MYTH: "Your signal-to-noise performance is determined by the first gain stage."

[Hayabusa]

See my reply to RexrothPigeon.

​

I saw it, I dont get it, it seems we all agree that SNR can only get worse (lower SNR) in next stages?

 

[GK.]

I saw it, I dont get it, it seems we all agree that SNR can only get worse (lower SNR) in next stages?

I am trying to get people to properly appreciate how the stage(s) following the volume control attenuator can start to negatively impact the SNR when you start to turn the volume control down from maximum.

If you are at all serious about low-noise analogue signal chain design, then this is something which you can't ignore to take for granted.

 

[Hayabusa]

I am trying to get people to properly appreciate how the stage(s) following the volume control attenuator can start to negatively impact the SNR when you start to turn the volume control down from maximum.

If you are at all serious about low-noise analogue signal chain design, then this is something which you can't ignore to take for granted.

I fully agree. only in special cases like coherently combining multiple channels you could enhance SNR. So mixing a L+R to mono could in theory positively effect SNR.

 

[kyuu]

Any of the members here range from technically adept to subject matter experts. NTK in particular is definitely quite knowledgeable in this field. The fact that nobody seems to understand what you are getting at should tell you that you are either making a fallcious argument or you aren't getting your point across well if in fact you do have one that everyone seems to be missing.

If you don't care if anyone understands what you're trying to get across, then why make the thread? I came in expecting to learn something but instead I'm merely annoyed by your behavior.

 

[solderdude]

Consider the following example (which is just one amongst countless hypothetical constructions):

We have an audio system and the majority of the system gain comes from the first stage - that being a MM phono pre-amplifier. This phono pre-amplifier has a voltage gain of 40dB (that is times 100) and an input-referred voltage noise of 3.5nV rt/Hz (for a [unweighted 20kHz BW] S/N ratio of about 80dB ref. 5mV). So, fairly typical figures for a good design.

That 3.5nV rt/Hz of input-referred noise is amplified to a whopping 350nV rt/Hz signal at the output of the phono pre-amplifier by its 40dB of gain.

Yeah, not really simply because of the RIAA correction which basically provides a lower gain for the higher frequencies.

350nV rt/Hz is obviously a huge amount of noise that will easily swamp out noise contribution of the following stages....

except that would only be there for the lowest frequencies. You see the downwards slope ?

Next thing to realize is the surface noise (also equalized) and noise in the recording which also contributes to the noise heard (not equalized)

BUT


Well, that 40dB of attenuation will effectively nullify the gain contribution of the phono pre-amplifier.

No the signal + noise of the input stage is lowered in total so the S/N ratio of the input stage remains the same.

That volume control attenuator doesn't magically attenuate just the music signal whilst ignoring the self-generated noise present at the output of the phono pre-amplifier - it attenuates both signal sources equally.

Correct so the S/N ratio of the input stage is not altered in any way.
In fact when the volume is turned up the S/N ratio is the same as from the input stage which thus limits the S/N ratio.

What you mean is that when the volume is lowered the signal + noise is lower which means the total noise is less and might disappear in the audibilty threshold of the human hearing and it may seem that the S/N ratio has improved. It didn't and is still determined by the pre-amp + self noise of the following stage.
That self noise should be ,depends on sensitivity of the transducer and hearing (age), below the hearing threshold.

Now if that volume control potentiometer is followed by a moderate amount of active signal processing circuity such as cascaded active crossover and/or tone-control/equalisation filter stages, then these are almost certainly going to be the dominant source of self-generated system noise almost all of the time by maybe an order of magnitude or even two.

Yep assuming the volume control is turned down the actual noise (which will differ in spectral content and be white noise which is more audible).
It will be dominant until the volume control is turned up to the level where the noise floor of the pre starts to dominate.
The thing is that the noise spectrum differs so will become more and more pink noise alike which is less objectional and less audible (equal loudness contours).

Well that is my rant for this evening.

Thank you for the rant.

The S/N ratio of the first stage starts to matter when the volume is turned up to such a level that it becomes dominant over that of the following stage.
In this particular case the spectrum of the noise is also important... pink vs white.

 

[antcollinet]

Here is a really basic example. Which stage is negatively impacting the signal-to-noise ratio the most? (specified noises are unweighted, input-referred).


