Fan cooling design

Do you plan start formatting posts correctly some time?

Not putting the attribution in the signature block would be a good start.

====================

------------------------------------------------------------------------------------

They next produced the fan free K series -

1. Class D mosfet design with toroidal tranny PSU.

1. Fully sealed aluminium cabinet acting as the heatsink.

Exceptionally reliable & long lived with no maintenance.

I opened one a while back after 15 years of service and the insides looked like new.

Also, when bench tested with a scope and THD meter it was barely possible to tell it was in fact Class D.

..... Phil

Well, (if memory serves) there were two to-3 pacs lying there. It could be that the screws came loose, fell out and then the transistors unsoldered. It was a long time ago... maybe only one transistor?

GH

=================== > >

You sure they were *soldered* in ???

FYI

I have seen power 2 bolt, flat pak transistors and their heatsink become so hot the legs became unsoldered from the PCB - a result of fan failure and no thermal cut off.

I have seen TO3 *lateral* mosfets mounted with with sockets where the two 3mm bolts became partially unwound. The resulting poor contact with the heatsink allowed them to get so hot the silver finish was badly discolored. The mosfets survived of course - laterals can do that.

See top pic of a H-H M900 - the heatsink is not grounded but carries audio output. There are no mica insulators too.

..... Phil

Grin... Yeah Well then maybe not a TO-3 pac. I did replace the fan, solder pass transistors back in place, and it worked fine. circa 80's.

George H.

We always soldered TO-3 transistor base and emitter, with wires or preferably directly to a PCB. The sockets were mostly terrible.

==================================

Early examples often had gold plated pins while others had nickel. For reliability, nickel needs to be soldered. Other and later examples had solder coated pins, which is problematic if not soldered in place.

Incompatible combinations, like nickel pins and tin plated sockets could corrode and develop high resistance in the emitter link.

Smart makers knew which combinations worked but others did not and repairers had no idea it mattered and used whatever was to hand.

Phase Linear amps had lots of issues with this.

.... Phil

Is this funny?

Could be. Calculations or thermometer?

The reasons there are so many articles on forced convection cooling in the 90s, is because that's when the telecom boffins finally caved - going for redundancy/service schedules to meet mtbf in the miniaturization of network hardware. With processors built for fan cooling, and hard drives that weren't, who can blame them? ( all aided and abetted by 'watts per cubic inch' module floggers.

That allowed multiple fans in multiple units. . . . and required soundproofing on top of air conditioning.

In an isolating scheme, ideally the parts with the lowest temperature index (caps and insulating magnetics) get preferential treatment.

In a critical forced conduction environment, the fins aren't at right angles to airflow, but then who really gets to do any designing, anyways? You use what's there and crank up the airflow.

RL

I had to simulate the circuit to estimate fet power dissipations. I tested the heat sinks in various air flows to estimate their theta. I tested that fan for flow patterns. Then did a little math.

There are no closed-form equations for this kind of thing.

But what I did is certainly design. Even better, it all works.

It looks like a harmonic balancer.

I agree, We use a plastic holder thing to hold a TO-3, panel mount. And then soldered wires to the pins.

George H.

A little. I had to add a fan to the bottom of my first box. Several class A drivers for B-field coils... Wasteful. In the 'worse' configuration temperature inside the box was bad... ~70C IIRC. I'd do better today. :^)

George H.

The machined spring pin sockets are reliable. They can solder onto a PCB, and the transistor pins can poke into them. If all the diameters are right and the current is reasonable.

We use them a lot, but mostly for pluggable options, like an expensive OCXO or something, or for things that could need to be replaced.

I don't miss TO-3 packages.

For fan cooling in the low watt range, I'm sort of fond of THM6263, and variants. They came out about the same time as fan use became permitted/popular. If you ask someone to bend metal nowadays, they look at you strangely.

Those extruded ones are cute, in that you can stick simple Al plates on top, in a pinch.

RL

It's not funny, it's beautiful. Each fan cools its four fets, the inductors, and a giant expensive toroidal transformer just beyond the end of the board.

The temp rise per watt dissipated by the toroid drops in half with that second-hand air flow. So the power rating goes up about 50%.

Transformer designers must default to assuming still air, and seem to be generally conservative at that.

None of that thermal design was very analytic. It was done with a bunch of simple bench testing, arithmetic, guesswork, and luck.

=============================

..... Phil

In a fixed/regulated voltage environment, a 50% reduction in surface rise of a transformer would permit a 40% increase in current (or throughput power). The simple VxI output generates I^2 copper losses, which dominate full-load losses.

They have a better idea of actual internal winding temperature rise, and an intimate knowledge of the thermal restrictions of the class of insulation system being employed.

You can duplicate their design limits by measuring the internal winding rise under normal use at high line/low line, and taking that number as a limiting factor in determining capability in your specific application.

Measuring rise of resistance is the simplest way, once you've identified the most-buried winding.

RL