Hi Nigel
What leads you to that thought: whilst steel is not as thermally conductive as copper, steel is a know entity - its thermal conductivity is stated and can therefore be accomodated in the heat dissipation calculations. Also, I have to admit that I have never come accross a copper-cased TO-3.
Kind regards
Mike
Last edited:
Mike, After years of not giving it a thought I had to say steel is a weird choice for T03 as copper is cheap and easier to make. If I can solder to the can top ( is it tin or nickel ? ) that's less of a problem. My job involves questions about metals so I feel a bit ashamed I never thought of this.
Anyone like to say why gold is less conductive than copper, it looks that gold should be better? The 32 electrons in one shell looks to be part of it ( band gap ). Mercury that has one extra electron more than gold ( and proton, neutron ) if memory is right is 43 times less conductive than gold. Silver is about 104% OFC and 106% ETPC, that is silver is the best.
Anyone like to say why gold is less conductive than copper, it looks that gold should be better? The 32 electrons in one shell looks to be part of it ( band gap ). Mercury that has one extra electron more than gold ( and proton, neutron ) if memory is right is 43 times less conductive than gold. Silver is about 104% OFC and 106% ETPC, that is silver is the best.
Copper is not cheap. Price comparison is hard because there are so many different types of each, but poking around the web, I see London copper futures are running about 12x more than iron futures. Switching to the scrap metal buyers and I find copper has 40x the scrap value of steel.
Not a big deal for a DIY build, granted, but when you're building transistors on an industrial scale it adds up quick.
The squished figure of 8 slug that sits between and beyond the pins is copper.
This applies to all the Onsemi, Toshiba and Motorola steel cased To3 that I have handled.
I don't know how the aluminium cased versions were assembled.
Last edited:
There is a half way house if you want. That is to use a 3.2 mm ring crimp connector - the red/blue/yellow type typically used on car electrics. These are made of tinned copper.
The way I connect to TO-3 cases is from the top and to use pvc over copper cable. Strip back the pvc, dip the exposed copper cable in liquid solder flux (use a flux pencil here) and then crimp the ring connector on.
Having got a good physical connection, use a soldering iron on the ring side and let solder run into the crimp space. The flux will aid the capillary action and make a good electrical connection.
Coming from the top of the TO-3 case on the heatsink, use a 3mm diameter stainless steel screw through the crimp ring with a shakeproof (crinkle) washer between the crimp and the TO-3 case to bite into both metals. I use a sandwitch of:
Screw head, plain washer, crinkle washer, crimp ring, crinkle washer, through the TO-3 case and mica washer, through the heatsink, through the isolation washer (which also aligns everything up) plain washer and locking nut. - hopefully that is clear enough
The plain washer compresses the crinkle washer to really bite into the case and crimp.
You could file the tinning off the crimp if you want to go a stage further. Copper screws would be too weak to really tighten up and probably would not add anything as you have made the most important joint which is wire-crimp-case.
It does mean that you need an extra hole in the heatsink to take the Collector wire through to meet up with those from the Base and Emitter, but that's not great problem. I use spare TO-3 isolation washer to form a plastic grommet.
Back on the PCB I use a similar system to connect theTO-3 wires to the PCB tracks. Use stainless steel nuts and bolts/scrimps with crinkle washers to bite into the PCB tracks.
Hope this helps
Mike
The way I connect to TO-3 cases is from the top and to use pvc over copper cable. Strip back the pvc, dip the exposed copper cable in liquid solder flux (use a flux pencil here) and then crimp the ring connector on.
Having got a good physical connection, use a soldering iron on the ring side and let solder run into the crimp space. The flux will aid the capillary action and make a good electrical connection.
Coming from the top of the TO-3 case on the heatsink, use a 3mm diameter stainless steel screw through the crimp ring with a shakeproof (crinkle) washer between the crimp and the TO-3 case to bite into both metals. I use a sandwitch of:
Screw head, plain washer, crinkle washer, crimp ring, crinkle washer, through the TO-3 case and mica washer, through the heatsink, through the isolation washer (which also aligns everything up) plain washer and locking nut. - hopefully that is clear enough
The plain washer compresses the crinkle washer to really bite into the case and crimp.
