I'm tired of hearing noise!

I support the point, that microphony often indicates high frequent self oscillations of a circuit. High permittivity ceramic capacitors in the signal path are another popular cause, because they generally show piezoelectric behaviour.

Reviewing the schematics, I see another possibly critical point. The OP is drving the output connector without a series resistor. Connecting long cables (e.g. > 5 m) is at risk to cause self oscillations by it's capacitive load.

I agree that a ceramic capacitor is a microphone when it is in the signal path. Also its value changes with voltage changes so it causes distortion.
Therefore I always use film capacitors in the signal path.

I also agree that many modern opamps oscillate when they try to directly drive the fairly high capacitance of a shielded cable. A series 100 ohm resistor between the opamp output and the cable fixes it.

I'm using very cheap ceramic chip SMD caps, might be an idea to buy some film caps for this purpose. I don't really like the idea of having a 100Ohm output resistor :O This has to cause some changes in the frequency response due to the varying impedance of a speaker or headset, right? Or are you saying that I should put the feedback path on the other side of the resistor?

I have to wait before I make another board, the simple reason being that I don't have any other boards left I'm also waiting for some C8 PCB AC connectors.

I'm going to try and sum up what I've learned though. If the amp is completely quiet when running off batteries, the noise (whatever the type) is caused by the psu and it's then about trying to prevent it from entering the amp just like I do with 50hz humming (star grounding, proper biasing etc.). If the noise is present even when battery driven, the noise is probably caused by the components themselves (like an unhappy op-amp). I should always use decoupling caps close to the op-amp pins to keep stability and to avoid noise, and keep noisy components (like a transformer) as far away as possible. I should also not bias my amp with zener diodes or use a voltage regulator as both of these can (will?) cause hissing noise. Am I thinking a long the right track here?

I'm still unsure about this last point though. For instance, my amps will always pick up noise from my soldering iron, but how do I prevent it from picking up noise from sources outside the amp? Having a metal casing and connecting ground to it? Also about using a transformer, I know that I would like to have the input connectors in the back of my casing in future designs, but this would require input wires going right past the transformer. Am I in these cases bound to use shielded cables? What if I'm using a toroidal transformer, can I in those cases run as many input wires wherever I want?

One last question about using a voltage regulator. Are regulator's really that noisy that they aren't usable in audio amplifier designs? I remember making an amp a while back using a wall-wart as a psu. I was struggling with a 2500hz ripple in the supply, adding a LM317 regulator completely fixed this issue, but I was left with a hideous hissing noise. Mind that I had a poor grounding on this design, but is it possible that the regulator itself caused this noise?

A regulator IN THE SUPPLY is a good idea, what we were advising against was using one to produce the half-supply to bias the op-amp input. Regulators are not designed for super low noise because the are almost universally surrounded by filter capacitors anyway but placing one at the input of an amplifier would have introduced noise where you have high gain following it.

Film caps are good but you really shouldn't have problems with ceramic types, it indicates you have too much gain and the tiny amount of microphony they produce is being over-emphasized.

As for output resistors, I think you are missing the point. The headphone impedance you are using is 'nominally' 32 Ohms but in reality, being an inductor and in a variable environment, the impedance probably varies from just a few Ohms to maybe 100 or more. To ensure you can drive all frequencies in to them with roughly equal volume level you need to have a fairly low impedance to drive them. In other words an amplifier that can deliver enough current to faithfully reproduce the signal voltage into a current hungry load. The other advantage of a low impedance output is it gives better 'damping factor', in other words it can hold the earphone diaphragm more precisely when it's own mass and return spring is fighting against the position you want it to be in.

The amplifiers you are using are only designed to drive high impedance loads, ideally in the K Ohms range. The idea of adding a series resistor is to protect the amplifier and preserve it's stability when being loaded so heavily by the headphones. You could add a series resistor and take the feedback from the headphone side of it which *may* improve stability but it introduces a further risk that the inductance of the headphones plays a greater part in the amount and phase of the signal being fed back. It potentially could make it even more unstable but without an exact plot of your headphone impedance vs. frequency it would be impossible to calculate.

What you should be doing is using the NE5532 as a pre-amplifier stage and coupling it's output to a power amplifying stage. This could be another IC or it could use discrete components. This would give you the low impedance your headphones would like to see and uses the pre-amp to best effect where voltage gain is needed.

Brian.

Your NE5532 is a preamp, not a power amp. It might oscillate at a high frequency if it is loaded with the capacitance of a shielded audio cable that connects to the power amplifier. The series 100 ohm resistor has no effect on the response into the 10k ohms input resistance of the power amplifier but it isolates the output of the preamp from the capacitance of the audio cable it is driving.
The power amplifier (that you do not have) drives a speaker or drives low impedance (32 ohms) headphones.

