135W Class AB amplifier

I started with this almost 10 years ago, and it started off as a 120W design. A friend of mine built it for an uncle (his one produced about 150W), and it lasted about 8 years until a massive grid surge sent thousands of volts into the amplifier (which was off), and that fried the amp and KEF IQ9s. Sad stuff. Anyway, the amp was abused daily and survived easily. My version has gone through a number of iterations, the last being awesome. I was in the process of constructing the box nicely when I connected the amp outputs to each other rather than the amp grounds. Bit of smoke and the project went into the cupboard. That was about 5 years ago. Since then I've gotten married, had a child and created the MAK, so the amp has been waiting. I've started on it now. The new design is based on the old one, except I use PNP input rather than NPN, but otherwise it's very similar. The amplifier stage has a bit more current, but that's about it. Here's the schematic:

TIP35s can be replaced with 2SC5200 and TIP36s with 2SA1943. The TIPs work perfectly though. Q4 and Q5 must have small heatsinks. Not that they need them, it just makes the bias current more stable. The output transistors must be mounted on a big heatink (0.6K/W ideally), and I suggest to mount using some sort of metal bar which hold the transistors down rather than screwing each one in. Then Q9 must be mounted on this bar. This ensures thermal runaway can't happen. Also, if you mount Q9 on the heatsink, then the bias current will drop to zero. Q7 must be on a 20K/W or better heatsink. If the heatsink is smaller, the DC offset can be affected, and the same cause will also make noise rejection worse.

Make sure the trimmer's resistance value is as small as possible (before turning on), then increase its resistance very slowly until you get to around 30-100mA. I use around 38mA for my amplifiers. It seems to be an easy value to get to and keep. Either way, if there's no bias current, I believe the amplifier is fast enough to take care of crossover distortion. I'll never know because I'll never test that. I don't have the time.

I like to use a simple non-inverting input op amp circuit for the input before the amplifier - I find this controls the amp nicely. I use about 2 gain on this, so it is a preamp. It makes volume control work better. I'll post how to do that sometime. And I definitely recommend NE5532 op amps - very very nice.

I'm building this at the moment. I'm using one 300VA transformer (toroidal!) per amplifier, each with it's own bridge rectifier and 2 10000uF 100V capacitors (snap in terminal type). One 500VA transformer will work fine too with 2 10000uF capacitors in total. The transformer must be 35-0-35. I say toroidal because it contains the flux. You can use standard transformers, just shield them from the amps.

30W Class AB amplifier

This amplifier has been simulated extensively, although I haven't built it yet. I am saving building this one for a time when I can spend good time building and making sure the wiring is going to be optimal. I am going to use this amp in a bi-amp situation with passive filters on the input. This will result in two amplifiers that amplify the highs and two amplifying the lows (for stereo) and thus I can achieve an effective power of 120W into each speaker.

If you want to do the same, you must design the filters yourself. I'm going to be using Celestion F20's which have a crossover frequency of 2500Hz, so that is where I'll set my design point.

This is the power supply I will use. It should do fine for four amps, but if not, double up on the capacitance. The earth circuit came from here:

I previously used a Class A amplifier (11W or so) but the amp wasn't stable enough for me. It sounded incredible, but I think that the lack of stability could have caused damage to my speakers in the long term. This amp theoretically produced 24W (CXI), but in practice it produced 11, which is loud, but considering the heat dissipated and the cost of everything, it wasn't worth it for an amplifier as unstable as that. Now I will use a Class AB amplifier with the same heatsinks, output devices and transformer and I will achieve lots more power, although at the expense of no Class A amplifier.

To set the above amplifier up, set R1 to max and R12 to 0. After doing this successfully, power on the amplifier. Set R1 so that the measured output offset is between 30 and 100mV. Once set, adjust R12 slowly to achieve a quiescent current of around 120mA. Keep checking the quiescent current as the amp heats up as it might change due to voltage drop changes in the output devices caused by heat. The heatsinks should be 0.6K/W or less for two amplifiers. If all is well after a while, enjoy. These steps are extremely important! If R12 is too high, the output devices will be destroyed. If R1 is too high, the offset could damage your nice expensive speaker.

Q10 and Q11 are for short protection. If you want to trade the risk of ugly clipping for the risk of a destroyed amp, you can leave them out. This will probably be wise if the amps are built in. The short protection will probably not affect the sound of the amp, but it could at extreme power. Just know that touching the output leads could destroy the amp if these transistors are left out. I estimate that you could short the output for 1s or less, so don't overestimate the protection; act as though it weren't there.

Frequency response: 10Hz - 100kHz flat
THD: Should be completely inaudible
Gain: 30.37dB (660mV input for 30W into 8 ohms)

10-14W Class A amplifier

I have built this amplifier and it does sound good. It requires a preamp as it hasn't got much gain. It doesn't sound very good when it distorts, but one would learn what level of sound would cause clipping. 10W is more than it sounds... It really is enough for everyday use. Class A also does sound very good. This isn't the most efficient amplifier ever - quiescent current is 1.6A.

Now it is inefficient and it requires big heat sinks and a large transformer and a great power supply and careful wiring, but in the end it is extremely simple and it sounds very good. The zener diode rejects any ripple coming from the power supply, But you still only want a ripple of 10mV max. The ripple reaching the input is amplified, so the zener gets rid of that, but whatever ripple there is will still reach the power stage.

If you want to build this, you're going to need to find a good power supply design. As I said before, I have built it and I am very satisfied. This amplifier is ideal for a bi amped system as the high driver's amplifier. It is stable (as any amplifier should be), but this does mean something to me as many of my old designs have been unstable. This is by far the easiest and most stable amplifier I have ever built.

To set it up, set the 470k trimmer to maximum resistance. Measure the current into the amplifier and slowly set the resistance of the 470k pot until a reading of between 1.65 and 1.7A is obtained.

21W Class AB amplifier

I designed this amplifier because I need an instrument amplifier for monitoring my music when on stage, etc. I sucked 21W out of this little design. Previously I had a decent 10W amplifier (RED Free Circuits), but we blew that one somehow. Now I will put this into the old box. I haven't built it yet, but the simulations say it works as I designed. In this design, wiring is important due to no differential amplifier. I might need to add ripple rejection with a zener as in the above design.

It is a simple design, and simple to set up. Set the 10k trimmer to about half way and the 470 ohm trimmer to 0 resistance. Power up and set the quiescent current to 30mA by trimming the 470 ohm trimmer. Once done, measure the offset voltage. Trim the 10k so that between -30 and 30mV is measured. The amplifier should now be able to pump out just over 21W.

Use heat sinks that are bigger than you would normally use for 20W. If you don't want to do this, omit one of the bias diodes and change the Darlington pair to a compound pair like this:

In fact this is probably better in all ways.

Here is the power supply...

Couldn't be simpler. If you put a switch on, then put it in series with the live wire.

Frequency response: 2Hz - 17kHz flat
THD: Once again, very low (I haven't measured it yet). It won't be audible
Gain: 29.87dB (600mV for 21W into 8 ohms)

The gain can be modified by changing the 100k feedback resistor to a 470k resistor and the 10k in series with the input to a 33k. This will increase the input impedance significantly and the new gain will be 21.58dB (~1.6V input for 21W into 8 ohms). This suits my application for the amp. Just remember that changing these resistors will change the output offset voltage, so if you're going to experiment by changing the feedback loop, do it with a cheap, single driver speaker if you're experimenting with a speaker connected. Also readjust the offset and quiescent current!

No comments: