AB-class amplifier schematic analysis & optimization

[Zero999]

The best IC for a headphone amp I've tried is the LM7171 , however it is not an easy opamp to deal with, requires a good layout and P/S decoupling, plus some careful filtering on the input and the output.

Cheers

Alex

That's a video/RF op-amp. Using it for a headphone amplifier just seems like asking for trouble. 

Speed is good, but having a lot of it just makes life harder, with no benefit.

It will also be very noisy, especially at bass frequencies around 50nV/√Hz and 60pA/√Hz, at 30Hz! The NE5532 with a suitable BJT follower on the output will be much less noisy and less prone to oscillation.

[magic]

If the NE5532 is used, as suggested, the polarity of C2 is wrong, it will be biased by the opamp input current (about -0.3uA for NE5532) x 22K feedback resistor or about -6.6mV on pins 2(6). It would be better to reverse the C2 polarity with NE5532.

Not a big deal given that elcos can safely take up to 500mV of reverse voltage (IIRC) but I am actually autistic enough that I would do the same

AFAIK it is not that uncommon to have some series resistor (100 Ohms range) at the output, so that a 600 Ohms and 32 Ohms speak could produce a similar sound level.

Certainly not common on modern low voltage portable devices driving modern low impedance headphones as it would unacceptably limit power output. And people say that headphones these days are designed to be driven from zero impedance, so there is that.

And it is not without side effects, because dynamic headphones and speakers are quite reactive due to diaphragm inertia and suspension compliance. Think of it as a DC motor: voltage controls speed (and air pressure), current is drawn in proportion to force required to move the diaphragm. At resonant frequency, force/current is close to nothing and impedance shoots to the moon.
Therefore voltage division between the driver and the series resistor varies with frequency and boom, unequal frequency response.

And yes, you are totally right that it is a problem with certain headphone amplifiers.

[Zero999]

If the NE5532 is used, as suggested, the polarity of C2 is wrong, it will be biased by the opamp input current (about -0.3uA for NE5532) x 22K feedback resistor or about -6.6mV on pins 2(6). It would be better to reverse the C2 polarity with NE5532.

Not a big deal given that elcos can safely take up to 500mV of reverse voltage (IIRC) but I am actually autistic enough that I would do the same

The problem is increased distortion, even if the capacitor isn't harmed, although it's probably negligible in this case.

[Alex Nikitin]

That's a video/RF op-amp. Using it for a headphone amplifier just seems like asking for trouble. 

Speed is good, but having a lot of it just makes life harder, with no benefit.

It will also be very noisy, especially at bass frequencies around 50nV/√Hz and 60pA/√Hz, at 30Hz! The NE5532 with a suitable BJT follower on the output will be much less noisy and less prone to oscillation.

The LM7171 is the best integrated solution I've heard. On the noise side - it is not a problem in a line level amp, it is not a phono stage! The current noise is about 7pA/√Hz at 30Hz, not 60, and 2pA/√Hz at 1kHz, so a 10K resistor on the input will only increase the noise by about 3dB on top of the voltage noise, which is fine. I've designed a respected headphone amp using this chip about 20 years ago, nobody ever complained about the noise level  .

Cheers

Alex

[Zero999]

That's a video/RF op-amp. Using it for a headphone amplifier just seems like asking for trouble. 

Speed is good, but having a lot of it just makes life harder, with no benefit.

It will also be very noisy, especially at bass frequencies around 50nV/√Hz and 60pA/√Hz, at 30Hz! The NE5532 with a suitable BJT follower on the output will be much less noisy and less prone to oscillation.

The LM7171 is the best integrated solution I've heard. On the noise side - it is not a problem in a line level amp, it is not a phono stage! The current noise is about 7pA/√Hz at 30Hz, not 60, and 2pA/√Hz at 1kHz, so a 10K resistor on the input will only increase the noise by about 3dB on top of the voltage noise, which is fine. I've designed a respected headphone amp using this chip about 20 years ago, nobody ever complained about the noise level  .

Cheers

Alex

Yes, I got the decimal point wrong when looking at the current noise graph. I must've been looking at the scale for the voltage noise!

