Classic 40W audio amp with discrete components

[soldar]

I have posted this already a couple of times and I am now making its own thread.  This is a 40W audio amplifier built with discrete components. Supplied at 60V (single rail) it will output 40W into an 8 ohm load. Extremely simple and works well. I built and sold many of these in the late 1970s.

There were many similar designs in those days. First they came out with a complementary pair (PNP & NPN) directly driving the speaker and later they added a pair of 2N3055 to provide even more power, as in this case.

The power components (to the right of the blue line) would be mounted on heat sinks on the back of the box while the PCB was mounted inside the box right next to the back of the box. Of course the stereo amplifier had two of these, one per channel.

I built so many that I was thinking of doing a modular design where the PCB would plug directly into a connector on the back of the box.  This way boards could be easily exchanged for testing and repair. I never got to do it though.

Unfortunately I cannot find now the design of the PCB but I will post it if I find it in the future.  It should not be difficult to do though and I invite others to design and post here.

The unregulated power supply had a 48Vac secondary which gave about 65 volts at no load and fell to about 58V at full load.  Today it would probably make sense to provide a regulated SMPS. 

[Zero999]

It could do with some overcurrent protection, but that should be fairly easy to add.

Ideally, the V_BE multiplier should also be on the same heatsink as the power transistors.

[soldar]

Elektor September 1975 carried a very similar schematic which it called "Edwin Amplifier" which did include it.

[soldar]

More recently Elektor published what they titled "Sixties-style 40W audio amplifier".

[Conrad Hoffman]

The later ones are improved, but we know how to do so much better now without increasing the parts count by much. I admit to building amps without protection, using the philosophy of "I built it, so I can fix it", but protection is still a good idea.

[schmitt trigger]

For true "sixties style" amps, it would use Germanium transistors. 

[soldar]

For true "sixties style" amps, it would use Germanium transistors. 

True. These designs are "late seventies-style".

The later ones are improved, but we know how to do so much better now without increasing the parts count by much. I admit to building amps without protection, using the philosophy of "I built it, so I can fix it", but protection is still a good idea.

I never built in any protection and do not remember ever having a problem. I may have had but I do not remember it. As you say, I built it and I fixed it.

This simple style with limited part count is as far as I am willing to go to take advantage of available components; once you get into designs which use over a dozen semiconductors then I start thinking it's not worth the trouble and I might as well buy something ready-built (unless it is something not commercially available).

Check out the attached design. May be good but it is over my limit.

[Gyro]

The later ones are improved, but we know how to do so much better now without increasing the parts count by much. I admit to building amps without protection, using the philosophy of "I built it, so I can fix it", but protection is still a good idea.

Yes, protection is a good idea. For me, it's not a case of protecting the devices in the amp but much more about protecting the speakers. I'm very wary about what I connect my big old Tannoys to.

My preference is for a separate DC offset detection circuit and output isolation relay rather than adding to the complexity of the amp. Worst case, the fuses can take care of the safety aspect, transistors are cheap (in comparison).

Check out the attached design. May be good but it is over my limit.

I've not seen an amp with a PTC thermistor in line with the output for a while.

[soldar]

Here's another one I found in the same line.

[planet12]

I always enjoy playing around with older or less usual amplifier designs - even if only inside LTspice, not necessarily on a PCB.

The OP's schematic has two feedback paths from the output to the input, one of them after the output capacitor (these being single-rail designs). My attempts at simulating it resulted in it being a quite efficient 500KHz sinewave oscillator until I removed the feedback from the point between the output capacitor and the speaker.

I suspect the transistor models I was using were... too fast. But I also suspect it is inherent in the design - look at the multiple RC time constants in the signal paths, looking to tame it. I would not want to manually do the AC analysis of that.

Once you remove that mess, it's not too dissimilar to #8 above - and it actually had some pretty good THD specs - surprisingly to me, it came in at around the 0.002% mark measured with 20W @ 1KHz into 8ohm (resistive).

The move to differential (long tailed pair) input seems to have happened pretty rapidly. Much easier to design with now that it's effectively a high power op-amp. You can then add extra refinements if you want to get fancy - current source LTP load, current mirrors in the collectors, buffered VAS, cascoded VAS, fully complementary throughout, etc... oops geeking out 

[dzseki]

I always enjoy playing around with older or less usual amplifier designs - even if only inside LTspice, not necessarily on a PCB.

