Q: TO-5/TO-39 Heat Sinks -- How tight a fit is needed

[Norm Dresner]

For a special project, I'd like to use paralleled TO-5/TO-39 transistors in
the output of a moderately low power amplifier -- say 15W -- much like is
done with TO-3 and TO-220 transistors for high-power amplifiers. The
intent is to get a higher slew rate and wider frequency range with the
faster transistors available in the smaller size cans. There are probably
a few TO-202 types I could use but the TO-5/TO-39 are more common and have
higher Ft and gain for less price, though in a 1-4 unit production for
internal use that's hardly a consideration, but I have to keep it reasonable
(i.e. under the radar).

In order to insure the best thermal tracking of the output, I feel I need a
common heatsink for each polarity of the output drivers separately (using
complementary feedback output stage topology). But to do this, the only way
I can think of is to take a massive piece of aluminum and drill holes for
each transistor case to fit in. Drawings I've found give the diameters as
TO-5 .320" +-.015"
TO-39 .325" +-.010"

If I drill holes that are .328" (21/64 IIRC), I'd pretty much be at the
nominal diameters for both sizes but the TO-5's could be as much as .023"
smaller than the holes. My intuition says that's too much for a good fit
for efficient heat transfer but I have absolutely no experience in this
regard and I'm hoping that someone else would know. I also have no feeling
for how hard or easy it would be to force oversized cases into unyeilding
holes.

Information and alternate suggestions appreciated.

TIA
Norm

 

[Tim Wescott]

For a special project, I'd like to use paralleled TO-5/TO-39 transistors in
the output of a moderately low power amplifier -- say 15W -- much like is
done with TO-3 and TO-220 transistors for high-power amplifiers. The
intent is to get a higher slew rate and wider frequency range with the
faster transistors available in the smaller size cans. There are probably
a few TO-202 types I could use but the TO-5/TO-39 are more common and have
higher Ft and gain for less price, though in a 1-4 unit production for
internal use that's hardly a consideration, but I have to keep it reasonable
(i.e. under the radar).

In order to insure the best thermal tracking of the output, I feel I need a
common heatsink for each polarity of the output drivers separately (using
complementary feedback output stage topology). But to do this, the only way
I can think of is to take a massive piece of aluminum and drill holes for
each transistor case to fit in. Drawings I've found give the diameters as
TO-5 .320" +-.015"
TO-39 .325" +-.010"

If I drill holes that are .328" (21/64 IIRC), I'd pretty much be at the
nominal diameters for both sizes but the TO-5's could be as much as .023"
smaller than the holes. My intuition says that's too much for a good fit
for efficient heat transfer but I have absolutely no experience in this
regard and I'm hoping that someone else would know. I also have no feeling
for how hard or easy it would be to force oversized cases into unyeilding
holes.

Information and alternate suggestions appreciated.

TIA
Norm

I would chuck this up in my lathe and bore the holes until they're just
right; I assume you don't have access to the proper equipment, however.

Suggestion #1: Use surface-mount parts with thick copper, heatsink the
board. This probably blows your budget, though.

Suggestion #2: Use two plates; a thick one and a thin one, and clamp
the transistor flange between them. If you can, get fancy and
counterbore the big part ever so slightly for the flange (and file out a
notch for the tab, etc., etc.). The usual construction of those dumps
the heat into the flange first, then it migrates up the wall of the case.

Suggestion #3: Make your heatsink, drill the holes in a line, then
split the plate lengthwise through the holes. Now bolt the plate back
together with the transistors in between. Go tight enough and you'll
have the only oval-cased TO-5 cans on the block!

 

[Norm Dresner]

I would chuck this up in my lathe and bore the holes until they're just
right; I assume you don't have access to the proper equipment, however.

Suggestion #1: Use surface-mount parts with thick copper, heatsink the
board. This probably blows your budget, though.

Suggestion #2: Use two plates; a thick one and a thin one, and clamp
the transistor flange between them. If you can, get fancy and
counterbore the big part ever so slightly for the flange (and file out a
notch for the tab, etc., etc.). The usual construction of those dumps
the heat into the flange first, then it migrates up the wall of the case.

Suggestion #3: Make your heatsink, drill the holes in a line, then
split the plate lengthwise through the holes. Now bolt the plate back
together with the transistors in between. Go tight enough and you'll
have the only oval-cased TO-5 cans on the block!

