What are these are called and where I can get some from? (type of heatsink)

[SingedFingers]

They are used to thermally bond TO-92 and TO-18 packages together. I want to stick a couple of matched 2n4416's in a DIY active probe (Tek P6046 clone) together:

[David Hess]

Tektronix called it a "cover", "temperature stabilizer", or "thermal equalizer" which is not much help.  I have no idea where to find them already made.  I usually end up using epoxy to either bond the flat sides together or bond the flat sides to an aluminum bar if a symmetrical layout is important.

There are some ideas here:

http://www.angelfire.com/sd/paulkemble/sound7e.html

[SingedFingers]

Thanks for the info. Much appreciated. I may just steal one out of a dead Tek 465 I have lying around rather than buy or bodge one in that case.

[MrBungle]

Do you know anyone with a milling machine? (Model engineer etc) Wouldn't take long to knock one up.
If you don't, I could make one and mail it to you, no charge.

[SingedFingers]

Thanks for the offer but sorted now. I've just eviscerated one out of one of my spare scope parts mules lying around.

[BrianHG]

They are used to thermally bond TO-92 and TO-18 packages together. I want to stick a couple of matched 2n4416's in a DIY active probe (Tek P6046 clone) together:

Get the ones in a metal can:
http://www.digikey.com/product-detail/en/central-semiconductor-corp/2N4416A/2N4416ACS-ND/4806895
and just soldier the 2 metal cans together.  They will be even better thermally bonded than a strapped on shared heat-sink.
Also, there is a 4th case pin for shielding.

[Cerebus]

Do you know anyone with a milling machine? (Model engineer etc) Wouldn't take long to knock one up.
If you don't, I could make one and mail it to you, no charge.

You could manage it with a drill press and a hacksaw. Well, OK, a drill press and a lot of very boring hacksawing.

[MrBungle]

Thanks for the offer but sorted now. I've just eviscerated one out of one of my spare scope parts mules lying around.

No worries. I missed the 'dead' part of your earlier post, thought it would be a shame to cannibalise a Tek, but if it was beyond repair then it makes sense 

You could manage it with a drill press and a hacksaw. Well, OK, a drill press and a lot of very boring hacksawing.

Haha, yes those days are loooong behind me and never to return!

[Ian.M]

Been there, done that a good many years back.   No milling machine needed, just a pillar drill  and a vice. You need a good enough selection of drills that you can find one fractionally under the TO-92 case diameter, so that it would be a tight force fit. Some test holes may be required. then its just drill two through holes in the block, chamfer the hole edges for ease of insertion and slit the ends leaving the middle intact.  A dab of silicone heatsink grease and pressing the transistors in, relying on the Aluminum being springy enough to provide enough clamping pressure, finishes the job.   Its far less work than the original as you aren't cutting the middle and don't have to drill and tap a cross hole for the clamping screw.   

There were also similar isothermal couplers formed out of fairly thick springy sheet copper with the ends formed round a rod to match the transistor diameter. They wouldn't be good enough for metal can devices but TO-92s have enough thermal resistance to the case surface that the relatively small cross section and long path through the copper isn't a major issue.

[David Hess]

You need a good enough selection of drills that you can find one fractionally under the TO-92 case diameter, so that it would be a tight force fit.

Thermal conductivity of the plastic package is poor enough that the fit does not need to be that tight.  This is why attaching a huge heat sink to a plastic packages does not help much compared to small heat sinks; the plastic package contributes most of the thermal resistance.  Metal can and ceramic packages can take better advantage of heat sinks because of their lower junction to case thermal resistance.  Thermal compound is still worth using.

If you have the right kind of epoxy resin, the flat face or top of the TO-92 part can be directly attached to an aluminum block.  It is not as good as a clamp but may be sufficient.

There were also similar isothermal couplers formed out of fairly thick springy sheet copper with the ends formed round a rod to match the transistor diameter. They wouldn't be good enough for metal can devices but TO-92s have enough thermal resistance to the case surface that the relatively small cross section and long path through the copper isn't a major issue.

One of the examples in the Tektronix diagram I included looks like the S clip is thickened in the middle for better thermal conductivity.  I am not sure where they used that one and I have only seen the thin S clip.

[SingedFingers]

I'm using the Tek one now. The TO-18 packages fit snugly and they have thermal compound on them (arctic silver no less).

I've been measuring drift overnight and we have a miserably small 0.8uV which is pretty good for a pair of matched FETs and within spec for my requirements. The meter may have actually drifted more

If anyone is interested, I used a modified version of this (sub the BJTs for BD140/139 pair) to match the FETs: https://www.fairchildsemi.com/application-notes/AN/AN-6610.pdf - worked really well and was fun to breadboard it Jim Williams Style

[Gyro]

If you're using TO92s then the simplest solution is the best - glue them back to back (flat-side to flat-side). Close proximity will also improve thermal coupling through the legs and PCB. If you want to go one step further then wrap a bit of copper foil around them. The effect of anything more bulky will be nullified by the thermal resistance of the epoxy packages.

[Cerebus]

A dab of silicone heatsink grease and pressing the transistors in, relying on the Aluminum being springy enough to provide enough clamping pressure, finishes the job.

It's worth mentioning that anything that applies force to the package (press fit, tightening mounting screws etc.) can potentially apply force to the die inside. In precision circuitry the stress on the die can have significant effects on circuit parameters, drift and thermal response. People go to, sometimes quite extreme, lengths to minimise these effects on things like precision voltage references including stress relief cut-outs on PCBs, etc. etc. Not a major issue for everyday circuitry but if you're into an area where your normal unit of calculation is PPM then it's something to consider.

[Ian.M]

Very good point, and as the coefficient of thermal expansion of the plastic package and the metal isothermal block are different as well, if you need that PPM level precision, if possible choose a hermetically sealed metal can package, otherwise you are going to need to make the TO-92 a slip fit in the hole, probably with some sort of retention plate that does not compress the package, and rely on the thermal grease for coupling.  Some devices are more sensitive to mechanical stress than others, the obsolete Dallas DS1820 temperature sensor being a case in point. See the first few pages of [url]https://www.digitemp.com/docs/ds1820-report.pdf[url] for the issues that simple humidity changes caused, presumably by swelling the package slightly.