I agree, if it's not providing much voltage gain, then the noise is negligible. Still I think using a video amplifier IC for a headphone amplifier is asking for problems with oscillation. No doubt it can be done, but why bother? There are far more suitable solutions which are far more stable and are better suited to the application.

[Alex Nikitin]

No doubt it can be done, but why bother? There are far more suitable solutions which are far more stable and are better suited to the application.

The reason is very simple - the sound quality when used with various real world headphones.

Cheers

Alex

[Zero999]

No doubt it can be done, but why bother? There are far more suitable solutions which are far more stable and are better suited to the application.

The reason is very simple - the sound quality when used with various real world headphones.

Cheers

Alex

Too bad if there's a risk of RFI, especially with long cables.

There are far more suitable solutions, offering the same/better sound quality and a much lower risk of turning into an unintentinal radio transmitter.

[dietert1]

When i read claims that an amplifier sounds different, my first thought is: Maybe it got a problem. Any amplifier with low output impedance, a reasonably flat frequence plot and less than 0.1 % THD up to 10 KHz at somewhat elevated output levels sound the same even for a trained listener. That amplifier is easy to build using modern parts.
Most tube amplifiers sound "interesting" since they don't fulfil these criteria (higher output impedance and THD more like 1 or 2 %). Higher THD simulates higher sound level..

Regards, Dieter

[Zero999]

When i read claims that an amplifier sounds different, my first thought is: Maybe it got a problem. Any amplifier with low output impedance, a reasonably flat frequence plot and less than 0.1 % THD up to 10 KHz at somewhat elevated output levels sound the same even for a trained listener. That amplifier is easy to build using modern parts.
Most tube amplifiers sound "interesting" since they don't fulfil these criteria (higher output impedance and THD more like 1 or 2 %). Higher THD simulates higher sound level..

Regards, Dieter

I mostly agree but beware of THD figures which don't always tell the full story. Some types of distortion are noticeable at low levels of THD, such as crossover distortion, whilst others such as slew rate limit distortion aren't noticeable until much higher THD numbers. In this case, my suspicion is the LM7171 could be ringing at certain frequencies which could be seen as sounding brighter, boosting treble response.

[Alex Nikitin]

When i read claims that an amplifier sounds different, my first thought is: Maybe it got a problem. Any amplifier with low output impedance, a reasonably flat frequence plot and less than 0.1 % THD up to 10 KHz at somewhat elevated output levels sound the same even for a trained listener. That amplifier is easy to build using modern parts.
Most tube amplifiers sound "interesting" since they don't fulfil these criteria (higher output impedance and THD more like 1 or 2 %). Higher THD simulates higher sound level..

Regards, Dieter

Let's say my experience is different from yours.

Cheers

Alex

[David Hess]

I think the Rod Elliott circuit shown in the beginning is a little more rational. Depending on the OpAmp it will certainly outperform all those diamond circuits.

In this case, the diamond circuits are meant to be used within the feedback loop of an operational amplifier and act as a current amplifier.

My preference would be to use something based on the earlier examples I gave because full power bandwidth is increased, gain-bandwidth product is increased, and the output operates closer to the supply rails for increased available output power.  And this can be done with no significant increase in complexity compared to the class-ab output stage.

The disadvantage is that circuit operation is more complex and less applicable to learning how common audio power amplifiers work.


Distortion and noise will be higher on the LM7171 because of lower open loop gain and lower common mode and power supply rejection.  Perhaps more importantly, the output power is limited by limited output voltage range and limited power dissipation from a single 8 pin package.

Another solution which avoids using discrete parts is use several audio operational amplifiers in parallel to get the needed power dissipation.  If rail-to-rail output parts are used, then increased output power for a given supply voltage will be available also.  There are several ways to configure operational amplifiers in parallel.

[Audioguru]

I've looked at the OPA2134 and the fact the datasheet mentions 600ohm load and +/-18V VCC is very interesting.

However it mentions the voltage output to be max +/-2.5V at 600 ohm, which is I=U/R = 4mA. 4mA @ 2.5V is about 10mW, which seems low for those purposes.

