Some questions about the process of designing audio circuits

Hi !
i understand it is a very complex topic ... but i would like to ask when you design a circuit how do you know if it will work/be stable or not ?
Recently i have read the following statement that confused me a lot ...

I believe you need to get soldering and LISTENING, simulations tell you zip about the real world

ok ... but before actually soldering and the listening a prototype there must be a way to check if a circuit will work fine or not ...
Sometimes i see some ideas of circuits and then an expert jumps in saying that it will not work ... how can it be sure about that ?
Thanks for any advice.
 
Last edited:
Simulations work great provided the models for the various active components (transistors, opamps etc) are close enough representations of the real thing and you know how to account for parasitics (stray capacitances, inductances) if you're doing anything high speed. Simulations may get rather limited where wound components are used (inductors, transformers) as they have frequency-dependent losses which aren't normally accounted for.
 
Simulations give you an idea of how a circuit will perform electrically, and you can find the best behavior,

Hi ! then simulation is a very important tool to check important parameters like stability, distortion, etc.
actually I would like to start there with very simple circuits and then see ...

what not necessarily is the "best sounding" one. Electrical characteristics and good sound are not the same thing, and some people says they are opposite

Ok i see. What is not clear to me is taken a circuit good at simulation then a prototype is built ... then there is the listening test.
If the listening test fails to convince, what the designer will do ?
which is the next step ?
 
Simulations work great provided the models for the various active components (transistors, opamps etc) are close enough representations of the real thing and you know how to account for parasitics (stray capacitances, inductances) if you're doing anything high speed. Simulations may get rather limited where wound components are used (inductors, transformers) as they have frequency-dependent losses which aren't normally accounted for

Hi ! thanks a lot for the precious explanation. However simulations i understand remain a very important step.
But let's assume that a prototype passes the simulation and measurements but fails at the listening test ... what to do at that stage ?
what can be done then in order to increase the sound quality of a prototype ?
 
Last edited:
You seem to be once again asking for simple recipes or algorithms for solving potentially complex problems.

What you do to solve a listening test problem depends on what the problem is, and how well you understand the circuit you are dealing with.

Consider this analogy:
When first learning calculus we are taught that there are various techniques for performing differentiation - you just apply the appropriate technique and the answer pops out. Many people are then surprised to find that you cannot do integration in the same way. There are various techniques, but the only way to find the correct one (if one exists at all) is trial and error (which can be shortened by experience and lucky guesses). There is no rule you can follow to do integration. Sometimes you have to apply several techniques in the right order; other times you can use either of two quite different techniques; sometimes no technique works and the function you have is non-integrable by analytic means. You are asking a 'differentiation' type question about an 'integration' type problem.
 
Hi ! then simulation is a very important tool to check important parameters like stability, distortion, etc.
actually I would like to start there with very simple circuits and then see ...
Not necessarily. Many decades Engineers work without simulators. Also simulations not always give the best results, because some parasitic effects not well take into account in simulators: for example, the large electric field near the plate of final tubes to other ones in the circuits, and leak magnetic fields around inductors and traffos capable of coupling to previous stages.
Ok i see. What is not clear to me is taken a circuit good at simulation then a prototype is built ... then there is the listening test.
If the listening test fails to convince, what the designer will do ?
which is the next step ?

Listening test depends on the listener, and are subjective, so a good design doesn`t warranty a good listening test, and vice versa. I saw some people that enjoys listening almost only boom boom's in their cars.
 

MAAC0

Member
2010-05-02 10:00 pm
Say You design a pre-amp with large resistors and the simulation tells You it works fine..
Now You build the circuit and You hear Johnson noise.
Now You decrease the resistor values and the noise becomes less audible.

Build - Test - Listen - Tweak is You answer.

For example audiophile guys don't like tone controls, but I can't live without them because I was raised in an era where bass and treble were always accentuated.
 
But let's assume that a prototype passes the simulation and measurements but fails at the listening test ... what to do at that stage ?
what can be done then in order to increase the sound quality of a prototype ?

Failure at the listening test most likely means the simulation wasn't comprehensive enough - didn't involve enough stimuli or perhaps the wrong ones. Measurements in the first instance need to agree with simulation - if there's disagreement then either the prototype build incorporated some errors or the simulation wasn't close enough to reality.

As to things to improve the SQ of a prototype, normally I'd start out by wondering if the power supplies were stiff enough. Adding a cap or even a bank of caps to the supply rails potentially will give the answer.
 
You seem to be once again asking for simple recipes or algorithms for solving potentially complex problems.
What you do to solve a listening test problem depends on what the problem is, and how well you understand the circuit you are dealing with.
Consider this analogy:
When first learning calculus we are taught that there are various techniques for performing differentiation - you just apply the appropriate technique and the answer pops out. Many people are then surprised to find that you cannot do integration in the same way. There are various techniques, but the only way to find the correct one (if one exists at all) is trial and error (which can be shortened by experience and lucky guesses). There is no rule you can follow to do integration. Sometimes you have to apply several techniques in the right order; other times you can use either of two quite different techniques; sometimes no technique works and the function you have is non-integrable by analytic means. You are asking a 'differentiation' type question about an 'integration' type problem

Hi ! the question is ... are they really complex or are they made complex for whatever reason ?
Last comment i got ... it is too simple ... it cannot sound good.
Or even better ...
“Everything should be made as simple as possible, but no simpler. ”
is this simple as possible level that is difficult to determine and obsesses me.

