The volume pot - The hidden villain of preamp

[dirkwright]

Did you consider this arrangement? I only put up the SNR graph for entertainment purposes. Obviously, noise would be much worse than shown. The pot should be a wire wound type, but they are expensive.

 

[popilin]

OK, then I hope you realize that your spice models have limited utility when it comes to modeling circuit noise. Generally, it's best to use the lowest value potentiometer (or resistor) that you can get away with in any particular circuit if you want to minimize noise.

This thread is not intended to be a treatise on noise, its only contribution, if any, is to put the volume pot on a place where it does the less harm.

I already did say this somewhere, but I repeat it again

Valves are high impedance devices, instead of with semiconductors, you can't put ridiculously low value pots, especially on preamp input where them affects input impedance.

Even more, because of dielectric anisotropy, the constitutive relation

D = ε E
​

Needs that ε must be a tensor, and the curve D=f(E) is called dielectric hysteresis curve, then, in order to preserve linearity we must avoid the use of enormous capacitors, so a 100 K volume pot is a good compromise between noise and capacitor size for the proposed scheme of post#1.

Allowing a gain stage to be exposed to the full input voltage from a source such as a CD player is bad engineering practice because the distortion will be far higher and there is the potential for overloading the gain stage. This is why the volume control is at the front of the preamp.

As a designer, you must know voltage levels you work with, this is why you can compensate CD input level with a resistive divider, as a CDP/DAC has very low output impedance, this is the best place to put very low value resistors, simple right?

 

[popilin]

You'll have to talk to Merlin about whether or not "factory conditions" plays a large enough role to invalidate his equations. I have the article but I don't have time to find a particular reference to this issue in there right now.

Also, don't take everything you read in a book as gospel. Use your head and question everything. Authors are merely people who were able to convince a publisher that their work was worthy of publication. It does not mean that everything they say in their work is the absolute truth, or the be-all, end-all of discussion about the subject.

If you need a "gospel", here it is, one of our forum superheroes, the reference and the last resource for almost everything.

Flicker noise can't easily be calculated as it varies so much from sample to sample; it has to be measured.

 

[dirkwright]

Yeah, I'm not responding to this thread any more. I don't think you're interested in what I have to say. Good day.

 

[sbrook]

Harking back to an early post in this thread by ruffrecords where he commented on the concept that the goal is (by my interpretation ... it was lots of posts ago!) to accurately reproduce the original, not to outengineer each other to no effect.

We all tend to forget the one part of the listening chain that can make all this one-upmanship quite pointless ... It's the one part we can't do a lot about! Our ears and sinus cavities.

Most people's ears vary from hour to hour, day to day, year to year ... So much so that the chance of reproducing what you heard in your brain in a studio vs what you heard in your brain in your home two quite different things. My ears are particularly bad for that, even though as I near retirement age, I've lost very little top end. Some days I suffer from horrible "noise" ... the doctors call it tinnitus ... to the point it drowns out all manner of superfluous sound. My sinuses have different amounts of yuck in them which varies the resonances in them which can actually impact my listening pleasure and may actually cause "buzzing" at selected frequencies. All quite normal for most people according to the hearing specialist doctor.

So chasing down that last bit of noise, that last nuance of phase shift is a pretty pointless exercise when your own ears can upset the applecart. It's a part of the diminishing returns paradigm.

 

[Merlinb]

Even more, because of dielectric anisotropy, the constitutive relation

D = ε E
​

Needs that ε must be a tensor, and the curve D=f(E) is called dielectric hysteresis curve, then, in order to preserve linearity we must avoid the use of enormous capacitors,

Not sure what you're getting at here. It is well known that capacitor distortion is minimised by minmising the signal voltage across the capacitor. Therefore, enormous capacitors are actually better. Even the worst capacitor dielectric will introduce unmeasurable distortion in the audio band if it is enormous enough... Of course, they may be other reasons to avoid large capacitances.

 

[sbrook]

Pardon???? How does putting an enormous capacitor minimize the signal voltage across it?

And then you say that "even the worst capacitor dielectric will introduce unmeasurable distortion" ... That's would be like saying a capacitor in my amplifier remembers the entire 1812 Overture. And you'll ask what do I mean? And I'll say "Prove it doesn't"!

