Transistor Distortion

[Matt Syson]

The circuit is shown in one of Doug Self's books so it is probably from a Soundcraft desk. Making a class b output stage (to minimise crossover distortion requires a bit more effort and the option of reliability (freedom from thermal runaway and simplicity few parts as possible) was probably chosen by the 'bean counters'. The T072 is probably not a good choice because they get distinctly distorted when you try loading them with less than about 2500 Ohms. since the users of this circuit might plug anything from 8 Ohm to 2K ohm headphones into it, designing for 'best' performance across such a range is fairly tricky. Reducing the 4k7 to something like 470 Ohms and using a 5532 could be a good option because the 5532 is essentially happy driving about 500 Ohms on it's own so the transistors would start sharing when higher current is needed more readily than the TL072. MC33078 is another 'nice' op amp which takes less quiescent current than a 5532 and is OK driving 600 Ohms. The amount of 'signal' (half wave rectified audio) tht is appearing on the supply rails of the TL072 in guestion AND what might be appearing on previous stages mah have a bearing on the distortion heard. EXACTLY where the headphone ground currents are going is also another consideration. It suggests from that clip of schematic that it is to a pair of serial connected electrolytics whether that it actually true or not adds yet another variable. The reactance of the capacitors is not negligible if you are considering low impedance phones AND the ground is shared between channels.

 

[ruffrecords]

The circuit is shown in one of Doug Self's books so it is probably from a Soundcraft desk.

Hi Matt. If you are referring the schematic in post #1 then that is not Doug's circuit. His is quite different and much better. Attached is a poor photo I just took from the book.

Cheers

Ian

 

[blakeyboy]

I don't think we would learn anything from looking at the waveform in an oscilloscope in this case because the distortion is 40 dB down. Even on a 10 bit scope you probably wouldn't be able to see that.

And I think we've already figured it's crossover distortion so you're not going to see clipping of peaks anyway.

ok...I think you are confused ( and the two people who liked your post ).

I was pointing out that simple correct method will often lead to faster diagnoses.
Checks 1, 2 and 3 instead of jumping to steps 4,5 and 6 because you've spent a lot on an analyser....


Maybe, in this case, the fault couln't be seen on a scope....

You should bear in mind that my point still stands and is unchanged.
And correct.


I'm so bored now.
I put up stuff after taking photos etc, when I'm busy, one or two of the usual people like it,
and a pile of idiots pile in because the truth has bruised their litte egos.
I don't waste my valuable time posting nonsense, locking antlers, or showing off confirmation biases.

See ya.

This is my last posting, so you are all free now to look _really_ clever to newbies.
( A couple of members are not included since they are professionals, and professionalism is only an attitude.
You know who you are. )

59 posts in 17 years- where did I find the time?

 

[ccaudle]

Even on a 10 bit scope you probably wouldn't be able to see that.

Crossover distortion is easily visible if you know what you are looking at. Analog scope, 8 bit digital, doesn't really matter.

 

[JohnRoberts]

The circuit is shown in one of Doug Self's books so it is probably from a Soundcraft desk.

evidence?

Making a class b output stage (to minimise crossover distortion requires a bit more effort and the option of reliability (freedom from thermal runaway and simplicity few parts as possible) was probably chosen by the 'bean counters'. The T072 is probably not a good choice because they get distinctly distorted when you try loading them with less than about 2500 Ohms.

except even with a smaller resistor that load is limited by the output transistor turning on at +/- 0ne base junction drop... so even a 470 ohm resistor won't draw much more than 1 mA, easy peasy even for a wimpy TL072

since the users of this circuit might plug anything from 8 Ohm to 2K ohm headphones into it, designing for 'best' performance across such a range is fairly tricky. Reducing the 4k7 to something like 470 Ohms and using a 5532 could be a good option because the 5532 is essentially happy driving about 500 Ohms on it's own so the transistors would start sharing when higher current is needed more readily than the TL072. MC33078 is another 'nice' op amp which takes less quiescent current than a 5532 and is OK driving 600 Ohms. The amount of 'signal' (half wave rectified audio) tht is appearing on the supply rails of the TL072 in guestion AND what might be appearing on previous stages mah have a bearing on the distortion heard. EXACTLY where the headphone ground currents are going is also another consideration. It suggests from that clip of schematic that it is to a pair of serial connected electrolytics whether that it actually true or not adds yet another variable. The reactance of the capacitors is not negligible if you are considering low impedance phones AND the ground is shared between channels.

Class B crossover distortion is easily audible with low level wide band audio signals.