No. the datasheet says the voltage loss when driving a 600 ohm load is the negative supply plus 2.2V to the positive supply minus 2.5V. So if the supply is +/-15V the output swing into 600 ohms is 25.3V p-p which is 9V RMS which is 135mW which would be very loud. Without the positive swing clipping a little more than the negative swing then the output power into 600 ohms with extremely low distortion is 130mW. 

[Alex Nikitin]

Distortion and noise will be higher on the LM7171 because of lower open loop gain and lower common mode and power supply rejection.  Perhaps more importantly, the output power is limited by limited output voltage range and limited power dissipation from a single 8 pin package.

The LM7171 in a DIP8 package can output about 0.7W RMS into a 100 Ohm load continuously  with a +/-15V supply, do you even need that much for a headphone amplifier? On the other hand, driving a speaker coil is not equivalent of driving a resistive load, that is one of reasons why amplifiers with similar traditional measurements could sound different in real life (or put it differently, why measurements don't tell a complete story) and why so many different amplifier designs exist.

Cheers

Alex

[Zero999]

Distortion and noise will be higher on the LM7171 because of lower open loop gain and lower common mode and power supply rejection.  Perhaps more importantly, the output power is limited by limited output voltage range and limited power dissipation from a single 8 pin package.

The LM7171 in a DIP8 package can output about 0.7W RMS into a 100 Ohm load continuously  with a +/-15V supply, do you even need that much for a headphone amplifier? On the other hand, driving a speaker coil is not equivalent of driving a resistive load, that is one of reasons why amplifiers with similar traditional measurements could sound different in real life (or put it differently, why measurements don't tell a complete story) and why so many different amplifier designs exist.

Cheers

Alex

Any measurements should be performed with the speaker or headphones connected to the amplifier, via a suitable length of cable. I'd say this is even more important when you're using an IC out of specification, as would be the case for the LM7171 driving headphones, rather than the purely resistive load it was designed for.

Yes, different amplifier designs exist for different purposes, which is why using a video amplifier for audio is daft, even if some people like the colourful ringing and distortion it might add to the sound.

Distortion and noise will be higher on the LM7171 because of lower open loop gain and lower common mode and power supply rejection.  Perhaps more importantly, the output power is limited by limited output voltage range and limited power dissipation from a single 8 pin package.

Yes, the LM7171 isn't suitable for use as a headphone amplifier. There are much better alternatives. Just because it can be done, it doesn't mean it's the best way.

Another solution which avoids using discrete parts is use several audio operational amplifiers in parallel to get the needed power dissipation.  If rail-to-rail output parts are used, then increased output power for a given supply voltage will be available also.  There are several ways to configure operational amplifiers in parallel.

Yes, it also results in lower noise and distortion.

[magic]

driving a speaker coil is not equivalent of driving a resistive load, that is one of reasons why amplifiers with similar traditional measurements could sound different in real life

True that, but after all those years of hearing about it I would like to finally see a demonstration of two amplifiers which both show no perceptible noise, 0.1dB gain flatness from 10Hz to 50kHz and some low THD figure, all measured into the actual demo speaker, and yet sound different enough that somebody (bonus: everybody) can tell them apart.

I'm not aware of any instances of such demonstration actually being done. I suppose it would be all over the Internet if it happened, given the amount of controversy.

[Alex Nikitin]

Any measurements should be performed with the speaker or headphones connected to the amplifier, via a suitable length of cable. I'd say this is even more important when you're using an IC out of specification, as would be the case for the LM7171 driving headphones, rather than the purely resistive load it was designed for.

Yes, different amplifier designs exist for different purposes, which is why using a video amplifier for audio is daft, even if some people like the colourful ringing and distortion it might add to the sound.

Distortion and noise will be higher on the LM7171 because of lower open loop gain and lower common mode and power supply rejection.  Perhaps more importantly, the output power is limited by limited output voltage range and limited power dissipation from a single 8 pin package.

