The ultimate DIY? A Stax SRM-T2!

[Inu]

Morphsci is today's winner...

That went fast.

Wow! only 35min!!!

[kevin gilmore]

Oops, thats definitely a bug...

Will figure out how i messed that one up...

My transformers don't have 100 volt taps.

I hate messing up the simple stuff.

DEFINITELY DO NOT HOOK UP THE 100V transformer

windings if your transformers have these windings!!!

Thats what happens when i try to make it goof proof

Edited July 4, 2010 by kevin gilmore

[Voltron]

Big congrat to Jim for snagging the T2!

[luvdunhill]

Even better, although Scott just beat you to the punch on this one via PM. Apart from instant availability the really great thing is that these ebay sellers usually have incredibly low shipping fees. For instance this seller charges $3.90 worldwide, about 1/10th of Future Electronics' min delivery charge to Sweden

I think just purchasing the insulated package version directly from Mouser would be the way to go, seeing how you'll already be placing a larger order with them.

[kevin gilmore]

Insulating version of 8n80c definitely a good idea. I did not even know it was available.

Someone needs to verify that it can dissipate enough heat though...

Definitely do not hook up both the 100 volt and 120 volt windings on the filament transformer at the

same time, pick one or the other. You can hook up all the wires on the high voltage transformers if

you wish.

This error has been corrected on the layout so that if there is another board run, this won't happen again.

Doh.

[struts]

Insulating version of 8n80c definitely a good idea. I did not even know it was available.

Someone needs to verify that it can dissipate enough heat though...

Heat dissipation on the FQPF part (TO-220F) is specified at 59W vs 178W for the FQP (TO-220) package. I'll stick with the latter.

[Craig Sawyers]

Heat dissipation on the FQPF part (TO-220F) is specified at 59W vs 178W for the FQP (TO-220) package. I'll stick with the latter.

But that is down to the total thermal resistance. Junction to case is 0.7C/W for TO220 and 2.1C/W for the plastic case - a ratio of 1:3, which is the same ratio as the maximum power dissipation (ie 178/59). But for the TO220 you need the ceramic insulator - which is 2C/W.

So the total thermal resistance for the TO220 plus alumina insulator is 2.7C/W as compared with 2.1C/W for the plastic. So the plastic should actually dissipate more power than the TO220 in the T2. It is also stiffer than the metal tabbed TO220 (which is a nightmare package to bolt down and keep it in contact with the heatsink), so should stay in contact over its whole area with the heastsink.

[luvdunhill]

I reached the same conclusion as Craig. I didn't look for a TO-247 or TO-3P version, as I'd assume they wouldn't fit...

[Craig Sawyers]

I reached the same conclusion as Craig. I didn't look for a TO-247 or TO-3P version, as I'd assume they wouldn't fit...

The trick with getting good TO220 thermal contact is to apply the right fastening torque, particularly with the elastomer insulators (but also with hard insulators like mica - or alumina). The spec depends on manufacturer, 0.8 - 1.1Nm (7 - 10in/lb) in one case and 0.49 - 0.686Nm for Sanken. Beyond that the tab starts to deform enough to lift the pressure off the body - the package sort of pivots upwards. I though that the plastic TO220 could take more torque - but the specs do not indicate that - the same as for the metal tabbed version.

I bought a torque screwdriver a few years ago for precisely this purpose - and the correct torque is surprisingly slight. When hand tightening, I've found that the tendency is to massively overtighten. As you increase torque, the thermal resistance falls to a minimum, and then increases again as you tighten more.

The way around that is to use a hole-less insulator, and one of those spring clips that bear in the middle of the transistor body. Keeps voltage isolation high, and applies much more pressure since there is no risk of pivoting. The problem is that you need a heatsink extrusion with a feature into which the clips engage.

[struts]

But that is down to the total thermal resistance. Junction to case is 0.7C/W for TO220 and 2.1C/W for the plastic case - a ratio of 1:3, which is the same ratio as the maximum power dissipation (ie 178/59). But for the TO220 you need the ceramic insulator - which is 2C/W.

Craig,

Thank you for correcting me, I guess I skipped that class..

