Description

Solid-copper heat-spreader moves heat much faster than mounting a TO-3P directly to the heatsink.

Replace a TO-3 transistor with your choice of TO-3P or TO-247 transistor without compromising thermal performance! Test results here. Looks cool too.

Adapters are solid copper, 1.6mm thick. Transistor is soldered to the adapter, SMD-style. (You’ll need a hot-air soldering station. It’s easy!)

TO-3 transistors are becoming unobtanium, which is unfortunate, because they have superior thermal performance compared to any available modern device. A TO-3 has nearly three times the heat-spreader surface area of a TO-3P, and nearly twice that of a TO-264! That’s not to say they are three times better, it’s complicated… But they do have more ability to move heat.

TO-3’s are going extinct, basically because they are expensive to manufacture, and not well suited for automated manufacture. Modern circuit designs tend to use many smaller devices in parallel.

The only TO-3 audio transistors currently in production are the On Semi MJ2119X line. They are excellent devices, but relatively slow Ft of 4MHz (minimum spec, it’s probably more like 10MHz typical), and with relatively high Cib and Cob capacitance. Certain amplifier designs, especially those optimized for low TIM distortion, may require higher-speed transistors.

If the MJ2119X isn’t suitable for your device… well what to do? Take your chances with fakes on the auction site?

These adapters allow you to fit any TO-3P or TO-247 transistor you like. You might already have something in your parts bin!

Why a copper heat-spreader?

Copper conducts heat nearly twice as well as aluminum. A transistor’s ability to quickly move heat into the aluminum heatsink depends on how much surface area it is in contact with. The more the heat is spread out before it gets to the thermal pad, and then the aluminum heatsink, the better the heat transfer, and the cooler the silicon die. The cooler the die, the more power you can dissipate.

Consider the MT-200 package of the 70s and 80s. It’s literally a TO-3P heat-spreader soldered to an additional, larger heat spreader. Sanken rated these MT-200’s for an additional 50W dissipation versus the same transistor in a TO-3P package.

The thermal properties of good old TO-3:

The TO-3P package was intentionally designed to be compatible with TO-3 mounting holes. But as you can see, there is only 1/3 the surface-area in contact with the heat sink. This is fine in many applications with steady-state power dissipation, as long as the maximum die temperature is not exceeded. But for transient peak loads like audio, the TO-3P case will not move heat away from the die as quickly as TO-3.

Surface Areas of interest:

I measured and calculated the area of the heat-spreaders of output transistors in TO-3, TO-3P, TO-264 and MT-200 packages. The contact patches are not perfect rectangles; I had to account for the plastic area around the mounting holes, the beveled corners, and all the odd-shaped cutouts. These figures are pretty accurate.

TO-3 – 620mm² (MJ21193 and Hoppe’s Brain adapters)
MT-200 – 592mm² (2SA1169)
TO-264 – 313mm² (MJL1302)
TO-3P – 207mm² (NJW1302)

Out of all these, the MT-200 is the best performing package. It has slightly less surface area than TO-3, but its heat spreader is pure copper, whereas most TO-3’s are steel, or steel with a copper plug heat-spreader. The TO-3’s transistor die is not centered, making the heat-spreader area opposite the die less effective. And the two holes in the case for the base and emitter leads make for an interruption in the heat-spreader, right next to the die.

Some Interesting Thermal Conductivity stats:

• Monocrystalline Synthetic Diamond – 3320 W/mK (The ultimate thermally conductive dialectric! A little expensive.)
• Naturally occurring Diamond – 2200 W/mK
• Elemental Copper – 401 W/mK
• Electrolytic Tough Pitch “ETP” Copper – 390 W/mK (Most common copper in electronics)
• Copper PCB substrate – 380 W/mK (These adapters)
• Elemental Aluminum – 237 W/mK
• 6063 Aluminum alloy heatsink – 201 W/mK
• 6060 Aluminum alloy heatsink – 166 W/mK
• Steel – 8-66W/mK, depending on alloying elements
• Tin/Lead 63/37 Eutectic solder – 50 W/mK
• Mica – 0.7 W/mK
• SilPad – 0.9 to 5 W/mK (Higher performing SilPads get expensive real quick.)

Thermal paste, Silpads and Mica do not actually conduct heat very well! They are down in the single digits. What can ya do? It’s just inconvenient physics; Materials that don’t conduct electricity well, usually don’t conduct heat well. Metals conduct heat via free electrons, the same phenomena that transports electron currents. Diamond conducts heat hugely better than any metal, via phonon waves, rather than free electrons.