View attachment 499922


At what level of attenuation is the contribution of the two stages equalized?

Now what if there is an active filter stage (tone control or active crossover) added between the volume potentiometer and the power amplifier, which will typically have much higher input-referred voltage noise than 10nV rt/Hz?

Here is another problem with your argument. By adding a 40dB attenuator to the output of stage one - it is no longer a gain stage. so of course it is not having as significant an impact on system noise.

It is also unrealistic. 40dB of attenuation in my system results in a volume level that is not even background music - it is the sort of level. you'd go to so you could have an "important" conversation without the sound interfering at all.

In my system, low background music for vinyl is around -20dB, and normal listening at about -10dB. So that 5mV noise on the output of the attenuator would become 50, to 158mV, which sort of changes the significance.

 

[GK.]

Yeah, not really simply because of the RIAA correction which basically provides a lower gain for the higher frequencies.

You're overthinking it. I specified input-referred noise. This means that the noise was measured at the output of the amplifier and then referred back to the input by dividing by the nominal gain. For a phono per-amplifier the nominal gain is universally specified at 1kHz.

If the phono amplifier has a specified nominal gain of 40dB and the input-referred noise is 3.5nV rt/Hz, then that means that a noise signal of 350nV rt/Hz was measured at the output.

That output noise signal is all that matters. At 1kHz, with a signal input of 5mV, the output voltage will be 500mV (40dB gain). If that 350nV rt/Hz was measured unweighted (20kHz BW), then the unweighted SNR of the stage is 500mV / (SQRT(20kHz)*350nV) = 80dB.

 

[solderdude]

If the phono amplifier has a specified nominal gain of 40dB and the input-refereed noise is 3.5nV rt/Hz, then that means that a noise signal of 350nV rt/Hz was measured at the output.

No you are under-thinking this.

The gain is specified at 1kHz but the gain is higher for low frequencies and lower at the (more audible as hiss) higher frequencies affecting audibility and spectrum of the noise.
That noise is simply the back ground noise of the circuit and what you see in the picture I posted above.
The actual S/N ratio is determined by the output voltage of the used cartridge.
This happens to be the first stage as the later stages have much lower gain and much lower self noise.

So the S/N ratio is determined by the input voltage of the circuit and the input voltage it receives when the volume control is turned up in such a way that the noise floor of the following stage is no longer dominant.
Fortunately this self noise is much lower and in in the majority of cases (depends on the sensitivity of the transducers, human hearing capabilities) below the human threshold.

 

[GK.]

No you are under-thinking this.

The gain is specified at 1kHz but the gain is higher for low frequencies and lower at the (more audible as hiss) higher frequencies affecting audibility and spectrum of the noise.
That noise is simply the back ground noise of the circuit.
The actual S/N ratio is determined by the output voltage of the used cartridge.

I know that the frequency response isn't flat and it is completely irrelevant!

If I tell you that the 40dB MM phono stage has a measured unweighted input-referred voltage noise of 3.5nV rt/Hz, then that means that it measures 350nV rt/Hz at the output with the input shorted. In a 20kHz noise bandwidth that is 49uV RMS. That delivers a SNR of 80dB with a 5mV 1kHz excitation.

 

[solderdude]

So what ?

The signal to noise ratio is important and that is cartridge dependent.

How much of that noise you will be hearing is another story and depends on the desired volume.

The self-noise of a well designed pre-amplifer is low enough to only become audible at a high volume setting and can be (and in most cases is) much better than that of the actual S/N ratio of the signal picked-up by the cartridge.

What is important are not the numbers but perception of noise (and thus also spectrum).

What's the fascination with the numbers ?

Below a certain volume control position the actual S/N level will be determined by the self-noise of the amplifier.
Above a certain volume control position the self-noise of the pre-amp will become dominant and determines the system S/N ratio.
The audibility of it is another matter (certainly when the needle is dropped).

 

[GK.]

Here is another problem with your argument. By adding a 40dB attenuator to the output of stage one - it is no longer a gain stage. so of course it is not having as significant an impact on system noise.

It is also unrealistic. 40dB of attenuation in my system results in a volume level that is not even background music - it is the sort of level. you'd go to so you could have an "important" conversation without the sound interfering at all.

In my system, low background music for vinyl is around -20dB, and normal listening at about -10dB. So that 5mV noise on the output of the attenuator would become 50, to 158mV, which sort of changes the significance.

