Building a rack mounted ESP Project 27 Power amp (heatsink?)

[stormbringer]

Hey guys, still researching and trying different solutions for my preamp (found in another thread, which have been on hold due to album recording and touring).

Meanwhile, i decided to buy a PCB from Rod Elliot for his 100W power amplifier, it was amazingly easy to build. now my plan is to put both the preamp and the power amp inside 2 separate 1U rack cases that i scavenged from old network switches at work.

Now the question i have is about the heat sink, i know rod has 0.5c/w written at the public page, but 1.0c/w at the locked page for buyers. i have bought this, as this was the biggest sink i could fit in the case, according to Rod's heatsink spreadsheet it's about 0.7c/w. But, i guess airflow wont be optimal due to the "layout" of the heatsink. You think i will get away with this one? or should i get an even bigger heatsink and put it in a 2U case? (i'm aware btw, that i will need to trim some of the sink off to actually make room for the toroid)

[Roly]

Murphy's Law of heatsinks rule #1 - they are never big enough.

Fins horizontal will seriously raise the temp rise per watt.  It's Sydney to a brick that 0.7ºC/W is the thermal resistance with the fins vertical, most likely the whole heatsink vertical.   This is more the sort of thing that normally gets used on rack cases.

Even a small air mover makes a very big difference, and can be temp controlled for minimum noise.

[stormbringer]

Thanx Roly, i'll have a look and see if i can find something similiar in europe/ebay. I do plan to add a fan, but i still need to be sure it doesnt BBQ if/when the fan gets bad, so going for a big enough sink stand-alone, and then adding the fan for extra cooling.

[stormbringer]

Seems difficult finding a proper heatsink for this Project. Currently thinking about splitting up the transistor pairs, mounting them off-board on 2 different heat sinks on each side instead as that would give me some headroom finding heatsinks that would actually fit the 1U case. The amplifier board is tiny, so i got loads of empty space in there anyway. I Think i need to mash up some form of test-rig to actually measure the heat sink's efficiency. (yeah, probably overkill, but Learning and experimenting is fun!)

[Roly]

going for a big enough sink stand-alone, and then adding the fan for extra cooling.

This is a very wise idea - maximise passive cooling as much as possible.

mash up some form of test-rig to actually measure the heat sink's efficiency

And so is this.  It will give you a good feel for what will, and won't, do what.

{One little sidebar with the Thermo-Fan is that the top LED position can also be used to operate a cutoff or mute for those occasions where some schmuck hangs their coat over your rig.}

[joecool85]

This will give you an idea on how I feel about heatsinks: http://www.ssguitar.com/index.php?topic=175.0

[stormbringer]

{One little sidebar with the Thermo-Fan is that the top LED position can also be used to operate a cutoff or mute for those occasions where some schmuck hangs their coat over your rig.}

Hm, maybe i should cover it in barb wire Jokes aside, that's a good idea. will definately check it out.

joecool85: Yeah, i've read that post alot, i've also been Reading Teemu's book, and all heat sink related stuff on Rod Elliot's pages.

I Think i got the general idea for it all, but i tend to sidetrack ALOT in my Projects, as i just can't resist Learning how stuff works. But i'd like to Think that i could actually make something quite good when done.

Edit: Just ordered a bunch of different heat sinks on ebay in all kinds of SHAPEs and sizes, will try Everything from sandwiching the transistors between 2 sinks to putting them on individual and what not when i get my test rig setup for them. will try to get some semi-scientific numbers/graphs and check back.

[joecool85]

Edit: Just ordered a bunch of different heat sinks on ebay in all kinds of SHAPEs and sizes, will try Everything from sandwiching the transistors between 2 sinks to putting them on individual and what not when i get my test rig setup for them. will try to get some semi-scientific numbers/graphs and check back.

Cool.  The more data the better!  We also love to learn.   :tu:

[phatt]

Are we all over thunking this? 
Heck how many millions of amplifiers exist with nothing more than a puny bit of alloy bolted to a case and some run for years like that :lmao:
Not that I suggest doing it  that cheap :trouble
Some Amp designs run hotter than others but I'd bet Rod Elliot is no fool and my bet is that it would be highly stable even under extreme conditions and in general use, run fairly cool.