Suggestion #3 parallels the shape of common TO-5 heat sinks that are
incomplete circles with just enough spring in the material to stretch over
the cans. I like it -- assuming I can split the plate with a thin-enough
saw to be able to maintain "contact" between the two halves when
reassembled. I think my band saw would probably work, though my radial arm
saw is definitely out. BUT ... do I really need contact between the two
halves? Wouldn't it be sufficient to have two heatsinks made of the two
halves bolted together? Hmm... probably so. I'll need to think about
that one.


Thanks
Norm

 

[John Woodgate]

(in <[email protected]>) about
'Q: TO-5/TO-39 Heat Sinks -- How tight a fit is needed', on Mon, 19 Jul
2004:

I also have
no feeling for how hard or easy it would be to force oversized cases
into unyeilding holes.

Don't. For a certain, but minimal, loss of heat-sink efficiency, use a
slab (can be finned) of heat-sink material (1/4-hard aluminium is good)
about the same thickness as the transistor case. Cut it down the middle
and have the cut edges machined flat.

Near the cut edge of one half, bore holes slightly undersize and have a
thin slot machined from the hole to the cut edge. That provides a
minimum amount of resilience, which you must have in order to allow for
thermal expansion and contraction.

Insert your transistors, put heat-transfer compound on the cut edges -
NOT TOO MUCH! - and clamp the two halves together.

 

[John Larkin]

For a special project, I'd like to use paralleled TO-5/TO-39 transistors in
the output of a moderately low power amplifier -- say 15W -- much like is
done with TO-3 and TO-220 transistors for high-power amplifiers. The
intent is to get a higher slew rate and wider frequency range with the
faster transistors available in the smaller size cans.

If you parallel a bunch of small transistors, the device and package
capacitances add, and the interconnects add inductance, so you
generally won't come out ahead. And TO-5 style cans pretty much imply
older-technology, slow parts these days. TO-5's are thermally awful,
anyhow; the part you really want to heatsink is right in the middle of
the leads!

What sorts of speeds/slew rates/power levels are you shooting for?

John

 

[mike]

For a special project, I'd like to use paralleled TO-5/TO-39 transistors in
the output of a moderately low power amplifier -- say 15W -- much like is
done with TO-3 and TO-220 transistors for high-power amplifiers. The
intent is to get a higher slew rate and wider frequency range with the
faster transistors available in the smaller size cans. There are probably
a few TO-202 types I could use but the TO-5/TO-39 are more common and have
higher Ft and gain for less price, though in a 1-4 unit production for
internal use that's hardly a consideration, but I have to keep it reasonable
(i.e. under the radar).

In order to insure the best thermal tracking of the output, I feel I need a
common heatsink for each polarity of the output drivers separately (using
complementary feedback output stage topology). But to do this, the only way
I can think of is to take a massive piece of aluminum and drill holes for
each transistor case to fit in. Drawings I've found give the diameters as
TO-5 .320" +-.015"
TO-39 .325" +-.010"

If I drill holes that are .328" (21/64 IIRC), I'd pretty much be at the
nominal diameters for both sizes but the TO-5's could be as much as .023"
smaller than the holes. My intuition says that's too much for a good fit
for efficient heat transfer but I have absolutely no experience in this
regard and I'm hoping that someone else would know. I also have no feeling
for how hard or easy it would be to force oversized cases into unyeilding
holes.

Information and alternate suggestions appreciated.

TIA
Norm

Think about it.
Where's the chip? On the bottom of the can.
Where's the heat sink? On the top of the can.
How are you gonna hook it all up? Lots of stray L and C.
Is this a good idea? No!

You don't say how wide "wider frequency range" is, but you can get
RF transistors with WIDE bandwidth and good heat sink characteristics.
mike

--
Return address is VALID.
Bunch of stuff For Sale and Wanted at the link below.
Compaq Aero floppy,ram,battery.
FT-212RH 2-meter 45W transceiver.
Toshiba & Compaq LiIon Batteries, Test Equipment
30pS pulser, Tektronix Concept Books, spot welding head...
http://www.geocities.com/SiliconValley/Monitor/4710/

 

[legg]

For a special project, I'd like to use paralleled TO-5/TO-39 transistors in
the output of a moderately low power amplifier -- say 15W -- much like is
done with TO-3 and TO-220 transistors for high-power amplifiers. The
intent is to get a higher slew rate and wider frequency range with the
faster transistors available in the smaller size cans. There are probably
a few TO-202 types I could use but the TO-5/TO-39 are more common and have
higher Ft and gain for less price, though in a 1-4 unit production for
internal use that's hardly a consideration, but I have to keep it reasonable
(i.e. under the radar).