What’s inside a TO-3?

Well it varies quite a bit! It depends on how much heat needs to move, and how fast.

AFAIK, all TO-3 cases are made of steel. Steel doesn’t conduct heat super well, but some alloys are better than others, so I’d guess the steel they use for transistor cases is at least 50W/mK. (Copper is 400W/mK)

High-power TO-3 transistors usually have a copper heat-spreader inside, like this ON Semi MJ21193.

ON Semi MJ21193 delidded. 5.5mm die, copper heat-spreader 1.6mm thick on 1.6mm steel base plate.

And this Toshiba 2SB554

Toshiba 2SB554 delidded. 5.5mm die, 2.8mm thick copper heat-spreader, all the way through to the other side.Some TO-3’s have a copper heat-spreader that goes all the way through the case, as on this Toshiba 2SB554. The rest is steel.

Toshiba 2SB554, copper heat-spreader is exposed to heatsink. Due to it’s higher copper mass, the Toshiba 2SB554 should have slightly better thermal performance than the On Semi. This isn’t the whole story though; The MJ21193 is actually rated for more power than 2SB554; 250W @200C junction temperature, and the 22B554 is rated for 150W @150C. They both have the same size die and the Toshiba probably has a lower thermal resistance. So it would seem that the extra 100W rating of the MJ21193 comes from allowing the junction temperature to go 50C higher. Is the On Semi die actually able to reliably withstand higher heat than the Toshiba? Maybe it is. Technology moves on. Maybe Toshiba and On Semi have different ideas of safe power levels. In any case, both of these transistors are famously robust devices.

Assembly:

Tools required:

• Hot-air soldering station
  *A temperature-controlled heat-gun should also work well enough. But you can get a basic hot-air soldering wand on Aliexpress for like $30. ¯\_(ツ)_/¯
• SMD solder paste, Eutectic 63/37 tin-lead recommended
• Two alligator clips, (Included)
• Helping Hands or similar work-holding apparatus

First, inspect the pad surface: To prevent corrosion, the adapters have a thin, transparent coating of OSP “Organic Solder Preservative”. OSP is an excellent surface finish for things with large SMD pads, like power transistors and LEDs. It’s a super thin coating at less than 0.5µM, so it leaves little residue behind and doesn’t contaminate the solder beneath the pad.

The OSP coating is fragile and easy to scratch. Scratches or nicks can cause corrosion of the copper. Time can cause corrosion. Inspect the component pad on your adapter, and if you see any corrosion, just buff it bright and shiny with steel wool or scotch-brite.

It’s normal for the coating to look a little spotted or cloudy. There may be scratches in the copper that have been coated at the factory with OSP. That’s fine as long as it’s not brown and corroded.

Next, inspect the transistor itself. If you are re-purposing old transistors, you might find there is corrosion on the heat-spreader. Hone the surface flat and shiny using a large flat file, a diamond hone, or wet sandpaper placed on a flat surface like glass.

Lead bending:

Be careful not to bend the leads too sharply. I like to use a conical wire-bending pliers. Bend the base and emitter leads down at a 90 degree angle. The bend occurs pretty close to the body of the transistor.

Optionally: Bend the collector lead backwards. (If you intend to use the collector connection terminal.) You can cut the collector lead off completely if you don’t plan to use it.

Spread a thin—emphasis on thin—layer of solder paste on the back side of the transistor. Don’t intentionally put any on the plastic parts but it’s OK if there’s a little.

Use two alligator clips to hold the transistor perfectly in position.

The alligator clips serve an important second function;

As the solder melts, the spring pressure from the alligator clips will squish the transistor down flat against the plate, pushing out excess solder, and pushing out voids. This will make the solder layer as thin as possible while still being contiguous.
Don’t let the transistor just float as one normally does in SMD soldering, that will make the layer of solder too thick and reduce heat transfer efficiency. Solder conducts heat really well, but it’s only about 1/4 as good as copper.  But if the solder layer is thin it contributes very little thermal resistance.

Soldering:

Set hot-air station to 350C, high airspeed. No tip.

Clamp the device in such a way that you can heat it from above and below. I use helping-hands with alligator clips.

We’re going to heat mostly from below. But first, pre-heat the top of the transistor for just 10 seconds or so. The prevents causing a large temperature difference between opposite sides, and reduces thermal expansion stress on the plastic body of the transistor.

Then, heat the adapter from below until the solder melts. The copper substrate will naturally heat extremely evenly, which makes it go nice and easy.

