Sub 1V SINAD?

[BKr0n]

I looked at one of the reviews a while ago ( https://www.audiosciencereview.com/...nchjim-luna-portable-hp-amp-dac-review.58272/ ) and one thing that stuck out to me was that at 50mV, sinad was way less than the best case scenario. What makes it difficult to get better sinad at sub 1V?

 

[Sokel]

I looked at one of the reviews a while ago ( https://www.audiosciencereview.com/...nchjim-luna-portable-hp-amp-dac-review.58272/ ) and one thing that stuck out to me was that at 50mV, sinad was way less than the best case scenario. What makes it difficult to get better sinad at sub 1V?

SINAD is relevant to the fundamental and at the reviews you see is expressed in dBr.

 

[voodooless]

Usually it’s noise that dominates at lower voltages.

 

[BKr0n]

Usually it’s noise that dominates at lower voltages.

Ah I see. Kind of surprised that no solution has been found for consumer audio. Then again, usually components that precise cost a few limbs and I would imagine things like heat noise would become an issue at that level as well. Definitely explains the price tag on the mola mola dac.

 

[MaxwellsEq]

I looked at one of the reviews a while ago ( https://www.audiosciencereview.com/...nchjim-luna-portable-hp-amp-dac-review.58272/ ) and one thing that stuck out to me was that at 50mV, sinad was way less than the best case scenario. What makes it difficult to get better sinad at sub 1V?

Good question!

As you lower the signal, noise measurement starts becoming the dominating factor. So imagine your test signal was -60dBFS and the device has a noise floor at -100dBFS, you only have 40dB in hand. If on the other hand your test signal is +10dBFS but the noise is still at -100dBFS, you have 110dB in hand, despite the device being identical.

It's a bit of a simplification, but as you increase the gain, or the input signal, you are more likely to encounter distortion, so distortion tends to characterize the upper limit of SINAD and noise the lower limit.

 

[staticV3]

Each dB in digital signal attenuation reduces SNR by one dB.

Amir measured 128.4dB SNR at 4.064Vrms.

To get to 50mVrms, he had to apply 20*log10(4.064/.05)=38.2dB of attenuation, which means SNR of 128.4-38.2≈90.2dB.

Lo and behold, 50mV SNR was measured at 90.8dB, which is within margin of the calculated value:


Since SNR and signal amplitude follow a linear relationship, I struggle to see the point in this exercise TBH.

All we need is the full-scale SNR and we can predict SNR at any arbitrary output level, like 1V (116.2dB), 0.1V (96.2dB), 10mV (76.2dB) etc etc.

 

[Sokel]

Simply put,the SINAD figure is not a fixed number but a ratio depending on the fundamental.
Even if noise could stay fixed for example the ratio would change because of the distance with the fundamental.

That's why you see higher voltage devices having an advantage (along with playing with AP's sweet spot near 4-5V)

 

[BKr0n]

Good question!

As you lower the signal, noise measurement starts becoming the dominating factor. So imagine your test signal was -60dBFS and the device has a noise floor at -100dBFS, you only have 40dB in hand. If on the other hand your test signal is +10dBFS but the noise is still at -100dBFS, you have 110dB in hand, despite the device being identical.

It's a bit of a simplification, but as you increase the gain, or the input signal, you are more likely to encounter distortion, so distortion tends to characterize the upper limit of SINAD and noise the lower limit.

Running out of bandwidth makes perfect sense!

Each dB in digital signal attenuation reduces SNR by one dB.

Amir measured 128.4dB SNR at 4.064Vrms.

To get to 50mVrms, he had to apply 20*log10(4.064/.05)=38.2dB of attenuation, which means SNR of 128.4-38.2≈90.2dB.

Lo and behold, 50mV SNR was measured at 90.8dB, which is within margin of the calculated value:
View attachment 420264

Since SNR and signal amplitude follow a liner relationship, I struggle to see the point in this exercise TBH.

All we need is the full-scale SNR and we can predict SNR at any arbitrary output level, like 1V (116.2dB), 0.1V (96.2dB), 10mV (76.2dB) etc etc.

Nah it makes sense. Noise can be a massive umbrella term so a standardized test method like this I think would be the best solution.

 

[antcollinet]

Ah I see. Kind of surprised that no solution has been found for consumer audio.

Because it doesn't really matter.

The signal is getting smaller, but the noise is staying the same. So the SNR gets smaller (Signal divided by noise)

However the noise, because it is staying the same, is not getting more audible. Assuming you can't hear it when there is no signal playing, or in the quiet parts of a song, then you also can't hear it when the signal level is reduced from full.

Definitely explains the price tag on the mola mola dac.

No, it doesn't. The SMSL SU-10 achieves better results at less than 1/10th the cost of the mola mola. So does the Topping D90E. The SMSL D0300 is as good as the mola mola at about 1/20th the cost. And the Topping D50 III is better again at 1/50th the cost. Yes the Mola Mola costs 50 times more for worse performance.


It doesn't need money, it just needs good engineering.

 

[MaxwellsEq]

Running out of bandwidth makes perfect sense!

You are probably aware that this is not bandwidth (which is the lowest frequency to the highest frequency and measured in kHz for audio). As you lower the signal the bandwidth doesn't change.

 

[unpluggged]

Kind of surprised that no solution has been found for consumer audio.

Which solution? If you can't hear any noise, then you don't have a noise problem, and SNR is sufficient regardless of its numeric value, so there is nothing to solve here.

And the other way around: if you do hear noise, it does not matter if the circuit has a 130 dB of SINAD ratio, it's still not transparent for your use case. So it's more a matter of gain structure and transducer sensitivity rather than that of a pure SNR/SINAD.