If it's unmeasurable how do you know it's there? You can hear it? I can hear the 1812 from my speakers ... even though they're unplugged!

Sorry but that just makes no sense to me!

 

[ruffrecords]

Pardon???? How does putting an enormous capacitor minimize the signal voltage across it?

Basic ac circuit theory. Any coupling capacitor forms a potential divider with the source impedance driving it and the load it is connected to. The larger the value of the capacitor, the smaller its impedance and hence the smaller the signal voltage across it.

Cheers

Ian

 

[sbrook]

D'Oh! Of course. I wasn't thinking that way! (puts his dunce cap on with the "of course I knew that" on it! )

Still it doesn't eliminate the second part of my argument. If it's not measurable is it there or is it significant.

 

[Merlinb]

If it's unmeasurable how do you know it's there?

You start with a smaller cap so it is measurable, then gradually use bigger and bigger capacitances. You will see the distortion trace slowly disappear off the edge of the graph as bigger capacitors are used, so when you get to really big capacitors it is safe to assume it is 'there' but buried below the noise floor.

 

[sbrook]

If you can't measure it, then your ear certainly isn't gonna hear it.

 

[Merlinb]

If you can't measure it, then your ear certainly isn't gonna hear it.

Yes. That's the point of using the enormous cap...

 

[popilin]

It is well known that capacitor distortion is minimised by minmising the signal voltage across the capacitor.

Of course, if you minimizes the signal voltage applied across the capacitor, you are minimizing the applied field E, then the dielectric hysteresis loop D=f(E) will be thin, the area inside it will be small too, and the device results quite linear.

Therefore, enormous capacitors are actually better.

No, this is a case on bigger isn't always better, let me explain

Can be proven that for two conducting parallel plates, filled with a dielectric of "constant" ε, capacitance is

C = (ε A) / (4πd)
​

Where
ε = Dielectric constant of insulator [cgs] dimensionless
A = Average winding area [cm²]
d = Insulator thickness [cm]
[C] = cm, 1cm ≈ 1.113 pF (I love cgs units)

For a given dielectric (ε), to increase capacitance, because d is already decreased, remember that manufacturers are stingy, so you must increase A, so you must increase the volume of the dielectric, then dielectric hysteresis losses increases, then the area inside the loop is bigger and this way you can destroy linearity.

The decrease on applied signal voltage varies linearly with capacitance, but losses doesn't, actually them varies exponentially with the volume of the dielectric, then exists a theoretical limit beyond that bigger is not better.

Even the worst capacitor dielectric will introduce unmeasurable distortion in the audio band if it is enormous enough...

No, there are horrible dielectrics out there, e.g. ceramic capacitors.

This can illustrate the point better than me.

The "Sound" of Capacitors

 

[hazard500]

Since this is a DIY thread I will ask a question - not a thread cap, but a serious question. Why on earth does any self respecting DIY person need a pre-amp?

I have designed and built my phono stage. It sits directly underneath my turntable and it outputs 2 volts. I own a Cary multi player (CD/SACD/DVD-A). It also outputs 2 volts. My hard drive is full of hi res down loads (I am a customer of HDTracks) and needle drops of my LPs. My sound card (24 bit/ 192k) also puts out 2 volts.

So my 'pre-amp' is rather simplistic. It consists of two source selectors (one for each channel) 6-position 2-way switches (my phono stage and sound card and power amps are balanced so I need 2 way switches). From there the output goes to my DIY mono block power amps. There is a balanced, switched attenuator directly on the input of my power amps (although I like DIY, I actually bought these from John Broskie. Build quality is superb).

SO what is the noise in my pre-am? Zero. What is the distrotion of my pre-amp? Zero. How many reasons do I have to build a pre-amp? Zero.

Now my solution is not for everyone. My wife hates that you have to walk to each mono block (each one is situtated directly behind it speaker) to change the volume. But she's not into DIY. She's into Downton Abbey and Supertramp. But if you are reading this thread I assume that you are into DIY and I ask you the question - why do you need a pre-amp? I do have a degree in electrical engineering and can write a text book on pre-amp design if I want to, but the true solution is this - the best pre-amp is no pre-amp.

