900W H-class PA Amp with Limiter

This circuit is for example not for DIY.

That doesn't mean it can't be done.... An experienced enough DIYer can follow an example and make it work. Usually, out of what he has on hand.

The circuit looks to me like a Crest front end, Dirty Harry's output stage, and QSC-type rail switches (they would have used 2117's if they were available way back when they did the MXa series).


The buffered 311's are plenty fast enough for rail switch drivers. QSC intentionally keeps the slew rate down to about 30V/us on the rails, because it's less prone to misfiring or other misbehavior. You can still do that with a 2117. A high speed circuit is much much much more sensitive to layout and produces weird unexpected failures. Dissipation goes up somewhat at 20kHz if you slow the switches down, but trust me you don't want to put 3kW into a compression driver. It's too loud, painful, and costs money.
 
hi wgski
greetings no i want to make the 900 watt one pure coincidence yesterday i found
your cascode zip looks very powerfull is there any pcb pattern for it even your amplifier
in casing looks powerfull one with 3 transformers and blue caps how many watts is it please
thanking you
andrew lebon
 
Like Crest front, and Crest output stage...
QSC MX3000a
 

Attachments

  • mx3000aa.jpg
    mx3000aa.jpg
    697.8 KB · Views: 8,852
hi wgski
greetings no i want to make the 900 watt one pure coincidence yesterday i found
your cascode zip looks very powerfull is there any pcb pattern for it even your amplifier
in casing looks powerfull one with 3 transformers and blue caps how many watts is it please
thanking you
andrew lebon

That's not a low-dissipation amplifier (just class AB not H). I built three of them, but the PCBs were hand-drawn directly on Cu-clad so not too reproducible. They were 1500W at 2 ohms and would melt the plugs off the extension cord. I have since been working on class H, TD, and D designs, with several class H in various states of completion. I have PCB designs for 2 and 3 rail class H amps - I have 8 of these board populated, but the rest of the chassis are still being built.
 

Attachments

  • KC_4kw.PNG
    KC_4kw.PNG
    48.3 KB · Views: 6,747
very interesting rail switching scheme, I was planing to employ your rail switching scheme on this amplifier, this one has a rail voltage of 0, +/-75VDC, +/-150VDC roughly 1000W at an 8 ohm load. too many parts I guess.

That one does have too many parts. A spearate commutating diode for each output????? That seems silly.

One other problem with that design is that the switch driver transistors are going to run out of gas. The voltage drop in the divider string will become excessive resulting in a loss of potential output at low Z. Either use a triple for more gain or boot strap it so it can go a little above the rail and use MOSFETs for drivers.
 
very interesting rail switching scheme, I was planing to employ your rail switching scheme on this amplifier, this one has a rail voltage of 0, +/-75VDC, +/-150VDC roughly 1000W at an 8 ohm load. too many parts I guess.
To many, and not adequte parts(2SA1943/2SC5200) and circuit for this power (Wg-ski) , see Crest 9001. IRF540 (Vds 100V) can rail 75V (Mackie SWA1801) but you must use all parts double at list and MJL4281A/MJL4302A in output stage.
 
but you must use all parts double at list and MJL4281A/MJL4302A in output stage.

No, you use MJL21193/4/5/6. Double the SOA of even the 4281. And no, I don't give a darn if the switchoff distortion is a little worse because of the "low" fT. It's a kilowatt PA amp, not something for a critical listening room. The commutation noise from the class H stages will far outweigh crossover distortion unless it's grossly underbiased. So use those big honkin' Motorolas and be done with it.

See Crest CA18. The original parts are 10 MHz fT (not 30,40,50 or 80) and can be replaced quite well with MJL21193's.

It is also typical of amps operating from +/-140V upper rails to use only six C5200's. See RMX5050.
 
FETs just don't seem to work well above +/-100V rails for a number of reasons. If you want "complementary", and need something bigger and badder than the IRFP240/9240, you're SOL. The driver circuits for all-N-channel outputs are more suited to switching than linear amplifiers. Paralleling more and more devices results in stability issues that just don't happen with bipolar emitter followers. Even with class D, there is a significant penalty in device performance going from 200V FETs to 250's, and above 250V it gets even worse.
 
That one does have too many parts. A spearate commutating diode for each output????? That seems silly.

At first it does, but done that way there is no need for emitter resistors for the upper rail transistors. So you get to choose - one diode per transistor pair or one emitter resistor per pair plus a big common diode that needs to be heatsinked. If each pair gets its own diode you should be able to get away with large axial types without extra heatsinking, like what is used in that circuit. Increase airflow until cold enough :D
 
It's a matter of "class G" vs "class H" rather than BJT vs MOSFET.

A high power "class G"/linear lifter will need more than one high rail transistor and emitter/source resistors (or separate diodes) irrespective of their type as the rail lifter needs to dissipate a lot of power (but the stress on the inner devices is lightened).

"Class H"/rail switcher circuits like yours get away with fewer high-dissipation transistors and can often use only a single switching FET in the rail switch so they are a bit more economical, especially for high powers. :) I like your use of IR2117 in the rail switches: simple and low component count! :)
 
Last edited:
I like to say that I have best relation power/square inch of PCB :devily: There is more space for PSU:)

You cheat. You used a jumper wire between the NPN and PNP banks. Just make sure you don't run it parallel to or close to your feedback resistors or input trace. :) And I wonder if the power traces need to be wider.... Remember how many amps are being pushed through them. Even if the average heating isn't an issue with 1/8 power pink noise, it's just evil to lose voltage swing in a PC trace.