I understand that the NE5532 is not designed to drive low impedance loads, but so far it has worked great even with long cables (5m). This wasn't really the issue of this thread though, unless the NE5532 is a source of noise when driving low impedance or high capacitive loads (which I don't think it is since it's noiseless when battery powered). I'm really just interested in the possible noise sources.

I've made a new design using the NE5532 as an input stage with BJT class AB output stage. It will be powered from a chassis mounted center tapped transformer. The input, output, rectifying caps and feedback/non-inverting input resistor/100nF IC capacitors all have their own ground path. It might not be ideal that R6, R14, R7, R12, C6 and C3 have their own ground path, but I was thinking that this path won't carry much current so it won't really be an issue. I've also increased the trace width of many of the traces. Can I hope for my first noise free non battery powered amplifier with this design?

NB: R15 and R16 are 0ohm jumper resistors.

That's probably a huge improvement but still has a few potential problems:

1. The output stage bias isn't fully stabilized, you are relying on the voltage drop in the two diodes being the same as the Vbe of the transistors to set the class A region of operation. The op-amp will pull the appropriate transistor into conduction and mostly cancel the distortion effects but for best performance I would use a circuit that stabilized it's own bias point and allowed you to adjust the quiescent current.

2. You may still have a noise problem! The output stage has current gain which is great for driving your headhones or even small loudspeakers but it has no voltage gain. You are still relying on the op-amp to boost the signal voltage as before so if the gain caused problems before, it still will.

If you add an extra stage before the output transistors you can fix the bias problem, isolate the op-amp from the feedback loop and give it some extra voltage gain in one step.

Brian.

I can see that there is an issue with the bias, but does it matter that much? Since this is a headphone amp, I don't really want it to enter class B operation so as long as I can find the right resistor values to give me the right bias current, I'm happy.

I just finished the amplifier and to my amazement, it's completely noise free!!! The input connector is connected to the board with about 20cm of unshielded cable, putting this cable closer to the transformer will produce a faint humming (as expected I assume). As I mentioned before, I want the input connectors on the rear end of the box so they have to go past the transformers. Are my two options to either use a toroidal transformer or to use a shielded cable?

I figured that I didn't want to mess with heat sinks just yet so I chose R1, R2, R8 and R9 to be 100k. I have to admit that I'm unsure of how to calculate the bias current here, perhaps it isn't possible without knowing the exact temperature and characteristics of the bjt's and diodes? This is the way I was thinking. Vs is 34V so that gives 34V-1.4V across R1+R2. 32.6/200k=0.163mA. In an ideal case, Ice=Id5=0.0815mA. With a Hfe of about 100 that will give a Ic of 8.15mA. The actual current that I measured was 0.0669mA through one amp and 0.0343mA through the other so the calculation was completely wrong. I kind of forgot about R3 and R4 in this case though (which are 10Ohm each by the way), but I have no idea of calculating the bias current with these two resistors present. The voltage across D5+D6 is ideally 1.4V which is the same as the voltage between Ice of both BJT's, this leaves 0V for R3+R4 thus no bias current I figured that R3 and R4 are there for protection in case of an output short circuit, but they seem to make it hard getting a decent bias current

You are using a very high supply voltage of 34V for your headphones amplifier to blow away your hearing. Your schematic shows only 10V which is fine but will still play headphones loud enough to destroy your hearing.

You are using enormous high power transistors when little transistors will be fine.
R3 and R4 should each be 0.33 ohms and the bias resistors should each be about 22k. Then the idle current of each amplifier is about 20mA (for class-AB with no crossover distortion) and they each dissipate only 200mW.

R3 and R4 reduce the hFE of the output transistors so that they are better matched and they reduce the effect of temperature change which changes the idle current. The diodes should touch the output transistors for better thermal stability.

+-12V transformer was the only thing I had

How do you calculate the bias current? And what effects would increasing or decreasing the value of R3 and R4 have?

I also noticed something strange. My NE5532 get's really hot which I can only assume is because of too high supply voltage. The datasheet says +-15V as recommended supply voltage, but +-22V as maximum rating so I'm not sure

Plecto said:

+-12V transformer was the only thing I had

Click to expand...

Do not use it because the max power will destroy your hearing and destroy the headphones.

How do you calculate the bias current?

Click to expand...

It is difficult unless you select transistors and diodes that have the same specs.

And what effects would increasing or decreasing the value of R3 and R4 have?

Click to expand...

R3 and R4 should be 0.33 ohms so that they do not affect the output current much.

I also noticed something strange. My NE5532 get's really hot which I can only assume is because of too high supply voltage. The datasheet says +-15V as recommended supply voltage, but +-22V as maximum rating so I'm not sure

Click to expand...

Your supply voltage is fine but should be reduced to limit the output power. Your opamps are probably oscillating at a high frequency due to a missing supply bypass capacitor, driving the capacitance of a shielded cable, driving a lower resistance than 1k ohms or a poor layout (input to output coupling).

