1
\$\begingroup\$

So I am experimenting with various signal generating, op-amp and power amplifier circuits. And I have limited access to technology. Where I am at, everything you buy is twice the price because of tariffs, so I try to make my own stuff with what I already have. I got a scope but no bench power supply.

I use TTL levels for my signal generation with a simple cell phone charger 5V rated 2.1A nominally. I use a second such charger at 5V rated at 1A for my negative supply to the op amp and the power amp stage. And that is where I think my troubles come from.

I get a nice sine wave with about 6.5V peak to peak (not entirely sure why I can't get the full 10V even if I crank the feed-back up to maximum distortion level, but let that be a different question.)

I have a power output stage constructed with a pair of BC547 npn and BC557 pnp transistors. It gives me a beautiful clean undistorted wave on the output. However, when I actually plug the little speaker in, that's rated at 8Ω 5W, the output waves are suddenly clipped hard at about 2V peak to peak, and all the waves are starting to look ugly.

I think that is because of the power supplies reaching their limits, so they cut off intermittently.

schematic

simulate this circuit – Schematic created using CircuitLab

Quick walk-through: I don't show the details of the sine wave generator, it's essentially a 555 astable multivibrator circuit, spitting out its 4V peak to peak, with a 3 stage low pass filters RC network to get a nice sine wave with minimal signal loss, around 2V peak to peak. Then that goes into OA1 as a simple high impedance input voltage follower to shield the finicky RC network from any further load. Then I have a "volume" potentiometer R1 which goes to a decoupling capacitor C1 of 100 nF, because then I bias that signal down with the potentiometer R2 nicely into the middle of the range, then OA2 is the pre-amp bringing the level to 6.5V peak to peak, and finally the end-stage with the pair of BC547 and 557, and the OA3 acting in this genial way to bring a totally clean undistorted wave into the output handling that "dead-zone" of the transistors.

Here I'm going to show you a few pictures of this weird clipping. The wave comes out of the end stage beautifully undistorted:

the wave comes out of the end stage beautifully undistorted

but as soon as I hook up the speaker, it gets terribly clipped

but as soon as I hook up the speaker, it gets terribly clipped

Since my suspicion is that the power supply just cannot deliver the power required here (I think it's around 1A which would bring it to its limits) I want to hook up a 12V and 10A rated DC power supply that I have spare. Here I am trying to think how I would do the 12V hookup and it should almost work like this:

schematic

simulate this circuit

I really am puzzled about the GND. Nothing seems right, even in the first schematics (my current design that breaks down under load).

  1. no obvious relationship between signal ground of the generator and OA1 input
  2. pointless connection of potentiometer R3 to this GND, why not connect it directly to V-?
  3. the speaker's GND goes to nowhere -- how does the power actually flow?

I guess in my first design the speaker's GND does flow directly into the common lead connecting the two 5V power supplies. So it did make sense that way. But now if I want to use that 12V DC power supply then where can I connect that GND now? I have seen people using another op-amp to simulate a GND, but I don't see how I could put power through this.

What else can I do? When I was young, I built a kit of a guitar amp, which had a husky power supply where two large capacitors connected positive and negative to ground. But the ground was also connected to the transformer from what I remember. It was like this (from memory):

schematic

simulate this circuit

I don't remember if there was a center tap from the transformer or if the ground was simply established by the two capacitors. They might have been even bigger that 100 μF, maybe 500 μF, they were about 1 inch in diameter and 2 inch height, all from memory, it's been like 35 years.

\$\endgroup\$
5
  • 1
    \$\begingroup\$ It's far more likely your opamp cannot drive the BJT output stage. You are looking to get over an amp output from the BJTs and they require quite a lot of drive current. Take a look at the datasheet and see what they (the opamp(s)) are capable of driving. I'm almost sure they are way, way, underneath the need. \$\endgroup\$
    – jonk
    Jun 15, 2020 at 2:15
  • \$\begingroup\$ It's more work than that. (You might also consider a Szlikai, too.) But that's in the right direction. \$\endgroup\$
    – jonk
    Jun 15, 2020 at 2:39
  • \$\begingroup\$ Thanks @jonk, sorry, I deleted my reply comment by accident, we were moving into the direction of a darlington pair. \$\endgroup\$ Jun 15, 2020 at 2:54
  • \$\begingroup\$ For power drivers you'll need (want) some emitter degeneration (to help address vagaries of BJTs and temperature effects) and also something to help avoid cross-over distortion, though you may be already okay with it given what I've seen from your design. Just make sure your opamps can drive the darned thing. \$\endgroup\$
    – jonk
    Jun 15, 2020 at 2:58
  • \$\begingroup\$ I confirmed with the data sheet that the LM324 spits out max 20 mA sourced or sunken. The BC547 base current max is 5 mA. So that means it would never be able to bring the op-amps to it's limits. However, it cannot produce more than 100 mA current, and I guess that may be why it is cutting off the peaks? \$\endgroup\$ Jun 15, 2020 at 4:05

