Sometimes, you learn the “proper” design principles of engineering from your electronics engineering school. Othertimes, you learn the “preferred” design philosophy from the articles of other great designers.Yet, in my mind, the most important place that I’ve learned to value great design philosophy is to let yourself become that dedicated super end user.
There are many great audio designers in the world, and there are many more great hi-end audio device users or audiophiles out there. But there are not many people who are lucky enough to be both. Having both mindsets will allow you to discover and see a far more balanced point of view. For example, if you are truly just a die-hard objective-only engineer, you might be thinking to yourself that it’s hard to prove a capacitor will change the sound characteristics. Even though there are some minor distortion distinctions, the average human ear won't be able to notice the difference.
On the other hand, if you are only an audiophile with demanding ears, sometimes you might be overly sensitive to think how the shape of the cable connectors will change the sound or how the PCB material will affect the sound quality without actually changing it.
Being a true “care-for-every-detail audiophile” and also a “prove-it-to-me engineer”, I poke around, experiment with, as well as research and mod a multitude of great audio devices to find how to make them sound at its absolute best. Half of the reasoning behind this is because I want get better music, the other half is because the journey is incredibly rewarding and interesting. With this, I would love to share some of the results and principles with you. Hopefully, we can also bounce around some ideas and design principles for future, especially there is more and more high definition music source out there. Some directions might need to change.
Where to get the most improvement of Sound Quality?
More money to buy expensive stuff? Well… with limited resource of money. That is not a good route although if you really need something good, at least you need to pay a reasonable cost. Experience has taught us there is a sweet spot. And after that, you will get the diminishing return on your investment of gear. So what you could do is pick the good device that has nice features with something you love in higher priority. For example, in my mind, buying a simple, but nice, topology Pass Labs XA power amp is definitely more rewarding than buying a Halcro amp that has extremely low distortion numbers.
As the designer, the same principle also applies. You have your design and your efforts budgeted, from there you decide where you put your efforts and best components on. Following, by orders, is usually the place I would focus and make my top priority.
Top 1: Circuit Topology
No matter how expensive each component is, if they are not able to work together perfectly or have synergy, the device won’t sound good. Dual mono, full balanced topology is one of my long time favorites. With a solid design in circuit topology in the beginning, it is a lot easier to make the system sound amazing.
On the other hand, single-ended has its strength as well, especially in tube circuit. Also, with or without negative feedback in which stage is another good thing to consider. How and where you put your power regulators and how you increase the slew rate is another. All these considerations MUST come first before you finalize the circuit topology.
Our design principle is simple: Always find the topology that fits the requirement best first! Sometimes, one simple change of topology will make the whole thing upside down, or vise versa.
Top 2: Lowest Sound/Noise Floor
A good system/audio device should be able to playback music details “vividly”. If you cover your mouth up and sing a song, that is exactly how a lot of low end system sounds when you hear their playback. To design a device have the lowest noise floor, so it won’t cover or bury any tiny music details here is always our second design goal.
In order to get this right (it’s harder than it looks like) you need to have a very quiet power supply circuit as a good beginning, also picking the right active components is crucial. From there experimenting with, testing and choosing key passive components is where I focus next. Especially if you have a DC blocking capacitor in signal chain, that is the place usually a lot of music details got swallowed. You can see one of the causes, here: http://en.wikipedia.org/wiki/Dielectric_absorption
Since digital to analog converters do not ‘create’ the music details, the best thing we could do is to “keep it”. Besides analog domain, in digital domain, Light Harmonic is one of the few companies not to use digital filter, insist on bit perfect input when we use R2R structure. On delta-sigma conversion circuit, usually I will pick the digital filter which won’t smear the details.
The other thing I want to distinguish a bit here too. Noise floor and Sound floor. Noise floor is straightforward, that means the intrinsic noises when even there are no music signal input. Usually, we are talking about the noise level below 0.1mV. But sound floor is a little bit more in my definition, it includes the noise floor, also the time jitter floor. We need to minimize the time error. This is important for clear and natural extended sound stage. It is important for you to hear that violin’s vibrato. Our design philosophy is not only to make sure the noise is lower enough, we want to make sure time error of each playback notes is low enough as well.
Again, frequency domain is important. But time domain, plays very important roles too!
Top 3: Active Components
After we have defined a solid topology and a superb sound floor, we then move on to picking the active components. In the digital domain it usually involves tons of tests! For example, when I designed Geek Pulse DAC, I tested roughly around two dozen clock distributors to make sure the transmission jitter is minimized. In the analog domain, my goal was to pick the components which are very linear in the beginning without you using other local or global feedback to correct.
In analog domain, we usually pick the one which is most linear in proper working condition. Good active components usually won’t be cheap. For example, the famous Toshiba JFETs that are out of production. So sometimes you need to go back to the top #1, pick the right topology again so you won’t get stuck with certain components…. Who said design was easy?!
In DAC design, usually you pick your DAC IC or design your own module in early stage. When we worked on the first prototype of Geek Out DAC, we used BB PCM1795 which sounded very nice. But we found one problem, the distortion is not good when it play PCM over 192K. Also, its jitter performance is not nearly as good as ESS9018. In the end, we made a big design change, switch to ESS9018. This IC is more sensitive to power supply with its performance, and the I2C settings inside is complex and sometimes troublesome. However, when you get it right. Sounds right!
Top 4: Power Supply Circuit
Power supply is kind of fundamental of everything. So maybe put it on No.4 is not totally right. Anyway, power supply is part of the whole amplification system, it is super important for any clock signal output/input, it is important for you to get the noise floor low. It’s the fundamental.
I still remember how I was amazed, about 10 years ago, by just changing one power supply in the oscillator circuit inside one of my higher end CD players. All of sudden, sound stage, smoothness, liquid warm mid and solid bass all appeared! It was seriously that amazing. One power supply circuit, which cost maybe $30 in material cost, definitely will kill $3000 retail priced analog cable’s improvement or even more.
In high level device, when it has the space, my preference is to use the discrete shunt regulators in very close distance. Why? Shunt regulator’s output part is working in Class A zone, which is always conducted. It isn’t like the normal serial regulator that will switch off when the circuit does not draw power. Also, shunt regulators stabilize the whole current drawing, the generated heat is constant which is another benefit.
Also, another principle of ours is to use more regulated power. Usually you will see two digits of power regulators inside our devices. One near the major MPU, several near DAC IC or module, dedicated pairs of each channels analog output, dedicated pairs of amplification. Some for accessories here and there… Why? Local regulators will give you the best quality and lower impedance of power supply. Also depending on which type of circuit you need to power up, you could pick the right regulators for it. Digital domain is definitely different than what you need in the analog domain. Sometimes, even the requirements are opposite.
In analog output stage, I also believe bipolar power supply is essential especially when you use complementary active components. Again, the topology of the power supply is an important part, although usually end users can not see it and vendors won’t provide the information on it.
One more thing I want to say here. Adding more and more capacitance in power supply filter caps won’t necessarily give you better sound. But the quality of power filter cap do have the effects on sound quality.
Well, before this gets too long. I think I would like to split the remaining parts into another blog.
In the next part, I’ll continue onto other important design considerations such as passive components, PCB layout and digital domain related design principles.