Eventually the solder will melt and go all shiny, and the transistor will sink down tight against the plate. Just a little solder should squish out the edges. Remove heat about two seconds after this happens, and let cool. Be careful not to overheat, or the transistor die’s bond to the case could be damaged. Most transistors specify a maximum die to case temperature of 260C for 5 seconds. Eutectic solder melts at 183C. That’s your window, no problem.

Really satisfying.

You should now have a super thin layer of solder holding the plates together with no voids.

I destroyed this MT-200 adapter to verify I wasn’t getting any voids. I filed it down at a shallow angle, then polished it to reveal a wide cross-section of the seam. You can barely even tell where the solder layer is, it’s so thin! I cut it in a bunch of spots, and couldn’t find any voids. It’s like one piece of metal. (Click to enlarge.)

Optional collector terminal:

Many circuit boards require that both bolts of the TO-3 package be connected to the collector.

Unlike a metal-can TO-3, The TO-3P has an electrically insulated mounting hole. This connection may be required in your circuit. The terminal provides this same connection to the collector.

Hold the terminal in place with an M3 screw and solder.

Installation:

Installation is like any  TO-3 device with a few caveats. You might need to get a little creative!

Length: The main issue you might encounter, is that your original bolts are too short because the adapter plus TO-3P transistor is 5mm thicker than the TO-3 flange. You may need longer bolts. Here are suggestions for parts from McMaster-Carr.

1. Socket-Head black-oxide Bolt, M3, 2.5mm Hex-drive. Pick the length you need.
2. M3 Split-ring lockwasher High-collar type specifically made for narrow head of socket-head bolt.
3. Flat washer
4. M3 star locknut

Width: Traditionally, TO-3 transistors are attached with #6 bolts, but it’s very common to find M3 bolts instead. #6 bolts are too wide to fit through a TO-3P transistor’s mounting hole, so you must replace them with M3. If you have TO-3 sockets with threads for #6 bolts, you may install an M3 bolt through the socket and install an M3 lock-nut as shown.

Too close for comfort:

Some TO-3 heatsinks are just packed with transistors right next to each other, and the corners of the TO-3P would physically interfere.

In that case, the edge of the TO-3P transistor may be sanded or filed down to match the profile of the TO-3, without cutting into the plastic case. Often, only one side needs to be sanded. (A belt or disc sander is ideal for the task.)

Thermal pads:

• • Mica pads actually work pretty good and you can re-use them if they are in good shape. Thermal conductivity around 0.7W/mK. Check for de-lamination or cracks. Clean the old thermal grease off with naptha or lighter fluid, it dissolves like magic.
  • Sil-Pads for TO-3 are readily available. I like Parker Cho-Therm T441
    60-11-4511-T441-08. 1.1W/mK, better performance than mica, very tough fiberglass mat prevents tearing or cut-through.
    

Depending on your socket, the base and emitter leads may fit more easily into the socket if you twist them 45 degrees. If the socket feels loosey-goosey around the transistor leads, just pinch them with a pliers, and that should restore the spring tension.

Insert the device into the socket. Alternate tightening each bolt a little at a time, until you reach the desired torque.

Check for zero continuity between the heatsink and the collector.

Uninstallation:

If you need to replace a transistor, the adapter can easily be re-used. Clamp the device in a vertical position. Heat at 350C from the back side. When the solder melts, the transistor will just slide right off, perhaps requiring a gentle tug. When the adapter cools, clean the surface with alcohol. You can now solder a new transistor to the adapter. You won’t need as much solder paste as the first time because there is now a layer of solder already present. It would be a good idea to put some extra flux on the pad before soldering.

Installed examples:

(More coming soon.)

A Marantz 1060 with ST 2SD1047 replacing the original Hitachi 2SC897.

I think the ST 2SD1047 is going to be a better fit for the 1060 than the usually recommended ON Semi MJ21194. The MJ21194 is a super high-power device, 250W, with a big 5.5mm die, and big Cob and Cib to go along with it. And only 4MHz. (Minimum spec, it’s probably more like 10MHz typical) The original Hitachi 2SC897’s are only 60W, with a much smaller die, therefore low Cob/Cib. This is only a 30WPC amp! So why use a 250W transistor? The ST 2SD1047 is 100W, 150pF Cob, and its 20MHz Ft is a closer match to the 15Mhz of the 2SC897.

The 1060 is a well-compensated amplifier, I’m not too worried that it will oscillate with these slightly faster transistors. But I have yet to actually test it, this amp still needs work!