In order to insure the best thermal tracking of the output, I feel I need a
common heatsink for each polarity of the output drivers separately (using
complementary feedback output stage topology). But to do this, the only way
I can think of is to take a massive piece of aluminum and drill holes for
each transistor case to fit in. Drawings I've found give the diameters as
TO-5 .320" +-.015"
TO-39 .325" +-.010"

Heatsink coupling TO-5

Wakefield
260-4T5E
200 SERIES

thermalloy
321128a00000
VIS - 034454
VIS - 034333

You'd be better off using more recent package types, even if
hermeticity is the aim.

evetr trtr]ied tryu[p]ing W]ITRH A wetr keyub atrd?

RL

 

[Norm Dresner]

If you parallel a bunch of small transistors, the device and package
capacitances add, and the interconnects add inductance, so you
generally won't come out ahead. And TO-5 style cans pretty much imply
older-technology, slow parts these days. TO-5's are thermally awful,
anyhow; the part you really want to heatsink is right in the middle of
the leads!

What sorts of speeds/slew rates/power levels are you shooting for?

It looked to me like the device capacitance for a typical Ft=100-200MHz TO-5
was on the order of 7-10 and for a single TO-3 I saw numbers in the 500+
range. Based on that I didn't think that paralleling, say 5-10 smaller
devices was that bad a tradeoff.

As I said earlier, about 15W. BW should run up to a few MHz, perhaps 2-4.
Slew rate adequate for these parameters.

And I was hoping to make it out of stock on hand rather than purchasing
expensive new parts.

Norm

 

[colin]

If you parallel a bunch of small transistors, the device and package
capacitances add, and the interconnects add inductance, so you
generally won't come out ahead. And TO-5 style cans pretty much imply
older-technology, slow parts these days. TO-5's are thermally awful,
anyhow; the part you really want to heatsink is right in the middle of
the leads!

What sorts of speeds/slew rates/power levels are you shooting for?

John

ive tried to make heatsinks for T05s by drilling holes in aluminium block
and then found the case size isnt realy that acurate, i would get the hole
just right for one transistor then found some other make (or older
transistor) wld be too tight or too sloppy, also like someone said you need
some resiliance, i found an aluminium strip bent round the transistor and
bolted to an aluminium bar worked pretty well.

there is a technique for paraleling transistors and avoiding adding the
capacitances together, ive seen it used in valve circuits particularly my
tek 585 scope, asuming you have a few hundred ohms source impedance you make
a transmision line out of inductors and capacitors but for the capacitors
you use the transistors base input capacitance, you terminate it at the last
transistor of course. for the output u must do the same thing, ie make a
transmision line starting with a termination at the first transsistor and
use the colector capacitance for the capacitors of course you need to match
your load to the impedance of the trans line.

however with the available transistors today its unlikly to be necesary to
go to this extent, as john said tell us what you are trying to acheive.
only probelm i find is that hi ft transistors tend to have low vce. ive used
many bfg591s in op stages of eqpmnt. its quite good how much power the SM
package can handle compared to T05.

Colin. =^.^=

 

[John Larkin]

It looked to me like the device capacitance for a typical Ft=100-200MHz TO-5
was on the order of 7-10 and for a single TO-3 I saw numbers in the 500+
range. Based on that I didn't think that paralleling, say 5-10 smaller
devices was that bad a tradeoff.

As I said earlier, about 15W. BW should run up to a few MHz, perhaps 2-4.
Slew rate adequate for these parameters.

And I was hoping to make it out of stock on hand rather than purchasing
expensive new parts.

Norm

Sounds like you could do that with a pair of $0.75 TO-220 bipolars or
mosfets. Again, TO-5/3 packaged parts are likely to be klunky old
stuff. 4 MHz isn't really very fast these days.

John

 

[Norm Dresner]

Sounds like you could do that with a pair of $0.75 TO-220 bipolars or
mosfets. Again, TO-5/3 packaged parts are likely to be klunky old
stuff. 4 MHz isn't really very fast these days.

John

I know I didn't say it before but we need very low distortion as well. Am I
really going to get that kind of performance out of a 15-30MHz TO-220? Or
are there parts I don't even suspect exist that I should know about?

Thanks
Norm

 

[Tim Wescott]

I know I didn't say it before but we need very low distortion as well. Am I
really going to get that kind of performance out of a 15-30MHz TO-220? Or
are there parts I don't even suspect exist that I should know about?

Thanks
Norm

The package doesn't make the distortion. If you use good circuit design
practices for the part at hand then yes, you should be OK.