 

[antcollinet]

And the other way around: if you do hear noise, it does not matter if the circuit has a 130 dB of SINAD

I'd put it to you that if you have 130dB of Sinad the noise is going to be inaudble, unless you are listening at levels that will instantly damage your hearing.

 

[unpluggged]

I'd put it to you that if you have 130dB of Sinad the noise is going to be inaudble, unless you are listening at levels that will instantly damage your hearing.

Not necessarily. For instance, there are sensitive IEMs that can render the noise floor of a digitally-attenuated output audible. Of course I would not set the volume too high in this case, but the example is valid: the SINAD of the electronic device is very high, yet the gain structure makes it useless in that particular circumstance. Could be mitigated with a passive attenuator, though.

 

[antcollinet]

Not necessarily. For instance, there are sensitive IEMs that can render the noise floor of a digitally-attenuated output audible. Of course I would not set the volume too high in this case, but the example is valid: the SINAD of the electronic device is very high, yet the gain structure makes it useless in that particular circumstance. Could be mitigated with a passive attenuator, though.

If the noise floor is 130dB below full scale, you'd need to have the volume set so high that the peaks would be around 130dB before the noise reached even the threshold of hearing, let alone audible above whatever noise gets past your IEM seals. If you are doing that, your hearing is not going to last long.

 

[staticV3]

If the noise floor is 130dB below full scale, you'd need to have the volume set so high that the peaks would be around 130dB before the noise reached even the threshold of hearing, let alone audible above whatever noise gets past your IEM seals. If you are doing that, your hearing is not going to last long.

I think what @unpluggged meant is that if you have a bad case of gain mismatch and are applying like 60+ dB of digital attenuation to your 130dB SINAD DAC, then it's totally possible to hear the DAC's noise floor as hiss, while listening at safe volume levels.

 

[unpluggged]

If the noise floor is 130dB below full scale, you'd need to have the volume set so high that the peaks would be around 130dB before the noise reached even the threshold of hearing, let alone audible above whatever noise gets past your IEM seals. If you are doing that, your hearing is not going to last long.

No-no, I'm not talking about the effective SNR with attenuated volume. I'm talking about the rated full-scale SNR of the device. Let's say, it's a DAP with digital volume control and 130 dB of SNR/SINAD. A consumer might assume that high SNR would invariably mean inaudible noise, but it's possible that the output voltage reference of the DAP is not optimal for sensitive IEMs, so the noise floor will be audible. Of course one will have to attenuate the volume greatly in real usage to bring it to reasonable levels, but the advertised full-scale SNR/SINAD will not be achieved in real-world use case. That's what I meant with this scenario. And in that case, say, a device with lower SNR but with much lower output gain would make the experience much more transparent.

In short, SNR and SINAD are 1) relative quantities; 2) as specs, often stated at full-scale signal amplitudes. I hope I managed to clarify myself

I think what @unpluggged meant is that if you have a bad case of gain mismatch and are applying like 60+ dB of digital attenuation to your 130dB SINAD DAC, then it's totally possible to hear the DAC's noise floor as hiss, while listening at safe volume levels.

Exactly!

 

[antcollinet]

No-no, I'm not talking about the effective SNR with attenuated volume. I'm talking about the rated full-scale SNR of the device. Let's say, it's a DAP with digital volume control and 130 dB of SNR/SINAD. A consumer might assume that high SNR would invariably mean inaudible noise, but it's possible that the output voltage reference of the DAP is not optimal for sensitive IEMs, so the noise floor will be audible. Of course one will have to attenuate the volume greatly in real usage to bring it to reasonable levels, but the advertised full-scale SNR/SINAD will not be achieved in real-world use case. That's what I meant with this scenario. And in that case, say, a device with lower SNR but with much lower output gain would make the experience much more transparent.

In short, SNR and SINAD are 1) relative quantities; 2) as specs, often stated at full-scale signal amplitudes. I hope I managed to clarify myself


Exactly!

Gotcha

But, a bit of an edge case.

 

[DonH56]

The Signal Chain: How do noise and distortion propagate through my system?

This article is to explore how noise and distortion change through our system, and how different components can influence the final output. A common question is “If I have DAC A, preamp B, and amp C, where should I set the volume and gain controls for the best performance?” Usually this...

www.audiosciencereview.com

 

[BKr0n]

I'd put it to you that if you have 130dB of Sinad the noise is going to be inaudble, unless you are listening at levels that will instantly damage your hearing.

I would argue that at that point its more about THD, harmonics and spurs. Klippel has a little hearing test on their website you can use to test how low you can hear noise. Usually after -60dB, even with really sensitive IEMs, you really can't hear anything. Distortion from further up in the band on the other hand, can bleed into the audible spectrum if its bad enough.

The Signal Chain: How do noise and distortion propagate through my system?

This article is to explore how noise and distortion change through our system, and how different components can influence the final output. A common question is “If I have DAC A, preamp B, and amp C, where should I set the volume and gain controls for the best performance?” Usually this...

www.audiosciencereview.com

Thank you for the article! I had to have it text to speech because of how long it was, but very informative. In that same vein, would you happen to know anywhere I can read about how companies like Topping or SMSL get good noise figures? (I see a lot of acronyms, but not a whole lot of theory)

 

[DonH56]

Thank you for the article! I had to have it text to speech because of how long it was, but very informative. In that same vein, would you happen to know anywhere I can read about how companies like Topping or SMSL get good noise figures? (I see a lot of acronyms, but not a whole lot of theory)

You mean the circuit theory? At what level? That is a pretty broad subject, but any college-level (senior or grad) text will have sections discussing noise and how to reduce it.