I hope we didn't scare off the OP with all this science...

JR

 

[swpaskett]

Back in post #11, @musipol asked if this ever sounded good. Did it? I can't imagine it did.
At low volumes, both the + and - drivers are in cutoff and you are driving the phones through that 4k7 resistor. Not going to hear much there. Increasing to higher volume, you get to the point that the drivers can start driving, but you still are going through about 1.4V of cutoff around 0V. Distortion should be pretty bad in this range, because the opamp is driving the phone through 4k7 part of the time, then the drivers come into play and can drive the phones satisfactorily. As you drive it harder distortion should decrease a bit because you are using the drivers for more of each cycle.
Sure, it's crossover distortion. Listen to JR, reduce the 4k7 to 470 and it will work a lot better. It will stress the 072 a bit so you may want to swap it for a 5532 as JR also suggested.
Bad designs can't be repaired. They can be redesigned and replaced with something better, but the only repair you can do with this is replace the 4k7 with a lower value and decide if it's good enough.
I use this topology in cheapo drivers quite a bit. It's ok if you understand its limits, but I have never seen the resistor set so high before to drive a low impedance. (I use 68 ohms when driving 600.) It's there specifically to allow the opamp to drive the load when the drivers are in cutoff. With a resistor valued as much as 578 times the load resistance this thing is theoretically able to drive, that opamp is going to have to slew to the rail before the load will begin to see a signal, and by then the drivers have probably turned on and the opamp is having to slew the other direction to keep them out of saturation. Get a scope and probe the output of the opamp; you'll see what I mean.

 

[Bo Deadly]

At low volumes, both the + and - drivers are in cutoff and you are driving the phones through that 4k7 resistor. Not going to hear much there.

You would hear it just fine. Feedback compensates for the 4K7 in series. The amp will just output a higher voltage at low levels.

Honestly I'm still not convinced that the problem is just crossover distortion. The op amp should compensate for that pretty well even with the 4K7.

As Matt Syson pointed out at the top of this page, the electrolytics connecting the hphone ground to the supplies has to be considered. If their ESR has gone high, that could be a problem.

 

[michinger]

YES!
I changed to a 200 ohm resistor i had here and it works without the distortion now!
Made a measurement and it looks how it should (like the output of an TL072).

I don´t know if it ever sounded good because I got the desk used. But now it is good enough

Thank you all for your answers and the time you invested!

 

[Matt Syson]

OK I had 'mis remembered' the source of the schematic as it is more likely Doug would have used the schematic in one of his books that has a pair of diodes to provide biasing and emitter resistors however others have used essentially the same circuit but with component values not expecting it to drive 8 ohm headphones. Using a pair of capacitors to provide an AC 'ground' is a relatively common trick after all the current comes from the rails and may be inconvenient if it goes to 'centre rail (ground) in which case shorting the 'ground' contact of the headphone socket (as it was shown) to 'real ground can force the half wave rectified audio into other places. You have to be far more careful when examining circuits in isolation. If the two capacitors are not equal capacitance and ESR that again can cause other side effects. The circuit as shown probably assumes that headphones will be used so 'floating' with respect to the common terminal of the power supply. If you were to consider the impedance, DC to many KHz between the power rails is very low then you effectively have the headphones 'seeing' a capacitance of around 440uf which in iteslf will lead to significant crosstalk left to right because headphone 'ground' is not quite the same as the ground of the 2 amplifiers driving the 2 earpieces.

 

[michinger]

OK I had 'mis remembered' the source of the schematic as it is more likely Doug would have used the schematic in one of his books that has a pair of diodes to provide biasing and emitter resistors however others have used essentially the same circuit but with component values not expecting it to drive 8 ohm headphones. Using a pair of capacitors to provide an AC 'ground' is a relatively common trick after all the current comes from the rails and may be inconvenient if it goes to 'centre rail (ground) in which case shorting the 'ground' contact of the headphone socket (as it was shown) to 'real ground can force the half wave rectified audio into other places. You have to be far more careful when examining circuits in isolation. If the two capacitors are not equal capacitance and ESR that again can cause other side effects. The circuit as shown probably assumes that headphones will be used so 'floating' with respect to the common terminal of the power supply. If you were to consider the impedance, DC to many KHz between the power rails is very low then you effectively have the headphones 'seeing' a capacitance of around 440uf which in iteslf will lead to significant crosstalk left to right because headphone 'ground' is not quite the same as the ground of the 2 amplifiers driving the 2 earpieces.