Another really grey zone is measurements. In every other field we can measure performance ... speed ... weight ... length ... hardness and so on
Not so with sound quality ... actually units that measure impeccably sound **** ... and viceversa ...
It seem like audio design is not fully understood ... it is easier to send a man on Mars than measuring sound quality ?
or maybe like someone say ... you cannot say anything about a wine taste just on the basis of chemical lab analysis ?
 
Last edited:
Not necessarily. Many decades Engineers work without simulators. Also simulations not always give the best results, because some parasitic effects not well take into account in simulators: for example, the large electric field near the plate of final tubes to other ones in the circuits, and leak magnetic fields around inductors and traffos capable of coupling to previous stages.

Hi ok i see ... but my point is that this should be measurable in some ways.
I reject the idea of designing by ear ... noise is noise ... and we can measure it. For instance let's take a basic part ... an attenuator.
Me and a friend did some tests and found that a discrete attenuator "sounded" better than a standard pot. More definition ... better soundstage ... cleaner voices ... and so on.
Every sound characteristic was clearly better at the point that i have become a fan of attenuators made with good quality resistors (our had Holco resistors ... quite good ones)
This should be made evident by measuring some specific parameters ...
measurements and ears should agree at a certain point.
While i am willing to accept that an equipment bad at measurements can sound nice it is very difficult for me to accept the viceversa ...
Then i see another attitude among audio designers.
Instead of optimizing at max simpler topologies they abandon them to go for much more complex topologies ... like a form of audio design perversion :D
I am quite sure, but i do not have competence and time now to show this, that exceptional sound can be obtained by simple circuits optimized to the extreme in terms of values and parts quality ...
Another feeling ... as it is so difficult to optimize a simple design just think how difficult can be optimize a very complex one with a huge parts count with every part introducing some kind of variables.

Listening test depends on the listener, and are subjective, so a good design doesn`t warranty a good listening test, and vice versa.
I saw some people that enjoys listening almost only boom boom's in their cars

:D Yes ! personally i have never thought that an excellent sound can be possible inside a car ... only a dedicated and treated listening room can give a chance.
 
Last edited:
Why it should be made as simple as possible depends on your priorities

Hi ! words were not mine ... :eek:
But i still believe in instruments more than ears...
Recently i had my earing system checked ... i cannot listen above about 14kHz maybe even less i did not get the full report. They mentioned a decrease of sensibility in the upper audio range ... :(
I am not talking about subliminal by the way ...
Anyway ... strangely enough i do not see a big interest/effort in developing not traditional lab tests targeting sound quality.
Maybe that would end the game ? :rolleyes:
I would like to add one last comment ...
With analog sources everything sound from decent to very very good.
Not so with digital ... sometimes it sounds very very bad indeed.
 
Say You design a pre-amp with large resistors and the simulation tells You it works fine..
Now You build the circuit and You hear Johnson noise.
Now You decrease the resistor values and the noise becomes less audible.

Hi ! i am not competent but i understand and trust your statement completely ... but my point is that noise can be measured.
Actually instruments can even provide spectra of noise down to -140/150 dB that no human ear can listen ... :rolleyes:

Build - Test - Listen - Tweak is You answer

my final word would be ... test again :rolleyes:

For example audiophile guys don't like tone controls, but I can't live without them because I was raised in an era where bass and treble were always accentuated
i have even by-passed the balance control on my preamp ... :eek: after having read that a balance pot can decrease channel separation in many cases. Taking it out from the signal path solves the issue.
Moreover i have seen an interesting video about the effect of the CD direct option on a commercial amp ... with the cd direct a square wave was less distorted. That test would be enough for me to decide against tone control
But i want to hear the most of a recording ...
Another very interesting result at least for me is that even bad recordings sound quite better if played back on high quality systems ... this motivates me in improving my system.
 
Last edited:
Failure at the listening test most likely means the simulation wasn't comprehensive enough - didn't involve enough stimuli or perhaps the wrong ones.
Measurements in the first instance need to agree with simulation - if there's disagreement then either the prototype build incorporated some errors or the simulation wasn't close enough to reality.

Hi ! thanks for the very important endorsement of the simulation analysis ... i agree and actually simulation it will be my next main object of interest/study.
Unfortunately now time is limited by my current job ... what a sad thing having to work to survive :(
(They say ... ill common half joy. I do not buy this at all ... )

As to things to improve the SQ of a prototype, normally I'd start out by wondering if the power supplies were stiff enough.
Adding a cap or even a bank of caps to the supply rails potentially will give the answer
Yes. I guess that simulation program use models representing a standard behaviour of parts. They cannot take into account important issue like thermal effects, board construction and parts placement, RF/EMI ...
I just wanted to make an apology of lab testing.
In every single scientific field instruments have provided a huge step-up in the quality of the results.
Just think in the medical field ...
Can audio design be an exception ? :rolleyes:
 
Last edited:

indra1

Member
2010-11-05 6:44 am
Bogor
Ginetto, measurements of output voltage of amplifiers have been shown many times to be indicative to acoustic output, but never correspond linearly. Near field acoustic measurement is possible in a quiet enough lab environment but far field measurement is a difficult undertaking. Seems to me large audio companies prefer to choose a listening panel rather than go through huge investment undertaking of complete far field acoustic measurement and interpretation.

I expect significant progress on correlation of measurement to perceived sonic quality once our understanding on interactions of electronics to the acoustic side improved.