Yes, the LM7171 isn't suitable for use as a headphone amplifier. There are much better alternatives. Just because it can be done, it doesn't mean it's the best way.

 

I personally find this amusing. I have shared some information which I consider valuable, as it is the result of many hours of work at the time I did that design. I am not trying to sell anything. I have a practical experience using that particular chip for that particular application in a production headphone amplifier which was produced for many years and in many thousands of units. And I get theoretical arguments from people who most likely never tried to work with that particular IC and not even read the data sheet carefully (as it clearly states the LM7171 can be used for driving ADSL lines and  transformers for instance). ADSL drivers in general are not a bad choice for headphone amplification (low distortion, high output current and ability to work on difficult loads) and there are designs which use them*.

Cheers

Alex

P.S. Worth noting that TI actually sells the THS6012 Dual Differential Line Driver as a "High Fidelity Headphone Amplifier" chip under a different name (TPA6120A2)  .

[Alex Nikitin]

driving a speaker coil is not equivalent of driving a resistive load, that is one of reasons why amplifiers with similar traditional measurements could sound different in real life

True that, but after all those years of hearing about it I would like to finally see a demonstration of two amplifiers which both show no perceptible noise, 0.1dB gain flatness from 10Hz to 50kHz and some low THD figure, all measured into the actual demo speaker, and yet sound different enough that somebody (bonus: everybody) can tell them apart.

I'm not aware of any instances of such demonstration actually being done. I suppose it would be all over the Internet if it happened, given the amount of controversy.

It is getting completely off-topic, however I will only ask a simple question: can you provide a link to an example of a statistically valid controlled DB listening test with a positive result?

Cheers

Alex

[Zero999]

I personally find this amusing. I have shared some information which I consider valuable, as it is the result of many hours of work at the time I did that design. I am not trying to sell anything. I have a practical experience using that particular chip for that particular application in a production headphone amplifier which was produced for many years and in many thousands of units. And I get theoretical arguments from people who most likely never tried to work with that particular IC and not even read the data sheet carefully (as it clearly states the LM7171 can be used for driving ADSL lines and  transformers for instance). ADSL drivers in general are not a bad choice for headphone amplification (low distortion, high output current and ability to work on difficult loads) and there are designs which use them*.

Cheers

Alex

P.S. Worth noting that TI actually sells the THS6012 Dual Differential Line Driver as a "High Fidelity Headphone Amplifier" chip under a different name (TPA6120A2)  .

I don't doubt that it's possible to build a headphone amplifier with the LM7171, just that it'll make life harder, with no benefits, over ICs specifically designed for it.

What "valuable information" have you posted? You've not posted any details of your design: schematic or layout. All you've done is said "I've used the LM7171, an IC which is not designed to be used as a headphone amp, as a headphone amp" and are wondering why people are questioning it. You haven't provided any solid reasons to use it, over the alternatives, in the face of valid arguments not to use it, as a headphone amp. This is not helpful, especially in the beginners section as someone with far less experience might try the LM7171 and wonder why it doesn't work!

driving a speaker coil is not equivalent of driving a resistive load, that is one of reasons why amplifiers with similar traditional measurements could sound different in real life

True that, but after all those years of hearing about it I would like to finally see a demonstration of two amplifiers which both show no perceptible noise, 0.1dB gain flatness from 10Hz to 50kHz and some low THD figure, all measured into the actual demo speaker, and yet sound different enough that somebody (bonus: everybody) can tell them apart.

I'm not aware of any instances of such demonstration actually being done. I suppose it would be all over the Internet if it happened, given the amount of controversy.

It is getting completely off-topic, however I will only ask a simple question: can you provide a link to an example of a statistically valid controlled DB listening test with a positive result?

Cheers

Alex

I'm not sure what you're asking? Double blind tests are a the gold standard for determining whether anything is detectable or effective. If your LM7171 amplifier is audibly transparent i.e. doesn't distort the sound, the difference between it and any other amplifier which is also audibly transparent, should be undetectable when subject to a double blind test.