These two 220µF had already been replaced before with 50V Panasonic´s.
But I understand your point.

 

[abbey road d enfer]

Stop.

Get a good clean variable oscillator, and a simple oscilloscope.
Try different resistive loads to start with.

Then you can begin to repair/change the circuit.
Everyone is trying to fix it from diagnostic steps 4, 5 and 6, having decided not to do diagnostic steps 1,2 and 3.

Simple stuff first. _Always._

For example:
If the waveform is clipped on just the positive peak, or just the negative peak, and would sound very similar to what you have described, and would look similar on a harmonic display,

none of you would know......

Someone has great appreciation of their own wits.

 

[JohnRoberts]

Someone has great appreciation of their own wits.

Hey you are supposed to be the kind one...

If he has something constructive to contribute I hope he comes back, he's new here and doesn't know the room.

JR

 

[FIX]

Generally i am more a reader then writing threads here, but this time I need some help.
Would be nice if the great people here could help me.

I try to repair a realtive simple headphone section of an Mixer:

View attachment 96925

The Signal on the HP jack is good until it reaches a middle loud volume, then you can hear distortion.
Not really bad, but not good enough either.

The signal is fine before the R31 and R32. After them the distortion starts, so it is about the 2SC2240 and 2SA970
I already changed that parts (also the resistors) but nothing changed.
The voltage there is 16,7V which should 16V but that can´t the problem I think.
Power supply is fine because everything else works, also other outputs of the mixer. So the voltages seems fine.

Would be nice if someone have ideas how to find the issue.
Thanks in advance!

One little tidbit, Always put the feedback compensation cap around the op-amp ONLY, as most output devices are slower than the op-amp, and you could have HF oscillation that you don't hear, eating at your headroom...

 

[JohnRoberts]

One little tidbit, Always put the feedback compensation cap around the op-amp ONLY, as most output devices are slower than the op-amp, and you could have HF oscillation that you don't hear, eating at your headroom...

Dominant LPF poles should be applied globally to deliver an accurate transfer function. Compensation caps to insure stability can be local to the op amp. It is unclear which that 47pF is. Caution, too much local compensation can compromise the overall transfer function.

I don't expect any of this is a major concern with this particular circuit.

JR

 

[Newmarket]

Dominant LPF poles should be applied globally to deliver an accurate transfer function. Compensation caps to insure stability can be local to the op amp. It is unclear which that 47pF is. Caution, too much local compensation can compromise the overall transfer function.

I don't expect any of this is a major concern with this particular circuit.

JR

tbh the original circuit seems compromised to the point where the compensation cap is moot ?
But - yes -:both directly around the opamp and around overall feedback capacitors have valid functions.

 

[abbey road d enfer]

If he has something constructive to contribute I hope he comes back, he's new here and doesn't know the room.

Well, him being a member only since 2005, should we give anothet 17 years to adjust?...

 

[abbey road d enfer]

Dominant LPF poles should be applied globally to deliver an accurate transfer function. Compensation caps to insure stability can be local to the op amp. It is unclear which that 47pF is.

With the transistors going from non-conduction to full conduction, their "speed" is extremely variable. In the x-over region, they are slow as hell, so the compensation would not really work if taken from the emitters.

 

[5v333]

What about when two conducting transistors goes to one conducting?

 

[JohnRoberts]

What about when two conducting transistors goes to one conducting?

if you are talking about the transistor pairs in parallel they will roughly turn on/off together give or take a little for Vbe matching. The npn/pnp class B output stage topology will reverse bias and cut off the opposite base emitter junction before turning on effectively eliminating any possible mutual conduction. This dead band between pulls-up and pulls-down is the specific cause of crossover distortion.

With the transistors going from non-conduction to full conduction, their "speed" is extremely variable. In the x-over region, they are slow as hell, so the compensation would not really work if taken from the emitters.

The TL07x series is unity gain stable so stability compensation is not an obvious concern. The 4.7k shunt resistor across the output stage is uncomfortably large. Any stray capacitance to ground at the final output would result in HF boost at the bases that would help turn on the transistors faster (but don't do that).

To repeat I strongly advocate substituting a lower value resistor in place of the 4.7k. (The OP has reported good results using a 200 ohm resistor there.)

JR

 

[abbey road d enfer]

What about when two conducting transistors goes to one conducting?

It would simply not be class B (actually class C in the schemo).
In the more common class AB arrangement, the transistor speed varies but less than going from non-conduction to conduction.
What you're suggesting would imply that the transistors on one side are significantly un-matched (or than one is duff).