80W HiFi Audio Amplifer

Nice amplifier

Similar projects worth following
I'm building a fancy high performance amplifier based off the examples from the book "High Power Audio Amplifier Construction Manual" by G. Randy Slone. It's gong to be my first HiFi design. It is expected to have THD of 0.001% while delivering 80W to 8 Ohm speaker

The HiFi Amplifier consists of three stages: input stage, voltage amplifier and power stage.

Input stage

Input stage is a differential pair of Q1 and Q5. The pair balanced through the current mirror made of Q2 and Q4 and degeneration resistors R1, R2, R5, and R6. The current of 0.65/150 = 4.5mA is provided by the current sink Q3. Since differential branches split the current equally, there is 2.25 mA on each branch. It means that transconductance of the differential pair transistor is gm = Ic/UT = 2.25mA/25mV = 90 mS.

Voltage amplifier stage

The VA stage is a classical Darlington pair with a current limiter through Q7 set to 0.65/47 = 13.8 mA. The current is supplied by the current source Q9 and is set to 0.65/100 = 6.5 mA.

The VA stage contains a double pole compensation through RC network of C7, C9, and R13. The role of the VA stage is to convert the current from the input stage into the voltage output that will be seen by the load. Therefore total voltage gain is input transconductance times Darlington transimpedance. And Darlington transimpedance is, err, something really high. Open loop voltage gain is actually very high and hard to straight forward calculate but luckily not that decisive. The overall gain is determined by the feedback loop, anyway (later in this post).

Output power stage

Output power stage is basic Class B amplifier made of high-power transistors Q16 and Q17. These transistors are pre-driven by lower power Q14 and Q15. Combined, they form something called Sziklai pair, that improves current gain of the output stage.

Power transistors are biased by collector follower Q10, R18, R19, R21, and RV1. Q10 has to be mounted on the heatsink together with power transistors to ensure thermal tracking. What? When the high power transistors warm up, their dynamic properties will change (i.e. output current). Bias must be able to follow this change, and not them slip out of control. So, when output current start increasing because of silicon warming up, Vce of bias will start dropping and reduce this current. This keeps transistors in the safe operating area.

Output current must be monitored and limited by power resistors R33 and R34 and transistors Q12 and Q13. Monitored voltage is scaled by voltage divider R26/R29 (or R27/R30) by factor of 0.3. When the current reaches the value of 0.65/0.22/0.3 = 9.6A, Q12 (or Q13 in negative half-period) shall tie the predriver transistors base to ground and keep the current limited.

R25 - R28 are here to provide something called single slope current limit. What is this? Our maximum allowed power dissipation is 150W, that means product of Vce and Ic must be 150 all the time. In the Ic - Vce diagram, this looks like the 1/x curve (Ic = 150/Vce). We're ignoring the secondary breakdown curve here. Current limit will keep transistor in the rectangular area under the Ilim. If Ilim is 9.6A then the maximum VCE allowed is  15.6 V.

But if VCE get's higher than 15.6, let's say that it climbs to the value of 20V, then max allowed current will be 7.5A, a value that current limiter will easily allow, and lead the transistor into the destruction. To avoid this, a bit of current is always fed to the bases of Q12/Q13 by resistors R25 and R28. This way, even for smaller output currents than 9.6V, the Q12/Q13 will start to shut down.

Rest of the circuit

To match the speaker's impedance, a Zobel network (R35, C12) and air coil inductor L1 was put.

Global feedback is fed from the output to the negative input of the differential pair via network of R10, R11, R12, C6, C8, and D1. In the middle of audio spectrum, C6 is short and C8 is open, what means that feedback gain (or beta) is 330 / (10k+330) = 0.032. Overall gain, determined as 1/beta is therefore 31.3 or 30 dB.

The HiFi Amplifier consists of all discrete, through-hole parts. It's basically...

Read more »

  • 1 × 2SA1302 Discrete Semiconductors / Power Transistors and MOSFETs
  • 1 × 2SC3281, Discrete Semiconductors / Power Transistors and MOSFETs
  • 4 × 2SD669 Discrete Semiconductors / Transistors, MOSFETs, FETs, IGBTs
  • 2 × 2SB649 Discrete Semiconductors / Transistors, MOSFETs, FETs, IGBTs
  • 5 × 2N5401 Discrete Semiconductors / Transistors, MOSFETs, FETs, IGBTs

View all 6 components

  • Power supply

    Pero7 days ago 0 comments

    What would be an audio amp project worth if we weren't playing some music on it. I hooked up my newly acquired cheap loudspeakers and the best type of preamp (a single potty) and got some crystal clear sound out. No humming, no distortion, no screeching. Not even those sound blasts during the power up. Even my more audiophilic friends said the sound is pretty good. Here's some screenshot from Rigol's scope.
    Read more »

  • Biasing and oscillations

    Pero03/08/2019 at 16:15 0 comments

    I've found a cheap loudspeaker which I could use to test my amplifier. It is 8 Ohm/50W speaker, so not exactly what I designed my amp for, but it was cheap and worth trying out preliminary tests. 

    When I plugged it in, and powered the amp, I could see somethin weird. On top of my clean sinewave, there was some kind of parasitic hf signal, not seen before. When I zoomed in, I could see it is a 3MHz component. 

    Zoom in on HF glitch:

    At first I was confused - where did this come from? Can the speaker introduce some resonance into the amp. Why is all of a sudden so instable. 

    Then I remembered that high frequency oscillations usually originate from low value parasitic L and C elements. And one serious parasitic capacitance between BJT base and collector usually contributes to the amplifier instabilities. As someone here on Hackaday said in comments, these kind of oscillations are preventale by the proper bias of the output stage. Indeed, when I turned the potentiometer, I could see the oscillations going away. Amazing!