You could file the tinning off the crimp if you want to go a stage further. Copper screws would be too weak to really tighten up and probably would not add anything as you have made the most important joint which is wire-crimp-case.
It does mean that you need an extra hole in the heatsink to take the Collector wire through to meet up with those from the Base and Emitter, but that's not great problem. I use spare TO-3 isolation washer to form a plastic grommet.
Back on the PCB I use a similar system to connect theTO-3 wires to the PCB tracks. Use stainless steel nuts and bolts/scrimps with crinkle washers to bite into the PCB tracks.
Hope this helps
Mike
![IMG_20180207_185954[1].jpg](/community/data/attachments/601/601574-d65091e85d5788086928771d0743e558.jpg)
I don't think there is anyone here at the moment who is strong on atomic physics or materials science. Really, there isn't any further need apart from curiosity, than to simply look up the properties of common metals and alloys in an engineer's table. This will be more accurate than figures for pure metals as it will refer to specific grades and alloys in common use.
Copper btw, has a unique and closely controlled manufacturing process where in the final melt, it is doped with phosphorus to maintain a constant conductivity for electrical applications like wire, bus bars etc. You can see from just this example that the textbook numbers you crunch need to thought through before assuming they can be applied to anything containing the pure metal, no mater how close to pure it is.
jstott made the point that copper is far from cheap and it probably forms a major production cost in the cheaply produced power semis we generally use. TO3P, 247, 220, 126 types may have solid copper backplates but there's less material there than is wasted in a TO3, TO66 or TO39 package. I think we'll be aware that Sanken's MT200 package with plenty of copper will soon disappear from large scale audio manufacturing anyway.
Motorola once made masses of Aluminium cased TO3s covering many types , general catalogue products and house codes. Not sure about mil. spec. I have worked on many old amps and power supplies that were full of aluminium case TO3s. (There's a pair of MJ15003s here in front of me for certs). There never was a problem or much distinction between the steel types as long as you checked the ratings and didn't make assumptions when your datasheets didn't cover them.
It doesn't answer you question directly, but for interest, I found this reply to the general question: why do some metals conduct better than others?
"In ordinary metals, the same electrons which carry electrical current are also responsible for most of the heat conduction. The reasons for the variation in thermal conductivity are the same as the reasons for the variation in electrical conductivity. Roughly speaking, there are three factors which determine these conductivities:
1. the density of conduction electrons
2. the typical speeds of conduction electrons
3. the typical distances that the electrons travel before they bump into something and change directions.
In ordinary metals, the third factor is the most variable one. One of the biggest reasons that it varies is that the electrons actually travel as waves which don't bounce off the atoms in good crystals, the same as light waves don't bounce off atoms as they travel through a good transparent crystal. Some ordinary metals (say brass) are alloys of different types of atoms, so the electrons bounce off the irregular patterns of different types of atoms. Others, like copper, silver, and aluminum, are pure elements. There will be a little electron scattering even in perfect crystals of pure elements, since the atoms are always jiggling a little out of position because of the energy connected with temperature. This 'thermal' scattering varies substantially between different metals. The scattering off alloys, defects, etc. can be much larger.
mike w
University of Illinois - dept of physics 10/22/2007
"In ordinary metals, the same electrons which carry electrical current are also responsible for most of the heat conduction. The reasons for the variation in thermal conductivity are the same as the reasons for the variation in electrical conductivity. Roughly speaking, there are three factors which determine these conductivities:
1. the density of conduction electrons
2. the typical speeds of conduction electrons
3. the typical distances that the electrons travel before they bump into something and change directions.
