Relatively high power amp: any designs?

If, say, I wanted to build a 300W-ish @ 4 ohms AB power amp for bass guitar applications, are there any proven designs I could look at? My requirements are:

  • headroom and good (graceful, gradual) clipping behaviour
  • reliability and ability to sustain long periods at full power
  • passively cooled

My idea was to replicate the Gallien Krueger 800RB power amp, which satisfies these criteria, but I'm OK with looking for alternatives.
You might look at this:
Lots more on that site too!

Anyway, that design will push up to 500W continuous into 4 ohms if you build the "full" version with lots of parallel output transistors. Clipping behavior is probably not what you want, but if you build a higher power than you need and put a line-level clipper before it...

Passive cooling simply comes from having a massive heatsink, regardless of the circuit. Class AB is about 70% efficient at exactly full power, and reduces from there. Not constant input power (that's Class A), but not a close match to output power either.
I use Elliot's site as a educational reference , but the amp projects are quite on the edge.
For instance project 68 is overcompensated , no base stoppers to promote sharing on such a large output stage. The 68 input stage is
nowhere near hi-fi and would not clip gracefully. No fancy clip-proof VAS , no decoupling on either the EF3 or CCS ......
And just about every other "no-no" that I learned not to do a decade ago. Yikes.
PS - project 68 would also clip (very) asymmetrically ... a bootstrapped high power amp ??? Double yikes !!
It probably does have the proper clipping behavior if it’s properly built. Or at least similar to that GK amp. You don’t need perfectly symmetrical clipping down to the mV or total lack of microsecond-long rail sticking. What is needed is a truly flat top without a bunch of intermodulation and ripple on top of it, and maintaining a low output impedance when clipping. The first is a function of the POWER SUPPLY, not the amplifier circuit. Don’t undersized the trafo and use 22,000 uF caps. This will maintain a low impedance down to about 8Hz at 4 ohms. Any supply dips will be long duration with respect to a clipped 41 Hz half sine, so they add nothing audible. For the amplifier itself to maintain a low impedance output when clipping, the supply itself needs to AND use a high current gain output stage. Use an EF3 or equivalent, problem solved. If following project 68, ditch the CFP drivers and just use EF3 with the same parts - one less reason to go slinging things into the wall when it oscillates.

When I use a PA amp for mobile disco and it’s got the proper characteristic, it can be run 6dB into clip continuously and it doesn’t sound “dirty” at all. You don’t even realize it’s clipping at all till you see the clip lights. Use something of ”equivalent” power and undersized caps, output stages with voltage gain (ie current source output) and it sounds like a mistuned FM radio station by comparison.
Hmm...I thought I'd understood all of it, but it looks like you're right. Class AB would have an adjustment pot (or tight tolerance parts) to set the quiescent current, and P68 just has the two diodes to roughly account for the drivers' Vbe. Since this is DIY, what happens if you slip a trimpot in with D2 and D3 anyway? (Figure 1 on that page) Slightly spread them apart so they conduct a little bit more.
Joined 2017
Paid Member
Rod's designs aren't for everyone. He favor's simplicity above all else. This will appeal to some, but not everyone. I believe his P68 project is designed for subs. If you like his "keep it simple" approach, the P101 might be better.

On the other end of the spectrum, the Wolverine might interest you: DIY Class A/B Amp The "Wolverine" build thread

It is an ultra-low distortion EF3 design. It's far more complicated - but also very well designed. The feedback from builders has been very positive. There's also a very active community supporting it.
  • Like
Reactions: 1 user
I advise you to modify the bass amplifier for more powerful bass.
Designers consider the ground to be zero impedance and it isn't. The positive input being referenced to the ground makes it still much higher impedance due to positive feedback, that weakens the bass. This why you need armada of reservoir capacitors, not to keep the supply voltage sufficient, only to lower the impedance. A simple modification by getting the positive input non referenced to the ground, decreases dramatically the impedance, and gives full bass as tube amps do.
To modify the above gallien-krueger amp, the resistors R1 and R28 instead of being grounded, they get each virtual ground made of 2*10k resistors from the +/- supplies along a 100 ohm adjust for the offset. The virtual grounds get bypassed by 2 100nF capacitors to each +/- supply.
Dc output protection needed if there is a power device failure.
Short circuit protection for output shorts in field operation.
600 watt power supply needed for 4 ohms which needs soft start
from surge current.
Would recommend many common PA amps with triple Darlington
Most MI amps start using forced cooling over 200 watts.
If you want steady 300 watts no forced cooling. Be many
output devices and large heatsink.
Agree Class D is garbage for bass and plenty of shut down
amps can be had used.
Amp clipping is amp clipping. Is it stable and wont latch is the question.
Usually why most MI amps have a limiter on input.
I would look at Hartke designs. If you gonna copy
a class G like GK. then it needs a limiter since class G or
a rail hopper is not very stable clipping. the so called GK grind
is the limiter.
Last edited:
The trick to getting a class G to clip cleanly is to size the power supply as you would a normal class AB which has clean clipping. Almost all class G or H use further undersized power supplies (even more so than class AB) because the average current drawn in normal operation is less - usually about half. In full sine wave operation, the reduction is only like 10%. Undersized supplies on class AB amps result in the same “grind”. Done right, you use the same size transformer and caps as you would a Levinson, Krell, or old BGW of equivalent wattage. And the same number of output transistors. But a smaller heat sink CAN be used, even without a fan. Get the power supply such that it’s voltage can only change at a single-digit-Hz RATE, and clipping will be clean. If the supply voltage can change by more than a volt or two at an audible RATE, it introduces audible artifacts on a clipped waveform - and that’s in addition to the normal odd harmonic spectrum. The worst offenders I’ve heard are subwoofer plate amps - cheap ones make it sound like the woofer is bottoming out and hitting the backplate when they clip.

The common PA amp with the triple darlington output is the right topology. It’s easy, it works, and has the right characteristics. You can put in any amount of refinement you see fit, but even the most basic is .05%-capable. Bias the outputs at 3 mV across the emitter resistor and crossover distortion is low enough NOT to be a problem. Done. And due to the high local feedback factor, will maintain a low output impedance (for loudspeaker damping) even if the global loop is NOT in control. Which is what happens at clipping.
No class B. Put a small bias current in the output stage. 10 miserable mA per transistor (3mV, 0.33 ohms) is sufficient if using an EF3. That’s also part of the reason TO use an EF3 - it takes less mA to make the crossover distortion disappear compared to EF2. And you can get away with even less if you use a high-feedback topology with an op amp input stage.