3 Answers 3

3
\$\begingroup\$

I get a nice sine wave with about 6.5V peak to peak... power output stage constructed with a pair of BC547 npn and BC557 pnp transistors. It gives me a beautiful clean undistorted wave on the output. However, when I actually plug the little speaker in, that's rated at 8Ω 5W, the output waves are suddenly clipped hard at about 2V peak to peak

Consider that the current through 8 Ω at 3.25 V is ~400 mA. Your transistors are only rated for 100 mA absolute maximum, so this is well over their rating. To keep peak output current within the ratings of the transistors you must reduce the peak voltage to less than 8 Ω * 0.1 A = 0.8 V.

At 2V the output current is 2 V / 8 Ω = 0.25 A, well below the capacity of your power supply. The real cause of the clipping is that the op amp cannot supply the required drive voltage to get the output up to 3.25 V, due to internal voltage drop in the bipolar output stage of the op amp (~2 V) and the Base-Emitter junction of the external transistor (~1 V).

This is a common problem faced by low voltage audio power amplifiers. Voltage drop can be minimized by using a quasi-complimentary output stage, and drive voltage can be raised by 'bootstrapping' it from the output. However to do this you would need a fully discrete power amplifier stage.

The simplest way to solve your distortion problem is to just reduce the drive amplitude, which will also get the peak current below the transistor ratings. If you don't care about overloading your transistors then you could use a 12 V supply to get an extra 2 V of peak 'headroom', but you will need to create a 'virtual earth' at half the supply voltage and AC couple the input and output. The circuit would look something like this:-

schematic

simulate this circuit – Schematic created using CircuitLab

\$\endgroup\$
7
  • \$\begingroup\$ Thanks for adding a single-supply schematic, though the output stage will not compete with a bridge-type car stereo output IC with 25 watts into 4 ohms, or about 15 watts of clean power per channel. OP will have to accept some compromise in performance. \$\endgroup\$
    – user105652
    Jun 15, 2020 at 5:57
  • \$\begingroup\$ This is definitely not a 'hi-fi' circuit. The output stage is pure class B so there will be audible crossover distortion \$\endgroup\$ Jun 15, 2020 at 6:05
  • \$\begingroup\$ Thanks, so you showed me to create a virtual ground (vGND) with a simple resistor based voltage divider, and you avoid any current to be actually pushed through those resistors because the load references the power supply negative terminal, and the final decoupling capacitor removes any bias. That's a solution indeed. Other things I am seeing in your schematics: (1) pull-down resistor to vGND with R3 and decouple right away, even before the voltage follower. (2) grounding the "volume" pot. (3) you keep OA4 feedback refering to vGND, (4) end-stage unchanged except for final C4 and spkr to V-. \$\endgroup\$ Jun 15, 2020 at 16:41
  • \$\begingroup\$ @BruceAbbott , I found that the op-amp driven class B is actually free of the corss-over distortion, it is better than the class AB approach with the two diodes. On the (analog) scope I can even see a slight wiggle where the positive and negative half wave are joined. But it's just like one "pixel" (as far as that even exist on an analog scope), nothing like a distortion, just to remind us of the magic that that final OP-amp is doing! \$\endgroup\$ Jun 15, 2020 at 16:45
  • \$\begingroup\$ Play a low frequency (eg. 100Hz) pure sine wave through the speaker and listen - you will hear the crossover distortion. The human ear is very sensitive to higher order harmonics in the audio frequency band. And remember that the crossover distortion is a fixed amplitude, so at lower output level it is a greater proportion of the signal. If you can see it on the scope it's much too large! \$\endgroup\$ Jun 15, 2020 at 20:14
1
\$\begingroup\$

These op-amps have bjt outputs, so you suffer a Vdrop of 1.2 volts at the op-amp output. Now when you add more transistors to boost the drive current that is another 0.6 volt drop, or 1.2 volts if you use Darlingtons. 5 volts minus 1.8 leaves just +/-3.2 volts to drive the speaker, +/-2.6 volts if you use Darlingtons. Adding a 8 ohm load bogs this down to a couple of volts at the speaker.

