English Amiga Board


Go Back   English Amiga Board > Support > support.Hardware > Hardware mods

 
 
Thread Tools
Old 08 November 2018, 20:11   #1
bitter
Registered User

 
Join Date: Oct 2018
Location: Fort Mill, SC, US
Posts: 47
Heatsink height for Alice in A1200

Does anyone know the maximum height of a heatsink for Alice in A1200 that will clear the keyboard? I'm thinking 6mm will be too high, but don't have any to try...
bitter is offline  
Old 08 November 2018, 21:41   #2
Mr.Flibble
Registered User

Mr.Flibble's Avatar
 
Join Date: Jun 2015
Location: UK
Posts: 454
6mm might be okay, just checking now
Mr.Flibble is offline  
Old 08 November 2018, 21:45   #3
Mr.Flibble
Registered User

Mr.Flibble's Avatar
 
Join Date: Jun 2015
Location: UK
Posts: 454
5mm to be safe from the looks of it
Mr.Flibble is offline  
Old 08 November 2018, 23:44   #4
bitter
Registered User

 
Join Date: Oct 2018
Location: Fort Mill, SC, US
Posts: 47
Thanks for checking!

Well, crap. I guess I need to find 4mm heatsinks, or shave down a 6mm one. Since I have the Indivision AGA installed, I don't have the HD bracket that supports the keyboard, and I'm afraid the back has a little more give. I don't want to glue the heatsink, but with paste it might get pushed around...
bitter is offline  
Old 09 November 2018, 11:58   #5
Mr.Flibble
Registered User

Mr.Flibble's Avatar
 
Join Date: Jun 2015
Location: UK
Posts: 454
6mm with just the first couple of mm sanded down a little should be okay, it’s right at the very edge of the chip where there is the least room
Mr.Flibble is offline  
Old 09 November 2018, 14:20   #6
Mrz
Registered User

 
Join Date: Dec 2016
Location: I
Posts: 267
the alice chip in the A1200 does not need any heatsink
adding one there is stupid/retarded
Mrz is offline  
Old 09 November 2018, 15:19   #7
bitter
Registered User

 
Join Date: Oct 2018
Location: Fort Mill, SC, US
Posts: 47
Quote:
Originally Posted by Mrz View Post
the alice chip in the A1200 does not need any heatsink
adding one there is stupid/retarded
Thank you for the constructive, well thought-out argument.

bitter is offline  
Old 12 November 2018, 12:17   #8
SMF
Registered User

 
Join Date: Dec 2013
Location: Sweden
Age: 36
Posts: 14
Send a message via MSN to SMF
Quote:
Originally Posted by Mrz View Post
the alice chip in the A1200 does not need any heatsink
adding one there is stupid/retarded

Thank you.
Please tell it to my alice, she doesn't read this forum and for some reason she gives graphical glitches when running high resolutions without a cooler there.
Would be great to be able to remove it!
SMF is offline  
Old 16 November 2018, 00:30   #9
Krashan
Hardware Designer

 
Join Date: Aug 2018
Location: Bialystok/Poland
Age: 45
Posts: 22
Heatsink is in fact not needed. I'm also an owner of Indivision AGA and had the same problem. I've described working fix here.
Krashan is offline  
Old 16 November 2018, 03:12   #10
bitter
Registered User

 
Join Date: Oct 2018
Location: Fort Mill, SC, US
Posts: 47
Quote:
Originally Posted by Krashan View Post
Heatsink is in fact not needed. I'm also an owner of Indivision AGA and had the same problem. I've described working fix here.
While I appreciate the fact that the fix for trashing is in fact a hardware issue (Jens just answered my question on his support forum, and I'll be doing his fix), that's not the point. The point is that long-term, heat cycles are the enemy of any IC. Lowering the heat helps with it. Not for tomorrow, maybe not even for the next 5 years, but long-term. These are old computers, at the very least, what harm can little heatsinking do?
bitter is offline  
Old 16 November 2018, 09:22   #11
Jope
-
Jope's Avatar
 
Join Date: Jul 2003
Location: Helsinki / Finland
Age: 38
Posts: 6,945
The mere act of heatsinking anything does no harm to anything than perhaps your wallet, if you go overboard with them.