EDIT: I forgot. Your huge very old output transistors are very slow so they add phase shift that might cause oscillation. Smaller and newer transistors are much faster with very little phase shift.

There is a further reason why the NE5532 might get hot: they have parallel diodes across the +/- input pins which limit the voltage across them to about 0.6V. In that design, there is a potential for both transistors to be held in a cut-off state, depending on the transistor types and the diode types. A normal op-amp would, through it's feedback, shift it's output to compensate for the lack of bias (although that may not be desirable) and bring the appropriate transistor into conduction. If the feedback resistors are too low in value, there is a risk of the limited input voltage being exceeded and the transistors failing to turn on. It's a very slight risk but there nevertheless. Personally, I would keep the power amp and pre-amp stages completely apart from a DC point of view but if you want to stay with that design, I suggest changing the NE5532 to a 'normal' op-amp such as the TL072/TL082 and keep the NE5532 for use as a pre-amp stage.

You are getting there....

Brian.

Thanks again for the great replies.

Do not use it because the max power will destroy your hearing and destroy the headphones.

Click to expand...

Just because the psu voltage is high doesn't mean that I HAVE to turn the volume to 11 I want the amplifier to work fine with 650Ohm headsets as well.

There is a further reason why the NE5532 might get hot: they have parallel diodes across the +/- input pins which limit the voltage across them to about 0.6V. In that design, there is a potential for both transistors to be held in a cut-off state, depending on the transistor types and the diode types. A normal op-amp would, through it's feedback, shift it's output to compensate for the lack of bias (although that may not be desirable) and bring the appropriate transistor into conduction. If the feedback resistors are too low in value, there is a risk of the limited input voltage being exceeded and the transistors failing to turn on. It's a very slight risk but there nevertheless. Personally, I would keep the power amp and pre-amp stages completely apart from a DC point of view but if you want to stay with that design, I suggest changing the NE5532 to a 'normal' op-amp such as the TL072/TL082 and keep the NE5532 for use as a pre-amp stage.

Click to expand...

I've looked at the TL072 and TL082. Comparing them to the NE5532 they are half the price and seem completely superior when looking at parameters like slew rate, input impedance, input bias current, offset current etc. Why am I using NE5532's exactly? :S Also, what is the difference between TL072 and TL082, they seem identical to me.

EDIT: I forgot. Your huge very old output transistors are very slow so they add phase shift that might cause oscillation. Smaller and newer transistors are much faster with very little phase shift.

Click to expand...

Could you recommend some TO-220 BJT's that are suitable?

I have to say that I really appreciate the help that I've been getting. I actually managed to make a noise free amplifier powered from a transformer which is a huge step for me

A TL07x is a TL08x that is selected for low noise audio circuits. A TL08x "might" be low noise (maybe all of the last production is low noise or maybe none of them) but it is called "general purpose". There are many modern fast TO-220 transistors that are much faster than the old slow 2N3055. Sorry, I do not know their part numbers.

The TL07x series is almost identical to the TL08x series but they are guaranteed to have a lower noise figure. It's still 18nV/root Hz whereas the NE5532 is typically 5nV.

Brian.

The responses in this thread has been so great so I don't want to let it go just yet. I've made like 50 of these Cmoy amps before (with NE5532's) and they have worked out great. Today I encountered a weird problem, I got distortions like these: https://obrazki.elektroda.pl/9654657000_1378587690.jpg. I then tried to make another amplifier, but to my amazement, the exact same thing happened. The only real difference was that I started using 220uF 16V output caps because I ran out of 220uF 10V caps, but replacing them with 330uF caps made no difference. I tried to get some help at an IRC channel, but it was difficult to get help fixing a circuit with an op-amp that is really not meant to drive loads of such low impedance's. I then decided to make one with a bjt output stage, but guess what, the exact same thing happened! This new amplifier can output more power and I have to turn the volume up higher to hear the distortion more clearly, but it's still the exact same thing. Every component is the same as the 50 other amps I've made, the NE5532's are even from the same reel. Here's a schematic and parts list of the new amplifier I made:

https://obrazki.elektroda.pl/9993570300_1378587719.jpg

It's not actually clear under which operation conditions the said distortions occur (load impedance, supply voltage level and voltage source impedance). Previously mentioned deficiencies of the circuit (e.g. missing bias divider bypassing) might also cause distortions.

Sorry, I guess I was in too much of a hurry. The load is a 62Ohm headset, the supply is a 9V battery (thus no bias capacitor). The distortion isn't that audible at lower volumes and it's also just audible in the lower frequencies.

The output above half supply seems to be lower than below half supply. Are both channels showing the same distortion? You have the output transistor connected differently from one channel to the other.

Brian.

The original Cmoy headphones amp used newer and better opamps and used an 18V supply with a Mickey Mouse virtual ground circuit.
Why are you using older opamps and only ONE 9V battery?

What is a "bias capacitor"?

Maybe one or both of your 9V batteries is/are dead.

Please post your detailed schematic.