Relation between THD+N and Hiss

[Genzopir]

I noticed that in many Class D amps, THD+N ratio gets better as volume is turned up but what I listen in real world is that hiss noise gets louder when I turn up the volume.
Another thing I noticed is as you can see above, THD+N ratio drops by 10dB as power goes up by about 10 dB which indicates the absolute value of THD+N is not changing.
Thus, any noise including hiss shouldn't get louder regardless of volume but it is not.
Can anyone tell me what's going on?

 

[McFly]

Because power output is not similar to you turning up the volume with nothing playing, well not exactly. The ratio on those graphs is the distance between an actual signal being played and its level relative to the noise floor. So a 1khz test tone being turned up gets way louder / further away from the noise floor. The noise floor is always the same absolute level irrespective of what volume level you are playing. Integrated amplifiers are odd because the preamp stage amplifies the noise floor of itself, and the source connected to it. Then you're adding on the noise floor of the power amp stage.

Take power amplifiers for example, to take preamplifiers + varying volume out of the equation. If you connect a speaker driver to its output terminals, you will always hear the same level of hiss (noise floor) during silence, no matter what level you put into the inputs of the power amplifier. It's then if you put a full-scale signal into those inputs the output at the speaker driver will be so fkn loud that its output is so many dB (as per above graphs) above the fixed noise floor.

Someone more mathematically inclined will probably have a better way of explaining it than I.

 

[mdsimon2]

I noticed that in many Class D amps, THD+N ratio gets better as volume is turned up but what I listen in real world is that hiss noise gets louder when I turn up the volume.

What are you using for volume control?

Another thing I noticed is as you can see above, THD+N ratio drops by 10dB as input power goes up by about 10 dB which indicates the absolute value of THD+N is not changing.
Thus, any noise including hiss shouldn't get louder regardless of volume but it is not.

In general noise will dominate THD+N at lower power output levels and will be constant. At higher power output distortion tends to dominate. SNR or DR are better metrics for noise than THD+N.

Michael

 

[dorakeg]

Power amps themselves can be very quiet, however, they are still dependent on the noise from other sources (like preamp, DAC etc).

So when you turn up the volume, this noise gets amplified from the preamp and it's fed into the power amp. End result is the hiss becomes louder.

Even integrated amps also the same, they are just pre/power sections housed in the same chassis.

 

[Waxx]

But some class D amps that should be silent, still have that hiss. A SET on a compression driver (that is in general very sensitive) is mostly silent, while we all know that those are not low distortion at all. But when i connect most TPA3118 based class D chipamps to the same compression driver, it has a constant hiss, even if no source is connected to the that power amp and they measure a lot cleaner than any SET ever will be. Top level class D amps like NCore or Purifi don't have that, and not all cheaper class D amps have that. But many do.

For someone with very sensitive ears like me (i can hear very silent noises) it's very irritating and a deal breaker. I would also like to know why this is and how we can measure that. I did not see any indicators in the measurements done here that show me if it is there.

 

[Blumlein 88]

But some class D amps that should be silent, still have that hiss. A SET on a compression driver (that is in general very sensitive) is mostly silent, while we all know that those are not low distortion at all. But when i connect most TPA3118 based class D chipamps to the same compression driver, it has a constant hiss, even if no source is connected to the that power amp and they measure a lot cleaner than any SET ever will be. Top level class D amps like NCore or Purifi don't have that, and not all cheaper class D amps have that. But many do.

For someone with very sensitive ears like me (i can hear very silent noises) it's very irritating and a deal breaker. I would also like to know why this is and how we can measure that. I did not see any indicators in the measurements done here that show me if it is there.

Can you give specific models that do this? I have some experience with well made class D amps from a few manufacturers and they are silent as the tomb. Also what other gear is connected from source to speaker. If you have a source that is noisy then the amp is amplifying that noise.

Measurements here would find amps that are in themselves too noisy. SNR should give you the info you need. Sometimes Amir shows THD+N vs power for amps and this too should show if something is getting near audible at upper frequencies.

 

[restorer-john]

I would also like to know why this is and how we can measure that. I did not see any indicators in the measurements done here that show me if it is there.

