Proving amplifiers are underpowered

[Kyron]

I'm new to Audio Science Review so I'm not really sure where to post this.

I hear and see a lot of claims that even a few watts is not enough to get the best out of some headphones. How can someone prove that an amplifier is underpowered for that headphone?
Would an underpowered amplifier give problems with a tone burst apart from obvious clipping in that it doesn't produce a constant output?
If you don't have access to expensive equipment what would be a good way to check the waveform?
Consider a set up where the amplifier is driving a headphone with a test signal and split the signal to the headphone with the other side feeding into a audio interface to capture the waveform. Is there a better way to do this?

 

[abdo123]

If you don't have access to expensive equipment what would be a good way to check the waveform?

an osciloscope can be as cheap as 20$.

 

[audio2design]

There is a process here:

1) Listen to a piece of music at the max volume you want to.
2) Analyze the music file for the peak value.
3) Use a program to generate tones at that peak value (making sure you are not screwing up volume controls etc.)
4) Analyse distortion at various frequencies at this level. Ideally you would do IMD tests to but start simple.

A scope is crude for doing this test. A loop-back box will give much better test capability but depending on the amp and how grounded and your sound card may not work. Either way you can accomplish much without a high end instrument. Even a cheap ADC with an isolated USB can allow you to do pretty extensive testing.

 

[levimax]

This will work better than an Osciliscope and not expensive https://sound-au.com/project191.htm

 

[raindance]

Use your ears. If you listen at a high volume and there's a reduction in bass and/or a lot of harshness added relative to lower volumes, maybe the amp is underpowered. Of course maybe the listener has hearing damage and they're simply trying to listen louder than the headphone is capable of going...

Knowing the sensitivity rating, impedance curve and max spl capability of the headphone could be useful also, but only if you have the ability to measure the acoustic level that's giving you issues.

Typically headphones have a pretty benign impedance curve and sensitivity is the biggest variable. Amplifier output impedance is something else to look out for as a high output impedance amp into a low impedance headphone acts as a voltage divider and effectively "throws away" volts which may result in distortion occurring due to clipping before the amp even reaches full output (from the headphone's perspective - read this carefully ).

As with all things audio, YMMV.

 

[audio2design]

This will work better than an Osciliscope and not expensive https://sound-au.com/project191.htm

This will not work. There are too many other variables.

 

[levimax]

This will not work. There are too many other variables.

OK you are refuting a very knowledgeable person who took time to write a long explanation of why it will work. Please elaborate why it wont work.

 

[anphex]

As far as I've seen so far about any amp Amir has tested shows that the performance regarding SINAD, IMD, crosstalk etc keeps getting better right up to the clipping point where a sharp noticable decline in quality happens. When looking at the average sensivity, a few hundred milliwatt will drive even the most picky headphones while the amp stay in it's "happy range". I'd even say that using an overpowered amp and driving it at the 0,01-0,1% range of it's max power is a very bad idea because you'll just driving it at its noise and distortion floor.

It's best to match the amp to your headphone because unlike speakers where we talk about impedance minimum and maximum gaps of a few Ohms usually, with headphones the jumps are far more dramatic. Amir shows how much power you need to reach 94dbSPL (which is already really really loud and defintely not healthy damn!) in graphs. That's a great baseline to start looking for a amp/headphone pair.

But honestly, 1 Watt is already more than enough for 99,99% of all headphones for human listening levels. I swear before you reach 94dbSPL on any music you'll rip the headphones off your head and throw them away.

 

[DVDdoug]

This will work better than an Osciliscope and not expensive https://sound-au.com/project191.htm

If you were going to build something yourself, and if you can tap-into the amplifier's power supply, you can easily make an op-amp comparator (which compares the output voltage to the power supply voltage). When the output voltage gets close* to the power supply voltage, you turn-on an LED. Then, you would want something additional to "stretch" the pulse. I built this into an amp that I built long ago and I've forgotten the details but there was probably a 555 timer to hold the LED on for a second or so. A comparator is one if the easiest things to make with an op-amp, but op-amps are typically limited +/-15V, so you'd need some resistors for voltage dividers (for both the reference and the audio output). Plus, you need a low-voltage power supply for an op-amp.

It would be cheap and easy for the manufacturer to add a clipping indicator... A lot easier than adding one to an existing amplifier...


* There is a voltage drop across the MOSFETs/transistors so the output can't exactly hit the supply voltage, and there may be some non-linearity at the limits so it my start to soft-clip before it hard-clips.

 

[NTK]

OK you are refuting a very knowledgeable person who took time to write a long explanation of why it will work. Please elaborate why it wont work.

For example, an amplifier may be able to output, say, 40 V into a speaker at its impedance peak of 40 ohm, but may clip well below at the speaker's impedance minimum of 2 ohm.

 

[audio2design]

This will not work. There are too many other variables.

The circuit works, the methodology does not. Capturing peak current and voltage provides you some detail but not enough to do what the op wants. They need to be simultaneous measurements.