Once I had a Fergusson receiver (made in UK), which had PNP transistors on its output, the whole unit had a negative rail compared to ground, now that was an oddball design

[soldar]

I always enjoy playing around with older or less usual amplifier designs - even if only inside LTspice, not necessarily on a PCB.

If you have done the LTSpice models of any of these it would be nice if you could post them and we could see them.

[planet12]

I always enjoy playing around with older or less usual amplifier designs - even if only inside LTspice, not necessarily on a PCB.

If you have done the LTSpice models of any of these it would be nice if you could post them and we could see them.

Did a bit of work tidying it up, and in the process realised my memory of the distortion etc. figures were waaaay off - it's an order of magnitude higher.

Simulation attached measures it at just under 0.03% THD at 20W/1KHz/8ohm.

AC analysis shows 44dB peak loop gain, 0dB @ 400KHz with 10 degree phase margin; gain margin = 14dB. "barely stable" IMO.

The main schematic uses the potentiometer symbol and subcircuit - put these in the same directory as the main .asc file and rename to remove the .txt extension (needed as this forum won't accept .sub / .asy files).

I've played pretty fast and loose with BJT models - basically just "something close that LTspiceXVII has built-in"

[soldar]

Thanks! I will take time to study it as I am not very proficient with LTSpice.

My first guess was the feedback from the speaker would reduce both gain and distortion but looking at it I now think it is positive feedback which would increase both. I do not understand it now.

I built (and sold) quite a few of these in my day and I do not remember having any problems with stability.

[planet12]

Thanks! I will take time to study it as I am not very proficient with LTSpice.

There's a few "tricks" in there to get a decent THD analysis - mainly changing mathematical precision settings and turning off data file compression. Hopefully the comments in my SPICE directives are clear enough.

My first guess was the feedback from the speaker would reduce both gain and distortion but looking at it I now think it is positive feedback which would increase both. I do not understand it now.

I built (and sold) quite a few of these in my day and I do not remember having any problems with stability.

I can see both a negative and positive feedback path occurring with the feedback from after the output capacitor.

• The negative feedback path is via the 1k8 resistor and the 160uF capacitor to the emitter of the input BC157
• The positive feedback path is via the 1k8 resistor to the middle of the divider created by the 33ohm resistor in combination with pot P1

I suspect the negative feedback from after the output capacitor was an attempt to use negative feedback to eliminate distortion caused by the output capacitor.

The positive portion of the feedback - if the levels are right (below unity across the entire operational range) this will act as a "bootstrap" to increase the input impedance, in a similar fashion to the circuits described here: https://electronics.stackexchange.com/questions/268944/effect-of-bootstrapping-in-amplifier-circuit

I might see if I can track down models for the older transistors as described in the original schematic; it'd be interesting what difference this makes to the behaviour of the amplifier.

[soldar]

I can see both a negative and positive feedback path occurring with the feedback from after the output capacitor.

• The negative feedback path is via the 1k8 resistor and the 160uF capacitor to the emitter of the input BC157
• The positive feedback path is via the 1k8 resistor to the middle of the divider created by the 33ohm resistor in combination with pot P1

I suspect the negative feedback from after the output capacitor was an attempt to use negative feedback to eliminate distortion caused by the output capacitor.

The positive portion of the feedback - if the levels are right (below unity across the entire operational range) this will act as a "bootstrap" to increase the input impedance, in a similar fashion to the circuits described here: https://electronics.stackexchange.com/questions/268944/effect-of-bootstrapping-in-amplifier-circuit

I might see if I can track down models for the older transistors as described in the original schematic; it'd be interesting what difference this makes to the behaviour of the amplifier.

I had assumed right at first but then I had analyzed it wrong. The path through the capacitor is negative feedback as I had assumed so it would decrease gain and distortion.

As I say, I never had problems with instability or oscillations but if the model shows any instability maybe putting some filtering so that bandwith is cut way down maybe that would solve the problem in the model?

[planet12]

I had assumed right at first but then I had analyzed it wrong. The path through the capacitor is negative feedback as I had assumed so it would decrease gain and distortion.