Normally low voltage speakers are 16 ohm to 32 ohms so output has full voltage swing, even if it is only a few volts. Think of the low voltage in cell-phones, yet the ear-buds can get fairly loud. Ear-buds are typically =>16 ohm and 32 ohms is common.

If you could build a 'bridge' type output you could pump 4 times the power into the same load, but as mentioned in comments the output stage must be properly designed, especially if you want low distortion. Look up 'power amplifier schematics' and study the output stages. Even though you have only +/- 5 volts and not +/- 50 volts, it can work, but directly driving 8 ohm loads at a loud volume is not possible due to Vdrop per each output stage.

\$\endgroup\$
6
  • \$\begingroup\$ This voltage drop issue probably explains why I can't get the full +/- 5V range filled with the signal even out of the LM324 op-amps. Now, currently with my little BC547/557 transistors, I do not get any further voltage drop, i.e, without load, I still get the entire 6.5V that I gained from my pre-amp OA stage. But I guess with Darlington I would lose more. Will try tomorrow. \$\endgroup\$ Jun 15, 2020 at 3:52
  • \$\begingroup\$ Your biggest limit is low supply voltage combined with a low speaker impedance. To get more bang for the buck the supply voltage must increase, or the speaker impedance must increase. Adding a 10 ohm or 20 ohm resistor in series with the speaker divides the amp output, so no gain there. \$\endgroup\$
    – user105652
    Jun 15, 2020 at 4:03
  • \$\begingroup\$ To avoid the extra voltage drop of the Darlington configuration, I guess the Sziklai configuration would be better, wouldn't it? I also confirmed with the data sheet that the LM324 spits out max 20 mA sourced or sunken. The BC547 base current max is 5 mA. So that means it would never be able to bring the op-amps to it's limits. However, it cannot produce more than 100 mA current, and I guess that may be why it is cutting off the peaks? \$\endgroup\$ Jun 15, 2020 at 4:05
  • \$\begingroup\$ Use a 16 ohm or 32 ohm speaker, and most of these issues will go away. With such low supply voltages, you have little choice. An 8 ohm speaker would never work with a cell phone or tablet. It would have a weak and tinny sound. \$\endgroup\$
    – user105652
    Jun 15, 2020 at 4:08
  • \$\begingroup\$ this actually leads me back to my original question. If I want to use a 12 V 10 A power supply, how would I create a middle ground? \$\endgroup\$ Jun 15, 2020 at 4:45
1
\$\begingroup\$

I have it figured out, thanks to the hint that the op-amp couldn't feed enough base current. It wasn't the power supply after all -- aside from the fact that it isn't enough voltage to drive that little 5W speaker to full use.

Since everyone was saying that the Darlington configuration will get me an extra base-to-emitter voltage drop, I thought I'd try Sziklai, and then I did a quasi-complementary push-pull pair instead because I looked it up on Google images and got the wrong schematics. But lo and behold it worked beautifully!

schematic

simulate this circuit – Schematic created using CircuitLab

And I did all that not before I cleaned up my entire bread board to be really neatly arranged, for that I had to rotate the LM324 chip by 180 degrees to get V+ on the upper rail and V- on the lower, I am pedantic that way. Upper power rail of breadboard has V+ on + and GND on the - above it. Lower power rail has V- on the - rail and GND on the + rail (because GND is positive of V-). That way I don't get it confused and added plenty of capacitors all over to get rid of current starvation, everything nice and steady.

tidy breadboard

I also want to show you just how genial the op-amp driver for the end-stage is, because it really gets rid of the cross-over distortion beautifully. As I mentioned in a comment here somewhere, you can even see where the upper wave and the lower wave are glued together as the op-amp jumps during the cross-over moment through the dead-zone. It's beautiful.

op amp controlled nearly perfect correction of cross-over distortion as we pass through the dead-zone

I am showing still the clipped version from the base current starvation by the op-amp, that way you can clearly see which one is the signal on the input to OA3 and what is the signal on the output of the end stage. You see these "pixel" effects on the scope, it shows as 3 dots or a little wiggle on the output but nothing major, and on the input you see a single dot which might be from the switching of the neighboring op amp in the same chip or from the sudden current change.