15W, fast and low distortion sounds like quite a challenge -- are you
allowed to run Niagara Falls through the heat sink?

 

[Ken Smith]

[...]
I'll add:

Suggestion #1: Use surface-mount parts with thick copper, heatsink the
board. This probably blows your budget, though.

If the circuit is not too complex, consider using a 0.015" PCB with top
side traces only and putting it down on a heatsink with thermal compond.

Copper sheet for flashing or roofing can be found in some hardware stores.

I've used plumbing parts as heat sinks on home projects.

Suggestion #2: Use two plates; a thick one and a thin one, and clamp
the transistor flange between them. If you can, get fancy and
counterbore the big part ever so slightly for the flange (and file out a
notch for the tab, etc., etc.). The usual construction of those dumps
the heat into the flange first, then it migrates up the wall of the case.

Also, you can gain some heat sinking by how you screw the parts
together. I often use brass threaded standoffs in place of
nuts. Depending on how long the standoff is, it can add a fair amount of
surface area.

 

[Norm Dresner]

The package doesn't make the distortion. If you use good circuit design
practices for the part at hand then yes, you should be OK.

15W, fast and low distortion sounds like quite a challenge -- are you
allowed to run Niagara Falls through the heat sink?

No, but we can put it in a wind tunnel. Force air cooling is perfectly
acceptable. This isn't high quality audio, it's part of an industrial
measurement setup and noise is no object.

Hmm .. I wonder if I could use a circulating oil bath ... ;-)

Norm

 

[Rich Grise]

Suggestion #3 parallels the shape of common TO-5 heat sinks that are
incomplete circles with just enough spring in the material to stretch over
the cans. I like it -- assuming I can split the plate with a
thin-enough saw to be able to maintain "contact" between the two halves
when
reassembled. I think my band saw would probably work, though my radial
arm
saw is definitely out. BUT ... do I really need contact between the two
halves?

Not at all. Just from the can to whatever aluminum is wrapped around it.
A gap or two won't hurt a bit.

Wouldn't it be sufficient to have two heatsinks made of the two
halves bolted together? Hmm... probably so. I'll need to think about
that one.
See above.
--

Cheers!
Rich

 

[colin]

For a special project, I'd like to use paralleled TO-5/TO-39 transistors in
the output of a moderately low power amplifier -- say 15W -- much like is
done with TO-3 and TO-220 transistors for high-power amplifiers. The
intent is to get a higher slew rate and wider frequency range with the
faster transistors available in the smaller size cans. There are probably
a few TO-202 types I could use but the TO-5/TO-39 are more common and have
higher Ft and gain for less price, though in a 1-4 unit production for
internal use that's hardly a consideration, but I have to keep it reasonable
(i.e. under the radar).

In order to insure the best thermal tracking of the output, I feel I need a
common heatsink for each polarity of the output drivers separately (using
complementary feedback output stage topology). But to do this, the only way
I can think of is to take a massive piece of aluminum and drill holes for
each transistor case to fit in. Drawings I've found give the diameters as
TO-5 .320" +-.015"
TO-39 .325" +-.010"

If I drill holes that are .328" (21/64 IIRC), I'd pretty much be at the
nominal diameters for both sizes but the TO-5's could be as much as .023"
smaller than the holes. My intuition says that's too much for a good fit
for efficient heat transfer but I have absolutely no experience in this
regard and I'm hoping that someone else would know. I also have no feeling
for how hard or easy it would be to force oversized cases into unyeilding
holes.

Information and alternate suggestions appreciated.

TIA
Norm

im reminded of a philps signal generator wich i looked inside long time ago
and at the circuit and found it had a nice output stage just consisting of
several TO5s think they were something like 2n2219/2n2905. might be a good
place to get design idea.

Colin =^.^=

 

[John Larkin]

I know I didn't say it before but we need very low distortion as well. Am I
really going to get that kind of performance out of a 15-30MHz TO-220? Or
are there parts I don't even suspect exist that I should know about?

Thanks
Norm

The nice thing about fets is that they don't have an Ft, so that's one
less thing to worry about. At your speeds, the only significant
parasitic is capacitance.

John

 

[Norm Dresner]

im reminded of a philps signal generator wich i looked inside long time ago
and at the circuit and found it had a nice output stage just consisting of
several TO5s think they were something like 2n2219/2n2905. might be a good
place to get design idea.

I've got a whole drawful of 2N2102 NPN and about 1/4 that many 2N5323 PNP
which might not be too bad either. Also 2 dozen 2N3637's for the PNP side
as well.

Thanks
Norm