside of the heatsink spreader, it looks like there were 2 adhesives.
One (1 cm sq) for the central die area and another (3cm sq minus the central 1cm sq) for the remaining BGA area.

Page 1-172 of this preliminary document seems to suggest operating temp of up to 70C (with max a little higher), so I'm not too worried about it myself.
https://www.manualslib.com/manual/1589205/Broadcom-Bcm7405.html

For anyone worried about using the dental flossing method - try it on something else first.
Eg anything that as has been mounted with strong adhesive tape.
 

Attachments

  • humax-soc1.jpg
    humax-soc1.jpg
    181.8 KB · Views: 17
  • humax-soc2.jpg
    humax-soc2.jpg
    150.1 KB · Views: 18
  • humax-soc3.jpg
    humax-soc3.jpg
    153.3 KB · Views: 16
Last edited:
Super! That explains a lot. So what I thought was a thick adhesive layer holding the spreader to the PCB is actually the BGA substrate.
 
Page 1-172 of this preliminary document seems to suggest operating temp of up to 70C
The Spec. also states it's maximum non operating temperature in 125 Deg C, suggesting that it will still function at 70 Deg. C after being at this higher temperature
 
Last edited:
The Spec. also states it's maximum non operating temperature in 125 Deg C, suggesting that it will still function at 70 Deg. C after being at this higher temperature
Non-operating is without voltage applied, and is specified for manufacturing processes. There are no guarantees what will happen if you run it at over 70°C (according to that quote), but we know it must be running higher than that for the spreader to reach that temperature.

IIRC commercial grade logic devices are 0-70 and military -50 to +125.
 
Back
Top