I used 40dB in the example to illustrate the concept because at that level with some simple sums it is clear that the input stage is no longer the systems dominant noise source by a considerable margin. It is therefore factually incorrect to ascribe the first stage to be the dominant source of the systems self-generated noise.

If you are designing a low-noise signal chain and you actually want to keep the first stage solidly noise-dominant for most practical use cases then it should be clear that the noise performance required of the stages following the volume control attenuator do start to become quite stringent at much less than 40dB of attenuation.

 

[GK.]

Below a certain volume control position the actual S/N level will be determined by the self-noise of the amplifier.
Above a certain volume control position the self-noise of the pre-amp will become dominant and determines the system S/N ratio.

I can't believe that we have come this far. Now explain to me the controversy.

 

[Hayabusa]

I can't believe that we have come this far. Now explain to me the controversy.

guys! get a room!

 

[Waveform Fidelity]

Why would you build a fixed gain stage and use a potentiometer to control the volume, rather than making the volume control variable gain? I think a lot of folks think volume controls are designed as a simple pot on the output of a fixed gain stage - they are not… at least not in high performance designs…

 

[solderdude]

I can't believe that we have come this far. Now explain to me the controversy.

controversy is perception, hearing threshold, sensitivity.
Numbers beyond a certain point have no relation to what is perceived.
They are just numbers.

 

[solderdude]

I think a lot of folks think volume controls are designed as a simple pot on the output of a fixed gain stage - they are not

They are in the vast majority of pre-amp cases or the potmeters are in front of fixed gain stages (integrated amps, headphone amps).
Volume controls (when used as a voltage divider) should be buffered, why not use that to add a little gain.
Just not in input stages of wide input range pre-amps as used in mixing consoles etc where the input must be able to handle everything from small microphone signals to line level sources without clipping the input stage.
There are trade-offs, certainly when it comes to power amps where instability is right around the corner when applying variable gain.

 

[antcollinet]

It is therefore factually incorrect to ascribe the first stage to be the dominant source of the systems self-generated noise.

The discussion is about gain stages. By attenuating the output of the first stage by an amount equal to it's gain - it is no longer a gain stage.

 

[GK.]

The discussion is about gain stages.

The discussion is about the general applicability and veracity of the popular belief quoted in the thread title.

By attenuating the output of the first stage by an amount equal to it's gain - it is no longer a gain stage.

Like I said already:

The elephant in the room is that the volume control attenuator when introduced into a signal chain breaks down the gain distribution that previously made the first stage (well, the net sum of all stages preceding the attenuator, to be pedantic) noise dominant.

 

[antcollinet]

The discussion is about the general applicability and veracity of the popular belief quoted in the thread title.




Like I said already:

And like you are failing to understand - is that the introduction of the attenuator negates the need for the prior gain stages. If you are going to attenute thier ouput, you don't need to include them in the first place. They become irrelevant. The first gain stage is the one that actually provides amplification into subsequent stages. Due to your attenuator, yours does not.

You are also fundamentally misunderstanding the purpose of the "rule" "theorem" "rule of thumb" or whatever it is.


Simply put:
If you are designing a system with multiple stages of amplification, it is likely that given your standard amplifier design, the stages will have a similar level of noise. If the resulting noise level is too high - you are guided to first focus on the first stage, because the noise out of that is amplified by all the subsequent stages. The effect of the work you do at the front end is multiplied by the total system gain through all stages.

If on the other hand the second stage generates 20 times the noise of the first stage, but only has gain of ten - then yes - you are going to want to fix that one first.

 

[TurtlePaul]

Why would you build a fixed gain stage and use a potentiometer to control the volume, rather than making the volume control variable gain? I think a lot of folks think volume controls are designed as a simple pot on the output of a fixed gain stage - they are not… at least not in high performance designs…

For many volume controls it is desirable to have a maximum attenuation which would exceed the stability threshold of the amplifier circuit. For many amp circuits the feedback loop needs to be short and clean because it sets the performance and stability of the amp, so it can be tough to incorporate a user accessible pot. Also, a dirty pot could make an amp self-destruct.

I would wager that a majority of volume controls are either signal path variable resistor networks (pots, resistor ladders, some digital chips) or are digital domain in DACs. Continuously variable negative feedback (gain) settings are much less common.