[Alex Nikitin]

I'm not sure what you're asking? Double blind tests are a the gold standard for determining whether anything is detectable or effective. If your LM7171 amplifier is audibly transparent i.e. doesn't distort the sound, the difference between it and any other amplifier which is also audibly transparent, should be undetectable when subject to a double blind test.

So, you can not provide a link to a statistically valid DBT with a positive result (meaning that there is a difference in the sound confirmed in the test) ? So, you have a "gold standard" with negative results only, don't you think it is funny?

Cheers

Alex

[Zero999]

I'm not sure what you're asking? Double blind tests are a the gold standard for determining whether anything is detectable or effective. If your LM7171 amplifier is audibly transparent i.e. doesn't distort the sound, the difference between it and any other amplifier which is also audibly transparent, should be undetectable when subject to a double blind test.

So, you can not provide a link to a statistically valid DBT with a positive result (meaning that there is a difference in the sound confirmed in the test) ? So, you have a "gold standard" with negative results only, don't you think it is funny?

Cheers

Alex

I still don't know what you're going on about. Please provide a more detailed explanation of what you're asking.

[Alex Nikitin]

I'm not sure what you're asking? Double blind tests are a the gold standard for determining whether anything is detectable or effective. If your LM7171 amplifier is audibly transparent i.e. doesn't distort the sound, the difference between it and any other amplifier which is also audibly transparent, should be undetectable when subject to a double blind test.

So, you can not provide a link to a statistically valid DBT with a positive result (meaning that there is a difference in the sound confirmed in the test) ? So, you have a "gold standard" with negative results only, don't you think it is funny?

Cheers

Alex

I still don't know what you're going on about. Please provide a more detailed explanation of what you're asking.

OK, let's go back to your earlier post:

I'm not sure what you're asking? Double blind tests are a the gold standard for determining whether anything is detectable or effective. If your LM7171 amplifier is audibly transparent i.e. doesn't distort the sound, the difference between it and any other amplifier which is also audibly transparent, should be undetectable when subject to a double blind test.

The DBT is a tool. To use any tool you should understand its capabilities and limitations, otherwise you can not trust the results. My question is very simple - could you provide an example of a statistically valid DBT with a positive result, i.e. where the difference between the sound from two pieces of equipment was registered. Clearly, if you have 50% distortion on one of the sources it should give a good positive result, but no one will bother with this kind of test. However unless we have examples of positive results it is not clear where the test sensitivity border line is. Saying that it is "industry standard", "gold standard" and such doesn't prove anything, it is only a convenient smoke screen. DBTs in audio are notoriously insensitive, that is why their use is essentially a biggest can of a snake oil in this industry, IMHO. Well, again, could you provide a link to an example of a statistically valid audio DBT with a positive result?

Cheers

Alex

[ratatax]

I know your suggestions here may be using higher quality parts, but what about this one :

http://www.ti.com/lit/ds/symlink/rc4580.pdf

It's cheap, has some oomph to be used as a headphones amplifier, has low enough THD/noise... and the implementation seems dead easy.

I'm a musician but I like when choices are made with measurable, scientific proofs, since audio as it is perceived by our ears is way too much subjective. Some opamps with thousand of V/uS slew rate and hundred of Mhz bandwidth may sound very good but I feel it's just unneeded complications (unneeded costs too), like you can buy a race car to drive at 30mph in a city...but WHY ? 

[Zero999]

I'm not sure what you're asking? Double blind tests are a the gold standard for determining whether anything is detectable or effective. If your LM7171 amplifier is audibly transparent i.e. doesn't distort the sound, the difference between it and any other amplifier which is also audibly transparent, should be undetectable when subject to a double blind test.

So, you can not provide a link to a statistically valid DBT with a positive result (meaning that there is a difference in the sound confirmed in the test) ? So, you have a "gold standard" with negative results only, don't you think it is funny?

Cheers

Alex

I still don't know what you're going on about. Please provide a more detailed explanation of what you're asking.

OK, let's go back to your earlier post:

I'm not sure what you're asking? Double blind tests are a the gold standard for determining whether anything is detectable or effective. If your LM7171 amplifier is audibly transparent i.e. doesn't distort the sound, the difference between it and any other amplifier which is also audibly transparent, should be undetectable when subject to a double blind test.

