How does the THD change when turning down the volume?

[asr6576]

As I know, the SINAD will grow when turning down the volume because the level of the noise floor remains almost the same.
So how about the THD?
If we have a 0dBFS signal and a -100dBFS second harmonic, then what happens when the signal goes down to -10dBFS?

 

[pma]

There is no simple answer, as it depends on amplifier's topology and circuit design. Usually the distortion may get a bit lower, though noise (in relation to signal) a bit higher. The question does not make much sense.

 

[doodlebro]

Yeah I'd look at the range of possibilities across random reviews Amir has done. Some devices never over-saturate with distortion and noise while others do it as low as -20dB.

 

[asr6576]

Yeah I'd look at the range of possibilities across random reviews Amir has done. Some devices never over-saturate with distortion and noise while others do it as low as -20dB.

Can you provide a link for example? Thx

 

[restorer-john]

If we have a 0dBFS signal and a -100dBFS second harmonic, then what happens when the signal goes down to -10dBFS?

In a digital system, the distortion goes up as the levels drop. Generally the complete opposite in an analogue system.

(In a digital system, you have a finite number of values to describe the waveform. The most accurate representation of the waveform is produced with the most values. Lower levels by definition use less discrete values and therefore produce a less accurate representation of the original waveform. ie. more distortion)

 

[A800]

There are basically 3 possibilities.
The distortion will decrease.
The distortion will increase.
The distortion will stay the same.

 

[mansr]

In a digital system, the distortion goes up as the levels drop. Generally the complete opposite in an analogue system.

(In a digital system, you have a finite number of values to describe the waveform. The most accurate representation of the waveform is produced with the most values. Lower levels by definition use less discrete values and therefore produce a less accurate representation of the original waveform. ie. more distortion)

That is only true if you're not using dither. With proper dither, distortion goes away. However, since the noise level is constant, the SNR drops as the signal level goes down.

 

[Digital Mastering System]

I heard Vanderkoy himself (SP?) demonstrate proper dither and no dither in a 12bit system. With no dither 12 bit sounded awful, but with dither it sounded fine, albeit a little noisy. Remember anything recorded before Dolby (1968?) was <70db S/N, so about 12 bit.

 

[RayDunzl]

As I know, the SINAD will grow when turning down the volume because the level of the noise floor remains almost the same.
So how about the THD?

Is this question in relation to measuring the electrical signal or listening to a transducer?

 

[suttondesign]

There are basically 3 possibilities.
The distortion will decrease.
The distortion will increase.
The distortion will stay the same.

Thus sayeth the economist.

 

[Asylum Seeker]

There are basically 3 possibilities.
The distortion will decrease.
The distortion will increase.
The distortion will stay the same.

You are my nominee for the Fields Medal, sir. The insight you have provided is indeed transcendental.

 

[A800]

You are my nominee for the Fields Medal, sir. The insight you have provided is indeed transcendental.

Thanks.

 

[bennetng]

Anyway here are some of my measurements: Realtek ALC892's output on ASRock B85m Pro4 recorded by X-Fi Titanium HD PCIE. RMAA uses -3dBFS in THD and IMD tests, I played the reference signal in foobar2000, and used ReplayGain to adjust the output level. "default" means -3dBFS. As you can see at 0dBFS (+3), there is no hard clipping, but THD is higher. On the other hand the IMD+N slots don't indicate a clear relationship between signal level and values, at least within a rather narrow +/-3dB range.

@pozz feel free to add these results if you like.

 

[jasonhanjk]

Anyway here are some of my measurements: Realtek ALC892's output on ASRock B85m Pro4 recorded by X-Fi Titanium HD PCIE. RMAA uses -3dBFS in THD and IMD tests, I played the reference signal in foobar2000, and used ReplayGain to adjust the output level. "default" means -3dBFS. As you can see at 0dBFS (+3), there is no hard clipping, but THD is higher. On the other hand the IMD+N slots don't indicate a clear relationship between signal level and values, at least within a rather narrow +/-3dB range.

@pozz feel free to add these results if you like.

There are 2 unknowns. The best level THD+N for the recording and the best level THD+N for the playback.

 

[bennetng]

There are 2 unknowns. The best level THD+N for the recording and the best level THD+N for the playback.

Any test without using a standardized analyzer like AP will have these limitations anyway. Read this if you need more details:
https://www.audiosciencereview.com/...-sample-rate-is-enough.4037/page-3#post-95240

 

[JohnYang1997]

There are basically 3 possibilities.
The distortion will decrease.
The distortion will increase.
The distortion will stay the same.

So you assume it changes monotonically?

 

[A800]

So you assume it changes monotonically?

There are basically 5 possibilities.
The distortion will decrease.
The distortion will increase.
The distortion will stay the same.
The distortion will change monotonically.
The distortion will not change monotonically.

 

[JohnYang1997]

There are basically 5 possibilities.
The distortion will decrease.
The distortion will increase.
The distortion will stay the same.
The distortion will change monotonically.
The distortion will not change monotonically.

Alright