Multi-kilowatt blaster in development

[wg_ski]

This is the high power blaster I've been working on designing. It's a relatively straightforward 3-tier class H, intended to deliver
something approaching 2kW/ch at 2 ohms. Why? Amps of this nature cost $1500 and up, and still don't really like 2 ohm loads so they're really more like 3 grand. It's intended to run on power supply components I got cheap in quantity - three 32.5-0-32.5 900VA trafos in series, with 68kuFx6 in reservoir capacitance. Top rail voltage is +/-138 similar to RMX4050, but with 2.5X the VA capacity in the supply it ought to do quite well. The circuit uses 9 pairs of MJL21193/4. Schematic shows one, but check the pcb layout. The topology is a triple darlington driven by a conventional Lin inside of an op-amp, with values I've tuned in over the years in several big class AB amps and recently verified with a smaller 2-tier version of this one. Rail switches are the same ones found in the QSCs and the clones, but with beefier commutation diodes and TO-247 packaged hexfets. Again, they use stuff I have on hand. Stability problems on the parent circuit have been identified and corrected - the circuit WILL NOT work with a quasi-complementary OPS, and fake MPSA56's should be
avoided :-).

The problems I've been tackling recently on the design have been in manufacturability. Not that this is ever headed for an assembly line, but I have plans to build four this summer, and if I can get more trafos, up to four more. I managed to get it to fit in 14" so the pcb cost would be reasonable. I just got them in last week - about $27 each from Sunstone. The design had to use a heat sink arrangement that can be obtained surplus and machined on a table saw. And it all has to fit in 4U. It is intended to be used with a daughterboard front end which couples to both channels and does the housekeeping - filtering, clip limiting, bridging, and the balanced inputs. This board is not yet fabbed, but will be shared with some other designs I've been working on. Another board, to be designed, will handle all the protection with the exception of SOA limiting which is on the main board. Main signal and ground go through the center of the board, while the HS's at rail potential go down each side. The thermal compensation is the same as Leach used, with two diodes on one heat sink
and two on the other, so differences in temperature do not cause runaway. I have found this to be essential on previous designs. The two boards (one for each channel) will be stacked with insulators between them, and between them and the chassis. The chassis will use two fans - one to suck air down one side and the other to exhaust it out the opposite side with flow directors forcing the airflow along the heat sinks. I have used this arrangement before with good results.

Note: This project is under construction and subject to change if there are problems. It is not for first time builders as the results of mistakes can be hazardous to persons or property. When all is said and done, final schematics and gerbers will be made available. Bigger and badder things are in work, but this stepping stone will be crossed first.

[wg_ski]

Main amp schematic - some may recognize this....

[wg_ski]

High side switches....

[wg_ski]

Low side switches.

Going home now to work on some of the manufacturability/chassis issues....

[astouffer]

Do you plan on listening inside a football stadium? You might as well look into a 3 phase feed...

[Corax]

Hi wg_ski,

this baby turns me on

I'm also interested in building such a 'mean machine'.
A rail switcher was also what I admired, 'cause of the relatively low wast of power or, in other word, the efficiency.

And to "look into a 3 phase feed" (what astouffer mentioned) was anyhow my way of powering it, but still with the option to use 2 phase feeds.
Well this would require three transformers (as you already mentioned), but hey, no work no fun.


And now arises a question:
When I take a glance at your main schematic, I see darlingtons made up of three transistors: 2SC4793 + MJL21194 + MJL21194 and 2SA1837 + MJL21193 + MJL21193, which give 6 BE-diodes in series. For the idle current stabilization circuit I see just 6 diodes in series which are in parallel to 4 BE-diodes of the 2SC4793s + 1k + 2k trimmer.
What worries me are the 6 diodes in series?!?
Considering a necessary voltage drop across the emitter resistors of the power MJLs, there will be no chance that an idle current will flow through them. Without those 6 diodes it would work - but right now I have my doubts. However it's still under construction, as you said yourself.

Furthermore I'm going to dig in my QSC schematics and let's see what I can figure out myself.

[wg_ski]

The six diodes (1N4148's) are the failsafe in case the vbe multiplier opens up. They are mounted to the PCB directly. It may take seven, to be determined by experiment. The idea is to have the the multiplier set the current and if something goes wrong the stack will clamp it. The ones in the collector circuit (diode connected transistors) of the vbe multiplier get mouted to the heatsinks - one at each end of each sink. Wires can get broken off, and if that happens the bias current should get railed short of two amps per transistor. A lot, but it can be shut down before anything really nasty happens.

You won't find this in the QSC circuits (but the rail switches are).

[MJL21193]

Hi wg-ski,
You have been up to no good! More power than me...!

Neat. For power supply, I don't even know if this would work properly, but it something I've been thinking about: have you considered a full bridge doubler? A single, high VA 50-0-50 would give the same voltage as your 3, 32-0-32. Have to be 1.5-2kVA, no?

A thought.

[Corax]

@MJL21193:

Are you serious about considering a full bridge doubler?
At these high power reatings a 'doubler' isn't such a good idea. For auxiliary voltage supplies I wouldn't mind but not in a high current power path.

The use of 3 transformers is (in particular) a good idea when using 3 phase star connected feeds, with the star point connected to 'neutral'. This way you can switch between 2 phase and 3 phase feeds. For two phase feeds all three transformer are paralleled at the primary side.


@wg_ski:

Besides using heavy-duty toroidal transformers I'm considering to use (for simplicity reasons) a (no-feedback) SMPS with PFC.
This way it might be possible to keep the whole amp below a weight of 30 pounds. And more interesting, for global use the input voltage would be of no interest (wide range input, i.e. from 90...240VAC). I've a particular design in mind and if you (and other too of course) are interested I'm going to share my thoughts and we can discuss the (final) design based on wishes, ease of design, reliability, and so on.

[peranders]

Just curious: What will be the use of the amp? 2000 W in the living room or as PA? If in the living room, how high audio level do you want to achieve?