In ordinary metals, the third factor is the most variable one. One of the biggest reasons that it varies is that the electrons actually travel as waves which don't bounce off the atoms in good crystals, the same as light waves don't bounce off atoms as they travel through a good transparent crystal. Some ordinary metals (say brass) are alloys of different types of atoms, so the electrons bounce off the irregular patterns of different types of atoms. Others, like copper, silver, and aluminum, are pure elements. There will be a little electron scattering even in perfect crystals of pure elements, since the atoms are always jiggling a little out of position because of the energy connected with temperature. This 'thermal' scattering varies substantially between different metals. The scattering off alloys, defects, etc. can be much larger.
mike w
University of Illinois - dept of physics 10/22/2007
Last edited:
Have built with SD525 (TO-220). The case is importanter than the common specifics. The smaller the better: definition of current and resonances of materials and more: less noise, less audible distortions, less stress.-)
I will try next days TE13003 and ST13003.
And tell;-)
Psu too!!! Do not forget!!!
I will try next days TE13003 and ST13003.
And tell;-)
Psu too!!! Do not forget!!!
TIP42 (!) I had built too.-))) I love PNPs.
But three active parts do sound a little bit slurred - compared with a single-part-amp. But endless dynamic (quiet-loud) and tempo and much much finer colours, 10.000 times better, than a bigger part - result of "cleanness", "noiselessness"-) No doubt: SOT-32 or similar better than TO-220 better than TO-247 better than TO-3P better than TO-3. This the direction.
But three active parts do sound a little bit slurred - compared with a single-part-amp. But endless dynamic (quiet-loud) and tempo and much much finer colours, 10.000 times better, than a bigger part - result of "cleanness", "noiselessness"-) No doubt: SOT-32 or similar better than TO-220 better than TO-247 better than TO-3P better than TO-3. This the direction.
Thanks everyone for the ideas on thermal/electrical conductivity. As Ian says the process is kinetic which of the process ignoring resistance implies no loss of energy. That flies in the face of bandgaps a bit which do have potenetial differences. Gold analysis described by bandgaps when it's conductivity investigated.
The most awful conductor in a T03 transistor is the transistor wafer. Certainly steel will do the job. Thermal and electrical are related.
Steel Electrical 10 to 100 E-8 ohms.m, note 10 is as low as it goes.
Steel Thermal 21-31 W/m K ( typical carbon steel ).
Iron Electrical 9.6 E-8 ohms.m
Iron Thermal 80.4 W/m K ( note most steel nowhere near iron )
Copper Electrical 1.68 E-8 ohms.m
Copper Thermal 386 W/m K
As the range of steel is vast I added iron to say what the base line is. As far as I know no metal on Earth improves by being an alloy in terms of thermal/electrical conductivity. Here are things from memory which if I get them wrong, sorry. If copper is 99.99% pure it has about 99% the conductivity of 100% pure copper. Most copper is called ETPC ( electrolytic tough pitch copper, about 51% of all copper sold I seem to remember ). ETPC has been commonly produced since 1914 when copper became the reference material. Why would we want that purity? The answer is very much more interesting than you might think. Generally the refining uses methane gas( these days )which converts most of the copper to a near pure form and very cheaply. The residual is hydrogen or more correctly water. Water can only get inside copper at very high temperatures which is typical of the refining. The hydrogen is at very high pressure. The hydrogen can be driven out using for example phosphorous. 99.94% pure copper 0.06% phosphorous will have conductivity of about 85% pure copper at best. Why would this matter? ETPC is easilly damaged if used for telegraph cables in high winds like Tornados, the science started about 1820. The more ideal alloy uses silver, silver is unique in doing little harm to the conductivity whilst making the alloy less brittle. It is often used in motors. Most telegraph cables would more ideally be steel and aluminium groups ( not alloys ), the science started with copper.
If zinc is added to copper to make brass a parabola conduction curve is seen. 1% zinc 99% copper shows circa 85% conductivity of ETPC, at 30% the alloy conducts as zinc having mechanical strength in it's favour. Not as bad as phosporous which ironically makes many NPN transistors work.