However experience has shown, that the internet tends to turn many harmless acts into mandatory things that everyone must blindly perform no matter what. First we had caps, which must ALWAYS be replaced, then the heatsink craze started to spread all across the C64 scene, now spilling over to the Amiga. We've gone quite a long time without heatsinking these chips that run warm to the touch. They don't really _need_ heatsinks, if they operate normally without one.
Jope is offline  
Old 17 November 2018, 02:28   #12
Stedy
Registered User

Stedy's Avatar
 
Join Date: Jan 2008
Location: United Kingdom
Age: 41
Posts: 625
Quote:
Originally Posted by Jope View Post
The mere act of heatsinking anything does no harm to anything than perhaps your wallet, if you go overboard with them.

However experience has shown, that the internet tends to turn many harmless acts into mandatory things that everyone must blindly perform no matter what. First we had caps, which must ALWAYS be replaced, then the heatsink craze started to spread all across the C64 scene, now spilling over to the Amiga. We've gone quite a long time without heatsinking these chips that run warm to the touch. They don't really _need_ heatsinks, if they operate normally without one.
The C64 custom chips made by MOS technologies are susceptible to Thermal runaway, something that afflict any device made using NMOS/PMOS fabrication technology. MOS semiconductors (as a technology, not the company) and some old school BJT transistors, can self destruct. When they get hot, they start conducting more electricity, which causes them to get hotter, they then conduct more electricity, this cycle keeps going until the device fails, sometimes spectacularly. The Intel 8080 was known to do this back in the day. The original 6502 was MOS but the later 65C02 was CMOS, so not susceptible to this self destruction and used less power.

In the early to mid 1980s, CMOS technology came along and the Amiga custom chips use this technology (as do the vast majority of chips today).

Microchip reliability as well as that of electronic components, doubles with every 10C temperature drop, this is part of the Arrhenius equation . For a device running at 125C, the lifetime for an automotive part is 2000 hours use. Assuming the amiga internals are at 55C or 70C less, such a device would have a lifetime of >500,000 hours continuous use or 29 years. There are many other parts more likely to fail in that time than the CMOS silicon devices.

Solder joint fatigue will play a part with thermal cycling but for a 20-30C variation, the impact will be low. I'm working on a project that operates from -40C to 100C and needs a lifetime of 30 years. The solder joints are more of an issue than the microchips which have no heatsinks.

Krashan has posted a link to an interesting article on the Alice chipset, I would recommend reading that if you have issues and inspecting the solder joints as well. If you suspect a heat problem with an Amiga, get a can or airduster or freezer spray. When the problem starts, hit the chips with the spray, see if the problem goes away. Heat causes solder joints to expand and cold causes them to contract.
Stedy is offline  
Old 17 November 2018, 08:52   #13
Jope
-
Jope's Avatar
 
Join Date: Jul 2003
Location: Helsinki / Finland
Age: 38
Posts: 6,945
Quote:
Originally Posted by Stedy View Post
The C64 custom chips made by MOS technologies are susceptible to Thermal runaway, something that afflict any device made using NMOS/PMOS fabrication technology. MOS semiconductors (as a technology, not the company) and some old school BJT transistors, can self destruct. When they get hot, they start conducting more electricity, which causes them to get hotter, they then conduct more electricity, this cycle keeps going until the device fails, sometimes spectacularly. The Intel 8080 was known to do this back in the day. The original 6502 was MOS but the later 65C02 was CMOS, so not susceptible to this self destruction and used less power.
Now this is interesting. Thanks for the information, up to now the reasoning behind this in the C64 is "because everyone knows they run hot". My experience has been that they do not run particularly hot, so that has not been a very compelling reason to invest time and money into this exercise.