Sure if you where to run this as a bass rig and run 2 Ohm speaker loads it will get darn hot but sounds like *Stormbringer* is careful and unlikely to kill it that way.

In my Observation it is highly likely a small well designed heat sink could outperform a larger one.

From what I've read,,, Power transistors need to dissipate heat instantly on large transients otherwise they will over heat and self destruct.

Worst case;
With no heatsink the heat has nowhere to run. :duh

Take a careful look at heat sinks notice how some have a *large Mass* directly under the power unit,, then the fins run off that at intervals.
(That is what you want)
Often that block mass at the base of the fins is 6mm thick (the thicker the better).

Whereas others only have a 3mm block mass and then many fins.. OK they work but not so good under stress. Usually the fancy ones (Cheap) 8|

Some SS Fenders have no real heatsink but at least they have a large bar of alloy about 12 x 30mm thick where the power devices are bolted. then heat is dissipated through the outside case which is thin sheet alloy and acts as a heat sink fin.

The mass at the base of the fins gives the power units somewhere to dump heat instantly and then the heat travels to the thinnest part and into the air.

You can get very deep into fizzHicks of heat transfer if you want but I doubt you will need to.
Just know that it takes very little heat to heat up thin metal and a lot more to heat up a solid mass. Now just invert that and you get how heatsinks work.

"The heat in the mass runs to the faster transferring thin fins"

Some Heat sinks on CPU units actually have a block mass of copper embedded into the bottom of those fancy cooling fins as copper transfers heat much faster than alloy alone.

I guess if you are designing something to run on the moon for 50 years then go all out and use everything you can throw at it to run cool but for guitar Amp use I think your first long heat sink will do just fine.

As You're going to test it first then any problems will soon surface before a meltdown.
Phil.

[Roly]

Power transistors need to dissipate heat instantly on large transients otherwise they will over heat and self destruct.

This is nothing like as big a problem with audio transistors as it is with SCR's in power control, such as theater lights.  Mullard used to make a range of heatsinks, 35D, 45D, 55D and (I think) 65D.  The latter was a fat base with tall fins and designed for use with SCR's in power control situations.

Thermal circuits can be modeled just like electrical circuits in tools like LTSpice, but finding the values of the resistances and capacitances requires a bit of experimentation.  Thermal resistance in degrees rise per watt is a direct analog of resistance as volts per amp.

We all tend to suck it and see, but it is very common when dealing with heatsinks by intuition to grossly underestimate what you need, to grossly overestimate the ability of a given heatsink to carry heat away to the surrounding air.

Doing the sums is not at all difficult really, you can find the thermal resistance of the chip to base, thermal washer, heatsink to air; it just gets a bit depressing when you realise just how much heatsink you really need to keep the chip temperature below 100ºC when the on-stage temperature under the lights during a summer gig can be over 40ºC.  Then somebody drapes their towel over it.

From scrounging old computers I have accumulated quite a pile of CPU heat dissipators, many with fans, and generally these should be useful for use with chip amps.

Even a small amount of air movement makes a big difference.  Telecom labs ages ago did some experiments and discovered that 90% of the cooling effect with a fan occurs in the first 10% of the rev range; the difference between a light draft and a roaring gale isn't much at all (except in dB).  Experiences with my ThermoFan controller show the same thing.

{and I hate to be picky, but heatsinks don't work on the moon.}

[J M Fahey]

{and I hate to be picky, but heatsinks don't work on the moon.}

I don't want to be picky squared  :lmao: , but I bet Phatt knows that , the idea was to give an example of *ultra* reliability requirements.
FWIW one of the 70's Voyager spaceships is already outside the solar system .... and still works 

Back to this amp, the main problem is trying to fit it inside a 1" rack case, which DEMANDS fan cooling .... or a NASA approved heat sink.

While in 2U height (what I use, by the way), Roly's example is more than enough http://www.jaycar.com.au/productView.asp?ID=HH8555

In fact, for 100W amps I use this one:

5 1/4" wide, 2" high , 1 1/4" fins , base thickness 5/16"
1,5 ºC/W .

The transistors are mounted on the inside of the 2U high aluminum back panel and the flat heat sink is sandwiched against it "on the other side".
It's very efficient because it's outside, augmented by the aluminum back panel and fins run vertical.
Of course, I use grease between it and the back panel.