And then finally here is the input and output signal even under load of the speaker with the end-stage using the TIP31 in this quasi-complementary Darlington pair setup, and lo and behold I do not get any problems with any further voltage drops, I guess that's all thanks to the op-amp driver stage that is able to compensate all of the problems while not clipping the peaks. It's pure op-amp magic. (I don't know if the guy from the "Simply Put" YouTube channel shouldn't have applied for a patent of that circuit, because it looks like nobody here knew just how well it worked!)

perfect sine wave under load of the 8Ω speaker

Can you tell which curve is the input signal and which is measured at the speaker while it's powered? I can't either, but we can see the 3 little spots on the down slope which we saw on the clipped curve.

Couple of more different frequencies. The above was about 1.5 kHz (scope at 0.1 ms per division). Now to max the highest frequency I can get, 25 kHz. You see the gain is awful. I think from the wave form generator which filters I have to adjust for every frequency, I get about the same 2V peak to peak. But now I get only 3V peak to peak from the end stage, and I think this is an op amp limitation as it clips heavily at a much smaller range. So this is the best I can do.

At 25 kHz (or even 33 kHz?) upper wave is from the initial voltage follower output, lower from the end stage:

at 25 kHz (or even 33 kHz?) upper wave is from the initial voltage follower output, lower from the end stage under load

Same 25 kHz (or even 33 kHz?) upper wave is the end stage at Q3 emitter, lower is at OA2 output. Under load, ripply-pixely on the down slope.

Same 25 kHz upper wave is the end stage at Q3 emitter, lower is at OA2 output. Under load, ripply-pixely on the down slope

Without load, the OA3 in and end stage out are pretty similar, the curly switching stuff and somehow this affecting the OA2 output also. I suppose for good Hi-Fi performance, I should not use a 4x op amp package but each separate with their own capacitor for clean power on each.

Without load, the OA3 in and end stage out are pretty identical

Finally at 13 kHz a decent gain is back and the same ripply-pixel down-slope under load.

at 13 kHz a decent gain is back and the same ripply-pixel down-slope under load

but not without load. If you want to see it tell me. Enough scope pictures.

So now this project is a full success. Tomorrow I will try he Sziklai configuration also. And I will get a better op amp, a rail-to-rail MOSFET based one, TLV2374 if I can source it.

UPDATE: Now I have separated the LM324 from a variety of higher quality op amp packages that I tried out, the NE5533 and the LM4558. And the result is much better. I get a better gain at high frequency, pretty much those same 6V peak to peak and less of these ripples in the slopes. Not sure about those dotted ripples on the down-slopes, because I did see those still when no load was applied.

schematic

simulate this circuit

This is at my maximum frequency and with load, much better gain and less ripples: at my maximum frequency and with load with NE5532, much better gain and less ripples

There is another tracing I found interesting. This here is from the output pin of OA3 (the rectangle wave is from the 555 before filtering, you can ignore that). This is at some 1.5 kHz where I would get 6V peak to peak on the end stage. And we can see here how the OA3 needs to "over-shoot" to correct for the cross-over distortion.

Finally for now, I have started trying the Sziklai configuration, which is just changing the high side to proper Sziklai, the low side of the quasi-complementary was already Sziklai.

schematic

simulate this circuit

The disappointing thing is that I can't get it stable. Sure, it's breadboard stuff with all sorts of flaky capacitance and lose contacts, but there is clearly a problem with the stability of the high side. Here under no load:

Sziklai, no load

and now with the speaker load:

enter image description here

There is clearly a problem getting this side stable. Not sure why. Anything wrong with my parts? TIP32? The other BC547 I used?

But if one may disregard the ugly fuzz, it seems like the gain is bigger with Sziklai on the high side.

\$\endgroup\$
2
  • \$\begingroup\$ Asking questions in an answer will not land you a solution. I suggest that you ask a new question focused on the stability problem only. Add the full schematic and link back to this question. Also, did you try putting small emitter resistors? \$\endgroup\$ Jun 18, 2020 at 20:05
  • \$\begingroup\$ @SredniVashtar, sure, I know. I just finished up the work that was brought up here. Darlington vs. Sziklai. I didn't really intend to figure out that Sziklai problem, I only stated it as pertinent observation. \$\endgroup\$ Jun 20, 2020 at 22:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.