I think Ian your answer is as near as we will get. Electrons per shell 2, 8, 18, 1 = Copper, 2, 8, 18, 18, 1 = Silver, 2, 8, 18, 32, 18, 1 = Gold. And then Aluminium 2, 8, 3. Just when it looks a safe to bet 1 final electron tells the story Aluminium makes it complex. It could be argued Alumimium is no worse than Gold at 2.82 verses 2.21 E-8 ohms.m ( Copper 1.68 )! 1 electron best, 2 less good, 3 better than 2? Very complex. Sodium is neither good nor bad. Standard theory is the electron of note is too close to the atom centre. That makes gold a bigger puzzel.
I have read explanations of gold that I suspect would not have been instantly obvious to Richard Fyneman. They still don't answer the question. If someone said gold is too heavy it would be equally valid. What I do know is copper with silver plating would make ideal T03 devices. And they wouldn't cost 10 cents more. To solder the collector to the can would be easy. I would guess they would be 5% happier on overheating.
One ironic twist in this story is water. It has very poor thermal conductivity at 0.6W/m K and yet is has the highest ability to absorb heat at circa 4.2kj/Kg K, Sodium is good. The reason nuclear reactors use water is this one. If the reactors is made very tough as it would be the temperture can be taken way above 100C and water remains liquid. The steam is produced via a heat exchanger to keep the reactor water away from the generator water. The trick with water is keep it moving.
In a very indirect way Silicon and Phosphorous make N type transistors work.
The most awful conductor in a T03 transistor is the transistor wafer. Certainly steel will do the job. Thermal and electrical are related.
Steel Electrical 10 to 100 E-8 ohms.m, note 10 is as low as it goes.
Steel Thermal 21-31 W/m K ( typical carbon steel ).
Iron Electrical 9.6 E-8 ohms.m
Iron Thermal 80.4 W/m K ( note most steel nowhere near iron )
Copper Electrical 1.68 E-8 ohms.m
Copper Thermal 386 W/m K
As the range of steel is vast I added iron to say what the base line is. As far as I know no metal on Earth improves by being an alloy in terms of thermal/electrical conductivity. Here are things from memory which if I get them wrong, sorry. If copper is 99.99% pure it has about 99% the conductivity of 100% pure copper. Most copper is called ETPC ( electrolytic tough pitch copper, about 51% of all copper sold I seem to remember ). ETPC has been commonly produced since 1914 when copper became the reference material. Why would we want that purity? The answer is very much more interesting than you might think. Generally the refining uses methane gas( these days )which converts most of the copper to a near pure form and very cheaply. The residual is hydrogen or more correctly water. Water can only get inside copper at very high temperatures which is typical of the refining. The hydrogen is at very high pressure. The hydrogen can be driven out using for example phosphorous. 99.94% pure copper 0.06% phosphorous will have conductivity of about 85% pure copper at best. Why would this matter? ETPC is easilly damaged if used for telegraph cables in high winds like Tornados, the science started about 1820. The more ideal alloy uses silver, silver is unique in doing little harm to the conductivity whilst making the alloy less brittle. It is often used in motors. Most telegraph cables would more ideally be steel and aluminium groups ( not alloys ), the science started with copper.
If zinc is added to copper to make brass a parabola conduction curve is seen. 1% zinc 99% copper shows circa 85% conductivity of ETPC, at 30% the alloy conducts as zinc having mechanical strength in it's favour. Not as bad as phosporous which ironically makes many NPN transistors work.
I think Ian your answer is as near as we will get. Electrons per shell 2, 8, 18, 1 = Copper, 2, 8, 18, 18, 1 = Silver, 2, 8, 18, 32, 18, 1 = Gold. And then Aluminium 2, 8, 3. Just when it looks a safe to bet 1 final electron tells the story Aluminium makes it complex. It could be argued Alumimium is no worse than Gold at 2.82 verses 2.21 E-8 ohms.m ( Copper 1.68 )! 1 electron best, 2 less good, 3 better than 2? Very complex. Sodium is neither good nor bad. Standard theory is the electron of note is too close to the atom centre. That makes gold a bigger puzzel.