    Now, I plan to make a simple pre-amp to test some serious music on this loudspeaker.  

  • Preliminary power test

    Pero02/28/2019 at 16:06 0 comments

    Instead of loudspeaker, it is better idea to try the amp out with  lower cost power resistors. Luckily, in my spare parts room, I've found some of those thick 30W resistors with a nice golden colored casing, I've combined them together to 4.7 Ohm and 9.4 Ohm load and powered the amp on. I was still using Rigol DP832A bench supply at +/-30V. 

    Read more »

  • Getting rid of counterfeits

    Pero02/27/2019 at 10:48 0 comments

    Apparently everyone in the world knew that 2SC3281/2SA1302 are most likely fakes. Toshiba doesn't produce these as of the year 2000 and originals are nowhere to be found. Again, ESP has a beautiful post about debunking counterfeits:

    Following the advice of many good people on Internet, I went on and purchased MJL3281/MJL1302 by On Semiconductor, from a reliable source (RS). This morning, I've tested the circuit with the new transistors and it worked like charm! Next step, finding a 100W speaker and testing out power performance.

  • ... and first fails

    Pero02/20/2019 at 15:44 7 comments

    After assembling output power stage, things went wrong. Somehow, I managed to burn high power transistor Q16 (2SA1302). After plugging in the circuit to +/-30V, my bench power supply indicated shorted and I knew immediately that something's wrong. After turning the power off, I measured 0Ohm between pins of the Q16. It was dead. For some reason all other transistors were well and alive.

    I stared at the symbol, schematic and layout for one hour. I could't see any obvious error. Could it be that I just received bad chips from ebay?

    As a comfort, here's a nice photo of the fully assembled board:

  • First tests

    Pero02/20/2019 at 09:54 0 comments

    I assembled the Amplifier board up to the power stage. Why? Before plugging high current in, I wanted to make sure that circuit is stable and nothing will burn when I power it on. I found a bench power supply that can deliver +/-30V (Rigol, two channels, forgot the model name), a function generator (also Rigol) and an oscilloscope (you're right, a Rigol one).

    Read more »

  • Simulation

    Pero02/18/2019 at 10:51 0 comments

    I used LTSpice to examine inner workings of the circuit. I've made a block diagram of every main part of the circuit to simplify the schematics. There you can see an input diff pair stage, voltage gain stage, a current source that feeds them both, output power stage and a bias for the output power stage. I've even put a passive feedback network into the separate block.

    Read more »

  • Power calculation

    Pero02/14/2019 at 16:49 0 comments

    Let's see what kind of trafo and heatsink to I need for this amplifier.

    • As stated before, the average output power is expected to be 80W at 8 Ohm load. During one cycle, power will be delivered from positive supply rail half the time, and from negative supply the other half, meaning that each rail will have to deliver at least 40W.
    • This implies effective current of 3.16 A, or 4.5 A peak , if we pretend that signal is pure sine (which is not). Power supply at 42V delivering 4.5 A peak half the time, gives away on average 42*3.16/3.14 = 60.2 W.
    • Transistors will, therefore dissipate 60.2 - 40 = 20.2 W each.
    • We re talking about efficiency of 66%, what is reasonable for Class B/AB amplifier
    • Knowing that maximum junction temperature is 150°C, the maximum allowable thermal resistance is (150-25)/20.2 = 6.18 K/W. This means, that Rth of junction-to-case, case-to-heatsink and heatsink-to-ambient combined must be less than 6.18 K/W.
    • Transformer I need for this amplifier is then 2x60.2 = 121 VA minimum, with a +/- 45V secondary voltage and a center tap.

  • Orders placed

    Pero01/21/2019 at 14:57 0 comments

    Finished layout and ordered a PCB from the JLCPCB. The PCBs are two layered, 10x16 cm and 5 pieces cost about 13€ all together.

    The power BJTs are not available from standard distributors, so I ordered them from e-bay.

View all 9 project logs

Enjoy this project?



Bharbour wrote 02/27/2019 at 13:20 point

You mentioned finding a 100W speaker for testing. Getting some non-inductive resistors and putting together an 8 ohm load that will handle 100W would be a good first step. I have set speaker cones on fire when things went seriously sideways with a power amp that I was working on. Resistors are a lot cheaper than speakers... Even a clothes iron (120v) can be used for some test loading. Good Luck in any event!

  Are you sure? yes | no

Pero wrote 02/28/2019 at 15:10 point

Yeah, makes sense. Luckily, I did find some power resistors in the spare parts room. So far, nothing has burnt, but I'll need to find a better heatsink!

  Are you sure? yes | no

roelh wrote 02/15/2019 at 17:37 point

Hi Pero, in the schematic in your gallery (main project picture), it looks if Q9 has its collector and emitter swapped ! But perhaps it is an older version ?

  Are you sure? yes | no

Pero wrote 02/18/2019 at 10:02 point

Hi, it seems you're right, I indeed oversaw it, arrgh! Hopefully, with  TH components, it will be easy to intervene and swap the pins

  Are you sure? yes | no

Richard Dudley wrote 01/26/2019 at 13:51 point

When using two pole compensation its worth checking out the clipping behaviour. You might find a need for some kind of clamp to minimize 'sticking'.

  Are you sure? yes | no

Ted Yapo wrote 01/24/2019 at 03:34 point

I'm watching this with great interest! I've made a few designs based on Douglas Self's books. It's great to see people still designing amps like this in our current class-D dystopia :-)

  Are you sure? yes | no

Pero wrote 01/24/2019 at 21:01 point

Thanks! I hope it'll turn out good!

  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates