Optimizing
Optimizing for best musical
performance
Some audio electronics engineers lack an understanding of
what audio electronics is all about: to give the listeners an
enjoyable moment with the music they love. The most
common flaw is the lack of timing, and the only brand I know that
always deliver in this respect is Audio Research. To me it
seems like many audio engineers do not even understand the
concept of musical timing, no wonder their gear cannot deliver.
A second common flaw is the lack of a natural grain-free mid range. A well known high-end brand (no names) has a preamp that is absolutely terrible. One wonders if the designer ever listens to music. However, I should state that it is really difficult to get it 100% right, only one of all my designs have a near perfect midrange.
Non of these two things can be put right through measurements alone. So what to do then ? Well, continue reading and you will get some hints on how to get it right. I will describe some important parameters to play around with during design and optimization, and also explain about making a structured subjective evaluation.
The amount of negative feedback
The single most important parameter to get the timing right is by
optimizing the amount of negative feedback. Every circuit
topology has an optimal amount of negative feedback where
the musical timing is dead right. There is no absolute
truth about how much is right, in one design it is 10dB and in
another design maybe 25dB, or even zero in some cases. All
my designs are made so that the amount of negative feedback can
be changed without altering the bias currents of the different
stages, tweaking a few resistors are the powerful tool to reach
the same level of musical timing as Audio Research. The
optimum is very narrow, just a few dBīs off will reduce the
timing significantly. If you are not sure how close you are
try changing an order of magnitude up and down to see the
tendency, and take it from there.
The bias currents
The bias current of each stage needs to be carefully optimized.
In small signal Class-A circuitry the clarity and
resolution often improves if the bias currents are increased but
at the same time the smoothness is reduced a bit, so a trade-off
is necessary. You must decide what you prefer. Note
that the front end of power amps are small signal circuitry, so
it applies here also.
The transistors
If you have problems with finding the right sonic signature, play
around with the transistors. No other single component or
parameter alters the sound more than the transistors. A
jfet and a bipolar sounds VERY different, and the difference
between the two jfets 2SK30AGR and 2SK170GR are significant.
Often a jfet gives a more smooth natural sound, but a
bipolar gives a closer and more open soundstage. Again a
trade-off is needed.
Local linearization
To introduce degeneration of a bipolar transistor (by adding an
emitter resistor) will improve the resolution. When the
degeneration is above the optimum you will start to lose
dynamics, so this is the stop criteria. Noise is another
limiting factor. Even you use a jfet with a low Gm, try a
little degeneration here also - often it improves the resolution
significantly.
The bandwidth
Sometimes you can alter the brightness slightly, less than 0,2
grades on the subjective scale (more about this below), by
altering the bandwidth of the amplifier, but donīt forget about
the phase margin....
Subjective evaluation
In the automotive business there are proven methods to make
subjective evaluations in a structured way to give the
development teams feedback on the expected customer satisfaction.
One good example is the seat comfort with a number of subjective
parameters to evaluate when choosing the final design of the
seat. A person participating in a test is prompted to grade
the feel on a 10-grade scale, with 10 being "best in the
marketing segment" and 7 being "average in the
segment". When the grades fulfill the requirement
specification on all single parameters the design is finished.
It is difficult to do structured product development of
subjective parameters without a scale to discuss around, and this
apply to audio as well since listening to music is very
subjective indeed.
When I listen to gear I try to judge the impression on the 10-grade scale. The next step is the most tricky part of the development process, and that is to select the corrective action to reach a certain goal. Say that the requirement for treble clarity is 9, and your current design reached 8. Can you gain 1 grade by increasing the bias current ? Can you gain 1 grade by altering the degeneration of an amplifying stage ? After a while you gain knowledge and get a sense of what is possible with each alteration. Sometimes you instantly hear that you cannot reach the goal and you must change the design all together. As an example I know that my CD modification has one weakness in pure hi-fi terms and I cannot correct it with optimization, I really need a redesign.
Finally I want to mention a common mistake sometimes also made in automotive business, namely to ignore the flaws because you already put so much effort in the design. "It is not a problem" the designer shouts when the evaluation team says grade 6. Do not fall into this, else you will have problems to consistently deliver high class audio equipment. Be critical to your own designs.
Tweaks ?
Tweaks are a psychological phenomenon that gives its users a
sense of joy to have made an improvement at low or no cost. As
means to make people happy tweaks can be justified, but for
significant sonic improvements tweaks are not the right track to
go.
Want to try an experiment ? If you are listening to music and you feel hungry, eat an apple or something else you like. The sound will improve ! The truth is your gear is the same but you changed (you are more relaxed when you are not hungry). This psychological effect is the base for all tweaks - if you feel better it sounds better.
Try another experiment. Play a quiet piece of classical music for a few minutes, switch to loud hardrock for 2-3 minutes and then go back to the quiet classical piece. Was the classical piece more relaxed the second time ? Probably yes, but not because of some strange demagnetizing effect (!!!). The truth is your ears adapted to the higher level of the hardrock so when the level went down again your ears were more unsensitive. Nothing magic, just pure science.
The sonic impact of all tweaks in the world put together is ten times smaller than finding the best transistors for the input differential stage. I can only advice to put the effort where it counts, tweaks gain at most 0,1 grades on the subjective scale.
Big changes in the sound can instead be made by e.g. altering the speaker placement or improving the room acoustics. With improved room acoustics you can easily gain one grade on the subjective scale. How to do that is beyond the scope of this site, but I am sure you can find useful information somewhere on the net (if I stumble across something I will link it here).
Links |
All rights reserved