I have read explanations of gold that I suspect would not have been instantly obvious to Richard Fyneman. They still don't answer the question. If someone said gold is too heavy it would be equally valid. What I do know is copper with silver plating would make ideal T03 devices. And they wouldn't cost 10 cents more. To solder the collector to the can would be easy. I would guess they would be 5% happier on overheating.
One ironic twist in this story is water. It has very poor thermal conductivity at 0.6W/m K and yet is has the highest ability to absorb heat at circa 4.2kj/Kg K, Sodium is good. The reason nuclear reactors use water is this one. If the reactors is made very tough as it would be the temperture can be taken way above 100C and water remains liquid. The steam is produced via a heat exchanger to keep the reactor water away from the generator water. The trick with water is keep it moving.
In a very indirect way Silicon and Phosphorous make N type transistors work.
Steel can transistors as Andrew T noted have large copper slugs inside. The copper spreads the heat out so that it can get through the steel. The more recent devices with plastic encapsulations generally use copper plates with solder or maybe silver coatings.
I'd hesitate to solder the devices because the temperature reached MAY unsolder the chips! Maybe low temperature solders would be OK but the thermal resistance of lower temp solders tends to be higher.
I've never seen problems connecting TO-3's with M3 or M3.5 brass screws and taking the connection from the underside next to the E/B leads. Actually 4BA was the best size for this. M3.5 does not quite have as large a washer.
I'd hesitate to solder the devices because the temperature reached MAY unsolder the chips! Maybe low temperature solders would be OK but the thermal resistance of lower temp solders tends to be higher.
I've never seen problems connecting TO-3's with M3 or M3.5 brass screws and taking the connection from the underside next to the E/B leads. Actually 4BA was the best size for this. M3.5 does not quite have as large a washer.
That was where I was until one of us said T03 no way. As much as Audiophile stuff is to be taken with a pinch of salt steel does seem wrong.
If anyone is interested in metals I would take gold, mercury and aluminium to be interesting. More interesting is why is gold rare and mercury not. Whilst that is a simplification mercury should be rarer than gold if atomic number matters. Someone said the electron pairing in mercury is what might more easilly happen where the metal was made in a Super Nova ( I say Black holes ). It also ruins it as a conductor.
I notice very few at DIY Audio are Audiophiles. Or if they are they keep it to themselves. I have often found Audiophile ideas often have some truth in them especially if a way of looking at cheaper things. Rean RCA plugs being a nice example.
If anyone is interested in metals I would take gold, mercury and aluminium to be interesting. More interesting is why is gold rare and mercury not. Whilst that is a simplification mercury should be rarer than gold if atomic number matters. Someone said the electron pairing in mercury is what might more easilly happen where the metal was made in a Super Nova ( I say Black holes ). It also ruins it as a conductor.
I notice very few at DIY Audio are Audiophiles. Or if they are they keep it to themselves. I have often found Audiophile ideas often have some truth in them especially if a way of looking at cheaper things. Rean RCA plugs being a nice example.
Agreed, I have also never had a problem connecting to TO-3 ddvices. Just on simple comparison of the thermal connectivity area and the copper slug of TO-3 versus T220 implies they are more efficient.
I am also coming from the point of view that to get a good electrical connection, you start with a good physical one which means good metalic contact - that why I use crinkle washers and connect to the TO-3 case explicity and not via a screw. The Base and Emitter leads are something I haven't mastered yet
Cheers
Mike
Hi Nigel, but why did you say "TO-3 no way". What was the rationale and evidence. Solid gold or even Silver as transistor heat conductors would be prohibitively expensive - "No way". And mercury is just "No way". Aluminium - since it is one of the most abundant metals available, there must be a reason for the semiconductor industry not adopting it.
Regards
Mike
Last edited: