Headphone amp with 0.000021% distortion

[tk@halmi]

[quote author="mcs"]I just noticed this bit: "0.000021% distortion"
Mikkel C. Simonsen (who uses amps with a lot more than 0.000021% distortion :green[/quote]

It is one of the few objective figures they can publish that buyers will pay attention to. See they got your attention in a way. Would you buy an opamp that states on the datasheet that: "This product sounds good", period?
There has not been much progress in defining what sounds good since it depends so much on taste, IMHO.
How would I express through quantitive analysis that a paiting looks good to me? Haven't a clue. :grin:

 

[mcs]

[quote author="tk@halmi"]Would you buy an opamp that states on the datasheet that: "This product sounds good", period?[/quote]
Why not? I don't really worry if the distortion is 0.1 or 0.000000001%. Other things like noise and bandwidth seem a lot more important.

Best regards,

Mikkel C. Simonsen

 

[tk@halmi]

[quote author="mcs"]Why not? I don't really worry if the distortion is 0.1 or 0.000000001%. Other things like noise and bandwidth seem a lot more important.
Mikkel C. Simonsen[/quote]

Aha! Specs ARE impotant to you. If not distortion something else you deem important. Some care about distortion numbers more than others. Get my drift?

Cheers,
Tamas

 

[mcs]

Yes. And I do also care about the distortion - but more about the level of the different harmonics. If the distortion is mostly 2. harmonic, a lot of it doesn't matter much. SE tube amps with a lot of 2. harmonic distortion (and not much else) sound very nice...

Best regards,

Mikkel C. Simonsen

 

[Guest]

[quote author="PRR"]> According to many studies everything below 1% in inaudible...
Many studies are wrong.[/quote]
W R O N G ? It is not academical word. What about distortion
of transducers. You must use some transducers to make psychoacoustic
study. And what about aural harmonics, distortion in the ear,
read von Bekesy. And von Bekesy is wrong, some about interspike
intervals and cochlear implantates ? Von Bekesy, stand up from
the coffin and reject Nobel price !!!
And how to measure 0,000... distortion? self oscilation of amplifiers
and air coils to reject fundamental sinusoid ? , what about residual noise,
shielding ?
Julian Dunn s designs of Audio Precisions apparatus are wrong ... .

?
?
?

Nothing in the science is wrong. We must only know WHAT is observed.
And must have comparations, normals and so on... .
Yes, in some papers can be observed, that 5 percent of distortion in
the amplifier is nearly inaudible.

That we must know, that it was pentode amplifier, testing record was
picked-up from vinyl-plate and distortion of speaker was 8 percent.

The paper is not wrong, but we gets 5 percent and with out - of context
we make some norm, i.e. DIN 45 500.

No scientific papers, but putting data out of context is WRONG.

xvlk

 

[Steve Jones]

Per, we don't see much surface mount here at the Lab, what you have done looks great. Can you tell us some of your opinions, pro's and cons of working in this format in terms of having boards made, finding parts, soldering and desoldering, all that stuff?

Sorry if I am picking your brains, but if you have time to give us some insight and advice it would really be appreciated!

 

[PRR]

> putting data out of context is WRONG.

Yes.

And any single-number data for "THD" is outside the context of human hearing. THD is convenient for technicians, but inappropriate for suggesting what it sounds like.

I will add that this point-zero-zero-zero-percent number does suggest that whatever distortion is present is "inaudible". The ear can be acutely sensitive to very small amounts of high-order distortion, but some modern gear (including this one) really does make distortion a non-issue.

 

[owel]

I'd love to hear how you handle those tiny resistors, what soldering eqpt you use... I assume you work with magnifying glasses?

Your project looks very pro!

 

[dale116dot7]

That amp also has gain and bandwidth up to well over 100 MHz from the datasheets. Part of the ultra-low THD measurement is interesting as it points to other characteristics of the design. It's an ultra-fast, ultra high-gain setup and that is interesting from a 'transparent' point of view in other ways than just a low THD number. It actually looks like a made-for-audio version of a high-speed video line driver sort of amplifier.

Wrong science example #1: Polywater?

 

[peranders]

[quote author="mcs"]I just noticed this bit: "0.000021% distortion"

Why should that matter? According to many studies everything below 1% in inaudible...

Best regards,

Mikkel C. Simonsen (who uses amps with a lot more than 0.000021% distortion :green[/quote]
Wow, this was a hot statement!

I personally believe that under 0.1% is desirable. The 0.000021% dist was only interesting from a technical point of view. It's an achievement to measure this!

The only thing I care about is that the TPA6120 is very good IC for driving headphones.

 

[peranders]

[quote author="dale116dot7"]It actually looks like a made-for-audio version of a high-speed video line driver sort of amplifier.
[/quote]
Indeed, what was the difference really? Unfortunately I have forgot the name of the driver which was nearly exactly the same as TPA6120.

 

[peranders]

[quote author="Steve Jones"]Per, we don't see much surface mount here at the Lab, what you have done looks great. Can you tell us some of your opinions, pro's and cons of working in this format in terms of having boards made, finding parts, soldering and desoldering, all that stuff?

Sorry if I am picking your brains, but if you have time to give us some insight and advice it would really be appreciated![/quote]
At work I use SMD parts since many years back and so this is everyday stuff. Mosts parts are nothing special.

The pcb itself is equally easy to make as a regular holemounted one. This pcb I have soldered by hand 1-2 hours and it's pretty easy if you have two soldering irons, just warm the 0805 parts at both ends at the same time. When I move away the irons the part is sucked into place from strong fources in the melt tin.

The reason why I like SMD is becuase it cute and also very unusual in DIY circles. It's also pretty stupid if I want to sell any pcb's. The advantage may be if I wanted to manufacture the boards.

 

[Guest]

perandersthe TPA6120 is very good IC for driving headphones.

what headphones? open ribbons?
And space-shuttle is very good tool for everyday commuting to the school.
xvlk

 

[peranders]

hm... what do you mean?

 

[mcs]

[quote author="peranders"]I personally believe that under 0.1% is desirable. The 0.000021% dist was only interesting from a technical point of view. It's an achievement to measure this![/quote]
Yes, it is interesting from a technical point of view - I just wouldn't have mentioned it. All audio chips on the market have a distortion thats "low enough" - unless they are made for portable AM radios or toys... I would just have measured it, found out it was below 0.05% or so, and concluded that the circuit works.

The only thing I care about is that the TPA6120 is very good IC for driving headphones.

That is the important part!

Best regards,

Mikkel C. Simonsen

 

[dale116dot7]

I have (yea, yea, I know) an Apex AHA4 headphone amp. The specifications say something about 0.01% THD and 0.01% IMD, and maximum output of 20 dB.

But I have a couple of customers that could hear distortion, and after going through everything (microphone, preamp, converter, etc), I replaced the headphone amp with something real (a 60 watt monitor amp). They were much happier!

After doing some digging, I found that the AHA4 used TL072 op-amps to drive the headphones directly. No power transistors, no buffers. Despite the specifications of maximum output level at 20 dB (relative to what?) I could scarcely get 2 volts P-P out of the thing. Maybe relative to 100 mV? No wonder the customer could hear distortion - it was clipping! The culprit? TL072 op-amps are not designed to drive headphones. They just can't do it.

I can see another DIY project here...

 

[mcs]

A TL072? They can't even drive a 600 Ohm load properly, can they?

Best regards,

Mikkel C. Simonsen

 

[PRR]

> That we must know, that it was pentode amplifier, testing record was picked-up from vinyl-disk and distortion of speaker was 8 percent.

Consider this: can you hear the difference between a bassoon and a power-saw, both playing the same note?

Now have an orchestra playing, then add either one bassoon or one power-saw.

The mellow bassoon can probably play at 8% of total loudness with little difference in sound. Its frequencies blend with the orchestra.

0.1% of power-saw loudness behind the orchestra, even if on-pitch, will be very annoying. Its frequencies include high partials which do not blend with the orchestra.

No-feedback tubes, phono-disks, and speakers mostly make "mellow" distortion. 1%, even 5% of that is not annoying.

Medium-feedback amplifiers suppress their own distortion, making the total errors small, but leaving a residual of harsh grating errors at frequency far removed from the original signal. When you put in 220Hz, and get out 220Hz plus a 2,860Hz tone, your ear notices, even if the 2,860 tone is extremely small. There is very little annoyance from a small (1%) 440Hz tone behind a 220Hz tone.

5% 2nd harmonic is small distortion. 0.5% of 11th harmonic is very annoying.

We also need to know the distortion frequency response. Many classic "0.01%THD" amps had a good spec at 1KHz, but rising to 0.1% or more at 10KHz. While the musical power above 1KHz is small, the -rise- of harmonic spectrum is very un-natural to the ear, and much more annoying than the 1KHz THD spec suggests.

So Total Harmonic Distortion is a very poor indicator of sound, unless the -shape- of the distortion is specified or known.

The chip Peranders is using has both an exceptionally low THD number (0.000021% of anything is still very small!) and a bandwidth far beyond the audio band (it is not likely to have a rise of distortion in the treble).

It is possible (though unlikely) to have an amp with low single-tone distortion (THD) and high multi-tone distortion (IMD). Since music is multi-tone, that would be bad. But most of the things you do to reduce THD also reduce IMD (though not as much). Also we know this chip is really a modified data line driver, an application where IMD is commercially bad (high IMD limits the number of channels you can put on one line without crosstalk). In fact IMD will be low far above the audio band, so any supersonic garbage that leaks into the audio will not alias into the sound.

 

[PRR]

> A TL072? They can't even drive a 600 Ohm load properly, can they?

Not to Pro line level, and not at lowest THD.

They do work fine for driving many 300Ω-600Ω headphones. TL072 can put over 6V peak in 600Ω, 40mW, which is ample for many headphone uses.

They suck bad in typical modern 32Ω headphones: about 0.5V RMS or about 8mW. Enough for casual background listening but inadequate for critical listening in most headphones.

> I can see another DIY project here...

www.headwize.com is full of such projects.

Several cheap chips blow-away the TL072 in low-Z loads. Even the antique 5532 is a worthy headphone driver. A few newer chips will deliver the 100mA or more needed for really-ample level in 32Ω phones. And then of course there are discrete designs, or bigger chips, or truly "generous" designs like single-ended Class-A resistance-coupled. At headphone levels you can even do wasteful things with vacuum tubes and not end up TOO big or hungry.

 

[Guest]

[quote author="PRR"]
We also need to know the distortion frequency response. Many classic "0.01%THD" amps had a good spec at 1KHz, but rising to 0.1% or more at 10KHz. While the musical power above 1KHz is small, the -rise- of harmonic spectrum is very un-natural to the ear, and much more annoying than the 1KHz THD spec suggests.
[/quote]
It is efect of dominant pole to the in feedback gain.
Operational amplifiers was originaly designed for automatisation,
anolog computers, and so on. They are not intended to use in audio
circuits. They distortion then MUST be unnatural.
Why we still use op amps. If there are transimpedance amplifiers, current
conweyors and other devices, which uses integrated circuit technology more effectively than op amps. Superior performance of the diamond buffer is an example.
Operational amplifiers are like a relict. We must stop buying it and
buy something never. And feed producers to stop produce really
obsolete devices based on obsolete principles.
Or to be hybernated ....
xvlk




So Total Harmonic Distortion is a very poor indicator of sound, unless the -shape- of the distortion is specified or known.

The chip Peranders is using has both an exceptionally low THD number (0.000021% of anything is still very small!) and a bandwidth far beyond the audio band (it is not likely to have a rise of distortion in the treble).

It is possible (though unlikely) to have an amp with low single-tone distortion (THD) and high multi-tone distortion (IMD). Since music is multi-tone, that would be bad. But most of the things you do to reduce THD also reduce IMD (though not as much). Also we know this chip is really a modified data line driver, an application where IMD is commercially bad (high IMD limits the number of channels you can put on one line without crosstalk). In fact IMD will be low far above the audio band, so any supersonic garbage that leaks into the audio will not alias into the sound.[/quote]