 

[sbrook]

And at what level does your ear actually HEAR this negligible distortion? Given that the ear is a very imperfect audio instrument.

Hazard ... I think you have a preamp ... it's just spread across different devices.

One of the normal uses of a preamp is to take signal sources at various levels and different impedance levels and bring them to a common denominator before feeding them into the amplifiers feeding your speakers. In days of yore, many devices actually fed raw signal levels to the preamp where either multiple interface amps, or multiple correction networks defined the output to be linear and a comparatively constant amplitude.

 

[popilin]

Where
ε = Dielectric constant of insulator [cgs] dimensionless
A = Average winding area [cm²]
d = Insulator thickness [cm]
[C] = cm, 1cm ≈ 1.113 pF (I love cgs units)

Erratum

ε = Dielectric constant [cgs] dimensionless
A = Plate area [cm²]
d = Dielectric thickness [cm]

That is !

 

[fdegrove]

Hi,

Of course, if you minimizes the signal voltage applied across the capacitor, you are minimizing the applied field E, then the dielectric hysteresis loop D=f(E) will be thin, the area inside it will be small too, and the device results quite linear.

Conversely, as MerlinB explained, by increasing C in the voltage divider you automagically reduce the field E as well. Hence the DA effect is reduced making the cap less audible.
This seems counter-intuitive but that's how it is.

For a given dielectric (ε), to increase capacitance, because d is already decreased, remember that manufacturers are stingy, so you must increase A, so you must increase the volume of the dielectric, then dielectric hysteresis losses increases, then the area inside the loop is bigger and this way you can destroy linearity.

Another counter-intuitive measure to reduce audibility of a cap is to use one with a higher voltage rating.

If you can't measure it, then your ear certainly isn't gonna hear it.

If measurement tools were designed the way the human brain works (ears are just mechanical transducers) then perhaps more audible differences would be measurable.

Ciao,

 

[fdegrove]

Hi,

Why on earth does any self respecting DIY person need a pre-amp?

You've just provided the answer to your own question.

I have designed and built my phono stage. It sits directly underneath my turntable and it outputs 2 volts.

Now imagine you do not have this phono pre with its built in line preamp, but a CDP instead. You'd still want a preamp to set the volume, no?

Ciao,

 

[popilin]

Conversely, as MerlinB explained, by increasing C in the voltage divider you automagically reduce the field E as well. Hence the DA effect is reduced making the cap less audible.
This seems counter-intuitive but that's how it is.

This is not counterintuitive at all, that's why I said

The decrease on applied signal voltage varies linearly with capacitance, but losses doesn't, actually them varies exponentially with the volume of the dielectric, then exists a theoretical limit beyond that bigger is not better.

Making a great simplification, ignoring dielectric anisotropy, we can consider

ε = ε' + i ε"
​

Then, dielectric loss

P = Po exp (- δ k z)
​

Where

tan (δ) = ε" / ε'
​

BTW, δ can be found on capacitor datasheets.

Another counter-intuitive measure to reduce audibility of a cap is to use one with a higher voltage rating.

Maybe it is not counterintuitive, if you can explain why, we can discuss it, if not, it will stay in the nebula of "is said out there"

 

[gingertube]

Since you are talking about capacitors - my experience (FWIW)

Once you restrict your discussion to film caps (no ceramics or electrolytics which both have crap dielectric absorbtion) then the most significant problems with capacitors can be traced to mechanical stuff, compressability of the dielectric etc. The problems manifest themselves particularly due to hystersis effects as the electric field reverses with signal.

If you design such that there is always significant DC bias across the cap, such that the electric field never reverses (dielectric is always stressed in the same direction), then you will not be able to tell any audible difference between a good but garden variety $1 polypropylene and a $100 "audiophool jewelry" capacitor.

If you have no DC bias across the cap and the electric field is constantly reversing with signal then any capacitor "warts" will stand out out like "canine testicles".
Redesign the circuit to make sure you have a DC bias on the cap sufficient to accommodate the full peak to peak signal swing.

Cheers,
Ian