I have noticed that the early VIC-II chips were heatsinked from the factory, and will fail if the heatsink is left off, but later on they stopped installing the heat sinks, and the VIC-IIs that left the factory without don't seem to be very failure prone for me.

Based on the PLA Dissected article: (page 26 of the pdf)
Quote:
CSG evolved several generations of NMOS manufacturing. Before about 1983, the chips were numbered with a 6xxx scheme. Their manufacturing process was commonly referred to as NMOS and had a critical dimension (CD) of about 28um for the 6502, over the years they seem to have been reduced, The VIC-II (MOS 6567) was implemented with 6um channel length.

The 7xxx chips appeared in about 1983. This process was called HMOS1. Only one year later, the next generation called HMOS2 was in use already, which was numbered 8xxx. The design rules had a CD of 5um. The numbering scheme was not used consistently in all cases, for example the CIA was still called 6526(A) when it was manufactured using the HMOS2 process.
And this Wikipedia article, The_HMOS_processes (sorry), I am left under the impression that I wouldn't have to worry too much about this thermal runaway with the vast majority of MOS Tech/CSG produced chips in the models/ages of Commodore computers I personally care about.

What is your opinion on this information? Seems the venerable 68k along with many of intel's 8086-80286 CPUs were HMOS and it wasn't common knowledge (at least to me until know) that they would suffer from such thermal runaways. However is CSG's "HMOS" the same as Intel's "HMOS", I do not know.

Oh yes, and it is true that the 7xxx HMOS1 process seemed to be suspect. Many C264 folks have had their 7xxx CPUs and TEDs fail due to overheating, but the 8xxx parts have fared better.

Last edited by Jope; 17 November 2018 at 09:03.
Jope is offline  
Old 17 November 2018, 16:13   #14
bitter
Registered User

 
Join Date: Oct 2018
Location: Fort Mill, SC, US
Posts: 47
This is very interesting reading, good to see some facts behind internet folklore. I can definitely speak to NMOS C64s eating VIC-IIs and PLAs from heat, and to lesser extend SIDs. I've gone through a lot of them until I started putting good heatsinks on them. 64Cs I never bothered, their HMOS chips run much cooler, and never had an issue. 264-series are probably the worst offenders though...

After doing a lot of reading, including this threat, my mind has definitely been put at ease about the A1200. While I ended up putting small heatsinks on Alice and 68030 on my acc card, once I make the hardware fix to Lisa, I'm done worrying about it. Thanks for a good discussion!
bitter is offline  
Old 16 December 2018, 02:08   #15
Stedy
Registered User

Stedy's Avatar
 
Join Date: Jan 2008
Location: United Kingdom
Age: 41
Posts: 625
Quote:
Originally Posted by Jope

Based on the PLA Dissected article: (page 26 of the pdf)
Quote:
CSG evolved several generations of NMOS manufacturing. Before about 1983, the chips were numbered with a 6xxx scheme. Their manufacturing process was commonly referred to as NMOS and had a critical dimension (CD) of about 28um for the 6502, over the years they seem to have been reduced, The VIC-II (MOS 6567) was implemented with 6um channel length.

The 7xxx chips appeared in about 1983. This process was called HMOS1. Only one year later, the next generation called HMOS2 was in use already, which was numbered 8xxx. The design rules had a CD of 5um. The numbering scheme was not used consistently in all cases, for example the CIA was still called 6526(A) when it was manufactured using the HMOS2 process.
And this Wikipedia article, The_HMOS_processes (sorry), I am left under the impression that I wouldn't have to worry too much about this thermal runaway with the vast majority of MOS Tech/CSG produced chips in the models/ages of Commodore computers I personally care about.

What is your opinion on this information? Seems the venerable 68k along with many of intel's 8086-80286 CPUs were HMOS and it wasn't common knowledge (at least to me until know) that they would suffer from such thermal runaways. However is CSG's "HMOS" the same as Intel's "HMOS", I do not know.