You provide an uncomfortable reminder that if I had paid more attention back at university I might have made it as an engineer and not have moved into sales!

So if the molded body only dissipates one third of the heat, where does the rest go?

[Craig Sawyers]

Craig,

Thank you for correcting me, I guess I skipped that class..

You provide an uncomfortable reminder that if I had paid more attention back at university I might have made it as an engineer and not have moved into sales!

So if the molded body only dissipates one third of the heat, where does the rest go?

It doesn't go anywhere as such.

The only thing that matters is the temperature of the silicon. That is usually taken to be 150C or thereabout. So if you are bolted onto an infinite heatsink, the thing that determines the temperature of the silicon is the package thermal resistance.

So, if the TO220 metal package dissipates 178W on an infinite heatsink with 0.7C/W, the silicon temperature rise is 178 x 0.7 = 125C, plus the heatsink ambient of 25C = 150C. Similarly for the plastic package, the silicon temperature rise is 59 x 2.1 = 124C - the same number (within the accuracy of the numbers).

If the heatsink is hotter - because in the T2 there are lots of devices attached to the same sink - you have to derate things further.

The total calculation involves adding together the thermal resistances, in this case package + heatsink-air, so 2.1 + 0.3 (guesstimated) = 2.4C/W. Assume heatsink temperature is 35C. So maximum device dissipation is (150 - 35)/2.4 = 48W.

For the metal device, we have to use a heatsink insulator - which is say 2C/W. So we need to add up package + insulator + heatsink-air, or 0.7 + 2 + 0.3 = 3C/W, giving a maximum dissipation of (150 - 35)/3 = 38W.

Of course, you would not operate a transistor die at 150C - you'd derate it to <100C for long term reliability. Which would mean max device dissipations of ~27W and ~22W respectively.

But anyway, the upshot is that the plastic packaged device gives better thermal performance. The only way to equalise them is to reduce the thermal resistance of the heatsink insulator for the metal package to ~1.4C/W - but we're stuck with alumina which is 2C/W for a TO220. Running some quick numbers the T2 devices run around 5W quiescent and perhaps 10W maximum - so safely inside the dissipation limit with an alumina insulator.

[luvdunhill]

So, if the TO220 metal package dissipates 178W on an infinite heatsink with 0.7C/W, the silicon temperature rise is 178 x 0.7 = 125C, plus the heatsink ambient of 25C = 150C. Similarly for the plastic package, the silicon temperature rise is 59 x 2.1 = 124C - the same number (within the accuracy of the numbers)

don't forget the thermal grease for the plastic package

[Craig Sawyers]

don't forget the thermal grease for the plastic package

I kind of reckoned that I'd defaulted to lecture mode quite enough for one day . I do have a habit of banging one once I get going.

But you're right - with the plastic package you get one layer of grease, and with the alumina you get two. So add another tenth or so per layer.

[struts]

It doesn't go anywhere as such.

The only thing that matters is the temperature of the silicon. That is usually taken to be 150C or thereabout. So if you are bolted onto an infinite heatsink, the thing that determines the temperature of the silicon is the package thermal resistance.

So, if the TO220 metal package dissipates 178W on an infinite heatsink with 0.7C/W, the silicon temperature rise is 178 x 0.7 = 125C, plus the heatsink ambient of 25C = 150C. Similarly for the plastic package, the silicon temperature rise is 59 x 2.1 = 124C - the same number (within the accuracy of the numbers).

If the heatsink is hotter - because in the T2 there are lots of devices attached to the same sink - you have to derate things further.

The total calculation involves adding together the thermal resistances, in this case package + heatsink-air, so 2.1 + 0.3 (guesstimated) = 2.4C/W. Assume heatsink temperature is 35C. So maximum device dissipation is (150 - 35)/2.4 = 48W.

For the metal device, we have to use a heatsink insulator - which is say 2C/W. So we need to add up package + insulator + heatsink-air, or 0.7 + 2 + 0.3 = 3C/W, giving a maximum dissipation of (150 - 35)/3 = 38W.

