Most mosfets get very unhappy at higher voltages and moderate currents, with power dissipation well below their ratings. You really have to read the SOAR curves, and then maybe test them, for linear use.
ftp://jjlarkin.lmi.net/ExFets.jpg
John
Bolting the heat source(s) to a 5-10 mm thick slab of copper the size of the aluminum heat sink, is quite effective in evenly delivering the heat into heat sink, since copper heat conductance is 2-3 times higher than aluminum or aluminum alloys.
Yes, I'd like to do that but we couldn't find any busbar copper at the time. The extra aluminium strip was good enough for the job at the time, only trying to get about 25W per transistor into the heatsink.
Grant.
On a sunny day (Thu, 29 Jul 2010 10:22:39 +1000) it happened Grant wrote in : I put couple photos of my eight transistor linear active load up
For a moment it occured to me that with 8 MOSFETS used as *switches* and 8 resistors in ratio 1, 2, 4, 8, 16, 32, 64, 128, you can make 265 load levels, and dissipate next to nothing in the MOSFETS. Then you need no shunt, as you know what you switched. Of course the '1' resistor would have to be the biggest one...
resistors
That's a nice architecture.
One of my products uses a high-power switching regulator that has a PWM-switching power fet in series with a resistor into the output caps... no inductor. It's no more efficient than a linear regulator, but the power is dissipated in the resistor, not the fet.
This same box has a dual 3-bit, 20 kilowatt DAC, implemented by switching binary-weighted three-phase transformer secondaries in and out, 25 volts/LSB.
John
The two shiny strips on this amp
ftp://jjlarkin.lmi.net/Amp.jpg
are nickel-plated copper heat spreaders. We buy the copper strip from McMaster and machine it then have it plated. The alloy is machinable - pure copper is gummy and hard to machine - so the thermal and electrical conductivity aren't as good as pure copper, but still better than aluminum alloy.
John
John Larkin a écrit :
resistors
He he...
and are multiplying DACs too, but won't like a DC reference...
-- Thanks, Fred.
That looks interesting, Can you give me a Welwyn part number. (I was on their website but they have a lot of resistors.)
George H.
resistors
That's where I'm heading with a new design, but top 3 bits are thermometer code to ease the large resistor switching problem.
7/8 is thermometer code, to 63/64 with weighted resistors, remaining 1/64 (minus a smidge smaller than a whisker, got to be precise here :) by power opamps fed by dual 8 bit DAC. 16 bit ADC for feedback.PIC in there too ;) There's an 8 way resistor bank switched in with P-channel FETs to add high/low range. 63 x 47R 17W plus 24 x 6R8 17W, plus a couple LM675T Opamps. Hmm, and a fan, or two, and . . .
Nine teeth left, dentures next month, ouch... Posting before coffee.
Grant.
resistors
But there's a win somewhere? Less noise?
Lovely, glad I'm not going there :) Getting close to a kW is enough for me this year. Plus have the same thing detect a mV/minute drift.
Aim high, see what can be achieved.
Grant.
Looks good!
Is that heatsink live? I can't tell if there's an insulator.
If it is live how did you mount it in the enclosure?
What package are the FETS in? They don't look like D2PAK's they look almost like TO-247's only SMD.
It does look nice though.:-)
Yes. There are 32 fets, 16 P-channels and 16 N-channels, and all their drains are hard-clamped to the heatsink.
Two black Delrin blocks on the bottom, visible in the pic, are bolted to the bottom of the chassis. The holes in the top of the heatsink engage three plastic pins on the top cover, to add a little stability.
They are basically TO-247s without holes, clamp mounted. We just clipped the leads short and "surface mounted" them.
It's a beast. I prefer to not make anything that could break your toe if you dropped it. The power transformer in this thing weighs more than one of my employees.
ftp://jjlarkin.lmi.net/PP5.JPG
John
I thought that was just part of the heatsink. My solution probably won't be so eloquent. I'm probably going to have to use some high temp automotive type adhesive to mount a non-conductive block of whatever I can find to the heatsink. Then use that to bolt onto the chassis.
Once I see the heatsink I may come up with a better idea.
They sold seven more of the same heatsink today.
Yep thats a back breaker alright.
The one we actually use in production is WDBR2-12RKT, 12 ohms, 200 watts. It looks like the one in my pic. What's slick is that the springy steel substrate is curled so that it really hugs the heat sink when you torque it down. They seem to handle high pulsed power very well. We used them to replace mil-type metal-case wirewounds that were dying from cycling fatigue.
John
Drill & tap heatsink and mount with M5, 3/16", or 1/4" nylon bolts and spacers? Depends if you doing one off, or production.
Depends how heavy the assembly is. That one I put photos of up uses
1/8" pop rivets and washers to connect PCB-like material as insulator to aluminium angle, easy for a prototype live heatsink, but it would break if dropped or pushed too hard.Must be good then :)
Good to know, thanks.
Grant.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.