top of page
Search
  • Writer's pictureGianluca Sperti

The subtle game of balancing

Updated: Mar 20, 2018



In audio, balance is always associated to specific circuit topologies and to fierce debates around the quite sterile technical subject if a balanced circuit works or sounds better than an unbalanced circuit. Interesting stuff? Maybe it is or maybe it is not but I am not going into this. I am going into how a proper balance of ideas must be reached to develop a new product and I’ll explain how the smart amplifier was born under this perspective.


The ancient guiding principle from Occam I have always followed - embrace the simplest solution - has certainly been the number one design criteria also for the smart amplifier. The initial and moving idea was to have an easy to use, complete system sporting the dynamic of much more expensive and complex audio systems. Something capable to connect the music on your preferred LP records or files stored in your NAS to your headphone or a pair of decent speakers to make your feet tap on the floor.


That is the idea. Simple, isn’t it?


Once you have the idea, you start slicing the idea into thinner pieces and study every bit on its own. So I started laying down some data about headphones: how much power, how much voltage drive, how low the residual noise must be. Look at inner fidelity and you’ll quickly find the power needs of dynamic headphones are scattered all over with most of them requiring very little power but a few drawing some serious wattege to reach 90dB SPL. Ok then, this is the first balance … how much power? Low power amplifiers will require smaller heat sinking that results in a much larger freedom in designing the chassis and also the noise can be tamed without brutal force. But low power means low sensitivity headphones would be a no-go, let alone speakers. Let’s twist the balance knob towards the high power range. Then it comes to voltage drive and gain: the smart amplifier uses a unity gain power buffer to drive the headphone and moves the task to provide gain to the preceding stage. This sets the boundary conditions for the power buffer and also for the gain stage.


Things were getting more interesting. I experimented a while ago with a differential stage capable of using any tube and I loved it: I liked the idea of just swapping in an ECC82 and a few minutes later comparing it with NOS E88CC and just enjoying the sonic signatures of different tubes. I studied a lot around this topology and determined that distortion mainly depends on the bias point (ie how much current is flowing in the tube) rather than the specific tube. THD for different tubes were pretty much in the same league and once the circuit was optimized around a precise working current, it offered stable performances. This circuit forces the balance and brings the tubes to that sweet spot. Balance again? Yes, but a different one. This is again setting quite a strong direction for the smart amplifier as this stage is far more expensive than a traditional common cathode with the classic vintage sound most people associate with tube amplifiers. And this introduces a second balance knob … cost and performance.


Then came the power supply and then the RIAA preamplifier and the DAC where steering decisions were taken on the main design criteria and solutions.


So, I had a clearer picture of how each single brick of the smart amplifier was going to look, the main performance in terms of gain and power consumption, the interface to the preceding and following stage and how much of the total cost and room in the chassis they were aiming for. Were those bricks sticking together? Too much gain? Too low power? Size was ok to have everything in a normal chassis? What about directional costs? Everything looked reasonable on paper so I drove a first balance line and entered the design phase.


What I typically do is look at what others have done, understand why and what are the decisions they took while designing that particular amplifier or preamplifier, the performances and instrumental measurements, listening impressions. Then I just start from a blank sheet connecting the dots from the previous preliminary design phase for each basic building brick I identified. I turn my simulator program on, the almighty spice, and draw a first idea. Then run a couple of simulations, see what’s the power and gain structure and other more specific parameters. There is always a voice in the background asking the same question over and over … ‘can you make it? I mean, can you actually solder the parts you are thinking to use on the PCB and make that thing work as expected?’ … so I start understanding sensitivity of the circuit to parts selection and the compromises it will require to the whole project. I abandoned quite a few ideas as they were simply under powered but sounding great or overpowered but would need heavy heatsinks. This is the most creative phase, the most rewarding and the most fatiguing. Then comes the really boring and no-one- wants-to-do stuff: finalize the design of the circuit, simulate, simulate and simulate again, select parts on mouser, simulate again. Repeat the process for each block of the project: the RIAA, the gain stage, the DAC and the power supply. And then verify if the whole circuit is behaving as planned, so I feed the performance from the power supply (ie regulation and noise figures) into the power buffer simulation, gain stage and RIAA and see if performances at the output are guaranteed. The same for gain, noise, impedance mismatch, bandwidth, distortion etc… etc… etc… And all those different aspects must be in balance and close to the target performance.


Once I am happy with the simulation of the subsystems and the whole amplifier, I draw a second balance line and start with the prototyping phase. You don’t need countless hours on the CAD drawing the PCB. Prototypes are easy … you can even make ugly PCB with lower than ideal routing and messy PCB-to-PCB interconnections because you need to validate the simulation quickly and see if the design reaches the planned performance. But, boys, when the first batch of PCB’s are delivered and you start populating them with parts, you get more and more exited and you have just one and only one thing in your mind … ‘I want to listen to this thing right now’. I must confess a little secret habit of mine: when I finish with a PCB or a main part of a circuit I just leave it as it is for one day and I power it up the next day. I prefer to recheck in cold blood before proceeding.


Fast forward to the moment I turn on the first prototype. Check the voltages around, turn it off. Turn it on and check other stuff. Turn it off. Smell around and inspect parts for overheating. Fix a few things around. Turn it on and see what happens with an oscilloscope. Turn it off. Connect a speaker, volume down. Turn it on. Volume up and a big smile! How rewarding. Then I just sit down and listen to it and the better the amplifier sounds the longer I keep it hooked up. If I fall in love with the sound, I start measuring the prototype on various parameters (bandwidth, noise, in/out impedance, distortion and all the stuff you usually find on magazines). If not, I measure a few things and modify the circuit here and there. Now, after many years I know what I like and based on the experience I can spot things in a circuit that I associate to ‘good sound’ and ‘bad sound’ but this is still a very experimental process.


The prototyping phase is not over yet. Auditions and critical listening sessions allow one to collect more impressions on what is good and what simply doesn’t sound good. This is very important as measurements reassure the design intent is reached in terms of performance but there is always something they cannot capture just like the nuances of good wines cannot be chemically described. Conclusions of the listening tests may suggest a little tweaking of the circuit, tuning of bias and selection of different parts and ideally an independent panel of experts should audition the prototype in different conditions, playing different records and with different associated pieces of equipment. Once the design is frozen, I keep the prototype running for days in my main system to both enjoy the sound and stress test all the components.


The tough game will come next. Going from the ugly prototype to a fully industrialized product takes countless hours. Nerves will be attacked by requests to reduce costs, to move things around and to redesign the PCBs to accommodate the wishes of the designer and ‘his’ chassis. This is a cooperative phase were multiple minds work together to design every single detail of that thing that will be on your desk connected to your ears. The game of balancing forces pulling and pushing in opposite directions is really now. It is ridiculous how irrelevant become the hours spent to find the perfect horizontal symmetry of the input PCB, designed a few days earlier, when facing the need of a different layout of the back panels with vertical arrangement of connectors. The solution? Just rework the PCB.

So, what are the balance points we hit during the development?


Great performance but poor flexibility and match with other pieces of equipment


Great performance but costs out of control


The ideal circuit is there but it has a different than expected sonic signature


Power as much as you need but noise with the most efficient speakers or headphones


Higher gain comes with higher noise


Power flowing freely in class A but too much heat to dissipate


Better parts always come with larger footprint that just don’t fit on the PCB


Optimized PCB layout never fit into good looking chassis


Simple human interface typically has a complex control logic


Quality is no friend with weight


Best materials usually cost more than planned


Good looking chassis are difficult to manufacture


Great material that cannot be worked

47 views0 comments
bottom of page