The DBT is a tool. To use any tool you should understand its capabilities and limitations, otherwise you can not trust the results. My question is very simple - could you provide an example of a statistically valid DBT with a positive result, i.e. where the difference between the sound from two pieces of equipment was registered. Clearly, if you have 50% distortion on one of the sources it should give a good positive result, but no one will bother with this kind of test. However unless we have examples of positive results it is not clear where the test sensitivity border line is. Saying that it is "industry standard", "gold standard" and such doesn't prove anything, it is only a convenient smoke screen. DBTs in audio are notoriously insensitive, that is why their use is essentially a biggest can of a snake oil in this industry, IMHO. Well, again, could you provide a link to an example of a statistically valid audio DBT with a positive result?

Cheers

Alex

Yes, a DBT is a tool. If the results are that no one can differentiate between two amplifiers, any better than chance, then the two amplifiers both sound the same. This is a scientific fact. If no one can tell the difference between the two amplifiers, then saying the they are sonically different is lying. The same is true of a quack claiming their new spiritual healing treatment cures cancer: without a DBT it's total lies.

No doubt two amplifiers can sound different and will give a positive result when subject to a DBT. This is likely to be the case if the amplifiers are over-driven or genuinely do produce distortion in normal operation. Some people might like certain types of distortion, so keeping THD figures down isn't always the goal.

The idea people can hear things which can't be measured is snake oil. It's true the human ear is more sensitive to some types of distortion than others, so THD figures alone aren't always that useful, but everything which is audible is measurable and is the whole point in having expensive test equipment.

No doubt tests have been done to determine what level of distortion is audible, tolerable, and immediately noticeable, in order to develop efficient lossy audio compression algorithms and determine the bare minimum standard for an audible transparent audio amplifier.

To answer the question: no, I've not seen an example statistically valid DBT with a positive result, comparing two amplifiers, but that doesn't mean to say that DBTs aren't valid. It most likely means that all decent amplifiers quality sound the same, as long as they're operated within their specifications.

I have watched Youtube videos demonstrating what levels of THD, of different types of distortion, people can hear and it would be interesting to see some more scientific DBTs.

[Alex Nikitin]

Yes, a DBT is a tool. If the results are that no one can differentiate between two amplifiers, any better than chance, then the two amplifiers both sound the same. This is a scientific fact. If no one can tell the difference between the two amplifiers, then saying the they are sonically different is lying. The same is true of a quack claiming their new spiritual healing treatment cures cancer: without a DBT it's total lies.

No doubt two amplifiers can sound different and will give a positive result when subject to a DBT. This is likely to be the case if the amplifiers are over-driven or genuinely do produce distortion in normal operation. Some people might like certain types of distortion, so keeping THD figures down isn't always the goal.

The idea people can hear things which can't be measured is snake oil. It's true the human ear is more sensitive to some types of distortion than others, so THD figures alone aren't always that useful, but everything which is audible is measurable and is the whole point in having expensive test equipment.

No doubt tests have been done to determine what level of distortion is audible, tolerable, and immediately noticeable, in order to develop efficient lossy audio compression algorithms and determine the bare minimum standard for an audible transparent audio amplifier.

1) Please note that you "have no doubts" but you couldn't provide even a single link, only generic blah-blah. You should try to find an example I asked for, you might be in for a surprise.

2) Nowhere I said that "people can hear things which can't be measured" , that is you putting words in my mouth  , and I disagree with this idea. I like measuring things, moreover, my background in electronics is in the test and measurement .

Cheers

Alex

P.S.

To answer the question: no, I've not seen an example statistically valid DBT with a positive result, comparing two amplifiers, but that doesn't mean to say that DBTs aren't valid. It most likely means that all decent amplifiers quality sound the same, as long as they're operated within their specifications.

And that is not science, but a pure act of faith  .

[dietert1]

Can't you just continue that discussion in private communication? I am afraid you are trolling this thread with your audiophoolisms.

Regards, Dieter