Oh yes, and it is true that the 7xxx HMOS1 process seemed to be suspect. Many C264 folks have had their 7xxx CPUs and TEDs fail due to overheating, but the 8xxx parts have fared better.
Sorry it has taken so long to respond, have had to do some research to fully answer the questions. It's not just down to the Fabrication technology, the packaging also has an influence. The 'speed governors' or thermal shutdown functions added to later devices definitely make more modern devices more robust, we have to remember some of these older devices do not have these safety features.

I started by gathering information on chips from the C64, from 6502.org, a few Amiga chipset documents (LISA and buster chips) and the Motorola 68000 series processors. The first bit of information I needed was the maximum power figures, the list below details the results:
Code:
6502     @ 2 MHZ   800mW    (HMOS technology)
65C02    @ 2 MHz   35mW     (CMOS technology)
6510     @ 2 MHz   650mW    (HMOS technology)
R6502    @ 2 MHz   700mW    (NMOS technology)
W65C02S  @ 2 MHz   15mW     (Modern CMOS)
6560 (VIC)         750mW    (NMOS technology)
6567 (VIC II)      1600mW   (NMOS?)
6581 (SID)         1000mW   (NMOS technology)
391227-01 (LISA)   1500mW   (CMOS technology)
MC68000  @ 8 MHz   1300mW   (HMOS technology)
MC68020  @ 16 MHz  900mW    (HCMOS technology)
MC68030  @ 25 MHz  2500mW   (HCMOS technology)
MC68040  @ 25 MHz  5000mW   (HCMOS technology)
MC68060  @ 50 MHz  3900mW
Looking at the list above, a couple of things stand out.
1) Moving from HMOS to CMOS dramatically reduced the power, compare the 6502 and 65C02.
2) The VIC II uses a lot of power for something in a 40 pin DIP package
3) Surprisingly the 68060 uses less power than the 68040, more advanced process technology helps.

I have the power figures, next we need to know how hot they can get and what is the acceptable limit. Silicon semiconductors will operate upto a junction temperature of 125C. A safe de-rated limit is 100C, this gives a good balance for a long life.

There is a simple formula, not easy to post in a forum, that allows you to calculate the junction temperature:
Tj = (R(theta)Ja x Pdiss) +Ta

Where
TJ = Junction temperature
R(theta)Ja = Junction to ambient thermal resistance
Pdiss = Power dissipation, examples in the previous table
Ta = ambient temperature

Finding R(theta)Ja for 40 pin DIP and 84 pin PLCC packages was the tricky part. Modern datasheets always have thermal data included, the MOS technology and Amiga ones were lacking information. A bit of research found some typical package thermal resistances, from Texas Instruments (http://www.ti.com/lit/an/snva509a/snva509a.pdf), Microsemi ()

So for the 40 pin DIP devices, a RJa (got bored of typing Theta), it is 40 C/W.
For the 28 pin DIP devices Rja is 50 C/W.
For the 84 pin PLCC it is 22C/W.

A simple explanation, if a device has a Tja value of 22C/W, for every 1 watt of power, the junction temperature will rise by 22C. Clearly the lower the Tja value, the cooler the device runs and more reliable it is. Adding a heat-sink, to the top of the package, lowers the Tja number. Likewise, applying forced air cooling (typically 200 feet/minute) also lowers the Tja value.

For the MC68000 series devices, the thermal data is specified in the datasheets:
Code:
MC68000 DIP 30 C/W   (Used in A500)
MC68020 PGA  32 C/W  (Used some accelerator cards)
MC68020 PQFP 42 C/W  (Used on A1200)
MC68030 PGA  15 C/W (Tjc)
MC68040 PGA  3 C/W (Tjc)
MC68060 PGA  2.5 C/W (Tjc)
For the MC68040 and MC68060, they specify the Rjc (Junction to case) thermal resistance, as you must heat-sink these parts. Using an application note from Micron (https://www.micron.com/-/media/clien...ermal_apps.pdf) and using the formula,
Tj = Tc +(Pdiss x Pc x Tjc)
Where
Tc 50C (our ambient)
Pc = 50% through case, 50% through board.
Tjc = value stated above.

