Oil-Immersion Cooling of TO-3 metal transistors Mineral oil USP flash point 112 deg C, it can vaporize into air enough to explode. Commonly used as a fuel for candles, 95% of gel wax is mineral oil and 5% is polymer resin. Mineral oil is a product from petroleum, a long-chain alkane of carbon & hydrogen only. From MSDS, DOES NOT REQUIRE ANY HAZARD WARNING LABEL UNDER OSHA HAZARD COMMUNICATION STANDARD. If use generates a mist, do not breath it. Laxative for over 6 yrs old, up to 3 tablespoons per 24 hrs, it is little absorbed in the intestines. Baby oil is perfumed mineral oil. Mineral oil is an insulator. Specific heat is 1.67 kJ/(kg deg K) vs. water is 4.19. Half a liter of mineral oil will heat up 9 deg C in 60 sec when 100W is applied (6kJ). .8g/milliliter, .4kg/half liter. May need to stir the mineral oil with a little propeller & small DC motor, especially when an ice bath cools the oil, maybe in proportion to voltage applied to load.
Ap 1, 2015: Power as f(t) is more & more important as I think about protection. AD633 is $8, Atmel CPLD might handle SOA lookup based on 3-bit I & V, or Mega2560 at $55 from Jameco might be better for SOA because it might be able to factor in time effects & oil temp. But as a matter of fact, if speaker-terminal short-to-ground protection is provided (by sensing base current of the paralleled output transistors), SOA seems to be no problem, if some resistance is provided down onto the output transistor collectors with 15,000 uF bypass, & if a circuit across speaker monitors for overload (voice coil overheating) & uses a relay to put in some resistance in series with speaker. But Arduino might be useful to model voice coil temperature & do the speaker protection. Or the voice coil might be simulated by one of the 4uH stability coils in series with the voice coil, made of thinner wire & placed in a confined place (limited air flow) to model the voice coil temperature & temp variation with time, with thermistor measuring the coil temp & providing a signal to pull in the relay & protect the voice coil. This really is a neat idea. See more comments on the web page https://sites.google.com/site/solderandcircuits/home/teachers--topics-in-common-with-your-teaching/big-speaker-idea-for-2016-career-day/low-level-circuits-description at the March 30 note. But by Ap 13, all plans for multiplying (to get power) are on the back burner, see notes of Ap 14.
Ap 14, 2015: Order Day was yesterday, plus the Marlin P. Jones Associates order came in yesterday with sales on 20 amp protection/sequencing relays, 16VDC supply to power all the relays (after series regulation down to 12V) & fans. Order to Jameco is the big one, for the two 24V 8.3A supplies, PNP power transistors, wire, 15000 uF caps, bright yellow LEDs for cone strobe, line filter, switches, & potentiometers. Order to Mouser is for numerous power resistors and the semiconductors: comparator, LM3886 power IC for midrange, high-voltage OPA551 op amp that goes before the woofer power transistors, inrush current limiter.
A protection-circuits breakthrough was concerning the dilemma about audio overload mimicking output-transistor shorting. I figured out a neat soft-clipping scheme for the OPA551 gain stage, where audio peaks to .3V come through unattenuated (to be amplifed to 12V peaks on the left of the woofer & -12V on the right of the bridge amplifier), but between .3V & .6V, peaks are soft clipped to be amplified to only 15.5V, & above .6V it sticks at 15.5V. With this clipping plan, it opened the way to declaring an output transistor short only if Vce is low when there isn't an audio peak coming through. In combination with voice-coil temp sensing & 24V supply overcurrent limit, this should give good protection.
The voice-coil temp sensing is inferred from the temp of the left 4uH stabilization inductor, It is to be wound with two, parallel runs of #26 magnet wire, plus a little PCB with thin traces & a SMD thermistor. The coil will heat up from speaker current & give a way to roughly sense the woofer voice-coil temp. The plan is to put some light insulation around the 4uH to mimic the heat entrapment that exists for the voice coil (which is down in the magnet gap), & even have a little fan onto the 4uH to mimic the voice-coil cooling that comes from cone motion. This plan opens my thinking to front-panel meter the voice-coil temp instead of voice-coil power. That is a really neat idea.
The voice-coil attenuation, putting 4.7ohm 50W in parallel with voice coil's 3.55ohm, plus 2.43ohm 120W (a homemade resistor with nichrome wire) in series with that paralleled combination, with merely one SPDT relay, is a very effective way to protect the woofer but let the audio program continue with attenuation.
Design of the gain-of-one, 100% local feedback circuit to go between voltage amplification of OPA551 and power transistors, is proceeding. Four choices for the dominant-pole cap are planned, with jumper selection to balance between low crossover distortion & closed-loop oscillation. The calculated cap to center on is .017uF, & it is buffered by an emitter follower just ahead of the bias spreader.
Intervening time is for supply box, see a web page for DC Power Supplies +-24V
Nov 29, 2015: Fabrication of the amplifier box is well underway. Plywood base. A 6"x6" PCB with 1/4" male quick disconnects for power distribution is key to managing all the supply wires from the assemblies. It is mounted in a steel frame that is pinned to the plywood base, & is easily removable. PCB 1 and PCB 2 are piggybacked and contain the diff amps that provide feedback control of the oil-cooled transistors; will soon be powered up & checked out, first with just the left channel to check feedback loop stability and selection of the dominant-pole cap. But real use with 12" or 15" woofer awaits low-pass filter (80Hz or 140Hz) so the woofer isn't subjected to "cone breakup," where high frequencies shake the voice coil but the bulk of the cone doesn't follow.
Other assemblies are done: 24V to 7.5V/5V buck converter, 8ohm headphone amp, midrange amp with LM3886, .3V peak leveling amp, 65.4Hz sine oscillator with amplitude control, light strobe. Control panel is pretty much settled.