Residual noise is the issue. It should be specified as a number in V (uV) and the bandwidth it is specified over. The trouble is, the spectrum of that residual noise is never mentioned.

 

[restorer-john]

SNR should give you the info you need.

You can tease out the residual noise metric, but it depends on the spectrum. Is it white, pink, dominated by mains spuriae we can't hear, or are there idle tones/HF noise? Like tape hiss, it's not white in its spectral distribution. Some noise is way more offensive than other noise.

 

[Blumlein 88]

You can tease out the residual noise metric, but it depends on the spectrum. Is it white, pink, dominated by mains spuriae we can't hear, or are there idle tones/HF noise? Like tape hiss, it's not white in its spectral distribution. Some noise is way more offensive than other noise.

Yes all true. I don't think any class D amps from Hypex, Purifi or ICE modules are going to be a noise problem unless something seriously incompetent has been done to them. Some of those do list output idle noise in microvolts over 22 khz.

 

[dshreter]

View attachment 255986View attachment 255991

I noticed that in many Class D amps, THD+N ratio gets better as volume is turned up but what I listen in real world is that hiss noise gets louder when I turn up the volume.
Another thing I noticed is as you can see above, THD+N ratio drops by 10dB as power goes up by about 10 dB which indicates the absolute value of THD+N is not changing.
Thus, any noise including hiss shouldn't get louder regardless of volume but it is not.
Can anyone tell me what's going on?

What equipment are you using where you hear this? DAC, pre-amp, and amp? And which box is being used to adjust volume?

 

[solderdude]

I noticed that in many Class D amps, THD+N ratio gets better as volume is turned up but what I listen in real world is that hiss noise gets louder when I turn up the volume.

It may have nothing to do with class-D (or any class).
Perhaps what you are hearing is recording noise in music which is almost always louder than amplifier noise. You also amplify the noise as it is part of the recording.
When you hear more noise (without any music) when turning up the volume may well be background noise of the electronics before the volume control being amplified.

 

[Sokel]

Some of those do list output idle noise in microvolts over 22 khz.

Yes and it's useful to people with very sensitive speakers:

 

[Sokel]

Residual noise is the issue. It should be specified as a number in V (uV) and the bandwidth it is specified over. The trouble is, the spectrum of that residual noise is never mentioned.

Like this or you mean something more?

 

[Waxx]

Can you give specific models that do this? I have some experience with well made class D amps from a few manufacturers and they are silent as the tomb. Also what other gear is connected from source to speaker. If you have a source that is noisy then the amp is amplifying that noise.

Measurements here would find amps that are in themselves too noisy. SNR should give you the info you need. Sometimes Amir shows THD+N vs power for amps and this too should show if something is getting near audible at upper frequencies.

There was no gear connected, it was a Sure board with a TPA3118 (don't remember the exact partnr) based amp straight too a B&C DE250 driver in a XT1086 horn with a 47uF film cap in series to protect the driver. It was just to measure the driver in the horn that we connected it that way, but before we connected the laptop as source, the hiss was already there and rather loud. The Sophia Electric 300B SET amp the owner uses for his horns don't have that. Nor did my marantz PM5004 (class AB Transistor) that we used to measure at the end...

 

[pjug]

I noticed that in many Class D amps, THD+N ratio gets better as volume is turned up but what I listen in real world is that hiss noise gets louder when I turn up the volume.

Forgive me if I am not getting what you are saying, but I think you may have a misunderstanding about what the power vs THD+N curve shows. With integrated amplifiers or little desktop amplifiers with volume control, Amir usually sets the volume to give constant 29dB gain. So the power vs THD+N curve is made by varying the signal level produced by the analyzer and sent to the amplifier, not by changing the amplifier volume. If you turn up the volume knob on an integrated amplifier then the noise will be amplified.

The TPA3118 does not appear to be a very low noise amplifier. In the review of the Sabaj A1 the signal to noise ratio is about 80dB at 5 watts.

Sabaj A1 Desktop Amplifier Review

This is a review and detailed measurements of the Sabaj A1 desktop (small) amplifier. It was kindly sent to me by a member. Not sure if it is still in production but you can still find it for $60 or so. Main interest in testing this unit is the use of the TPA3118 class D amplifier which is...

www.audiosciencereview.com