We are nearly ready to start calculating that all important junction temperature. We need an ambient temperature. In a modern PC, with many fans, it may be 25-35C at 25C ambient. For these retro computers, that rely on convection to cool, I would use a slightly pessimistic 50C ambient.

So based on the number above and assume everyone reading this is still awake, the potential temperatures are:
Code:
Device     Speed   Power     Junction
6502     @ 2 MHZ   800mW     82.0C
65C02    @ 2 MHz   35mW      51.4C
6510     @ 2 MHz   650mW     76.0C
R6502    @ 2 MHz   700mW     78.0C
W65C02S  @ 2 MHz   15mW      50.6C
6560 (VIC)         750mW     80.0C
6567 (VIC II)      1600mW    114.0C
6581 (SID)         1000mW    100.0C
391227-01 (LISA)   1500mW    83.0C
MC68000 (DIP)      1300mW    89.0C
MC68020 (PQFP)     900mW     87.8C
MC68020 (PGA)      900mW     78.8C
MC68030 (PGA)      2500mW    68.8C
MC68040  @ 25 MHz  5000mW    57.5C
MC68060  @ 50 MHz  3900mW    54.9C
In this list, 2 chips exceed my de-rated 100C limit.
The VIC-II and SID are widely reported (in my searches) as needing heatsinks and some C64s had them fitted at the factory. If fitting a heatsink reduces the themal resistance by 30%, it will make a dramatic improvement and would reduce the junction temperatures to 79.8C for the VIC II and 70C for the SID device.

I should quickly, touch on, hot surfaces. Sometimes when testing, you mistakenly touch a chip and it feels very hot, I should know, I do this a lot. How hot is it though?
There is a touch safe limit for electronics, it is 60C, this is a temperature, that when exceeded, is too hot to hold for more than a few seconds and will leave red marks due to light burning of the skin. In severe case, if a part is >100C, it really, really hurts! Burns from hot microchips, will normally heal themselves in a week or two. The infrared thermometer guns are good to get an idea of the case temperature. You then have to use some more maths to calculate the junction temperature, but you also need to take into account the colour of the device for the emissivity of the package. It's a bit fiddly.

The information I've posted here, use a simple model to estimate the temperature of common Commodore microchips. Other factors can influence the temperature rise and the cooling effects. I've not touched on the effects of hot devices, heating up the board and causing the ambient temperature to rise in localised areas. I've assumed little to no airflow, many people add fans* to cool the system and the PCBs do provide some heat-sink ability, if they have power planes.

*Where I work a Fan is the 'F-word' as they are unreliable, prone to mechanical failures and generally have a lifetime less than the equipment they are there to protect.
Stedy is offline  
Old 16 December 2018, 14:45   #16
Jope
-
Jope's Avatar
 
Join Date: Jul 2003
Location: Helsinki / Finland
Age: 38
Posts: 6,945
Thank you for a very comprehensive response! A lot of data to digest, but I'll re-read it until I understand. :-)
Jope is offline  
 


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools

Similar Threads
Thread Thread Starter Forum Replies Last Post
Reliable source for A1200 Alice chip mpattonm support.Hardware 1 30 January 2018 14:13
D-Box A1200 not enough height? Bamiga2002 support.Hardware 20 07 January 2012 17:29
A1200 crashes (overheated Alice?) Nikolaj_sofus support.Hardware 8 19 November 2009 09:29
A1200 rev.2B motherboard: resistor in ALICE pin keropi Hardware pics 11 06 March 2008 01:38
Wanted: Paula and Alice PLCC IC's for A1200 ASAP chaoticjelly MarketPlace 4 05 March 2008 01:15

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +2. The time now is 08:46.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2019, vBulletin Solutions Inc.
Page generated in 1.03979 seconds with 15 queries