pasta

[Jan Kuba]

K indiu  a obecně k pájení polovodicu  kvuli odvodu tepla si dovolim
obsahlejsi info típnuté z jiné konference:

Date: Thu, 19 Mar 2020 13:02:48 -0000
From: "John Regnault" <G4SWX>
To: <moon-net na mailman.pe1itr.com>,
Subject: Re: [Moon-Net] MRFX1k8

Hi Neil,

Indium solder is not good for this sort of application.

In a long-gone, previous existence in the early 1980s, I used to be
the research engineer in charge of bonding and testing of
semiconductor lasers with very significant heat transfer and current
density issues. Initially for rapid  DC testing of chips we used to
use Indium soldering using a temperature cycled bonding machine. Our
failure rate was very high, often 30-50% after 24 hours at 60 degrees
C heatsink temperature.

There are a lot of references to the growth of whiskers and why Pure
Indium solder is not a good idea. See a few
references:https://nepp.nasa.gov/whisker/other_whisker/#indium

In the end we moved to using AuSn eutectic solder at a somewhat higher
temperature and plated diamond heat-spreaders!
https://www.indium.com/blog/eutectic-gold-tin-ausn.php

This required a fairly exacting temperature cycle which is not suited
to amateur 'hot-plate' technology.

It took many years work to get the solder and device metallisation
reliability right so that we could obtain >100K hours on an elevated
temperature life test. It took 2 more years to change from diamond
heat spreaders to a CuW alloy.....{we did try BeO along this path!!!!}
 Although the thermal conductivity is a bit higher than pure copper,
CuW alloys are far superior in terms of thermal expansion and
machinability. One of the higher thermal conductivity ones would be
well suited for LDMOS heat spreaders.

For the application of soldering LDMOS to copper heat spreaders one of
the common solder pastes, with a melting point 150-200C is in the
right ball park. Higher temperatures do induce more significant
take-up of the device metallisation into the solder as can also occur
if the solder remains molten for too long on a hot plate. [I must look
out my book with all of the metal eutectic tables} Rather than a hot
plate I prefer to get a friend to use his temperature cycling SMD
re-flow oven to do this task.

Generally the choice of solder does not induce the sort of failures
seen by most amateurs although I have seen tin whisker growth on the
collector pads of my TS2000 after 8 years hammering on EME! In this
case the Pb free solder paste used by the manufacturer was the cause.

In my view Michael's failure that started this thread was precipitated
by very high dissipation, coupled with a high drive level which led to
a chain of failures and catastrophe. Getting the heat away from the
heat spreader is essential which is why water cooling solutions are
probably the best. OZ1CT used to have some good pictures on his site
but I can't find them at the moment. The water cooling blocks from
PE1RKI also look pretty good. If you are going to use air cooling IT
WILL BE NOISY. If it is not noisy you will probably not be passing
enough air and will develop a hot spot under the heat spreader. I have
an Italab commercial SSPA which increases the fan speed as the
heatsink temperature goes up. When running full power it is far louder
than my biggest valve PA ever was. Happily I am not in the same room
when it is running.


73

John G4SWX
=====================================================================================

From: Neil <g4dbn>

To: moon-net na mailman.pe1itr.com <mailto:moon-net na mailman.pe1itr.com>

Subject: Re: [Moon-Net] MRFX1k8 failure--> water cooling?



The documentation says that using compound to fill the void is the
correct approach.? I guess there is a good reason for it, perhaps to
prevent fracture of the substrate.? It is easy enough to calculate the
temperature gradient across a 0.09mm thick layer of thermal transfer
compound. The void is 21.0 x 50.4 mm, so for Arctic Silver, at
8.9W/mK, the heat transfer rate for a 2 degree temperature difference
is about 21 watts, or inversely, a thermal resistance between the case
and heatsink of about 0.1 degree per watt.? Thermal interface sheet
with a conductance around 5W/mK is available, not sure if it is made
in thin enough sheets.? Only risk I can see is possible migration of
the compound over a large number of heat cycles, but this stuff is
horribly sticky, so I guess it would stay put. I think there is a
paper about the specific reasoning behind the relieved back, by my
memory fails me on the exact source of that paper.? I imagine that if
you want a low-cost assembly, any lapping or machining of the contact
area is a waste, so this may be a compromise between reasonably low
thermal resistance and ability to absorb the tolerances of an extruded
or die-cast heatsink.

An extra couple of degrees rise above heatsink temperature at 20 watts
seems reasonable to me for this specific application, especially as
the junction to case thermal resistance is often several degrees per
watt.

Neil G4DBN


středa 29. ledna 2020 Čihák Martin <cihak na dkv.cd.cz> napsal(a):

> Vím ještě o  jedné cestě. Dát pod tranzistor indiovou podložku (fólii).
> Lze ji koupit na e-bay. Osobně jsem to ale nezkoušel – taky tranzistory
> pájím.
>
> Indium je velmi měkké a ta fólie se pod tranzistorem rozmáčkne podobně
> jako vazelína a vyplní drobné nerovnosti povrchu tranzistoru a chladiče.
> Indium je dobře tepelně i elektricky vodivé takže funguje podobně jako cín,
> jen je spoj rozebíratelný.
>
>
>
>                 Martin.
>
>
>
> *From:* Hw-list [mailto:hw-list-bounces na list.hw.cz] *On Behalf Of *Jan
> Kuba
> *Sent:* Tuesday, January 28, 2020 11:40 PM
> *To:* HW-news
> *Subject:* Re: pasta
>
>
>
> U VF výkonových tranzistorů s výkony okolo 1kW je jediná cesta jak to
> teplo odvést tranzistor přiletovat. Pasty  nefungují...( špatně )
>
>
>
>
>
------------- další část ---------------
HTML příloha byla odstraněna...
URL: <http://list.hw.cz/pipermail/hw-list/attachments/20200321/128ca8c3/attachment-0001.html>