Of course, you would not operate a transistor die at 150C - you'd derate it to <100C for long term reliability. Which would mean max device dissipations of ~27W and ~22W respectively.

But anyway, the upshot is that the plastic packaged device gives better thermal performance. The only way to equalise them is to reduce the thermal resistance of the heatsink insulator for the metal package to ~1.4C/W - but we're stuck with alumina which is 2C/W for a TO220. Running some quick numbers the T2 devices run around 5W quiescent and perhaps 10W maximum - so safely inside the dissipation limit with an alumina insulator.

Thanks Craig. We are not worthy!

[luvdunhill]

But you're right - with the plastic package you get one layer of grease, and with the alumina you get two. So add another tenth or so per layer.

.. plus the bracket, plus the interface between the bracket and the heat sink, plus the anodizing (if any) on the heat sink, plus ambient differences say between Iceland and Texas, etc... I think that 22W/27W will end up a bit lower. Even 100C is pushing a TO-220 package, lets say 75C. Anyways, its good to see the calculations

[digger945]

If anyone is interested I would make a deal on some older EL-34s. A quad of xf3s and two matched pairs of xf2s.

[Craig Sawyers]

.. plus the bracket, plus the interface between the bracket and the heat sink, plus the anodizing (if any) on the heat sink, plus ambient differences say between Iceland and Texas, etc... I think that 22W/27W will end up a bit lower. Even 100C is pushing a TO-220 package, lets say 75C. Anyways, its good to see the calculations

Lifetime (MTTF) data on specific transistors is very difficult to pin down. Generically for silicon at a junction temperature of 150C it seems to be around 10^6 hours, or 100 *years*, assuming it is a chemical activation process - defects, contamination etc. People do elevated temperature testing to reduce MTTF to a few thousand hours and then extrapolate. There is an arcane paper here http://rel.intersil.com/docs/rel/calculation_of_semiconductor_failure_rates.pdf that sets out some detail.

[kevin gilmore]

Inu has a listenable unit. YEAH. I'll let him describe the sound.

In less than 2 weeks he will have the knob and custom piece...

Then he will be the first with a finished FINISHED T2

Better than my version of "i've died and gone to heaven"

Birgir, stop being a slaggard and get yours done...

Just got a raw piece of OUTSTANDING wood from craig in GB...

Lets see what kind of knob(s) i can turn this into...

Edited July 9, 2010 by kevin gilmore

[Voltron]

What a pioneer! Congrat Inu!

[spritzer]

It's not like I haven't been working on other projects... I have way too much stuff going on...

[Craig Sawyers]

Just got a raw piece of OUTSTANDING wood from craig in GB...

Lets see what kind of knob(s) i can turn this into... :D

Good it got to you in good shape, Kevin - massively looking forward to what comes out

[Craig Sawyers]

Inu has a listenable unit. YEAH. I'll let him describe the sound.

In less than 2 weeks he will have the knob and custom piece...

Then he will be the first with a finished FINISHED T2

Inu - tell us more, more more!

[Inu]

I added a wire between the GND terminal on the EMI filter and GND terminal on the PS board otherwise I get an electric shock every time when touch both at the same time.

Appearing voltage is about AC106V on the PS itself and it becomes DC208V when connect the AMP. When touched, it feels like a maximum position of OMRON massager.

And without connect the chassis to the circuit GND, couldn’t eliminate the hum and static noise.

I attached multi turn pots for temporary and play my favorite CD…

My first impression is “very atmospheric” and “extended low end”.

I couldn’t stop listing it, and extended my lunch brake to 2PM.

Next biggest issue for me is a timing to show these big heaters to my wife …

[kevin gilmore]

I couldn

Edited July 9, 2010 by kevin gilmore

[Craig Sawyers]

I added a wire between the GND terminal on the EMI filter and GND terminal on the PS board otherwise I get an electric shock every time when touch both at the same time.

Thanks for the heads up - I'll watch that one when I build mine up.

Great to hear that the beast lives, and sounds great! Can hardly wait.

Mrs S picked up the credit card bill today, which had lots of money to Mouser, Farnell, Dalbani etc on it. It sure was an interesting conversation (not)