As I say, I never had problems with instability or oscillations but if the model shows any instability maybe putting some filtering so that bandwith is cut way down maybe that would solve the problem in the model?

I put a little more work into this to validate if my "transistors are too fast" theory was correct, redoing things with different transistor models - I've subbed in BC547/557 for the BC148/BC157's, as my understanding is they pretty close, the BC1xx being the older metal can versions, and have switched to the Fairchild BD139/BD140 models for the VAS and output drivers.

It looks like I was right - using these have shifted things around enough that it no longer oscillates when adding back the 1k8 feedback from the output capacitor, and in fact performance is improved.

Closed-loop gain drops significantly, from 98 (20dB) to 28 (14.5dB) - you'll see I've increased the input signal level to maintain 20W into 8ohm output.

THD drops from 0.049% (distortion increased from 0.03% with updated transistor models but without 1k8 feedback resistor) to 0.006% - quite a significant improvement.

Input impedance changes a little, but not drastically:

Freq	With extra FB	Without extra FB
10Hz	86k	69k
100Hz	149k	135k
1KHz	148k	146k
10KHz	85k	84k
20KHz	49k	48k

Update LTspice schematic attached (you'll still need the potentiometer models from my post above).

[planet12]

Update LTspice schematic attached (you'll still need the potentiometer models from my post above).

Just a quick note on the simulation above: the addition of the extra feedback path breaks open-loop AC analysis (the V(output)/V(X) stuff).

I haven't yet figured out exactly how to do the open loop analysis given the circuit now has effectively three feedback paths (two negative, one postive).

[soldar]

Thanks planet12, that is outstanding work. You are obviously much more knowledgeable about LTSpice so I won't even try but here are some thoughts.

From your analysis it seems the performance is pretty good.

If we consider C4 and C10 as large enough to be a short-circuit to AC then R19 is in effect in parallel with R5 and you could (for analysis) remove R19 and replace R5 with 1K3. That would simplify things.

Regarding the other feedback, the C8 bootstrap, I dunno... maybe just remove it if it gets in the way.

[planet12]

From your analysis it seems the performance is pretty good.

If we consider C4 and C10 as large enough to be a short-circuit to AC then R19 is in effect in parallel with R5 and you could (for analysis) remove R19 and replace R5 with 1K3. That would simplify things.

Regarding the other feedback, the C8 bootstrap, I dunno... maybe just remove it if it gets in the way.

The performance is... okay for it's age I suppose. The loop gain is pretty average, and the loop bandwidth is pretty low. The phase margin is less than the standard recommended margin of 45 degrees.

I suspected the THD would rise significantly with frequency due to the low the loop gain and low loop bandwidth, and a resimulation at 10KHz shows an almost ten-fold rise to 0.04% @ 20W - again, not terrible, but better is achievable with a long-tailed-pair style input stage (along with a higher input impedance).

Considering C4/C10 as short-circuits works well enough for basic gain calculations, but doesn't help with phase response (and therefore stability analysis).

I can definitely see why designers pretty rapidly flipped to using LTP inputs - one extra transistor (or two-to-three if you want a proper current source load) provides significant improvements, and in my opinion simplify design and analysis. In the meantime, PNP power transistors have improved to the point the output stage can be truly complementary, rather than the quasi-complementary as seen in this design.

It was an interesting design to fiddle with - but it's not something I'd build today.

[asilva54]

I liked this circuit, primarily by offset and Bias adjustments. My power supply here is 33VCC from my Philips receiver. Will it be necessary to change something in the input stage? Those base resistors?

[mariush]

It's rarely worth the time to bother with such circuits, when you can get TDA2050 clones for 50 cents, and the circuit in the datasheet is stupid simple to make (see page 4) : https://www.circuitbasics.com/wp-content/uploads/2016/10/TDA2050-Datasheet.pdf

[David Hess]

It reminds me of the Dynaco Stereo 120.

These days I would use a current and voltage boosted operational amplifier design, or if I wanted something completely discrete, a standard input, VAS, and output design with the VAS replaced with a current feedback amplifier stage.  I did some work 20 years ago like this but I did not really know what I was doing at the time, although I still got something usable.