 

[levimax]

For example, an amplifier may be able to output, say, 40 V into a speaker at its impedance peak of 40 ohm, but may clip well below at the speaker's impedance minimum of 2 ohm.

The tester I linked to records both peak voltage and peak current so if both are under the capability of the amp it seems like you should be good but since you don't know "when" peak voltage vs peak current occurred it is kind of weird. The comparator mentioned by @DVDdoug would have the same issue you mention i.e. checks for voltage near the rails but not current. Even the procedure mentioned by @audio2design would have the speaker impedance issue.

 

[audio2design]

If you were going to build something yourself, and if you can tap-into the amplifier's power supply, you can easily make an op-amp comparator (which compares the output voltage to the power supply voltage). When the output voltage gets close* to the power supply voltage, you turn-on an LED. Then, you would want something additional to "stretch" the pulse. I built this into an amp that I built long ago and I've forgotten the details but there was probably a 555 timer to hold the LED on for a second or so. A comparator is one if the easiest things to make with an op-amp, but op-amps are typically limited +/-15V, so you'd need some resistors for voltage dividers (for both the reference and the audio output). Plus, you need a low-voltage power supply for an op-amp.

It would be cheap and easy for the manufacturer to add a clipping indicator... A lot easier than adding one to an existing amplifier...


* There is a voltage drop across the MOSFETs/transistors so the output can't exactly hit the supply voltage, and there may be some non-linearity at the limits so it my start to soft-clip before it hard-clips.

Think of it in reverse, not something that turns on when the voltage is close to the rail, but something that turns off and then float it from the upper rail with BJTs/FETs. Won't be as accurate as an op-amp, but the difference in loudness between 34V from a 38V rail and 35V from a 38V rail is very small in db and the circuit is simpler, though you can get op-amps with large common mode ranges.

 

[NTK]

The tester I linked to records both peak voltage and peak current so if both are under the capability of the amp it seems like you should be good but since you don't know "when" peak voltage vs peak current occurred it is kind of weird. The comparator mentioned by @DVDdoug would have the same issue you mention i.e. checks for voltage near the rails but not current. Even the procedure mentioned by @audio2design would have the speaker impedance issue.

Yes. Rod's circuit captures the peak voltage and peak current capability. But it is highly unlikely that an amplifier can output both simultaneously.

 

[audio2design]

The tester I linked to records both peak voltage and peak current so if both are under the capability of the amp it seems like you should be good but since you don't know "when" peak voltage vs peak current occurred it is kind of weird. The comparator mentioned by @DVDdoug would have the same issue you mention i.e. checks for voltage near the rails but not current. Even the procedure mentioned by @audio2design would have the speaker impedance issue.

Speaker Impedance is covered by,
"4) Analyse distortion at various frequencies at this level. Ideally you would do IMD tests to but start simple.
"

An excessive voltage peak is pretty much always bad. A current peak is much harder to interpret. Low feedback could add frequency dependencies.

 

[levimax]

Yes. Rod's circuit captures the peak voltage and peak current capability. But it is highly unlikely that an amplifier can output both simultaneously.

Since we are only trying to determine if the amp is clipping when playing music at a given level then isn't knowing that neither peak current nor peak voltage is exceeding the amps capability enough even though it is not as satisfying as knowing "peak power"? I don't know just trying to learn.

 

[levimax]

Speaker Impedance is covered by,
"4) Analyse distortion at various frequencies at this level. Ideally you would do IMD tests to but start simple.
"

An excessive voltage peak is pretty much always bad. A current peak is much harder to interpret. Low feedback could add frequency dependencies.

The only issue I see with #4 is that it would be a very loud or even possibly damaging test with speakers.... headphones per OP you would be OK I guess, or are you talking about measuring at this peak level into a dummy load? If a dummy load then we are back to the variable speaker impedance issue. Maybe I am missing something obvious.

 

[NTK]

Since we are only trying to determine if the amp is clipping when playing music at a given level then isn't knowing the neither peak current nor peak voltage is exceeding the amps capability enough even though it is not as satisfying as knowing "peak power"? I don't know just trying to learn.

I can think of a couple of ways.

1. If the source is digital, then the max source level into the amplifier is deterministic. May be we need to add a few dB of margin for inter-sample overs. All that is remaining is to calculate, based on the amplifier gain, what the output voltage would be, and see if the amplifier can output that voltage into the minimum load impedance (or whatever makes sense if the minimum load impedance is at a very high frequency where signal level is generally low).
   To determine the max source level, we can set the source's volume knob to the max level the listener will use, then measure the output voltage with a digital full scale signal.
2. Record the input and output signals to the amp. Normalize the levels and use DeltaWave to compare them. Increase the levels until we see a significant distortion (PK metric) increase. We can use real music for this test.

 

[Cbdb2]

Yes. Rod's circuit captures the peak voltage and peak current capability. But it is highly unlikely that an amplifier can output both simultaneously.

Of course it can and does, always into a resistive load, and at the frequency the phase of the impedance is zero in complex loads.