In "Marshallish" looking heads, which are taller, I use no visible heatsinks but a taller 1.5mm thick  back panel, augmented by an extra 2.5mm rectangular plate sandwiched to it, which is clearly visible here:

[stormbringer]

Thanx guys, really appreciate the tips and feedback here!

The more i Think about it, the more i lean towards going for a 2U case, but at the same time it would be awesome getting it to fit in a 1U. (going on europe tour november, and well, less space taken is Always nice, even if it's just 1U). On the other hand i really dont want it to fail either of course.

I'm still thinking about splitting up the 4 output transistors into 2 pairs on separate heatsinks, but also thinking about integrating the heatsinks in the enclosure. kinda like this:



On the other hand i guess there might just not be enough air around if i go with 1U, so i guess the tests are still on. just waiting for my Transformer to build the Power supply needed and a thermocouple-thermometer i ordered.

Again, thank you!

[phatt]

Power transistors need to dissipate heat instantly on large transients otherwise they will over heat and self destruct.

{and I hate to be picky, but heatsinks don't work on the moon.}

Touche Roly, Ya got me

I guess there is no point in trying to sell you my ash tray for a motor bike invention then? :lmao:

I just wanted to raise the point that it depends on how the circuit is designed.
Some Amps run exceedingly hot while others of same/similar wattage do Not.

I have an old Inkel Factory PA unit which claims 120Watts and has a huge heat sink and it runs like an oven while another Amp I have claims 100 Watts and it runs luke warm, both working from similar voltage rails and both units have similar sized heat sinks.

I did note long time back while learning to jump through hoops trying to make a simple DC power amp that depending how you setup the whole thing you can make the output devices run very hot while the emitter resistors remain only warm.

Change a few things at the front and the opposite might happen.
I've seen inside some poweramps where the emitter resistors are obviously running so hot they blacken the pcb,, even some de-laminate the pcb tracks.

I notice some designers focus a lot on the thermal stability of output devices while others not so much but I read somewhere long time back that the Mid drive transistor (the class A part) is often thermally cycling faster that the output units.
Makes sense as it's often the only device that is connected rail to rail so it has to cope with more voltage across it. Would it make sense to monitor that instead, or maybe both?

Maybe better if You explain that one, winky.

The only time the Inkel broke down was a 6 hour stint in the middle of summer and after some testing I realized the power amp was actually still working fine but some regulator chips for preamps where mounted on the very same heat sink  causing them to shutdown.

Simple fix,, just remounted the regs on a separate slab of alloy away from the main heatsink. Has not even hinted at stopping since. :tu:

I use the Inkel as it has many inputs and covers me for guitar and keyboard so very handy to have for bigger gigs.
Thanks, Phil.

[Roly]

Talking of "*ultra* reliability", how's the public transport and sports reform going over your way?  I can smell the tear gas from here.  I wish my fellow countrymen would get half so engaged with issues.  Just switched back to the previous Prime Minister, but everyone I talk to it's not a matter of who they want but who they dislike least, they don't want one, but they really don't want the other.  Difficult.

Yeah, Voyager is scary when these days you can't get a Set Top Box to last more then a few months before going into software meltdown.


@stormbringer - NOW your talking!  This reminds me of a rack amp I used at one time that was very similar, but had a really cute trick.  A fan in the middle of the back blew air in around all the innards, and there was a small gap each side under side heatsinks just like this case, the bottom having an "L" upward lip each side to direct the exhaust air up through the fins.  I thought it was very cunning.

I guess there is no point in trying to sell you my ash tray for a motor bike invention then?

You don't read American biker magazines?  The ones with the ad's for plastic tubes that you can fit a spliff into for a toke while ridin' your hog into the sunset?

running so hot they blacken the pcb,, even some de-laminate the pcb tracks

This is also common in industrial stuff, worst I saw was a hole charred right through a fiberglass PCB.

just remounted the regs on a separate slab of alloy away from the main heatsink.

Hope you didn't thereby disable their lurky protection idea.

Nah, no excuse for thermal instability, any more than for signal instability, just easy to overlook/ignore.

I think I mentioned the Gibson which was actually oscillating at about one cycle every ten minutes.  Replacing the distant bias compensation diodes with two BD139 E-B junctions bolted to the main heatsink made it totally stable - helps to place the sensor on what you want to control.